Hemin-induced Activation of the Thioredoxin Gene by Nrf2

Thioredoxin plays an important role in various cellular processes through redox regulation. Here, we have demonstrated that thioredoxin expression is transcriptionally induced in K562 cells by hemin (ferriprotoporphyrin IX) through activation of a regulatory region positioned from −452 to −420 bp of the thioredoxin gene. Insertion of a mutation in the antioxidant responsive element (ARE)/AP-1 consensus binding sequence in this region abolished the response to hemin. With electrophoretic mobility shift and DNA affinity assays, we have shown that the NF-E2p45/small Maf complex constitutively binds to the ARE. The binding of the Nrf2/small Maf complex to ARE was induced by hemin, whereas the binding of Jun/Fos proteins to ARE was induced by phorbol 12-myristate 13-acetate, but not hemin. Hemin induced nuclear translocation of Nrf2 but did not affect nuclear expression of Jun/Fos proteins. Overexpression of Nrf2 augmented the response to hemin in a dose-dependent manner. In contrast, overexpression of the dominant negative mutant of Nrf2 suppressed hemin-induced activation through the ARE. We show here hemin-induced activation of the thioredoxin gene by Nrf2 through the ARE and propose a novel mechanism of the regulation of the ARE through a switch of its binding factors.

Thioredoxin plays an important role in various cellular processes through redox regulation. Here, we have demonstrated that thioredoxin expression is transcriptionally induced in K562 cells by hemin (ferriprotoporphyrin IX) through activation of a regulatory region positioned from ؊452 to ؊420 bp of the thioredoxin gene. Insertion of a mutation in the antioxidant responsive element (ARE)/AP-1 consensus binding sequence in this region abolished the response to hemin. With electrophoretic mobility shift and DNA affinity assays, we have shown that the NF-E2p45/small Maf complex constitutively binds to the ARE. The binding of the Nrf2/small Maf complex to ARE was induced by hemin, whereas the binding of Jun/Fos proteins to ARE was induced by phorbol 12-myristate 13-acetate, but not hemin. Hemin induced nuclear translocation of Nrf2 but did not affect nuclear expression of Jun/Fos proteins. Overexpression of Nrf2 augmented the response to hemin in a dose-dependent manner. In contrast, overexpression of the dominant negative mutant of Nrf2 suppressed hemininduced activation through the ARE. We show here hemin-induced activation of the thioredoxin gene by Nrf2 through the ARE and propose a novel mechanism of the regulation of the ARE through a switch of its binding factors.
Thioredoxin was originally identified in Escherichia coli and is known to be a dithiol hydrogen donor for a variety of target proteins. The two cysteine residues of thioredoxin undergo reversible oxidation-reduction reactions catalyzed by an NADPH-dependent enzyme, thioredoxin reductase. The thioredoxin and glutathione systems constitute the major cellular reducing systems (1). Human thioredoxin was cloned as an adult T cell leukemia-derived factor (2). Several cytokine-like factors such as 3B6-IL-1 (3,4) are identical to thioredoxin, indicating that thioredoxin plays a multifunctional role in the eukaryotic system. Thioredoxin also modulates the activity of various transcription factors, including nuclear factor-B (NF-B), activator protein-1 (AP-1), hypoxia inducing factor 1 (HIF-1), and p53 (5,6) by the regulation of reduction and oxidation (redox regulation). Thioredoxin has radical scavenging activity (7) and can protect cells from tumor necrosis factor (8), hydrogen peroxide (9), and ischemic reperfusion injury (10). Overexpression of thioredoxin in transgenic mice attenuates focal ischemic brain damage (11). Thioredoxin expression is induced in vivo in ischemia followed by reperfusion (12). These observations indicated that thioredoxin physiologically protects cells against oxidative stress-related conditions (13). Therefore, it is important to elucidate the molecular mechanism of the regulation of thioredoxin gene expression by oxidative stress.
Previous studies of the isolation and characterization of the human and mouse thioredoxin gene (14 -17) showed that there are conserved SP-1 binding motifs in the gene regulatory region of the thioredoxin gene. In an erythroleukemic cell line K562, thioredoxin was reported to be induced transcriptionally by ferriprotoporphyrin IX (hemin) 1 (18). Although heat shock factor (HSF)-2 was indicated to be responsible for this activation, the responsive element to hemin in the thioredoxin promoter has not yet been determined. There is accumulating evidence that heme (or more accurately, hemin, which is the oxidized form) itself acts as an intracellular regulator of a wide variety of metabolic pathways (19). The mechanisms of transmission of the signals induced by hemin appear heterogeneous and remain to be elucidated. A report showed that the heme responsive element of the mouse heme oxygenase-1 gene is an extended AP-1 binding site, which resembles the recognition sequences for MAF and NF-E2 transcription factors (20). The optimal recognition sequence of v-Maf (21) and NF-E2p45 (22) is similar to the antioxidant responsive element (ARE) (23)/ electrophile-responsive element (EpRE) (24). The CNC-bZIP transcription factors including NF-E2p45 (22,25), Nrf1, and Nrf2 (26 -28) form heterodimers with small Maf proteins, binding to the ARE (29). Although the liberation of Nrf2 from Keap1 and subsequent nuclear translocation of Nrf2 is reported to be an important mechanism of activation through the ARE (30), the regulation of factor binding and activation of ARE remains to be elucidated.
These previous studies prompted us to investigate the mechanism of the regulation of thioredoxin gene expression by hemin. We report here that hemin activates the thioredoxin gene through the ARE. We also showed that the thioredoxin gene is regulated through the ARE by the binding of NF-E2p45/small Maf under unstimulated conditions, that of Nrf2/small Maf with hemin stimulation, and that of the Jun/Fos proteins with PMA stimulation. The binding of these factors to the ARE correlated well with their nuclear expression pattern. We also present evidence that Nrf2 plays a role in the hemin-induced activation of the thioredoxin gene. We here propose a novel mechanism of the regulation of the ARE by a switch of binding factors including CNC-bZIP/small Maf transcription factors and the Jun/Fos proteins, depending on different stimuli.

EXPERIMENTAL PROCEDURES
Materials, Cell Lines, and Cell Culture-Hemin and PMA were purchased from Sigma (St. Louis, MO). Hemin stocks were prepared as reported previously (31). PMA was dissolved in dimethyl sulfoxide (Me 2 SO). The final concentration of Me 2 SO was kept to 0.1%. K562 (erythroleukemic cell line) cells were cultured in RPMI 1640 medium, supplemented with 10% heat-inactivated fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin in 5% CO 2 at 37°C.
Transfection and Luciferase Assay-K562 cells were transfected with luciferase reporter expression vectors using DMRIE-C (Life Technologies, Inc.) according to the manufacturer's instruction. After a 4-h incubation, various doses of hemin or control were added. For controlling the efficiency of transfection, Renilla luciferase gene expression was monitored using pRL-TK, pRL-SV40, or pRL-CMV (Promega). Luciferase gene expression, normalized by Renilla luciferase activity was analyzed 24 h later using an assay kit (Promega). Assays were performed in duplicate. Magnetic cell separation of transiently transfected cells was performed according to the instruction of manufacturer (MACS, Miltenyi Biotec, Germany).
Electrophoretic Mobility Shift Assay-EMSA was performed as described previously (35). Nuclear extracts were prepared from exponentially growing K562 cells incubated with various amounts of hemin, PMA, or control. Aliquots of 10 g of nuclear extract was incubated with 32 P-end-labeled double-stranded oligonucleotides in a binding reaction buffer containing 20 mM HEPES, pH 7.9, 0.02 mM EDTA, 14% glycerol, 1 g of poly(dI-dC), 100 mM KCl, 1.5 mM MgCl 2 , 1.2 mM DTT for 20 min at 25°C. For specificity analyses, 100-fold molar excesses of unlabeled oligonucleotide competitors were added and were preincubated for 15 min. When indicated, reaction mixtures were incubated with antibodies for 20 min on ice before labeled oligonucleotides were added.

Identification of the Hemin Responsive
Region in the Thioredoxin Promoter-To investigate the mechanism of activation by hemin, we first tested the response of the thioredoxin gene regulatory region to hemin treatment by a luciferase reporter assay. An activation of the thioredoxin gene was observed in a luciferase reporter construct containing whole thioredoxin promoter (pTRX(Ϫ1148)-Luc). To analyze the region responsible for the response to hemin, we used luciferase reporter genes containing a series of deletions of the thioredoxin promoter region. As shown in Fig. 1A, 3.6, 5.5, 2.9, or 3.5-fold induction of relative luciferase activity was observed using reporter genes (pTRX(Ϫ1148)-Luc, pTRX(Ϫ980)-Luc, pTRX(Ϫ874)-Luc, or pTRX(Ϫ463)-Luc) in response to hemin, respectively. The response was dose-dependent (data not shown). In contrast, no significant induction was observed using a reporter gene (pTRX(Ϫ352)-Luc), suggesting that Ϫ463 to Ϫ352 of the thioredoxin upstream sequence is important for the hemin response (Fig. 1A). To further determine the region required for hemin responsiveness, we used reporter vectors containing various lengths of nucleotides derived from Ϫ463 to Ϫ352 of the thioredoxin upstream sequence. Hemin activated both pTRX(Ϫ463, Ϫ352)-Luc and pTRX(Ϫ452, Ϫ352)-Luc vectors. In addition, hemin activated pTRX (Ϫ468, Ϫ435)-Luc, but not pTRX (Ϫ463, Ϫ447)-Luc. These findings showed that the Ϫ452 to Ϫ435 region is important for hemin response. Indeed, experiments using pTRX (Ϫ452, Ϫ420)-Luc showed a 25-fold response to hemin treatment, revealing that the Ϫ452 to Ϫ420 sequence is necessary for the hemin response (Fig. 1B).
Marked Reduction in Responsiveness to Hemin by the Insertion of a Mutation in the Sequence Similar to the ARE Overlapping with the AP-1 Site-In the hemin responsive region, the typical consensus sequence for binding of HSFs is not included. However, the region contains a sequence TGCTGAG-TAAC, which resembles the optimal recognition sequence, TGCTGA(C/G)TCAGCA (21) and (T/C)GCTGA(G/C)TCA(C/T) (22) of v-Maf and NF-E2p45, respectively. The sequence is also similar to the ARE, (A/G)GTGACNNNGC (23) and the AP-1 consensus binding sequence, TGA(C/G)TCA. We therefore tested the involvement of the sequence in hemin responsiveness. Activation of the thioredoxin gene by hemin was markedly reduced in a vector (pTRX(Ϫ452, Ϫ420)-M-Luc), which has a mutation in the region (Fig. 2A). As the effect of the mutation was partial, we further analyzed the hemin response, using constructs with various mutations in the hemin responsive region. A 10-fold activation was observed with a reporter vector containing wild-type ARE (pTrxAREWT-Luc) in response to hemin. The response was reduced in the pTrxAREM1-Luc, pTrxAREM2-Luc, or pTrxAREM3-Luc vector and was almost abolished in the pTrxAREM4-Luc vector (Fig. 2B). Collectively, these results showed the involvement of the ARE sequence overlapping with the AP-1 site in the response to hemin. In addition, the deletion of the ARE resulted in decrease of the basal activity of the thioredoxin promoter, showing that the ARE also contribute to the basal activity of the thioredoxin gene (data not shown).

Involvement of NF-E2p45 and Nrf2 in Binding
Complexes to the ARE in the Thioredoxin Promoter-We next analyzed AREbinding proteins by the EMSA. A constitutively bound complex (complex I), hemin-induced complex (complex II), and PMAinduced complex (complex III) to the ARE were detected in nuclear extracts of K562 cells. The binding of these complexes appeared to be specific because the binding of complex I, II, and III was abrogated by the addition of an excess amount of the wild-type oligonucleotides (Fig. 3) but not by the M4 mutant of the ARE (data not shown). Faster migrating complexes seemed nonspecific because they were not competed by wild-type oligonucleotides. Complex I and II binding showed no competition by an excess of oligonucleotides encompassing the heat shock element (HSE), and no supershift was induced by either anti-HSF-1 antibody or anti-HSF-2 antibody. In contrast, hemininduced specific binding to the HSE was abolished by an excess amount of oligonucleotide encoding the HSE and was supershifted by these antibodies (Fig. 3A). We next tested whether the complexes include proteins known to bind to the ARE or the AP-1 site. Complex I was supershifted by the addition of anti-NF-E2p45 or anti-MafK antibodies, but not by antibodies against Nrf2 (Fig. 3B), Jun, Fos (Fig. 3D), Nrf1, or v-Maf (data not shown), suggesting that complex I includes NF-E2p45 and small Maf proteins. Binding of complex II increased after treatment with hemin (Fig. 3, A and C). The hemin-induced complex (complex II) was completely abrogated or supershifted by the addition of anti-Nrf2 or anti-MafK antibodies but not by antibodies against NF-E2p45, Jun, Fos (Fig. 3C), Nrf1, or v-Maf (data not shown), suggesting that complex II includes Nrf2 and small Maf proteins. We also observed a PMA-induced binding complex to the ARE. The PMA-induced complex (complex III) was abrogated or supershifted by the addition of anti-Jun or anti-Fos antibodies but not by anti-Nrf2 antibodies (Fig. 3D), suggesting that complex III includes the Jun/Fos proteins.
Analysis of ARE-binding Proteins Using DNA Affinity Purification Assay-We performed the DNA affinity purification assay to further analyze binding factors to the ARE. NF-E2p45  1-4) or treated for 6 h with 50 ng/ml PMA (lanes 5-9) were used. Oligonucleotides encompassing AREW (lanes 2 and 6) or aliquots of 100 ng of anti-Jun (lanes 3 and 8), anti-Fos (lanes 4 and 9), or anti-Nrf2 (lane 7) antibodies were incubated with reaction mixture prior to the addition of radiolabeled probes. AREW was used as a probe. was detected in eluates from affinity beads conjugated with wild type ARE (AREW), but not from those with mutated ARE (AREM4), using nuclear extracts from untreated or hemintreated cells. Nrf2 was only detected in eluates from AREW affinity beads using nuclear extracts from hemin-treated cells but not from untreated cells. We observed MafK in eluates from AREW beads using nuclear extracts from untreated or hemintreated cells. In contrast, Jun and Fos proteins were only detected in eluates using nuclear extracts from PMA-treated cells (Fig. 4).
Nuclear Expression of ARE and AP-1 Binding Factors-As nuclear translocation is considered to be an important mechanism of activation of transcription factors including Nrf2, we tested nuclear expression of Nrf2, NF-E2p45, small MafK, and the Jun/Fos proteins. The Jun and Fos proteins were detected in nuclei only after PMA stimulation. Staining with anti-NF-E2p45 or anti-MafK antibodies was shown in nuclei before and after hemin treatment. In contrast, Nrf2 proteins were stained in nuclei only after hemin treatment (Fig. 5A). Because thioredoxin is translocated into nucleus by PMA (37), we also tested nuclear expression of thioredoxin in hemin treatment. Thioredoxin translocated into nuclei after PMA or hemin stimulation in K562 cells (Fig. 5A). We then examined nuclear expression of these factors by Western blotting. NF-E2p45 expression was detected in nuclear extracts without stimulation. However, after hemin or PMA treatment, the expression decreased. Nrf2 expression was augmented by hemin. In contrast, expression of the Jun/Fos proteins was augmented by PMA, not by hemin (Fig. 5B). MafK remained in the nuclei before and after stimulation. Either hemin or PMA treatment augmented thioredoxin nuclear expression (Fig. 5B).
Activation of the Thioredoxin Promoter by Overexpression of Nrf2 and MafK-To further test whether Nrf2 and MafK was involved in the hemin-induced activation of the thioredoxin gene, we used transient transfection experiments in K562 cells using pTrxAREWT-Luc. Overexpression of both Nrf2 and MafK augmented hemin-induced activation of the thioredoxin promoter in K562 cells, dependent on the dose of Nrf2 (Fig. 6A). In addition, overexpression of the dominant negative mutant of Nrf2 suppressed hemin-induced activation of the thioredoxin gene (Fig. 6B). These data collectively showed that Nrf2 and the small Maf proteins mediate the hemin-induced activation of the thioredoxin gene. DISCUSSION In the present study, we have shown that a region from Ϫ452 to Ϫ420 bp of the gene regulatory region of the thioredoxin gene is required for hemin responsiveness and is also important for basal expression of the thioredoxin gene. The sequence from position Ϫ452 to Ϫ420 bp of the human thioredoxin promoter is highly homologous to that from position Ϫ607 to Ϫ575 bp of the mouse thioredoxin gene (15), indicating that the sequence in the thioredoxin promoter is conserved in both the human and mouse thioredoxin genes. We previously reported an oxidative responsive element (ORE), which is the sequence from Ϫ953 to Ϫ930 bp and mediates the stress response of hydrogen peroxide (14). The possible contribution of ORE to the hemin response should be further examined. Sistonen and co-workers (18)  gested that HSF-2 mediates hemin-induced activation of the thioredoxin gene. However, as observed in EMSA, specific binding complexes to the hemin responsive sequence were not affected by the addition of either unlabeled oligonucleotides encompassing HSE or antibodies against HSF-1 or HSF-2 (Fig.  3A). These observations showed that HSFs are not included among proteins that bind to the sequence. Indeed, no typical consensus sequence for HSFs was identified in the thioredoxin gene. The possible involvement of HSFs in hemin-induced activation of the thioredoxin gene should be further tested. In contrast, we found the TGCTGAGTAAC sequence, which resembles the ARE and AP-1 binding site, in the hemin responsive region of the thioredoxin promoter. Mutation in the ARE core sequence abolished the hemin response (M4 mutant in Fig.  2B), showing that the ARE is important for hemin-induced thioredoxin gene activation. We also observed a decrease of the hemin response with the M1 mutant (Fig. 2B), suggesting that the extended ARE sequence is necessary to be fully functional. This observation is consistent with a previous report (38), further indicating that hemin-induced thioredoxin gene activation is mediated by the ARE.
In the EMSA and DNA affinity binding assays, we showed that NF-E2p45/small Maf complex constitutively binds to the ARE and that Nrf2/small Maf complex is induced to bind to the ARE by hemin (Figs. 3, B and C; and 4). Other members of small Maf proteins may also contribute to these complexes, although the level of MafK is higher than MafG in K562 cells (39). The Jun and Fos proteins have been reported to be involved in the binding complex to the ARE (21, 29, 40 -42). In our experiment, we could not detect members of the Jun/Fos proteins in constitutively bound or hemin-induced binding complexes to the ARE (Figs. 3, C and D; and 4). In contrast, the binding of the Jun/Fos proteins to the ARE was detected only after PMA stimulation (Figs. 3, C and D; and 4). Taken together, we here propose a model that the ARE of the thioredoxin gene is regulated by a switch of its binding proteins (Fig. 7).
This change of binding proteins seemed to be regulated by the control of nuclear expression of ARE and AP-1 binding factors, because their nuclear expression pattern (Fig. 5) correlated well with the binding to the ARE (Figs. 3 and 4). Nuclear expression of these factors may be regulated by several mechanisms. Cycloheximide treatment abolished hemin-induced binding to the ARE, 2 suggesting an involvement of protein synthesis. Induction of small Maf protein synthesis is reported in ␤-naphthoflavone-induced activation of the ␥-glutamylcysteine synthetase subunit gene (43). In our data, the nuclear expression of MafK was unchanged or slightly augmented by hemin treatment. In contrast, Nrf2 nuclear expression was significantly augmented by hemin. The hemin-induced binding of Nrf2 to the ARE of thioredoxin gene was not preceded by augmentation of Nrf2 mRNA level (data not shown). Our results using confocal microscopy showed hemininduced up-regulation of Nrf2 expression in nuclei. These data are consistent with a previous study that reports exposure to electrophilic agents does not change the Nrf2 steady-state mRNA level and liberates Nrf2 from Keap1, leading to Nrf2 nuclear translocation (30). Thus, the control of translocation and turnover of Nrf2 protein seems to be an important mech-2 H. Masutani, unpublished observations. anism for hemin-induced augmentation of nuclear Nrf2 expression. Meanwhile, the reason for the decrease of nuclear expression of NF-E2p45 after hemin or PMA treatment (Fig. 5) is currently unclear. Regulation of nuclear export of Bach2 by oxidative stress has been reported (44). Nuclear expression of NF-E2p45 may also be regulated at the level of nuclear export.
Overexpression of both Nrf2 and small Maf protein augmented hemin-induced thioredoxin gene activation, which was dependent on the dose of Nrf2. In contrast, overexpression of a dominant negative mutant of Nrf2 suppressed hemin-induced activation of the thioredoxin gene (Fig. 6). Therefore, the major stimulatory effect on the thioredoxin gene by hemin seemed to be mediated by Nrf2 and small Maf proteins. Meanwhile, the role of the Jun/Fos proteins in the regulation of the ARE remains to be clarified. PMA stimulation did not change the luciferase activity using the pTrxAREWT-Luc vector (data not shown), suggesting that the PMA-induced switch from NF-E2p45/small Maf to the Jun/Fos proteins does not change the activation status of the thioredoxin gene in K562 cells. In other tissues or developmental stages, however, the Jun/Fos proteins might be important for basal and inducible thioredoxin gene activation. In addition, we can also speculate that in a pathological condition, perturbation of signal to the ARE by dysregulated activation of the AP-1 system causes dysregulation of differentiation process, resulting in oncogenesis.
It is important to note that the phase II enzyme genes such as the human and rat NAD(P)H: quinone oxidoreductase genes, the rat and murine glutathione S-transferase Ya genes, and the human ␥-glutamylcysteine synthetase subunit gene all contain the ARE. Nrf2 has been shown to be a regulator of the phase II enzyme genes (30,43,45). Thus, thioredoxin and these redox enzymes all have a common regulatory mechanism and may have co-ordinated roles against oxidative stress. Moreover, thioredoxin may be involved in cytoprotection against heme-and iron-related oxidative stress. Hemin, iron, and hemoglobin are promoters of free radical formation (46,47). Heme release from hemoglobin has been implicated in the pathogenesis of reperfusion injury (48). Previously, we showed that thioredoxin expression is induced in ischemic reperfusion (13). Thioredoxin is also reported to facilitate the induction of the heme oxygenase-1 (49). Collectively, study of the regulation of hemin-induced thioredoxin gene activation may lead to the clarification of the molecular basis of host defense against heme-induced oxidative stress-associated conditions such as reperfusion injury. In addition, because thioredoxin has a potent reducing activity, the current study indicates a role of thioredoxin in the protection against oxidation of hemoglobin in erythroid differentiation. Furthermore, our results show nuclear translocation of thioredoxin after PMA or hemin stimulation (Fig. 5). CNC-bZIP transcription factors and the Jun/ Fos proteins have conserved redox-sensitive cysteine residues (40). Because the thioredoxin system has an important role in the redox regulation of transcription factors, ARE-mediated thioredoxin gene activation may contribute to the regulation of ARE binding factors.
Finally, we have previously reported that thioredoxin negatively regulates p38 MAP kinase activation (50). We observed suppression of hemin-induced activation of the thioredoxin gene by p38 MAP kinase inhibitors in K562 cells, suggesting the involvement of the p38 MAP kinase system in the activation of the thioredoxin gene. 2 It is possible to speculate that ARE-mediated thioredoxin gene activation is a negative feedback mechanism. Meanwhile, a recent report showed that the heme oxygenase-1 gene is activated through the ARE by MEKK1, TAK1, and ASK-1, but not by p38 MAP kinase (51). In addition, another report showed that protein kinase C-medi-ated Nrf2 phosphorylation is involved in the PMA-induced activation of the ARE (52), although we observed only slight augmentation of nuclear expression of Nrf2 after PMA treatment (Fig. 5). Our results and these studies indicate that several distinct signaling pathways lead to the ARE, depending on individual stimulus and cell types. Further work is in progress to elucidate the upstream pathway to the ARE of the thioredoxin gene.