Abstract
The transcription factor nuclear factor erythroid-derived 2–related factor 2 (Nrf2) regulates induction of an extensive cellular stress response network when complexed with the cAMP-responsive element binding protein (CBP) at antioxidant response elements (ARE) located in the promoter region of target genes. Activating transcription factor 3 (ATF3) can repress Nrf2-mediated signaling in a manner that is not well understood. Here, we show that ATF3-mediated suppression is a consequence of direct ATF3-Nrf2 protein-protein interactions that result in displacement of CBP from the ARE. This work establishes ATF3 as a novel repressor of the Nrf2-directed stress response pathway. [Cancer Res 2008;68(2):364–8]
Introduction
The nuclear factor erythroid-derived 2–related factor 2 (Nrf2)-regulated signaling pathway functions to detoxify electrophilic and reactive oxygen carcinogens (1) and involves transcriptional induction of >200 genes via heterodimeric binding at antioxidant response elements (ARE) located in the proximal promoter of target genes. Nrf2 heterodimerization is a sequence-specific consequence of leucine zipper–coiled coil domain interactions with transcription factors such as small Mafs, activating transcription factor (ATF)4, and the Jun family (1). Under homeostatic conditions, Nrf2 levels are maintained at low levels due to E3 ubiquitin ligase–mediated ubiquitylation and proteasome-dependent degradation of Nrf2 (2). Oxidative and electrophilic carcinogens can inhibit Nrf2 degradation, allowing newly synthesized Nrf2 to translocate into the nucleus and induce ARE-directed gene expression; examples include glutathione S-transferases (GST), NAD(P)H: quinone reductase, glutamate cysteine ligase, heme oxygenase, superoxide dismutase (SOD), and catalase (1). Nrf2-mediated gene expression can impact inflammatory responses (3), oxidative-mediated DNA adduct formation (4), and susceptibility to chemical carcinogens (5) due to phase II gene expression. Although significant attention has been directed toward understanding mechanisms of induction, suppression of Nrf2 transactivation is less understood but phenotypically as important. Currently, it is known that overexpression of Fra or Fos can repress Nrf2-mediated gene expression with titration being the hypothesized mechanism. Additionally, Nrf2 heterodimers can be displaced from ARE sequences by formation of Nrf3/maf heterodimers (6, 7).
ATF3, a basic leucine zipper (bZIP) DNA binding protein, is a member of the ATF/cAMP-responsive element binding protein (CREB) transcription factor family and can form both homodimers as well as heterodimers with bZIP proteins to repress transcription (8). Although basal expression is minimal, ATF3 can be induced, for example by transforming growth factor (TGF)β1(9). Recent work has shown that TGFβ-mediated ATF3 expression significantly repressed expression of Nrf2-mediated ARE-directed phase II genes such as glutamate-cytosine ligase catalytic subunit (GCLC), GSTs, catalase, SOD1, and selenoprotein P (10). However, molecular details are elusive. Here, we report on mechanistic studies involving ATF3 repression of Nrf2/ARE-directed gene expression. Immunoprecipitation experiments coupled with GST-Nrf2 binding assays indicated that ATF3 directly bound Nrf2. DNA affinity precipitation and chromatin immunoprecipitation assay (ChIP) experiments showed that the ATF3-Nrf2 interaction occurred at the level of the ARE under conditions in which ARE-directed reporter activity was attenuated. Small hairpin RNA (shRNA) knockdown of ATF3 showed that the CREB-binding protein, CBP, known to associate with two transcriptional activation domains within Nrf2, was displaced in an ATF3-dependent fashion. Thus, these experiments identify ATF3 as a novel repressor of Nrf2-mediated gene expression.
Materials and Methods
Cell culture, plasmids, and antibodies. NMuMG cells (mouse mammary epithelial cells; a gift from Dr. C Arteaga, Vanderbilt University, Nashville, TN) were maintained in DMEM supplemented with 10% FCS and insulin (10 mg/mL). In some experiments, the serum concentration was decreased to 5% and cells were exposed to TGFβ1 to increase the expression of endogenous ATF3. The ATF3 expression vector (pCG/ATF3) was a gracious gift from Dr. T. Hai (Ohio State University, Columbus, OH). pcDNA/V5mNrf2 was a gracious gift from Dr. M. McMahon (University of Dundee, Dundee, Scotland, United Kingdom). The following antibodies were used: Actin (Sigma), ATF3, and Nrf2 (Santa Cruz).
Immunoblotting and immunoprecipitation. Protein extracts from NMuMG cells were analyzed as described in ref. 10. Detection was performed using Enhanced Chemiluminescence kit (Amersham Biosciences). For immunoprecipitation, cells were washed twice in ice-cold PBS and solubilized in radioimmunoprecipitation assay buffer [RIPA; 150 mmol/L NaCl, 10 mmol/L Tris (pH 7.2), 1% Triton X-100, 1% Na deoxycholate, 0.1% SDS, and 5 mmol/L EDTA plus 1 mmol/L 4-(2-aminoethyl) benzenesulfonylfluoride hydrochloride]. Solubilized protein was cleared with protein A/G and then immunoprecipitated with antibody. The washed pellet was solubilized in 5× SDS sample buffer.
GST/Nrf2 ATF3 interaction. Mouse Nrf2 was PCR amplified and inserted into pGEX-4T1 vector between SalI and Not I sites. The following primers were used to amplify mNrf2: 5′-CCC GGG TCG ACT CAT GAT GGA CTT GGA GTT GCC ACC G-3′ and 5′-CAG CAG CGG CCG CCT AGT TTT TCT TTG TAT CTG GCT TC-3′. The resulting pGEX-4T1 vectors were verified by sequencing and then used to transform BL21 (PLyss) strain. Fusion proteins were purified from bacteria using glutathione-agarose resin. The eluted proteins were dialyzed in buffer [20 mmol/L Tris-HCl (pH 8.0), 100 mmol/L NaCl, 1 mmol/L EDTA, and 1 mmol/L mercaptoethanol] and stored in solution at 4°C. Empty vector was used to express GST to use as a control in the binding assays.
PCR-amplified ATF3 was inserted into a pcDNA3.1 vector and used in a rabbit in vitro transcription and translation kit (Promega) to express [35S]-labeled ATF3. GST-Nrf2 and GST (15–25 μg) were immobilized onto glutathione-agarose beads (30 μL of 50% slurry) in 300 μL NETN buffer [20 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 1 mmol/L EDTA, and 0.5% NP40]. Unbound proteins were removed by washing the agarose beads with NETN buffer. The washed pellets were resuspended in 400 μL binding buffer [20 mmol/L HEPES-KOH (pH 7.9), 2.5 mmol/L MgCl2, 100 mmol/L KCl, 20%(v/v) glycerol, 0.2 mmol/L EDTA, and 0.5 mmol/L DTT] and incubated at 4°C for 1 h with 40 μL of [35S]-labeled ATF3. The pellets were washed thrice with NETN buffer, boiled in 10 μL of 5× SDS buffer, and separated on 12% polyacrylamide gel. After drying, the gel was exposed to film.
DNA precipitation. NMuMG cells were treated with either vehicle control or 2 ng/mL TGFβ1 for 18 h. Cells were harvested and nuclei were isolated. Nuclear extracts were prepared by extracting with EB [10 mmol/L HEPES (pH 7.9), 420 mmol/L NaCl, 0.1 mmol/L EGTA, 25% glycerol, 1.5 mmol/L MgCl2, and 0.5 mmol/L DTT] containing protease inhibitor cocktail. An ARE4 element from the GCLC promoter region was synthesized with biotin on the sense strand at the 5′ end. Streptavidin agarose (30 μL) was incubated with 3 μg of Biotin ARE4 in 600 μL of NE3 buffer [20 mmol/L HEPES (pH 7.9), 2.5 mmol/L MgCl2, 100 mmol/L KCl, 20% glycerol, 0.2 mmol/L EDTA, 0.5 mmol/L DTT, and protease inhibitor cocktail] for 30 min with mixing at 4C. Approximately 500 to 700 μg of nuclear extract and 100 μg of Poly dI-dC were added and the volume was adjusted to 1.5 mL. The mixture was incubated at 4°C for 20 min with gentle mixing. The samples were centrifuged and the pellet was washed thrice with WB (20 mmol/L HEPES (pH 7.9), 100 mmol/L KCl, 2.5 mmol/L MgCl2, and protease inhibitor cocktail). The pellet was heated with 20 μL of 5× SDS buffer. Samples were subjected to immunoblot analysis.
ChIP assay. NMuMG cells were treated with either with vehicle control or 2 ng/mL TGFβ1 for 18 h. Cells were crosslinked with formaldehyde (final concentration 1%) for 10 min at room temperature and processed for ChIP analysis (ChIP-IT) according to the manufacture's instructions (Active Motif). PCR amplification of mGST Ya ARE (11) was accomplished using the following primers: 5′-GAA TCA GCT TGT GGG TGT GT-3′ (sense primer) and 5′-CAG TTA CTC CTG TGG GAA AG-3′ (antisense primer). The PCR products were separated on 3% agarose gels and stained with ethidium bromide. PCR products were confirmed by DNA sequencing.
ATF3 shRNA. NMuMG cells, growing exponentially in T-75 flasks, were infected with a shRNA retroviral vector (Open Biosystems) directed against ATF3, and clones were selected for puromycin resistance. Control cells were infected with a shRNA retroviral vector expressing a nonsilencing construct and clones were selected for puromycin resistance.
Transcriptional assays. Cells were cotransfected with the following plasmids: pGL3pro/ARE4 luciferase and pcDNA3.1/LacZ using Lipofectamine (Invitrogen), according to the manufacturer's protocol. Firefly luciferase and β-galactosidase activities in cell lysates were determined using a Reporter Assay System (Promega) according to the manufacturer's protocol. Luciferase activity was normalized to β-galactosidase activity to account for transfection efficiency.
Measurement of reactive oxygen species. Reactive oxygen species (ROS) production was measured in NMuMG cells after exposure to 100 μmol/L diethyl maleate (DEM) for 3 h at 37°C using the oxidative sensitive dye C-400 (Molecular Probes).
Results and Discussion
Nrf2 functions as a node for regulating the expression of >200 genes that encode a cellular stress response network. Although ATF3 can suppress the induction of this network (10), a molecular understanding of the mechanism is lacking. The data presented in Fig. 1A show that ATF3 can suppress the ability of Nrf2 to direct ARE-regulated gene expression (P < 0.05; ANOVA). If increased expression of ATF3 repressed Nrf2-dependent gene expression, then one may expect that loss of ATF3 expression would enhance Nrf2-directed gene expression. Figure 1B illustrates three clones stably infected with a retrovirus expressing a nonsilencing shRNA (clones 1–3) and three clones stably infected with a retrovirus expressing shRNA directed against ATF3 (ATF3 expression was knocked down by 70% in clones A & B and 50% in clone C relative to control). Expression of HO-1, a Nrf2 target gene (12), was increased 2.3- to 2.6-fold relative to control after shRNA-mediated ATF3 knockdown, consistent with expectations (Fig. 1B). Both the expression of GST and cellular glutathione (GSH) concentrations are Nrf2 regulated (13, 14). GST isoenzymes catalyze nucleophilic attack by GSH on nonpolar compounds such as DEM that contain an electrophilic carbon, nitrogen, or sulfur atom. We assessed GST/GSH-mediated DEM detoxification by quantifying production of ROS (15) and found that ATF3 expression suppressed detoxification, as measured by a 30% increase in ROS production (Fig. 1C). These results are consistent with ATF3-mediated suppression of Nrf2-directed gene expression.
A, ATF3 represses ARE-directed reporter activity. NMuMG cells were transiently cotransfected with pGL3pro/ARE4 luciferase, pcDNA3.1/LacZ, pcDNA/V5mNrf2, pCG/ATF3, and pcDNA3.1 (concentrations denoted in Figure). Luciferase (Lux) activity derived from pGL3pro/ARE4 luciferase is expressed relative to β-galactosidase activity. ARE4-directed luciferase activity increased 2.8-fold in cells transiently transfected with an Nrf2 expression vector compared with cells transfected with control vector (pcDNA3.1; P < 0.05; Student's t test). Cotransfection of an ATF3 expression vector repressed the ability of Nrf2 to induce ARE4-directed reporter activity (P < 0.05; ANOVA); B, immunoblot demonstrating that HO-1 expression was increased in NMuMG cells stably infected with retrovirus expressing shRNA directed against ATF3 (clones A–C). Control cells were infected with virus expressing nonsilencing shRNA (clones 1–3). The expression of ATF3 relative to actin and HO-1 relative to actin was determined using a quantitative digitizing software program. C, ATF3 represses detoxification of a xenobiotic. NMuMG cells were transiently transfected with either pcDNA3.1 (control vector) or pCG/ATF3. Forty-eight hours after transfection, cells were exposed to either vehicle control or 100 μmol/L DEM for 3 h at 37°C. ROS-mediated C-400 fluorescence was then measured (10).
A, ATF3 represses ARE-directed reporter activity. NMuMG cells were transiently cotransfected with pGL3pro/ARE4 luciferase, pcDNA3.1/LacZ, pcDNA/V5mNrf2, pCG/ATF3, and pcDNA3.1 (concentrations denoted in Figure). Luciferase (Lux) activity derived from pGL3pro/ARE4 luciferase is expressed relative to β-galactosidase activity. ARE4-directed luciferase activity increased 2.8-fold in cells transiently transfected with an Nrf2 expression vector compared with cells transfected with control vector (pcDNA3.1; P < 0.05; Student's t test). Cotransfection of an ATF3 expression vector repressed the ability of Nrf2 to induce ARE4-directed reporter activity (P < 0.05; ANOVA); B, immunoblot demonstrating that HO-1 expression was increased in NMuMG cells stably infected with retrovirus expressing shRNA directed against ATF3 (clones A–C). Control cells were infected with virus expressing nonsilencing shRNA (clones 1–3). The expression of ATF3 relative to actin and HO-1 relative to actin was determined using a quantitative digitizing software program. C, ATF3 represses detoxification of a xenobiotic. NMuMG cells were transiently transfected with either pcDNA3.1 (control vector) or pCG/ATF3. Forty-eight hours after transfection, cells were exposed to either vehicle control or 100 μmol/L DEM for 3 h at 37°C. ROS-mediated C-400 fluorescence was then measured (10).
We found that endogenous Nrf2 and ATF3 reside as an immunoprecipitable complex (Fig. 2A). A GST/Nrf2 fusion protein was constructed and used to determine whether Nrf2 physically interacted with in vitro transcribed/translated ATF3. A purified GST/Nrf2 fusion protein immobilized on GSH-agarose beads was found to capture [35S]-labeled ATF3, in contrast to purified GST immobilized on GSH-agarose beads (Fig. 2B). These results show that Nrf2 and ATF3 physically interact in vitro. Furthermore, we found that in vitro transcribed/translated [35S]-labeled ATF3 and [35S]-labeled Nrf2 formed a heterodimer (Fig. 2C), which bound to and retarded a representative ARE sequence (GST Ya ARE; ref. 16) in a gel mobility shift assay (Fig. 2D). This interaction was disrupted by the presence of an ATF3 antibody (αATF3).
ATF3 binds to Nrf2 at AREs. A, immunopreciptitation of endogenous ATF3 followed by immunoblotting for endogenous Nrf2. NMuMG cells were solubilized in RIPA buffer and ATF3 was immunoprecipitated. The resulting immunoblot was probed for ATF3 and Nrf2. B, ATF3 binds Nrf2 in vitro. In vitro transcription/translation with [35S]-labeled ATF3 binds GST-Nrf2. C and D, in vitro transcribed/translated ATF3 and Nrf2 associate and bind to GST Ya ARE, as shown by GMSA. In vitro translated [35S]-labeled ATF3 and [35S]-labeled Nrf2 were incubated together at 4°C for 1 h and immunoprecipitated (C) or incubated with a GST Ya ARE double-stranded oligomer and analyzed by GMSA. An αATF3 disrupted the ATF3/Nrf2/ARE complex (D).
ATF3 binds to Nrf2 at AREs. A, immunopreciptitation of endogenous ATF3 followed by immunoblotting for endogenous Nrf2. NMuMG cells were solubilized in RIPA buffer and ATF3 was immunoprecipitated. The resulting immunoblot was probed for ATF3 and Nrf2. B, ATF3 binds Nrf2 in vitro. In vitro transcription/translation with [35S]-labeled ATF3 binds GST-Nrf2. C and D, in vitro transcribed/translated ATF3 and Nrf2 associate and bind to GST Ya ARE, as shown by GMSA. In vitro translated [35S]-labeled ATF3 and [35S]-labeled Nrf2 were incubated together at 4°C for 1 h and immunoprecipitated (C) or incubated with a GST Ya ARE double-stranded oligomer and analyzed by GMSA. An αATF3 disrupted the ATF3/Nrf2/ARE complex (D).
Under basal conditions, ATF3 exhibits both a nuclear and cytosolic distribution, but expression is limited (10). Biochemical fractionation followed by DNA precipitation using a biotinylated double-stranded representative antioxidant response element oligonucleotide (ARE4; ref. 17) and detection by immunoblotting allowed us to show that ATF3 was associated with bound Nrf2 and resided at the ARE (Fig. 3A). Two Nrf2 species of slightly different molecular weight were found bound to the ARE oligomer. The nature of these two species is not known at this time but may be related to phosphorylation of Nrf2 (18). The observation that c-Jun resided with Nrf2 at an ARE (Fig. 3A) is consistent with the work of Jeyapaul et al. (19) who showed that c-Jun-Nrf2 heterodimers positively regulate ARE-directed gene expression. Once bound to an ARE, Nrf2 cooperatively associates with CBP to activate transcription of target genes (20). As expected, the DNA precipitation experiments showed the presence of CBP at the ARE (Fig. 3A).
A, DNA precipitation (ppt) experiments illustrating the association of ATF3 with representative AREs. NMuMG cells were exposed to 0 or 2 ng/mL TGFβ1 for 18 h, nuclear protein isolated, and incubated with biotinylated double-stranded oligonucleotides covering ARE4 of GCLC. DNA-associated proteins were precipitated by streptavidin-agarose beads and detected by immunoblotting using antibodies to Nrf2, ATF3, c-Jun, and CBP. B, ChIP assay demonstrating that CBP is associated with the ARE in the mGST Ya promoter in vivo. NMuMG cells were exposed to 0 or 2 ng/mL TGFβ for 18 h before crosslinking, DNA fragmentation, and immunoprecipitation with a CBP antibody. C, ChIP assay demonstrating that ATF3 is associated with the ARE in the mGST Ya promoter in vivo. Cells were exposed to 0 or 2 ng/mL of TGFβ for 18 h before crosslinking, DNA fragmentation, and immunoprecipitation with an αATF3. In two samples, the αATF3 was reacted with a blocking peptide before immunoprecipitation. Fold increase represents the intensity of the PCR-amplified GST Ya ARE relative to input.
A, DNA precipitation (ppt) experiments illustrating the association of ATF3 with representative AREs. NMuMG cells were exposed to 0 or 2 ng/mL TGFβ1 for 18 h, nuclear protein isolated, and incubated with biotinylated double-stranded oligonucleotides covering ARE4 of GCLC. DNA-associated proteins were precipitated by streptavidin-agarose beads and detected by immunoblotting using antibodies to Nrf2, ATF3, c-Jun, and CBP. B, ChIP assay demonstrating that CBP is associated with the ARE in the mGST Ya promoter in vivo. NMuMG cells were exposed to 0 or 2 ng/mL TGFβ for 18 h before crosslinking, DNA fragmentation, and immunoprecipitation with a CBP antibody. C, ChIP assay demonstrating that ATF3 is associated with the ARE in the mGST Ya promoter in vivo. Cells were exposed to 0 or 2 ng/mL of TGFβ for 18 h before crosslinking, DNA fragmentation, and immunoprecipitation with an αATF3. In two samples, the αATF3 was reacted with a blocking peptide before immunoprecipitation. Fold increase represents the intensity of the PCR-amplified GST Ya ARE relative to input.
ATF3 expression can be increased in a TGFβ-mediated manner (9). The amount of ATF3 present at the ARE increased (Fig. 3A) when nuclear ATF3 levels were elevated in TGFβ-treated cells (10) and (Fig. 3C) coincident with a decrease in the presence of c-Jun and CBP. Nrf2 binding was unaffected (Fig. 3A). Use of a ChIP assay and a representative ARE sequence (GST Ya ARE; ref. 11) corroborated the observation that CBP levels were diminished coincident with increased detection of ATF3 at the ARE (Fig. 3B).
Use of ChIP assays showed that TGFβ1 increased the amount of ATF3 present at a representative ARE in vivo (Fig. 3C). Under basal conditions, ATF3 was associated with only 35% of the GST Ya ARE. Exposing cells to 2 ng/mL of TGFβ1 for 4 h elevated nuclear ATF3 levels (10) and increased the amount of ATF3 residing at the GST Ya ARE by ∼2-fold, coincident with ATF3-mediated suppression of Nrf2 target genes (10).
Next, we used the NMuMG cells that stably expressed ATF3 shRNA from a retroviral vector or expressed a nonsilencing construct, as described in Fig. 1. In control cells, TGFβ produced a robust increase in ATF3 expression (Fig. 4A). Please note that under these detection conditions, basal ATF3 was not observed. In cells expressing ATF3 shRNA, TGFβ did not produce a robust increase in ATF3 (Fig. 4A). Knock down of ATF3 by shRNA retroviral expression abrogated TGFβ-mediated loss of CBP at the ARE, as shown in the ChIP assay shown in Fig. 4B. Specifically, cells expressing nonsilencing shRNA, exhibited a 50% loss of CBP associated with the GST Ya ARE. In contrast, there was only a 10% decrease in cells expressing ATF3 shRNA.
A, stable retroviral expression of an ATF3 shRNA knocks down ATF3. NMuMG cells were exposed to 0 or 2 ng/mL TGFβ1 for 24 h, and lysates were immunoblotted for ATF3 and actin. B, ChIP assay demonstrating that CBP binding at an ARE in cells expressing an ATF3 shRNA is independent of TGFβ. Ab, antibody.
A, stable retroviral expression of an ATF3 shRNA knocks down ATF3. NMuMG cells were exposed to 0 or 2 ng/mL TGFβ1 for 24 h, and lysates were immunoblotted for ATF3 and actin. B, ChIP assay demonstrating that CBP binding at an ARE in cells expressing an ATF3 shRNA is independent of TGFβ. Ab, antibody.
In conclusion we propose that formation of ATF3-Nrf2 heterodimers at AREs located in the proximal promoter of Nrf2 target genes suppresses recruitment of CBP, which in turn represses the ability of Nrf2 to induce target gene expression. These data identify ATF3 as a novel repressor of Nrf2-mediated signaling.
Acknowledgments
Grant support: NIH/National Cancer Institute grants RO1 CA115556 and T32 CA093240.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.