Down-regulation of Cytochrome P450 1A1 Gene Promoter by Oxidative Stress

Oxidative stress interferes with several cellular functions, in particular transcriptional regulation. We show here that the human cytochrome P450 1A1 (CYP1A1) is down-regulated at the transcriptional level by oxidative stress. Basal as well as 2,3,7,8-tetrachloro-p-dioxin-induced promoter activities are strongly impaired by H2O2 treatment or glutathione depletion with l-buthionine-(S,R)-sulfoximine. Tumor necrosis factor α inhibits CYP1A1 expression, and this inhibition is prevented by the antioxidant pyrrolidine dithiocarbamate. We show that these regulations depend on the integrity of the nuclear factor 1 (NFI) site located in the proximal promoter. We therefore examined the redox regulation of this transcription factor. Treatment of human HepG2 or rat H4 hepatoma cells with H2O2 orl-buthionine-(S,R)-sulfoximine inactivates the binding of the NFI transcription factor to its DNA consensus sequence. Furthermore, H2O2 treatment leads to a dose-dependent decrease of reporter gene expressions driven by promoters containing NFI binding sites. Glutathione depletion and catalase inhibition also repress a NFI-driven promoter. Under the same conditions, the CP-1 transcription factor activity is not affected by oxidative stress. Thus, NFI seems particularly sensitive to oxidative stress. This accounts, at least partially, for the regulation ofcyp1A1 gene expression.

Oxidative stress interferes with several cellular functions, in particular transcriptional regulation. We show here that the human cytochrome P450 1A1 (CYP1A1) is down-regulated at the transcriptional level by oxidative stress. Basal as well as 2,3,7,8-tetrachloro-p-dioxin-induced promoter activities are strongly impaired by H 2 O 2 treatment or glutathione depletion with L-buthionine-(S,R)-sulfoximine. Tumor necrosis factor ␣ inhibits CYP1A1 expression, and this inhibition is prevented by the antioxidant pyrrolidine dithiocarbamate. We show that these regulations depend on the integrity of the nuclear factor 1 (NFI) site located in the proximal promoter. We therefore examined the redox regulation of this transcription factor. Treatment of human HepG2 or rat H4 hepatoma cells with H 2 O 2 or L-buthionine-(S,R)-sulfoximine inactivates the binding of the NFI transcription factor to its DNA consensus sequence. Furthermore, H 2 O 2 treatment leads to a dose-dependent decrease of reporter gene expressions driven by promoters containing NFI binding sites. Glutathione depletion and catalase inhibition also repress a NFI-driven promoter. Under the same conditions, the CP-1 transcription factor activity is not affected by oxidative stress. Thus, NFI seems particularly sensitive to oxidative stress. This accounts, at least partially, for the regulation of cyp1A1 gene expression.
Cytochrome P450 monooxygenases are drug-metabolizing enzymes that play a major role in the detoxification and elimination of hydrophobic xenobiotics. Paradoxically, these enzymes also generate reactive metabolites, among them mutagens (1). CYP1A 1 is a ubiquitous member of the P450 superfamily that belongs to the aryl hydrocarbon (Ah) inducible gene battery. It is highly inducible by 2,3,7,8-tetrachlorop-dioxine (TCDD) and by planar aromatic hydrocarbons such as 3-methylcholanthrene or benzo(a)pyrene. The Ah receptor, once activated by such compounds, translocates into the nucleus, heterodimerizes with Arnt (Ah receptor nuclear translo-cator), and binds to a DNA sequence called the xenobioticresponsive element. Extensive studies on the murine cyp1A1 gene have shown that this regulation involves a cross-talk between enhancer and promoter sequences with changes in the chromatin structure (2). A nuclear factor 1 (NFI) binding site was shown to be critical for the activity of this promoter (3). The 5Ј upstream region of the human cyp1A1 gene contains several xenobiotic-responsive elements mediating the Ah response, a negative regulatory element (4), and a proximal promoter containing a NFI site. The activity of several cytochromes P450 including CYP1A1 is altered in vivo during inflammation. Some cytokines down-regulate cytochrome P450 activity levels in human hepatocytes (5). In addition, H 2 O 2 was shown to down-regulate CYP1A1 mRNA level in rat hepatocytes (6). The molecular mechanisms involved in this regulation were unknown, which led us to study the effect of oxidative stress on the human cyp1A1 promoter activity.
Oxidative stress plays an important role in cell regulation (7). Several studies show its role as a positive regulator of some early alert genes such as GADD153 (8) and antioxidant enzymes (see Ref. 9 and references therein) via the transcription factors NF-B and AP-1 (reviewed in Ref. 10) or the antioxidant-responsive element sequence (11). Oxidative stress causes a modification of the intracellular redox status and can chemically affect redox-sensitive moieties such as sulfhydryl groups. Any protein containing strategic cysteine residues may undergo functional alterations that will largely depend upon its conformation and the residues' redox potentials. Several in vitro studies have shown that the DNA binding ability of various transcription factors could be modified by oxidative reagents: AP-1, NF-B, Sp1, MTF-1, USF, p53, Egr-1, and Ets (for a review and references, see Ref. 12). The oxidation or alkylation of a thiol group located in the DNA binding domain may alter the DNA-protein interaction because of steric hindrance or lack of hydrogen bond. In addition, the variation of the intracellular redox status affects disulfide bridges and metal ion coordination (13). Thus, the oxidation of a cysteine even not located in the DNA binding domain can result in a conformational change of the protein or influence its dimerization, which will greatly affect its DNA binding ability or its transactivating function (14,15). Yet, in vitro studies are not sufficient to prove that a transcription factor activity is modified by oxidative stress within the cell. An alternative mechanism for the regulation of transcription factors by oxidative stress is the modification of enzyme activities involved in signal transduction. For example, kinase or phosphatase activities have been shown to be modulated by H 2 O 2 (16 -19). Since some transcription factors (e.g. NF-B, c-Jun, Ets, and Sp1) are targets of such enzymes (20 -22), this may be critical for transcriptional control.
Here, we report for the first time that NFI activity is redoxregulated in vivo and that this regulation strongly affects * This work was supported by INSERM, University Paris-Val de Marne, and Association pour la Recherche contre le Cancer Grant 6604. 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.
‡ A fellow of the Délègation Générale à l'Arnement. § To whom correspondence should be addressed: Tel.: 33 1 42 86 20 75; Fax: 33 1 42 86 20 72; E-mail: robert.barouki@biomedicale.univ-paris5.fr. 1 The abbreviations used are: CYP, cytochrome P450 monooxygenase; TCDD, 2,3,7,8-tetrachloro-p-dioxine; BSO, L-buthionine-(S,R)-sulfoximine; TNF␣, tumor necrosis factor-␣; Ah, aryl hydrocarbon; NFI, nuclear factor 1; EMSA, electrophoretic mobility shift assay; CAT, chloramphenicol acetyltransferase; CTF, CAAT transcription factor; PDTC, pyrrolidine dithiocarbamate. CYP1A1 expression. NFI plays an important role in viral (23) and cellular gene regulation. It was originally described as a factor required for the replication of adenovirus DNA (24). It was then shown to be part of a family of ubiquitous transcription factors encoded by four different genes (NFI A, NFI B, NFI C/CTF, and NFI X) that bind to the TGGC(N) 5 GCCA DNA consensus sequence after dimerization (25) . The highly conserved 220 N-terminal residues containing the DNA binding domain are not homologous to any well characterized class of DNA binding domains (26). NFI is widely involved in the control of constitutive or inducible gene expression. In the liver, it has been shown to be critical for the expression of collagen I (27), albumin (28), and aspartate aminotransferase (29) and for the specific expression of ␣-fetoprotein (30) and vitellogenin (31). Members of the NFI family have been reported as possible targets for phosphorylation or glycosylation. Yet, this does not seem to affect their abilities to bind DNA in vitro (26). The NFI family members possess four conserved cysteines (32). Three of them were critical for DNA binding in vitro, which was abolished by their alkylation or oxidation (26). These observations suggest that NFI could be sensitive to modifications of the redox status within the cell. In this study, we have used hepatoma cell cultures to assess the functional impact of oxidative stress on the binding of NFI to DNA and its transcriptional activity and the consequences for the cyp1A1 gene promoter activity.

EXPERIMENTAL PROCEDURES
Chemicals-H 2 O 2 was used from a 30% stock (Merck). Other chemicals were obtained from Sigma (unless otherwise stated), and oligonucleotides were from Eurogentec (Belgium) and Genset (Paris, France). Human recombinant TNF␣ was obtained from Tebu (Le Perray, France).
Cell Culture-The rat hepatoma cell line H4 II EC3 (33) and the human hepatoma cell line HepG2 (34) were maintained as described by Garlatti et al. (29). Cell viability was assessed using the trypan blue exclusion test. Cells from nonconfluent dishes were trypsinized and resuspended in 3 ml of phosphate-buffered saline. 1 ml of colorant (trypan blue 6.2 mM, NaCl 0.8 M) was then added and gently mixed. After 2 min, cells were counted. Viability was expressed as (number of non-blue cells/total number of cells) ϫ 100.
Nuclear Extracts-Confluent 10-cm dishes were trypsinized. All subsequent steps were performed at 4°C. Cells were centrifuged and then resuspended in a hypotonic buffer (10 mM Hepes, pH 8, 0.1 mM EDTA, 0.1 mM EGTA, 10 mM KCl, 0.5 mM phenylmethylsulfonyl fluoride) and allowed to swell for 15 min. Cells were lysed by adding 1.5% Nonidet P-40 and vortexing for 10 s. After centrifugation (7000 ϫ g, 1 min) the supernatant could be saved for cytosolic extract analysis and pellets were resuspended in a hypertonic buffer (20 mM Hepes, pH 8, 1 mM EDTA, 1 mM EGTA, 0.4 M NaCl, 1 mM phenylmethylsulfonyl fluoride). After a 15-min incubation, the mixture was centrifuged (13,000 ϫ g, 10 min), and the supernatant was aliquoted and stored at Ϫ80°C. The protein concentration was measured according to the method of Bradford (35).
Glutathione Assay-Approximately 20 ϫ 10 6 cells were harvested and allowed to swell for 15 min (cf. nuclear extract procedure). They were then homogenized with a Dounce homogenizer. The mixture was centrifuged (13,000 ϫ g, 7 min), and the supernatant was assayed for glutathione using the GSH-400 kit from Oxys (Portland, OR). Proteins were assayed according to the method of Bradford (35). Glutathione levels were normalized to protein content.
Western Blots-Whole cell or nuclear extracts were homogenized by vigorous mixing in loading buffer (62 mM Tris, pH 6.8, 10% (v/v) glycerol, 2% (w/v) SDS, 0.7 M ␤-mercaptoethanol, 2.5% (v/v) saturated bromphenol blue) followed by a 5-min boiling. Samples were run at 12 V/cm for 1 h in a 10% polyacrylamide gel. Proteins were then transferred to nitrocellulose (Hybond C; Amersham Pharmacia Biotech) by electroblotting (90 mA, 1 h). Blots were incubated with 1 g/ml antibody for 1 h, washed three times 15 min, and incubated with anti-rabbit IgG coupled to horseradish peroxidase (50 ng/ml) for 1 h at room temperature. Peroxidase activity was measured with the ECL kit (Amersham Pharmacia Biotech). The rabbit polyclonal nonspecific NFI antibody targeting the N-terminal part was obtained from Tebu (Le Perray, France). Rabbit antibodies raised against NFI/CTF1 and NFI X3 were generous gifts of Drs. N. Tanese (New York University Medical Center) and B. Corthésy (EPFL, Lausane, Switzerland).
Northern Blots-RNA preparation and Northern blots were performed as already described (36). Probes were synthesized from cDNAs with the Megaprime DNA labeling kit (Amersham Pharmacia Biotech) according to the manufacturer's instructions. Quantifications were performed with a PhosphorImager and ImageQuant software (Molecular Dynamics).
Plasmids-The Firefly and Renilla luciferase expression plasmids (pGL3 basic and pRL-SV40) were purchased from Promega. The pRSV-FL plasmid was a gift from Dr M. Aggerbeck (INSERM, Paris). The pTk-CAT plasmid, a gift from Dr. C. Forest (CNRS, Meudon, France), contains the thymidine kinase gene promoter (Ϫ105,ϩ51) of the herpes simplex 1 virus upstream of the chloramphenicol acetyltransferase (CAT) reporter gene. Recombinant plasmids were constructed as follows (see Fig. 1A).
The Ϫ1566,ϩ73 5Ј region of the human cyp1A1 gene (a gift of Dr. Maurel, INSERM, Montpellier, France), was subcloned into the KpnI-HindIII double-digested pGL3 basic vector, upstream of the firefly luciferase reporter gene to yield p1A1-FL.
The pmut1A1-FL is identical to the p1A1-FL except for a double mutation at positions Ϫ49 and Ϫ50 in the NFI site (GCCA 3 CGCA). It was obtained by site-directed mutagenesis on p1A1-FL, using a mutated oligonucleotide and the GeneEdit kit (Promega) according to the manufacturer's instructions.
A double-stranded oligomer deriving from the 3Ј-end of the human ␣-globin gene proximal promoter (Ϫ79,ϩ1) was synthesized with point mutations at the bases shown in italics, in order to change Sp1-like binding sites and BglII 5Ј and HindIII 3Ј ends: 5Ј-GATCTCCGCACC-AGCCAATGAGCGAATTCCGACCGAGGCTTCCCCTGCGTCACAGG-TATAAACCCTGACGCACTCACGGACCAGCAC-3Ј. It was subcloned into the BglII-HindIII double-digested pRL-SV40 vector (thus deprived of the SV40 early promoter) upstream of the Renilla luciferase reporter gene to yield p␣glob-RL, which was used as a control in transfection experiments. A NFI site, present in the human ␣-globin gene promoter upstream of position Ϫ79, was not included in this sequence.
The TriNFI double-stranded oligomer (5Ј-AGCTTGGATTGAAGC-CAATATTGGATTGAAGCCAATATTGGATTGAAGCCAAGA-3Ј, containing three adjacent NFI consensus sites and HindIII-compatible ends) was subcloned into the HindIII site located in the thymidine kinase promoter of pTk-CAT to give the pTriNFI-CAT plasmid.
All constructs were checked by DNA sequencing analysis. Transfection Experiments-Transfection experiments were performed in HepG2 cells as described previously (29,37). These cells were used because the endogenous cyp1A1 gene was regulated by H 2 O 2 (cf. Fig. 2) and because of their excellent transfection efficiency. Briefly, 1 day prior to the transfection, cells (0.5 ϫ 10 6 cells/5-cm dish) were seeded into the usual culture medium. The CAT plasmids (3 g of DNA) and the firefly and/or Renilla luciferase expression vectors (3 and 1 g, respectively) were introduced into the cells by the calcium phosphate coprecipitation technique followed by a 2-min glycerol shock. Cells were treated by adding various agents to the culture medium. After an overnight incubation, cells were homogenized for enzymatic assays. The CAT and firefly luciferase activities were determined as described previously (29,37). Dual luciferase assay (firefly and Renilla) was performed with a Promega kit. Renilla luciferase activity was used to normalize the transfection efficiency in all culture dishes (38). Blanks were obtained by assaying CAT or luciferase activity in mock-transfected cells.
The experiments described above have been performed using an internal control for transfection. In initial assays, we tested several promoters commonly used for this purpose. We noticed that some of those promoters, particularly SV40, were sensitive to even moderate amounts of H 2 O 2 (Fig. 1B) and thus cannot be used as internal controls. During the study of the effect of H 2 O 2 on several transcription factors, we found that CP-1 was not sensitive to such a treatment. 2 We therefore synthesized the plasmid p␣glob-RL, which contains a promoter derived from the proximal part of the human ␣-globin gene promoter upstream of the Renilla luciferase reporter gene. This promoter consists of a CP-1 site and a TATA box and is insensitive to the H 2 O 2 concentrations used here (Fig. 1B).
Statistics-Neumann-Kheuls and Student's two-tailed t tests were performed using analysis of variance software (PLC Inc.).

Oxidative Stress Down-regulates TCDD-induced CYP1A1
mRNA Increase-In both human HepG2 and rat H4 hepatoma cells, TCDD increased CYP1A1 mRNAs. The level of CYP1A1 mRNA in cells treated with TCDD is dramatically reduced by prior treatment with H 2 O 2 (Fig. 2). The use of L-buthionine-(S,R)-sulfoximine (BSO), an inhibitor of ␥-L-glutamyl-L-cysteine synthase (39), which causes a strong decrease of the intracellular GSH pool (cf. Table I), also caused a downregulation. The basal levels were weak. This effect was more potent in the human hepatoma cell line HepG2 than in the rat hepatoma H4.
Redox Regulation of the cyp1A1 Promoter-To determine whether the CYP1A1 mRNAs down-regulation by oxidative stress was accounted for by a modification of the gene promoter activity, we transfected the p1A1-FL plasmid containing the firefly luciferase reporter gene controlled by the human cyp1A1 gene promoter (Ϫ1566,ϩ73 fragment) in HepG2 cells. The treatment of cells by increasing concentrations of H 2 O 2 caused a dose-dependent inhibition of the reporter gene expression (Fig. 3A, curve a). A BSO-induced GSH depletion also impaired the expression driven by this promoter (Fig. 3B, open bars). Dioxin (TCDD) induced cyp1A1 promoter activity 20-fold (not shown). As shown in Fig. 3C (open bars), the induced activity was also strongly inhibited by H 2 O 2 , suggesting that the decrease in steady-state mRNAs levels are mainly due to a decrease in promoter activity. The effect of H 2 O 2 was more potent on the TCDD-induced activity than on the basal activity. In some cellular systems, the cytokine TNF␣ is known to increase cellular H 2 O 2 levels (40,41). In HepG2 cells, TNF␣ decreased cyp1A1 promoter activity by 40% (Fig. 3D, open bars). This effect is prevented in the presence of the antioxidant PDTC, suggesting that it is mediated by a redox mechanism. In the same experiments, TNF␣ did not affect the expression driven by a CP-1/TATA promoter (not shown).
Critical Contribution of the Proximal NFI Site to the Redox Regulation-Since both basal and induced activities were altered by oxidative stress, we hypothesized that the proximal NFI site may be implicated in the H 2 O 2 effect. In order to test this hypothesis, we mutated this site in the cyp1A1 promoter. We first verified by EMSA that this mutation inhibited NFI binding (not shown). The mutated promoter displayed decreased basal activity (8-fold decrease, Fig. 3A). It also displayed decreased TCDD inducibility, 11-versus 20-fold induction for pmut1A1-FL and p1A1-FL, respectively (not shown). The role of the NFI site in the basal and induced activities of the 1.6-kilobase pair-long human promoter is in agreement with previous studies using the mouse promoter (3).
Interestingly, the effect of H 2 O 2 and BSO on the basal activity was abolished by the mutation of the NFI site (Fig. 3A,    b and Fig. 3B, shaded bars). Similarly, the activity of the dioxin-induced mutated promoter was not altered by oxidative stress (Fig. 3C, shaded bars). We conclude that NFI site integrity is required for the inhibitory effect caused by oxidative stress. Furthermore, this site is also required for the inhibition of promoter activity by TNF␣, since this cytokine had no effect on the activity of the mutated promoter (Fig. 3D, shaded bars). NFI Activity Is Down-regulated by Oxidative Stress-In order to further characterize the contribution of NFI to the oxidative regulation of gene expression, the activity and regulation of short promoter fragments containing critical NFI sites were tested. We first tested a promoter containing a NFI site upstream of a TATA box (pNFI/TATA-FL plasmid containing the 60-base pair-long proximal promoter of the cyp1A1 gene). As shown in Fig. 4A, increasing concentrations of H 2 O 2 downregulated this promoter activity in a dose-dependent manner. When the NFI site of this promoter fragment was mutated, the inhibitory effect of H 2 O 2 was prevented (the basal activity of the mutated promoter was decreased but remained at least 3 times higher than the one obtained with the pGL3 basic vector). The effect of other compounds modifying the redox status within the cell was also tested. BSO inhibited this promoter activity (not shown). Aminotriazole, an inhibitor of the H 2 O 2 scavenging enzyme catalase, also decreased this promoter activity, which was further repressed by H 2 O 2 (Fig. 4B). Finally, TNF␣ down-regulated the NFI/TATA promoter activity in the absence, but not in the presence, of PDTC (Fig. 4C).
To study the influence of H 2 O 2 on NFI activity in a heterologous context, we transiently transfected HepG2 cells with a recombinant Tk-CAT plasmid. The pTriNFI-CAT construct (three NFI sites inserted upstream of the thymidine kinase promoter) displayed a basal CAT expression 4 times higher than the pTk-CAT. The reporter gene expression decreased dramatically when cells were treated with increasing concentrations of H 2 O 2 . The dose-dependent inhibition of NFI transcriptional activity was maximal at a 1 mM concentration (75% decrease) as shown in Fig. 4D. Under the same conditions, the CAT gene expression driven by the HSV-Tk promoter (pTk-CAT) was also decreased but to a lesser extent (see Fig. 4D, inset). It has been reported that the HSV-Tk promoter contains a CCAAT box (42), which could bind NFI, and two Sp1 binding sites, which could account for this observation. 2 mM range. However, we observed, by a spectrophotometric method (43), a rapid degradation of H 2 O 2 in the culture medium, in agreement with previous studies in liver cells (6,44). Thus, only a minor fraction of the initial H 2 O 2 introduced in the medium could reach the intracellular location. In our experiments, cells did not suffer significant irreversible cytotoxicity after exposure to H 2 O 2 (Fig. 5) or to BSO and aminotriazole (data not shown). These data are consistent with published studies on hepatoma cells undergoing H 2 O 2 treatment (44 -46). It should be noted that sensitivity to H 2 O 2 depends on the cell type and that liver cells appear to be particularly resistant to such a treatment.
Characterization of NFI Complexes and Effect of H 2 O 2 -The binding of NFI to DNA was evaluated by EMSA using nuclear extracts of hepatoma cells treated or not with H 2 O 2 . In H4 cell nuclear extracts, four NFI complexes were observed in gel retardation assays (labeled as a, b, c, and d in Fig. 6). These complexes are specific as demonstrated by competition experiments (Fig. 6, A and B). In order to further characterize these complexes, we performed supershift experiments. Using a NFI C/CTF antibody, we observed that the intensity of all four complexes was decreased (Fig. 6C). However, the effect of the antibody on the different complexes was variable, and only three supershifted complexes were clearly detected (labeled as A, B, and C in Fig. 6C). The NFI complex b was entirely supershifted. Interestingly, this complex was almost completely down-regulated in nuclear extracts of cells treated with H 2 O 2 (Fig. 6C). Using an anti-NFI X antibody, only a faint supershift was observed, and the intensity of the four NFI complexes was not modified (data not shown). Thus, several NFI/DNA complexes, particularly complexes a and b, contain the NFI C/CTF isoform. In addition, these complexes are the most sensitive to H 2 O 2 treatment. We show below that NFI C/CTF is the most abundant isoform.
In HepG2 cells, NFI complexes could not be as clearly resolved as in H4 cells. Nonetheless, we also observed the formation of four complexes that were displaced in a competition experiment (Fig. 7A). The NFI C/CTF antibody also decreased NFI/DNA interaction (not shown). Treatment of HepG2 cells with a 1 mM H 2 O 2 concentration led to a 50% decrease in NFI/DNA binding (Fig. 7A).
Effect of Antioxidants on the H 2 O 2 -altered NFI⅐DNA Complexes-Antioxidant molecules prevented in vivo the action of oxidative stress on NFI/DNA binding. N-Acetylcysteine is a precursor of GSH and contributes to raising its level in hepatocytes (6). Pretreatment of H4 cell cultures with 5 mM Nacetylcysteine or 10 mM dithiothreitol prevented the effect of H 2 O 2 (data not shown).
Effect of Intracellular GSH Content on NFI/DNA Binding-To focus on the role of thiols during oxidative stress, we induced GSH depletion with BSO (cf. Table I) in both H4 and HepG2 cells. An overnight treatment of H4 cells with BSO led to a dose-dependent alteration of NFI/DNA binding. The percentage decreases were 20, 33, and 64% at the doses 10, 20, and 50 M, respectively (Fig. 7B, lanes 5-7). H 2 O 2 alone had a similar effect (lane 3). We observed that a pretreatment of cells for 20 h by BSO strengthened the inhibition of DNA binding caused by H 2 O 2 . The BSO pretreatment effect occurred in a dose-dependent manner (Fig. 7B, lanes 1-3). With 50 M BSO and 2 mM H 2 O 2 (30 min), the formation of the protein-DNA complexes was almost totally blunted (under this particular condition, cells suffered a limited loss of viability of less than 20%). Thus, BSO-pretreated cells are more sensitive to H 2 O 2 , reflecting the predominance of a glutathione-based antioxidant mechanism. It is noticeable that among the four NFI complexes, the slowest migrating ones were the most sensitive to H 2 O 2 , since they disappeared first when the intensity of the oxidative treatment increased. This is consistent with experiments described above (Fig. 6C).
Differential Regulation of NFI and CP-1-To further investigate the specificity of the effects of oxidative stress on the DNA-binding abilities of NFI, we used the Duo probe (see "Experimental Procedures") in EMSA experiments. This se-  1 h. NS, nonspecific complex; a, b, c, and d, specific NFI  complexes; A, B, and C, supershifted complexes. quence, present in the human ␣-globin gene promoter, contains adjacent NFI and CP-1 sites. In control cells, the Duo probe preferentially bound NFI rather than CP-1. The latter binding was only fully observed when NFI was displaced with an excess of unlabeled NFI oligonucleotide (Fig. 8, lane 3). We observed that, under oxidative conditions, NFI binding dramatically decreased. CP-1 could then bind to its cognate DNA sequence near the vacant NFI site (Fig. 8, lanes 5 and 6). This clear switch shows that oxidative stress influences NFI but not CP-1 DNA binding. Other experiments with HepG2 nuclear extracts and separate NFI and CP-1 probes led to the same conclusion (data not shown).
NFI Protein Levels Are Not Affected by Oxidative Stress-To test whether the alteration of the NFI/DNA binding could result from a decrease of the protein content following oxidative stress, we performed Western blot experiments. We used three different antibodies. The antibody designated Anti NFI general (Fig. 9) targets the N-terminal portion of the protein, which is highly conserved in all NFI isoforms. This antibody thus reveals all NFI isoforms present in the extracts. Anti CTF and Anti NFI X (Fig. 9) are specific antibodies raised against the NFI C/CTF and NFI X isoforms. In H4 cell nuclear extracts, several proteins are revealed by the anti-NFI general antibody (Fig. 9, right panels). A 54-kDa protein appears to be the major isoform. An isoform displaying the same molecular weight is revealed by the anti-CTF antibody (Fig. 9, left panels). Two minor bands (50 and 67 kDa) correspond to the isoforms recognized by the anti-NFI X antibody (compare middle and right panels). Oxidative stress, generated by treatment with either 2 mM H 2 O 2 for 30 min or 50 M BSO for 20 h, did not affect the nuclear levels of any NFI isoforms (cf. the Anti NFI general panels). In particular, the major isoforms NFI C/CTF and NFI X showed stable levels (see left and middle panels). Furthermore, Western blots performed with total or cytosolic extracts showed no significant variations of the contents of NFI C/CTF and NFI X (data not shown).

DISCUSSION
Most studies of gene regulation by oxidative stress have focused on induction of immediate early genes (8,47) or phase II enzymes (9,11,46). Few studies have addressed the downregulation of genes. CYP1A1 mRNAs have been shown to be under negative control by H 2 O 2 in rat hepatocytes (6). Furthermore, these mRNAs are decreased in human hepatocytes by inflammatory cytokines (5,48) known to generate reactive oxygen species within the cell (40,41). We thus hypothesized that the cyp1A1 gene promoter could be negatively regulated by oxidative stress and that this may explain, at least partially, the biological effect of these cytokines. We show here that this is indeed the case and that the transcription factor NFI mediates this negative effect. The role of other transacting elements, such as the Ah pathway, in this redox regulation, although not excluded, seems unlikely. Indeed, the mutation that inhibits NFI binding to its cognate DNA sequence located in the proximal promoter totally abolishes the redox regulation on basal as well as TCDD-induced activity. In the mutated promoter, the fold induction by TCDD is decreased by half but remains potent. This inducibility is unimpaired by oxidative stress. This study establishes the role of NFI in gene regulation by reactive oxygen species, which could be relevant for the cyp1A1 gene promoter down-regulation by cytokines. We report that the cytokine TNF␣ down-regulates CYP1A1 expression through a redox mechanism involving NFI. This cytokine is known, at least in some cellular models, to elicit reactive oxygen species production within the cell (41). Furthermore, in hepatocytes, it has been reported to activate nitric-oxide synthase expression (49), which may cause endogenous oxidative stress. It can also lead to glutathione depletion (50).
In order to generate an oxidative stress, we used either H 2 O 2 , which diffuses through membranes, or BSO, which inhibits glutathione synthesis. The alteration of cellular mechanisms regulating the redox homeostasis, particularly through GSH, elicits an endogenous oxidative stress because the basal metabolism generates oxidant species (51). These endogenous species are essentially trapped by specific enzymes such as catalase, glutathione peroxidase, or superoxide dismutase. Depleting GSH thus leads to a modification of the intracellular redox status. We show here that a severe depletion of GSH following BSO treatment tends to mimic exogenous oxidative stress and down-regulate cyp1A1 gene promoter activity through NFI. Aminotriazole, which inhibits catalase, also down-regulates the NFI/TATA promoter. These experiments show that xenobiotics that are not oxidants themselves can induce oxidative stress and have the same consequence on NFI activity. This down-regulation elicited by different species such as H 2 O 2 , aminotriazole, BSO, or TNF␣ converge on a common target via a redox mechanism.
The ability of NFI to mediate a negative regulation by oxidative stress was not specific to the cyp1A1 gene promoter. It could also be observed in the context of the heterologous HSV-Tk promoter containing consensus NFI binding sites. Furthermore, EMSA experiments allowed us to show that oxidative stress induced in vivo by H 2 O 2 or glutathione depletion leads to an alteration of NFI binding to its cognate DNA sequence. We could identify the 54-kDa NFI C/CTF as an isoform that is particularly sensitive to oxidative stress. It seems unlikely that this effect involves a regulation of the amounts of the NFI proteins themselves. Indeed, Western blots showed that neither H 2 O 2 nor BSO treatment of the cells was followed by a loss of the NFI proteins in the nuclear (and cytosolic) location, suggesting that oxidative stress alters the efficiency of NFI/DNA interaction. In addition, the pretreatment of cells with antioxidants prevents the effect of H 2 O 2 . Those data are in agreement with in vitro studies that correlated the efficiency of NFI binding to DNA with a reversible oxidation of strategic cysteines (26, 52). However, our data suggest that the alteration of the DNA binding only partially accounts for the repression of NFI transactivation function by oxidative stress. Indeed, the transactivating function of NFI is more sensitive to oxidative stress than its DNA binding activity. Submillimolar H 2 O 2 concentrations (this work) as well as TNF␣ treatment (53) do not affect the binding, yet we show that they clearly inhibit the transcription driven by NFI. In the case of other ubiquitous transcription factors such as AP-1 or NF-B, the effects of reactive oxygen species on their transactivating functions (which are increased) and on their DNA binding abilities (which are impaired by the oxidation of a strategic cysteine) are opposite (10,54). As for NFI, both DNA binding and transactivating function are altered by oxidative stress, which makes this transcription factor particularly sensitive to changes of the redox status within the cell.
Reactive oxygen species play an important role in signal transduction (7,55). Cysteine thiol groups (the main redoxsensitive moiety) may react differently to a wave of oxidative stress according to the protein conformation and to their own redox potentials. Thus, a change of the intracellular redox status may alter some transcription factors, while others are unaffected. In this study, we report that both the transactivating function and DNA binding of CP-1, a ubiquitous transcription factor, remained unaffected by oxidative conditions that altered NFI function. The CP-1 DNA-binding domain contains two cysteines, but their mutation to serines does not affect DNA binding (15), contrary to the mutations in NFI (26). When CP-1 and NFI compete for two close DNA binding sites, as in the human ␣-globin gene proximal promoter, we observed an interesting NFI/CP-1 redox switch. NFI and CP-1 are known to compete for the binding to at least another promoter sequence in the gene of the ␣1b adrenergic receptor (56). Furthermore, NFI and AP-2 (which is activated by oxidative stress (57)) compete for the binding to the promoter region of the human growth hormone gene (58). Thus, further investigations have to be undertaken with other promoter sequences containing both binding sites to assess the potential biological relevance of redox switches involving NFI.
The differential response of various transcription factors toward the imbalance of the redox state within the cell may be a new regulatory pathway for gene regulation. We report here that NFI mediates the human cyp1A1 gene inhibition by oxidative stress. Furthermore, the CP-1 transcription factor has been reported to have a negative regulatory function on this expression (4). Since we show here that it is insensitive to oxidative stress, this could further explain the strong inhibition of CYP1A1 by oxidative stress. NFI plays an important role in the regulation of many genes (see the Introduction). The downregulation of its activity by redox mechanisms could have diverse physiological consequences. Interestingly, NFI has been shown to be a repressor of the expression of two enzymes (glutathione S-transferase and metallothionein) that are induced by oxidative stress (59).
As previously mentioned, cyp1A1 belongs to the Ah-inducible gene battery, which comprises other phase I and phase II drug-metabolizing enzymes. In all these genes, including cyp1A1, induction by TCDD is mediated by Ah receptor binding sites (xenobiotic-responsive elements) (2). Interestingly, four of the phase II genes of this battery (among them UDP glucuronosyltransferase, menadione oxidoreductase, and aldehyde dehydrogenase) have been reported to be up-regulated by oxidative stress through an electrophile response element/antioxidantresponsive element) (60). In this respect, the regulation of cyp1A1 is opposite, and it is mediated by the NFI site present in its proximal promoter (besides, the cyp1A1 promoter does FIG. 9. Influence of oxidative stress on the nuclear levels of NFI proteins. Western blots were performed with 10 g of H4 cell nuclear extracts from cells treated or not treated with H 2 O 2 (2 mM, 30 min) or BSO (50 M, 20 h). Three immunoblots were performed with anti-NFI C/CTF, anti-NFI X, and an antibody targeting the highly conserved N terminus of all NFI isoforms (Anti NF-1 general panels). not contain any electrophile response element site). Such a negative regulation could be metabolically relevant, since several cytochromes P450 generate H 2 O 2 as an effect of the reactions they catalyze (61). In the case of CYP1A1, it has been associated with oxidative DNA damage (62). It is thus conceivable that H 2 O 2 generation could result in a retrocontrol loop limiting the expression or induction of this gene. An alternative mechanism could be suggested by a study showing that TCDD increases TNF␣ mRNA levels (63), which could also lead to CYP1A1 down-regulation.