NF-κB Regulates Phagocytic NADPH Oxidase by Inducing the Expression of gp91phox*

The superoxide-generating phagocytic NADPH oxidase is an important component of the innate immune response against microbial agents, and is involved in shaping the cellular response to a variety of physiological and pathological signals. One of the downstream targets of NADPH oxidase-derived radicals is the ubiquitous transcription factor NF-κB, which controls the expression of a large array of genes involved in immune function and cell survival. Here we show that NF-κB itself is a key factor in controlling NADPH oxidase expression and function. In monocytic and microglial cell lines, the expression of the NADPH oxidase subunit gp91phox was induced by lipopolysaccharide/interferon γ treatment and was inhibited in cells constitutively expressing IκBα. Furthermore, inducible reactive oxygen species production was inhibited in IκBα overexpressing cells. gp91phox expression was very low in RelA-/- fibroblasts and could be induced by reconstituting these cells with p65/RelA. Thus, gp91phox expression is dependent on the presence of p65/RelA. We also found that gp91phox transcription is dependent on NF-κB and we identified two potential cis-acting elements in the murine gp91phox promoter that control NF-κB-dependent regulation. The findings raise the possibility of a positive feedback loop in which NF-κB activation by oxidative stress leads to further radical production via NADPH oxidase.

The superoxide-generating phagocytic NADPH oxidase is an essential component of the antimicrobial host defense system and is therefore highly expressed in granulocytes and macrophages. It catalyzes the reduction of molecular oxygen to superoxide (O 2 . ) by using NADPH as an electron donor. Superoxide is a ROS 2 intermediate, which is the starting material for generation of a variety of reactive oxidants including hydrogen peroxide, hydroxyl radical, singlet oxygen, reactive halogens, and peroxynitrite. Because of the destructive action of most of these radicals, NADPH oxidase activity must be tightly regulated. NADPH oxidase is composed of five subunits, p40 phox (phox for phagocytic oxidase), p47 phox , p67 phox , p22 phox , and gp91 phox . The latter two are membrane associated and together constitute the flavocytochrome b 558 , whereas the other components are located in the cytoplasm of resting cells. Upon activation, the cytoplasmic components translocate to the cell membrane where they bind to flavocytochrome b 558 , thus forming the active NADPH oxidase (reviewed in Ref. 1). In addition to the five classical subunits, the small GTPases Rac1 (2) and Rac2 (3) have emerged as important functional units that are integrated in the membrane bound enzymatic complex and are indispensable for NADPH oxidase activity. More recent evidence indicates that NADPH oxidase is also present in non-myeloid cells, such as myocytes (4), fibroblasts (5,6), as well as endothelial (7,8) and neuronal cells (9). However, in these cells gp91 phox (Nox2) might not constitute the dominant catalytic subunit, as Nox1 is the major form expressed in vascular smooth muscle cells (10) and Nox4 the most abundant isoform in fibroblasts (6) and endothelial cells (11). In most non-myeloid cells, NADPH oxidase seems to have a regulatory role in controlling or modulating several signaling cascades initiated by such molecules as low density lipoproteins (12), angiotensin II (4,13), and amyloid-␤ peptides (14,15). Therefore, in addition to diseases related to the immune system, NADPH oxidase has also been implicated in hypertension, atherosclerosis, Alzheimer disease, and stroke (14, 16 -18). At the transcriptional level, regulation of the gp91 phox encoding CYBB gene is best characterized. In cells of the myelomonocytic lineage CYBB expression is tightly regulated by transcriptional activators and repressors. A special role is attributed to the lineage specific transcription factor PU.1, which binds alone or in a complex with Elf-1, IRF-1 (interferon regulatory factor-1), and ICSBP (interferon consensus sequence-binding protein), and combined is termed HAF-1 (hematopoiesis-associated factor-1), to the CYBB promoter (19,20), and is also involved in regulating lineage-specific expression of NCF4 (p40 phox ), NCF1 (p47 phox ), and NCF2 (p67 phox ) (21)(22)(23). Additionally, CP1 (CCAAT-binding protein-1) and YY1 are recruited to the promoter (24,25). In immature myelocytes, binding of transactivators is inhibited by repressors CDP (CCAAT displacement protein), HOXA10 (homeobox protein A10), and SATB1 (special AT-rich sequence binding protein 1), which are down-regulated or deactivated during functional maturation (26 -28). Little is known regarding regulation of NADPH oxidase constituting genes in non-phagocytic cells.
The ubiquitously expressed NF-B is a key transcription factor in regulating expression of pro-inflammatory, immune modulatory, and anti-apoptotic genes (29). NF-B is a homo-or heterodimeric complex composed of Rel family proteins p50, p52, Bcl3, RelB, cRel, and p65 (30). Cellular activation by a broad array of stimuli including cytokines, bacterial lipopolysaccharides, viruses, and radiation results in liberation of dimeric NF-B from cytoplasmic inhibitory molecules (IBs). Upon nuclear import and binding to specific decameric recognition motifs, which are reflected by the consensus GGGRHTYYCC (R ϭ purine, Y ϭ pyrimidine, and H ϭ not G), NF-B dimers function as trans-acting elements in the promoter region of NF-B-dependent genes (30).
It has been shown that NF-B can be activated by NADPH oxidase through ROS intermediates (31)(32)(33)(34). In this study we investigated whether subunits of NADPH oxidase might be regulated by NF-B itself, thus establishing a potential self-perpetuating activation cycle. We show that in mouse monocytic and microglial cell lines, as well as embryonic fibroblasts, inducible expression of gp91 phox is dependent on the presence of the p65 subunit of NF-B.

Cell Culture and Reagents
The monocytic murine J774.A1 cell line was purchased from ATCC. BV-2 murine microglial cells, which were developed in the laboratory of Dr. Virginia Bocchini (35), were obtained from Dr. Sunghee Cho (Weill Medical College). MEF isolated from RelA Ϫ/Ϫ mice were kindly provided by Dr. A. Beg (Columbia University, New York) and have been described previously (36). All cells were grown at 37°C in Dulbecco's modified Eagle's medium (MediaTech Inc., Herndon, VA) supplemented with 10% fetal bovine serum, 100 units/ml penicillin G, and 100 g/ml streptomycin B (all Atlanta Biologicals, Norcross, GA) in a humidified atmosphere containing 5% CO 2 . Endotoxin levels of all cell culture reagents were Յ0.24 endotoxin units/ml. LPS was purchased from Sigma (L-7261) and recombinant murine IFN␥ was from Calbiochem.

Plasmid Constructs
Expression Vectors-The porcine IB␣ cDNA containing Ser to Ala substitutions at amino acid positions 32 and 36 (IB␣ AA) and therefore coding for a stabilized form of IB␣ was a kind gift from Rainer De Martin (Department of Vascular Biology & Thrombosis Research, Medical University of Vienna, Austria). The cDNA was cloned into the retroviral expression vector pQCXIN (Clontech). Retroviral vectors carrying the human p65 cDNA and its serine mutants have been described previously (37). The full-length coding region of murine gp91 phox was amplified from cDNA prepared from LPS/IFN␥-stimulated J774.A1 cells using Pfu polymerase (Stratagene, La Jolla, CA). The sequence of the 5Ј-primer, which contained a BsiWI restriction site, was 5Ј-ATTA-CGTACGCCAACACTTAACCTTTGCTAACAT-3Ј, and the sequence of the 3Ј-primer containing a BamHI restriction site was 5Ј-A-TTAGGATCCCATTTGGCAGCATACACTGG-3Ј, respectively. The amplicon was digested with BsiWI/BamHI and cloned into pQCXIP retroviral vector (Clontech).

Reporter Constructs and Transient Transfections
A DNA fragment spanning from ϩ38 to Ϫ1935 (relative to the transcriptional start site) of the murine CYBB promoter was amplified by PCR from C57/BL6 genomic DNA and cloned into pGL3 vector (Promega, Madison, WI) using SacI and NcoI restriction sites. Similarly, a reporter spanning ϩ38 to Ϫ1605 was constructed using NcoI and NheI restriction sites. Reporter constructs carrying a tandem repeat of respective B sites were constructed in analogy to previously published procedures (37). Inserted sequences of all plasmid constructs were verified by automated DNA sequencing using Dye Terminator chemistry. Reporter gene assays in MEF were performed as described previously (37). Raw 264.7 cells were seeded in 12-well plates and transfected 24 h later. One g of DNA was mixed with 3 l of FuGENE 6 reagent (Roche Biochemicals) in Opti-MEM medium (Invitrogen, Carlsbad, CA) according to the manufacturers suggestions and the mixture was added directly to the cells. A plasmid carrying ␤-galactosidase under control of the elongation factor-1␣ promoter (pEF4-V5-His-LacZ, Invitrogen) was used to normalize luciferase levels in Raw cells. To minimize endotoxin exposure, plasmid DNA used in transfections was cleared of contaminating endotoxin (EndoFree, Qiagen, Valencia, CA).

Stable Transfectants
Retrovirus containing supernatants were obtained after transiently transfecting EcoPack2-293 ecotropic or AmphoPak amphotropic packaging cell lines (Clontech) using Lipofectamine (Invitrogen). Vesicular stomatitis virus-G pseudotyped viral particles were generated by transfecting GP-293 cells with the respective construct and pVSV-G vector (Clontech). Viral titers were determined by titration on NIH 3T3 cells and were Ͼ1 ϫ 10 5 /ml. RelA Ϫ/Ϫ MEF were infected as described (37). J774.A1 and BV-2 cells were infected with amphotropic or vesicular stomatitis virus-G pseudotyped virus using centrifugal delivery (38). Briefly, cells were exposed to retroviral supernatants in the presence of 8 g/ml Polybrene and centrifuged at 2000 ϫ g for 2 h at 25°C. Cell pellets were resuspended in fresh medium and returned to the incubator. Transduced cell pools were selected 72 h after infection in medium containing 1 mg/ml Geneticin or 1.5 g/ml puromycin over 14 -21 days after which cells were used for experiments.

Nuclear Run-off
J774.A1 cells were treated with LPS (100 ng/ml) and IFN␥ (100 units/ ml) for 2 h. Cells were washed once in cold Hanks' balanced salt solution and collected by scraping. Cells were collected by centrifugation at 4°C and resuspended in hypotonic buffer (10 mM Hepes, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl). After 15 min incubation on ice, cells were lysed by applying 20 strokes in a Dounce homogenizer using the tight fitting pestle. Nuclei were collected by centrifugation, resuspended in storage buffer (50 mM Tris-Cl, pH 8.3, 40% glycerol, 5 mM MgCl 2 , and 0.1 mM EDTA) and stored at Ϫ70°C until use. Nuclei (2 ϫ 10 7 in 200 l of storage buffer) were thawed at room temperature and mixed with an equal amount of 2ϫ reaction buffer (10 mM Tris-HCl, pH 8.0, 5 mM MgOAc, 0.3 M KCl, 0.5 mM ATP, GTP, UTP, 2 mM dithiothreitol, 24 g/ml creatine phosphokinase, and 24 mM creatine phosphate). After addition of 120 Ci of [␣-32 P]CTP (800 Ci/mmol), nuclei were incubated at 30°C for 30 min with continuous agitation. Reactions were terminated by adding 10 l of 1 mg/ml RNase-free DNase I (Roche Biochemicals). After 10 min incubation at 30°C, 5 l of 10% SDS and 150 g of proteinase K (Worthington, Lakewood, NJ) were added and reactions were incubated for 1 h at 37°C. RNA was extracted using TRIzol LS (Invitrogen) according to the manufacturers suggestions. Preparation of membranes, hybridization, and wash steps were performed as described elsewhere (41). The following plasmids (5 g) were linearized with ScaI and immobilized to nitrocellulose membranes: ICAM-1 (pGEM-T containing a 685-bp fragment of the murine ICAM-1 coding sequence), gp91 phox (pGEM-T containing a 1565-bp fragment of the murine gp91 coding sequence), and hypoxanthine guanine phosphoribosyl transferase (pGEM-T containing a 1054-bp fragment of the murine hypoxanthine guanine phosphoribosyl transferase coding sequence).

Chromatin Immunoprecipitation (ChIP)
ChIP was performed by a combination of published protocols (42,43). Briefly, 10 7 cells were cross-linked by adding formaldehyde (1% final) directly to the growth medium. Plates were agitated on an orbital shaker for 10 min at room temperature. After two washes with cold phosphate-buffered saline, cells were collected in 5 ml of cell collection buffer (100 mM Tris-HCl, pH 9.4 and 10 mM dithiothreitol) and incubated on ice for 15 min and subsequently at 30°C for 15 min. Cells were collected by centrifugation and resuspended in 500 l of lysis buffer (10 mM EDTA, 50 mM Tris-HCl, pH 8.0, 1% SDS, 0.5% EmpigenBB). Lysates were incubated for 10 min on ice and sonicated 4 times for 15 s at 20% of maximal amplitude (Sonifier Cell Disruptor 250, Branson, Danbury, CT). This led to DNA fragments spanning from 0.3 to 1.5 kb as analyzed by agarose gel electrophoresis. Chromatin was cleared by a 10-min centrifugation at 16,000 ϫ g at 4°C and stored at Ϫ70°C. Soluble chromatin derived from 2 ϫ 10 6 cells was used for each immunoprecipitation. Samples were diluted with 10 volumes of dilution buffer (0.2 mM EDTA, 167 mM NaCl, 16.7 mM Tris-HCl, pH 8.1, 0.01% SDS, and 0.5% Triton X-100) and subjected to immunoprecipitation with 2 g of anti-p65 (sc-372, Santa Cruz Biotechnology) or 2 g of preimmune IgG overnight after a 1-h preclearing at 4°C with protein A-Sepharose (80 l of 50% slurry in 10 mM Tris-HCl, pH 8.1, 1 mM EDTA, 0.5% bovine serum albumin, and 19.2 g of glycogen). 1/100 volume of the soluble chromatin was removed after preclearing (input). Complexes were recovered by a 1-h incubation at 4°C with 80 l of blocked protein A-Sepharose slurry. Precipitates were serially washed for 5 min with 300 l of Wash Buffer I (2 mM EDTA, 20 mM Tris-HCl pH 8.0, 0.1% SDS, 1% Triton X-100, 150 mM NaCl), twice with Wash Buffer II (20 mM Tris-HCl, pH 8.0, 2 mM EDTA, 500 mM NaCl, 1% Triton X-100, 0.1% SDS), once with Wash Buffer III (1 mM EDTA, 10 mM Tris-HCl, pH 8.0, 1% Nonidet P-40, 1% deoxycholate, 0.25 M LiCl), and then twice with TE. All buffers were supplemented with protease inhibitors. Precipitated chromatin complexes were removed from the beads by a 30-min incubation with 50 l of elution buffer (1% SDS, 0.1 M NaHCO 3 ) on a rocking platform. This step was repeated twice, with a 10-min incubation. Eluates were separated from the remaining agarose beads by centrifugation over 0.45-m pore size spin columns. Cross-linking was reversed by 4 h to overnight incubation at 65°C. Immunoprecipitated and input DNA was purified with QIAquick purification columns (Qiagen) as described (42). DNA was analyzed for the presence of CYBB promoter sequences by real-time PCR using SYBR Green chemistry (Invitrogen) as described above using ChIP primer set 1 (5Ј-TTATGGGAACAGCCTT-TCAGTT-3Ј and 5Ј-TGTTGGTTCTGCCTCTCTTCT-3Ј; 138-bp amplicon) and ChIP primer set 2 (5Ј-GCTTCAGTGAGGAC-CCAATC-3Ј and 5Ј-CCACTTTTCCATCATCCATGT-3Ј; 139-bp amplicon). Input DNA was amplified for each sample in parallel and amounts of sequence-specific immunoprecipitated DNA were expressed as the percentile fraction of input DNA. Both primer pairs amplified a single band with the expected molecular weight as analyzed by agarose gel electrophoresis. PCR products were purified over silica gel columns (Qiagen) and cloned into pGEM-T (Promega) vector. Two individual clones were sequenced to verify the correctness of the amplicon.

Cellular ROS Production
ROS production was assayed by flow cytometry in 2Ј,7Ј-dihydrodichlorofluorescein diacetate (DHDCF)-labeled cells. Cells were seeded in 6-well plates. Subconfluent cells were pretreated for 4 h with LPS/IFN␥ as indicated. Cells were washed once with RPMI 1640 (without phenol red) and incubated in RPMI 1640 containing 10 M DHDCF in the presence or absence of 3.2 M phorbol 12-myristate 13-acetate for 30 min at 37°C. After two more washes, cells were collected in phosphatebuffered saline containing 5 g/ml propidium iodide and analyzed by flow cytometry on a Beckman-Coulter XL flow cytometer (Weill Medical College, Flow Cytometry Core Facility). Propidium iodide-positive cells were excluded from the analysis. Data were analyzed using Win-MDI version 2.8 software (The Scripps Institute, Flow Cytometry Core Facility).

Superoxide (O 2 . ) Production in a Cell-free System
␤-NADPH-dependent O 2 . production was measured in cell-membrane fractions by an enhanced luminol-based assay as described previously (44). Cells were washed twice in Hanks' balanced salt solution and collected by centrifugation. After resuspension in lysis buffer (10 mM Tris-HCl, pH 7.1, 300 mM sucrose, 1 mM MgCl 2 , 1 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, 5 M leupeptin, 1 M pepstatin A, and 10 g/ml aprotinin), cells were disrupted on ice by sonication for three 10-s bursts at 11% power output separated by 30-s intervals. Unbroken cells, nuclei, and large debris were removed by centrifugation at 10,000 ϫ g for 10 min at 4°C. The supernatant was centrifuged at 100,000 ϫ g for 60 min at 4°C. The pellet, which constitutes the membrane fraction, was dissolved in assay buffer (100 mM potassium phosphate, pH 7.

Data Analysis
All data were derived from at least three independent experiments. Data were expressed as mean Ϯ S.E. Two-group comparisons were analyzed by the two-tailed t test for dependent or independent samples, as appropriate. Multiple comparisons were evaluated by ANOVA and Tukey's tests. Statistical significance was considered for p Ͻ 0.05.

RESULTS
Regulation of gp91 phox mRNA Levels by NF-B-To test whether the expression of phagocytic NADPH oxidase subunits was dependent on NF-B, J774.A1 macrophages were retrovirally transduced to express a form of IB␣ that was missing two phosphoacceptor sites responsible for signal-induced degradation (IB␣ AA). Cells expressing IB␣ were less responsive to LPS/IFN␥ in up-regulating NF-B-dependent genes as Mn-superoxide dismutase or interleukin-6 ( Fig. 1A). Exposure of cells to LPS/IFN␥ resulted in an 11.3-fold induction of mRNA for the gp91 phox subunit. Induction was significantly reduced in IB␣ overexpressing cells (3.2-fold). Besides gp91 phox , p47 phox expression was also up-regulated in LPS/IFN␥-treated cells (3.3-fold) and significantly inhibited by IB␣ overexpression (1.2-fold induction). There was a minimal but significant induction of p22 phox (1.5-fold) and Rac2 (1.8-fold) after LPS/IFN␥ stimulation and this increase was blunted in IB␣ overexpressors. Rac1 and p67 phox expression were not affected. Protein levels were regulated in a similar manner. gp91 phox protein was readily induced in J774 cells after a 6-h LPS/IFN␥ treatment. In IB␣ AA expressing cells, however, basal and inducible protein levels were diminished (Fig. 1B). A similar regulation of gp91 phox expression was found in IB␣ AA transduced Raw 264.7 cells (data not shown). p47 phox protein was also induced by LPS/IFN␥ treatment as was to a lesser extend p22 phox . Induction of both genes was blunted in IB␣ AA expressing FIGURE 1. A, relative expression levels of indicated genes as quantified by qPCR. J774.A1 murine monocytic cells were transduced with empty retroviral vector (vector) or an IB␣ carrying retrovirus (IB␣ AA). Cells were stimulated with LPS (100 ng/ml) and IFN␥ (100 units/ml) for 6 h. Expression levels of untreated (control) empty vector-transduced cells were set to 1. Asterisk, p Ͻ 0.05 from stimulated vector, n ϭ 3, t test. B, gp91 phox , p22 phox , and p47 phox protein levels were detected by Western blotting. Immunodetection of ␤-actin serves as loading control. SOD, superoxide dismutase.
cells. In contrast to gp91 phox , however, IB␣ did not diminish constitutive protein levels.
Similar experiments were conducted in BV-2 murine microglia. In empty vector-transduced cells, gp91 phox expression was induced 6.6fold after treatment with LPS/IFN␥. The same treatment resulted in a 2.7-fold induction in IB␣ AA expressing cells (Fig. 2). However, we did not observe any significant changes in expression levels of other NADPH oxidase subunits and Rac1/2 proteins after LPS/IFN␥ treatment or NF-B inhibition.
The p65 Subunit of NF-B Is Needed for Basal and Inducible gp91 phox Expression-Because IB␣ has the potential to inhibit several NF-B proteins, we were investigating which NF-B subunit is indispensable for gp91 phox expression. From the six mammalian Rel proteins, p65 and cRel are thought to have the highest trans-activating potential. We therefore targeted the p65 subunit by RNA interference. Retroviral expression of a siRNA directed against the 3Ј-untranslated region resulted in efficient inhibition of p65 expression in J774.A1 cells (Fig.  3A). Delivery of a similar siRNA directed against luciferase had no effect on p65 levels. Examination of gp91 phox mRNA levels revealed that basal and inducible levels were significantly reduced in cells depleted of p65 as compared with cells expressing an unrelated siRNA (Fig. 3B). To ensure specificity of this approach, we transduced p65 siRNA expressing cells with a retrovirus carrying a human p65 that was not targeted by the mouse-specific siRNA. Transgene expression levels were comparable with endogenous p65 and did not interfere with siRNA-mediated repression of murine p65 (Fig. 3C). As shown in Fig. 3B reconstituting p65 in these cells completely reverted gp91 phox mRNA repression and expression levels became indistinguishable from ones observed in wild type cells. Western blot analysis of gp91 phox expression in these cells showed that basal and LPS/IFN␥ responsive protein levels were greatly reduced in p65 siRNA expressing cells (Fig. 3D).
To further emphasize the vital role of p65 in regulating gp91 phox expression, we examined gp91 phox expression in MEF derived from p65/ RelA Ϫ/Ϫ mice (36). This experimental system allows for direct assessment of the role of p65 in regulating NADPH oxidase subunit expres-sion. Basal mRNA levels were very low in RelA Ϫ/Ϫ MEF but could be induced by LPS/IFN␥ treatment (20-fold). Transduction of these cells with a p65 expressing retrovirus established a 40-fold higher basal level most likely by inducing a constitutive presence of p65 containing NF-B complexes in the nucleus. Upon exposure to LPS/IFN␥ for 6 h, gp91 phox mRNA levels were increased 350-fold over basal levels seen in p65deficient MEF (Fig. 4A). As in BV-2 microglial cells, expression of p22 phox , p47 phox , p67 phox , and Rac1 were not significantly changed by introduction of p65, whereas Rac2 was undetectable in these cells. NOX4, which is thought to be the predominant NOX isoform in fibroblasts and endothelial cells, was not regulated by LPS/IFN␥ treatment or NF-B p65 expression. Although gp91 phox expression levels were low in RelA Ϫ/Ϫ MEF (28.4 cycles to threshold (C T ), supplemental Table), expression levels in cells reconstituted with p65 and stimulated with LPS/IFN␥ (22.6 mean C T ) were comparable with the ones observed in resting monocytic cells (22.4 mean C T ). As in RelA Ϫ/Ϫ MEF reconstituted with p65, gp91 phox expression was highly inducible in wild type MEF (Fig. 4B). gp91 phox protein was undetectable in RelA Ϫ/Ϫ MEF. After reconstituting these cells with p65, gp91 phox could be detected by Western blotting and protein levels were further increased by LPS/IFN␥ treatment (Fig. 4C).

LPS/IFN␥ Inducible ROS Production in J774 Macrophages Is
Dependent on NF-B Activation-We next investigated whether the observed reduction in gp91 phox mRNA and protein levels in IB␣ overexpressing cells resulted in decreased ROS production. Cells were treated with LPS/IFN␥ for 4 h and a "respiratory burst" was induced with phorbol 12-myristate 13-acetate. J774 cells transduced with IB␣ AA expressing retrovirus failed to significantly increase ROS production, whereas empty vector-transduced cells showed a 5-fold increase in DHDCF fluorescence (Fig. 5, A and C), which was comparable with values observed in native J774.A1 cells (Fig. 5C). Because inhibiting NF-B can potentially affect other ROS generating systems such as xanthine oxidase (45), cyclooxygenase-2 (46), and inducible nitric-oxide synthase (47), all capable of producing superoxide, we tested whether IB␣ overexpressing cells would regain their ability to produce ROS after re-introduction of gp91 phox . IB␣ expressing cells transduced with a vector constitutively expressing gp91 phox , but not cells transduced with the empty retroviral vector regained the ability to produce ROS at levels comparable with wild type cells (Fig. 5, B and C) production after LPS/IFN␥ treatment (Fig. 5D). In contrast, membranes derived from LPS/IFN␥-treated IB␣ AA expressing cells did not show any increase over basal activity levels.   Identification and Characterization of NF-B-responsive cis-Acting Elements in the Murine gp91 phox Promoter-Although the experiments described above suggest a direct involvement of NF-B in the regulation of gp91 phox expression, it is not clear whether the observed increase in mRNA levels was initiated by a transcriptional event. Nuclear run-off experiments were performed to address whether LPS/IFN␥ treatment induced de novo gp91 phox RNA synthesis. De novo synthesis of ICAM-1, which is highly inducible in monocytic cell lines (49), as well as gp91 phox mRNA was clearly enhanced in nuclei derived from LPS/IFN␥-stimulated cells (Fig. 6B). In contrast, transcription of the housekeeping gene HPRT was not altered by the treatment. These findings establish a strong transcriptional component in the up-regulation of gp91 phox by LPS/IFN␥ treatment.
The murine gp91 phox promoter contains two potential NF-B binding sites at positions Ϫ1788 and Ϫ1819 relative to the transcriptional start site, herein termed B1 and B2. B1 (5Ј-GGGATTCTCC-3Ј) is reversely located on the complementary DNA strand and fits the conservative GGGRHTYYCC B consensus (Fig. 6A). B2 (5Ј-AGG-GACTTTC-3Ј) is a less conserved consensus featuring an adenine at its 5Ј-end.
To determine whether these sites are occupied by NF-B in vivo, we performed chromatin immunoprecipitation with antibodies directed against p65. RelA Ϫ/Ϫ MEF transduced with retroviral vector alone or with a vector expressing p65 were left untreated or were stimulated with LPS/IFN␥ for 1 h. Soluble cross-linked chromatin was precipitated with p65 specific antibody, which revealed a stimulation dependent and specific enrichment of sequences surrounding the two potential B sites (Fig. 6C, primer set 2) but not sequences close to the transcriptional start site of the CYBB gene (Fig. 6C, primer set 1). There was no enrichment of this sequence in ChIPs of RelA Ϫ/Ϫ MEF. In monocytic J774 cells, p65 binding was also evident in cells treated with LPS/IFN␥ for 2 h (Fig. 6D). There was no inducible binding of p65 in IB␣ AA expressing cells or in cells depleted of endogenous p65 by siRNA expression. As in MEF, binding was localized to promoter sequences surrounding putative B elements (primer set 2) but not to sequences close to the transcriptional start site (primer set 1). Hence, these experiments establish that p65 binds to the gp91 phox promoter in an inducible fashion and localizes p65 binding to a promoter region carrying two putative B consensus sites.
We next analyzed NF-B binding in vitro by performing EMSA with 21-meric (B2) and 24-meric (B1) double-stranded DNA oligonucleotides representing promoter sequences containing the putative B elements. Nuclear extracts from untreated J774 cells or cells treated with LPS/IFN␥ for 1 h were incubated with radiolabeled probes and complexes were separated by native electrophoresis (Fig. 7A). There was constitutive binding to the B element in untreated cells (lane 1), which could be supershifted with anti-p50 antibody (data not shown). Higher molecular weight complexes appeared after LPS/IFN␥ treatment (lane 2). An excess of corresponding unlabeled oligonucleotide but not mutated oligonucleotide was able to compete for DNA binding complexes (lanes 3 and 4). Supershift assays using antibodies against p65, p50, cRel, and p52 revealed the presence of p65 and p50 subunits in LPS/IFN␥-treated cells (lanes 5 and 6). Nuclear extracts from LPS/ IFN␥-treated J774 cells expressing IB␣ AA or p65 siRNA showed decreased binding activity of higher molecular weight complexes that have been shown to contain p50 and p65 proteins (Fig. 7B). To further characterize the putative cis-acting elements, we used luciferase reporter constructs spanning from the translational start site to the internal NheI site (NheI-Luc, ϩ38 to Ϫ1605), thus excluding the putative B sites, or a reporter that reaches to the internal SmaI site, those including these elements (SmaI-Luc, ϩ38 to Ϫ1935). When transfected together with p65 in RelA Ϫ/Ϫ MEF, the SmaI-Luc reporter was induced 9-fold, whereas only a 2-fold induction was observed with the NheI-Luc construct (Fig. 8A). Combining p65 with p50 or cRel gave essentially similar results. In contrast to p65, p50 or cRel were not able to induce the reporter constructs on their own (data not shown). To study the two B sites in a CYBB promoter independent setting, we generated luciferase reporter constructs that harbor a tandem copy of the respective B sites in front of a minimal promoter as described previously (37). Both B1 and B2 carrying reporters were strongly induced by p65 when transfected into RelA Ϫ/Ϫ MEF (Fig. 8B). Co-transfection of p50 together with p65 reduced reporter gene activity hinting that p50 might act as a repressor on these sites, whereas co-transfection of cRel did not significantly alter transcriptional activity as compared with p65. p50 or cRel were not able to induce the reporter constructs on their own (data not shown). In monocytic Raw 264.7 cells expression of the NheI-Luc reporter was induced 2-fold after LPS/IFN␥ treatment (Fig. 8C). This induction was not inhibited by co-transfecting IB␣ AA or p65 siRNA expression vectors. The SmaI-Luc reporter construct showed 10.5-fold higher activity after LPS/IFN␥ treatment as compared with NheI-Luc reporter activity in resting cells. Luciferase activities were significantly reduced by co-transfecting IB␣ AA or p65 siRNA expression vectors.
We recently described the importance of phosphorylation sites within the p65 Rel homology domain for targeting NF-B transcrip-tional activity to specific gene subsets (37). Although we could not predict the behavior of the B2 site because of its unusual composition, the B1 site matches the group II consensus (5-KGRAHWTYCC-3Ј). Therefore, we analyzed gp91 phox expression in RelA Ϫ/Ϫ MEF reconstituted with wild type p65 or p65 mutants S205A, S276A, and S281A. As predicted by our model, gp91 phox expression was activated strongest by wild type p65, followed by serines 205, 276, and 281 mutants with decreasing activity (Fig. 8D).

DISCUSSION
The transcriptional regulation of NADPH oxidase subunits has been studied mainly in macrophages and leukocytes. In these cells, most efforts have been directed toward understanding the increase in NADPH oxidase activity during myeloid cell maturation (50), a process mainly regulated by lineage-specific transcriptional activators and repressors (51,52). The importance of these transcription factors is highlighted by the fact that expression and activity levels of NADPH oxidase in myelomonocytic cells is 100 -1000-fold higher than in nonphagocytic cells (7). 3 In this study we provide evidence that expression of the catalytic NADPH oxidase subunit gp91 phox is controlled by NF-B in a cell-type independent fashion. Previous studies have shown that LPS and IFN␥ can induce gp91 phox mRNA expression in human monocytes, although the transcriptional mechanism has not been revealed (53,54). Here we identify two putative B elements within the murine CYBB promoter that are essential for translating LPS/IFN␥mediated gene induction. This is based on the finding that these elements bind p50 and p65 in vitro, and that inducible p65 binding to the promoter region harboring these sites is detected in vivo. Furthermore, deletion of the two putative B elements from a CYBB promoter fragment abolished the response to NF-B. We did not address whether one element is dominant over the other. EMSA studies revealed a stronger binding to the B2 site. When these B sites were placed upstream of an artificial minimal promoter, however, NF-B-dependent transactivation was stronger from B1 than from B2 sites. Because of their close proximity within the CYBB promoter, it is not unlikely that they exert a synergistic effect as seen with other genes carrying multiple B elements (47). We also identified corresponding elements in the human CYBB promoter located 3.5-kb upstream of the transcriptional start site. The region shows high sequence homology with the murine locus (Fig. 8E). Future studies will show whether these sites have similar functionality.
In this study we did not investigate the mechanism by which NF-B overwrites the inhibitory effect of repressors such as CDP or SAT1B that might be responsible for low gp91 phox expression levels in nonmyeloid cells such as fibroblasts. However, other genes that contain both CDP/SATB1 and NF-B binding sites are inducible upon NF-B binding despite the presence of repressors. For example, NF-B can induce c-myc expression in Pre-B cells (55) and p21 Cip1/Waf1 in keratinocytes (56). Because CDP and SAT1B both interact with CREB-binding protein (CBP) or p300 and require its acetyltransferase activity for efficient repression (57,58), whereas NF-B engages CBP/p300 as a transcriptional co-activator (59,60), it is possible that NF-B activation efficiently competes for CBP/p300 binding and therefore alleviates repression.
From an immunological standpoint, the reliance of gp91 phox expression on NF-B is not unexpected. Because NF-B is activated within minutes after exposure to adequate stimuli, this transcription factor is especially well suited to elicit a fast genetic response. This is evidenced by the fact that a vast majority of pro-inflammatory and immune mod-  , and anti-p52 (lane 8) antibodies. # marks non-NF-B related binding activities. Gels using B1 probes were exposed for 48 -72 h, whereas gels with B2 oligonucleotides were exposed for 16 -24 h. B, binding activity of nuclear extracts from retrovirally transduced cells carrying empty vector (vector), or IB␣ AA (IB␣ AA), luciferase siRNA (Luc siRNA), or p65 siRNA (p65 siRNA) expression constructs. Where indicated, cells were exposed to LPS (100 ng/ml) and IFN␥ (100 units/ml) for 1 h. FIGURE 8. A, NF-B inducible relative luciferase activity (RLU) of CYBB promoter is dependent on the presence of a 300-bp fragment carrying two putative B elements. Equimolar amounts of p65, p65 and p50, or p65 and cRel were co-transfected with SmaI-Luc or NheI-Luc reporter plasmids into RelA Ϫ/Ϫ MEF. A plasmid expressing ␤-galactosidase under the control of the SV40 enhancer was used to normalize for differences in transfection efficiencies. Asterisk, p Ͻ 0.001; #, p Ͻ 0.05 from NheI-Luc, n ϭ 8 from four independent experiments, t test. B, reporter plasmids carrying a tandem repeat of the B1 or B2 element were transfected alone or together with p65, p65 and p50, or p65 and cRel into RelA Ϫ/Ϫ MEF. Asterisk, p Ͻ 0.05 from p65 and p65/cRel, n ϭ 6 from three independent experiments, ANOVA and Tukey's test. C, raw 264.7 cells were transiently transfected with the indicated reporter constructs (500 ng) and empty expression plasmid (vector) or expression plasmids carrying IB␣ AA or p65 siRNA (all at 300 ng). Forty hours after transfection cells were treated with LPS (1 g/ml) and IFN␥ (100 units/ml) for 7 h. A plasmid expressing ␤-galactosidase under the control of the EF-1␣ promoter (200 ng) was used to normalize for differences in transfection efficiencies. Asterisk, p Ͻ 0.001 from all other groups; #, p Ͻ 0.05 from p65 siRNA control, n ϭ 6 from three independent experiments, ANOVA and Tukey's test. D, relative gp91 phox mRNA expression levels as quantified by qPCR. RelA Ϫ/Ϫ MEF were transduced with empty retroviral vector (vector), wild type p65, or the indicated p65 serine mutants. Cells were stimulated with LPS (100 ng/ml) and IFN␥ (100 units/ml) for 6 h. Expression levels of untreated empty vector-transduced cells were set to 1; n ϭ 3. E, alignment of murine and human CYBB promoter DNA sequences surrounding putative B elements. ulatory genes rely on NF-B for efficient expression (29). Phagocytic cells are part of the immune surveillance system and have to respond swiftly to pathogens. NADPH oxidase constitutes a very powerful bactericidal system that can be activated within minutes by plasmalemmal recruitment of its subunits. Exposure to bacterial cell wall components might further enhance the ability of phagocytes to inactivate these pathogens by inducing gp91 phox expression and therefore enhancing NADPH oxidase activity.
Regulation of gp91 phox in other cell types is also likely to be controlled by NF-B. We show that NF-B is essential for gp91 phox expression in fibroblasts. Because of the ubiquitous nature of NF-B we hypothesize that gp91 phox expression might be controlled in a similar way in endothelial and neuronal cells. Even though NOX4 was not regulated by LPS/IFN␥ treatment in MEF and expression was not dependent on NF-B, it will be interesting to see whether other gp91 phox homologues (NOX1, NOX3, and NOX5) are regulated by NF-B.
In many pathological conditions such as atherosclerosis, hypertension, organ ischemia, and neurodegenerative diseases, production of ROS and NF-B activation are intrinsically related. In these conditions, ROS might activate NF-B, which, in turn, induces gp91 phox expression to further enhance ROS production. Thus, we provide evidence for the existence of a positive feedback loop that might be important for sustained ROS production.