Transcriptional Roles of Nuclear Factor B and Nuclear Factor-Interleukin-6 in the Tumor Necrosis Factor -Dependent Induction of Cyclooxygenase-2 in MC3T3-E1 Cells

When a mouse osteoblastic cell line MC3T3-E1 was cultured in the presence of tumor necrosis factor α (TNFα), the release of prostaglandin E2 and the cyclooxygenase activity increased in a dose- and time-dependent manner. The increase of the enzyme activity was attributed mostly to the induction of cyclooxygenase-2 rather than cyclooxygenase-1 as judged by the inhibitory effect of NS398, Western blotting, and Northern blotting. In this system we attempted to elucidate the transcriptional regulation of the cyclooxygenase-2 gene. As examined by the luciferase assay, two positive regulatory regions (−186 to −131 and −512 to −385 base pairs) were found in the 5′-flanking promoter region of the mouse cyclooxygenase-2 gene in the TNFα-stimulated cells. The former included putative NF-IL6 (C/EBPβ) and AP2 elements, and the latter contained the NFκB motif. A DNA probe including the NF-IL6 and AP2 sites gave positive bands upon electrophoretic mobility shift assay using the nuclear extracts of MC3T3-E1 cells. The bands were supershifted by the addition of anti-NF-IL6 antibody but not by anti-AP2 antibody. A probe including the NFκB site also gave positive bands, which were supershifted by anti-NFκB p50 and p65 antibodies. Furthermore, when the motif of NF-IL6 or NFκB or both was subjected to point mutation, the luciferase activity was markedly reduced. These data suggested a potential role of both NF-IL6 and NFκB in the induction of cyclooxygenase-2 by TNFα.

boxanes from arachidonic acid. The enzyme is a bifunctional enzyme, and has the cyclooxygenase activity (from arachidonic acid to PGG 2 ) and the hydroperoxidase activity (from PGG 2 to PGH 2 ). Two isozymes referred to as cyclooxygenase-1 and -2 in this paper are found in the mammalian tissues (1,2). Cyclooxygenase-1 is generally considered as a constitutive enzyme, while cyclooxygenase-2 is rapidly and transiently induced by various cytokines, hormones, and tumor promoters (1,2). In view of such an inducible nature the cyclooxygenase-2 has recently been a subject of active molecular biological investigations. Promoter regions of the cyclooxygenase-2 genes of mouse (3), rat (4), and human (5) have been cloned and sequenced. Regardless of the animal species these promoter regions contained a canonical TATA box and various putative transcriptional regulatory elements such as CRE, NF-IL6 (C/EBP␤), AP2, SP1, NFB, and GATA box. Among these elements, CRE (6,7) and C/EBP␤ (8) were shown to act as positive regulatory elements for the cyclooxygenase-2 transcription.
An osteogenic MC3T3-E1 cell line was established from newborn mouse calvaria. The cells differentiate into osteoblasts, and show calcification in vitro (9,10). With this cell line we have been investigating the cyclooxygenase induction by epidermal growth factor (11,12), transforming growth factor ␤ (12), epinephrine (13), and PGs (14). More recently the cyclooxygenase-2 induction was demonstrated with the MC3T3-E1 cells by PGs (15,16) and by transforming growth factor ␤ (16). In the present work, we attempted to find out the transcriptional regulatory factors involved in the induction of cyclooxygenase-2 by TNF␣ in the MC3T3-E1 cells.
Cell Culture-MC3T3-E1 cells (7.5 ϫ 10 5 cells/dish) were plated in 150-mm plastic dishes with 30 ml of ␣-MEM containing 10% fetal bovine serum and 100 units/ml of penicillin G as described previously (11). The dishes were placed in a humidified 7% CO 2 , 93% air incubator * This work was supported by grants-in-aid for scientific research from the Ministry of Education, Science and Culture of Japan, the Japanese Foundation of Metabolism and Disease, the Japan Foundation for Applied Enzymology, Ono Pharmaceutical Company, Kissei Pharmaceutical Company, Sankyo Company, Takeda Pharmaceutical Industry, and the Japan Research Foundation for Clinical Pharmacology. 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 1 The abbreviations used are: PG, prostaglandin; CRE, cyclic AMP response element; NF-IL6, nuclear factor IL-6; C/EBP, CCAAT enhancer binding protein; NFB, nuclear factor B; ␣-MEM, ␣-modified Eagle's minimum essential medium; TNF␣, tumor necrosis factor ␣; bp, base pairs; IL, interleukin. at 37°C. The cells were subcultured every 3 days.
Determination of PGE 2 Synthesis-MC3T3-E1 cells (6 ϫ 10 4 cells/ dish) were plated in 35-mm dishes with 2 ml of ␣-MEM containing 10% fetal bovine serum. Confluent cultures were usually obtained on the 4th day. Then, the medium was changed to ␣-MEM supplemented with 2% newborn bovine serum, and TNF␣ was added. At indicated time intervals, the culture medium was removed and subjected to radioimmunoassay for PGE 2 (11).
Cyclooxygenase Assay-MC3T3-E1 cells (9 ϫ 10 5 cells/dish) were plated in 150-mm dishes with 30 ml of ␣-MEM containing 10% fetal bovine serum, and treated with TNF␣. The cells were scraped from the dishes at various time intervals, suspended in 300 l of 20 mM Tris-HCl (pH 7.4) containing 5 mM tryptophan, and sonicated twice each for 3 s at 20 kHz. The sonicates were incubated with 10 M [1-14 C]arachidonic acid (50,000 cpm in 5 l of ethanol) for 2 min at 24°C in a 100-l assay mixture containing 100 mM Tris-HCl (pH 8.0), 2 M hematin, and 5 mM tryptophan. The reaction products were separated by TLC (15), and detected by a BAS2000 imaging analyzer (Fujix, Tokyo, Japan). Protein concentration was determined by the method of Lowry et al. (17) with bovine serum albumin as standard.
Western Blotting-A polyclonal anti-cyclooxygenase-2 was prepared with a peptide corresponding amino acids 579 -594 of murine enzyme, and provided by Dr. Yoko Hayashi of this laboratory. The sonicates of MC3T3-E1 cells were subjected to 10% polyacrylamide gel electrophoresis in the presence of 0.1% SDS (10 g of protein/lane). The protein bands were transferred to an Immobilon-P membrane, which was incubated with the polyclonal antibody against cyclooxygenase-2 and then with 35 S-labeled anti-rabbit IgG whole antibody as the second antibody. Radioactivity of the immunocomplex was quantified by a BAS2000 imaging analyzer.
Northern Blotting-Total RNA was extracted from MC3T3-E1 cells using ISOGEN (a mixture of guanidium isothiocyanate and phenol) according to the manufacturer's instruction. Northern blotting was performed as described previously (15,18). The relative radioactivity of each band was estimated using a BAS2000.
Preparation of the 5Ј-Flanking Region of Mouse Cyclooxygenase-2 Gene and Construction of Luciferase Reporter Vectors-We screened a mouse genomic library constructed in EMBL3 to clone the 5Ј-flanking region of the mouse cyclooxygenase-2 gene by the method as described previously for 12-lipoxygenase (19). DNA fragments of mouse cyclooxygenase-2 promoter regions of various lengths or their point mutants were prepared from the clone by polymerase chain reaction, and were inserted into the plasmid pXP-1 as described previously for 12-lipoxygenase (20).
Transfection of Plasmids to MC3T3-E1 Cells and Luciferase Assay-MC3T3-E1 cells (1.4 ϫ 10 5 cells/dish) were cultured for 2 days in 60-mm dishes with 5 ml of ␣-MEM containing 10% fetal bovine serum. For transfection the subconfluent cells were treated with plasmid DNA (2.3 g) containing cyclooxygenase-2 promoter and luciferase reporter gene, standard plasmid DNA (1.2 g) containing the ␤-galactosidase gene, and 9.2 l of lipofectAMINE in 2.3 ml of serum-free ␣-MEM for 4 h at 37°C according to the manufacturer's instructions. The medium was changed to 5 ml of ␣-MEM with 10% fetal bovine serum, and the transfected cells were further incubated for 3 days. Then the cells were stimulated by 20 ng/ml TNF␣ in the presence of ␣-MEM supplemented with 2% newborn bovine serum. After 12 h the cells were scraped from the dishes, and the luciferase activity was measured by a Lumat LB9501 luminometer (Berthold, Germany) using a Pica Gene luciferase assay system according to the manufacturer's instructions. The luciferase activities were normalized on the basis of ␤-galactosidase activities which were assayed as described previously (20).
Preparation of Nuclear Extracts-MC3T3-E1 cells (9 ϫ 10 5 cells/dish) were plated in 10 150-mm dishes with 30 ml of ␣-MEM containing 10% fetal bovine serum, and stimulated with various concentrations of TNF␣. After 1 h the cells were harvested, and nuclear extracts were prepared as described previously (20).
Electrophoretic Mobility Shift Assay-We synthesized the five oligonucleotides as shown in Fig. 4. The complementary oligonucleotides were annealed to the corresponding oligonucleotides as described by Berger and Kimmel (21), and the double-stranded oligonucleotides were purified electrophoretically on 15% polyacrylamide gel, end-labeled with [␥-32 P]ATP, and used as probes. The binding of the probes (10,000 cpm, about 10 fmol) to the nuclear extracts (2.3 g protein) was performed in a 20-l mixture containing 5 g of poly(dI-dC)-poly(dI-dC), 15 mM Tris-HCl at pH 7.5, 1 mM EDTA, 100 mM KCl, 5 mM MgCl 2 , 12% glycerol, 0.5 mM phenylmethylsulfonyl fluoride, and 0.5 mM dithiothreitol. For the supershift experiment, each antibody was added to the mixture. The mixture was incubated at 25°C for 30 min, and applied to 4% polyacrylamide gel electrophoresis gels at a constant 150 V for 2 h. Distribution of the radioactivity on the dried gel was analyzed by a BAS2000 imaging analyzer.

RESULTS
Induction of Cyclooxygenase-2 by TNF␣-When MC3T3-E1 cells were incubated with various concentrations of TNF␣ for 12 h, there was a dose-dependent increase in the amount of PGE 2 released into the medium (Fig. 1A). After the addition of 20 ng/ml TNF␣, the PGE 2 release increased with a lag time of about 1 h, and reached a maximum at 9 h (Fig. 1B). As shown in Fig. 1, C and D, TNF␣ increased the cyclooxygenase activity dose and time dependently, and the enzyme activity continued to increase for 24 h to a specific activity of 0.8 -2 nmol/2 min/mg of protein.
NS398, a specific inhibitor of cyclooxygenase-2 (22), inhibited the enzyme activity almost completely at 15 M concentration ( Fig. 1, C and D), suggesting that most of the increased enzyme activity was attributable to cyclooxygenase-2 rather than cyclooxygenase-1. To confirm this finding we carried out Western blotting with an antibody specific for cyclooxygenase-2 ( Fig.  2A). Without TNF␣ the cyclooxygenase-2 protein was undetectable, and the addition of 20 ng/ml TNF␣ increased the enzyme amount time dependently up to 9 h. There was a slight but reproducible decrease around 12 h, followed by an increase again.
Furthermore, as shown in Fig. 2B, the change in mRNA level was followed by Northern blot analysis. The addition of TNF␣ brought about a biphasic increase in the cyclooxygenase-2 mRNA. First, there was a rapid 10-fold increase reaching a maximum at 2 h, followed by a decrease at 3 h. Then, the mRNA increased again at 6 -12 h. However, after the first transient peak the presence of 15 M NS398 reduced the second peak. The cyclooxygenase-1 mRNA was not detectable under these experimental conditions.
For deletion analysis of the promoter region, we constructed luciferase vectors of various lengths covering the region from Ϫ621 to Ϫ41 bp (Fig. 3). Each vector was transfected to MC3T3-E1 cells by the lipofection method, and the luciferase activity of the cell lysate was measured at 12 h after the addition of 20 ng/ml TNF␣. The addition of TNF␣ markedly stimulated the luciferase activity (closed column versus slashed column). The luciferase activity decreased when two regions (Ϫ512 to Ϫ385 bp, Ϫ186 to Ϫ131 bp) were deleted, and these regions were presumed to have positive response elements. We noted NFB consensus element (Ϫ401 to Ϫ393 bp), AP2 element (Ϫ150 to Ϫ142 bp), and NF-IL6 element (Ϫ138 to Ϫ130 bp).
Electrophoretic Mobility Shift Assay Targeting Positive Regulatory Elements-For further identification of these positive regulatory elements, we prepared five double-stranded oligonucleotide probes (g1, g2, gA, gB, and gC). As shown in Fig. 4, g1 (Ϫ409 to Ϫ385 bp) contained NFB and their vicinity. gA (Ϫ155 to Ϫ121 bp) contained AP2 and NF-IL6. g2, gB, and gC were mutant probes for NFB, AP2, and NF-IL6, respectively. We carried out electrophoretic mobility shift assay using the nuclear extracts of MC3T3-E1 cells preincubated with TNF␣ for 1 h. As shown in Fig. 5, when g1 probe was incubated with the nuclear extracts, a broad band of complex ␣ was observed with increasing density depending on the amount of TNF␣ (lanes 3-6). The complex ␣ was hardly detectable before the addition of TNF␣ (lane 2). Considering the subunit structure of NFB (p50 and p65), the g1 probe was incubated with the purified NFB p50, and a band appeared at the position of complex ␣ (lane 7). The complex ␣ was supershifted to complex ␤ by the addition of anti-p50 antibody (lane 10), and was supershifted to complex ␥ by the addition of anti-p65 antibody (lane 11). The binding was not observed using the probe g2 with a mutation in NFB motif (lanes 13 and 14).
When gA probe was incubated with the nuclear extracts, three complexes (a, b, and c) appeared with increasing density depending on the amount of TNF␣ (Fig. 6, lanes 3-6). These complexes were supershifted to bands d, f, and g by the addition of antibody against NF-IL6 (lane 8). The binding was scarcely observed by the use of the probe gC with a mutation in NF-IL6 (lanes 15-19). Moreover, the complex a was supershifted to complex e by the addition of anti-NFB p50 (lane 9). The bands a, b, and c were not supershifted by anti-AP2 antibody (lane 7), and the binding profile was not affected by the use of the probe gB with a mutation in AP2 (lanes 10 -14).
These data suggested that NFB and NF-IL6 were increased by the addition of TNF␣ in a dose-dependent manner.
Possible Interaction of Both NFB and NF-IL6 Binding Sites-For mutation analysis of the promoter region, we constructed luciferase vectors including mouse cyclooxygenase-2 promoter region (Ϫ512 to ϩ123 bp) with site-specific mutations (Fig. 7). Each vector was transfected to MC3T3-E1 cells, and the luciferase activity was measured as described above. As shown in Fig. 7, the luciferase activity decreased by 63% in a mutant of NFB, 82% in a mutant of NF-IL6, and only 19% in a mutant of AP2 site. By mutation of both NFB and NF-IL6, the luciferase activity was lost to the level of pXP1. These data suggested that both NFB and NF-IL6 in combination acted as transcription factors for the induction of cyclooxygenase-2.
Furthermore, expression vectors containing the genes of NF-IL6 and NFB p50, respectively, were cotransfected to MC3T3-E1 cells together with a luciferase plasmid containing the promoter region of cyclooxygenase-2 (Ϫ512 to ϩ123 bp). The luciferase activity increased depending on the amount of NF-IL6 cDNA (data not shown). A similar experiment using the expression vector for NFB p50 has so far been unsuccessful. DISCUSSION It was reported previously that TNF␣ increased the PGE 2 production and the cyclooxygenase activity in mouse osteoblastic cell line MC3T3-E1 (24 -26). We confirmed these findings (Fig. 1), and demonstrated that the increased enzyme activity was attributed mostly to the induction of cyclooxygenase-2 according to the inhibitory effect of NS398, Western blotting, and Northern blotting (Figs. 1 and 2). The mechanism of the biphasic increase of cyclooxygenase-2 protein and mRNA (Fig.  2) may be discussed as follows. First, TNF␣ activates phospholipase A 2 (as reported in Ref. 27) and induces cyclooxygenase-2, resulting in an increased production of PGE 2 . Then, the produced PGE 2 binds to a PGE receptor, and increases the cyclooxygenase-2 protein and mRNA again. Previously we reported the cyclooxygenase-2 induction by various PGs, including PGE 2 which was a major arachidonate metabolite in this cell line (15).
Since the addition of TNF␣ brought about such a typical and prominent induction of cyclooxygenase-2 in MC3T3-E1 cells, we attempted to elucidate the transcriptional regulation of the cyclooxygenase-2 gene in this system. The luciferase assay (Figs. 3 and 7) and electrophoretic mobility shift assay (Figs. 5 and 6) revealed two positive regulatory elements on the 5Јflanking region of mouse cyclooxygenase-2 gene: NFB motif (5Ј-GGGATTCCC-3Ј) located at nucleotides Ϫ401 to Ϫ393 bp and NF-IL6 motif (also referred to as C/EBP␤, 5Ј-TTGCG-CAAC-3Ј) at Ϫ138 to Ϫ130 bp. NF-IL6 was originally identified as a nuclear factor binding to the IL-1 response element of the human IL-6 gene (28). By cDNA cloning NF-IL6 was found to be a member of the CCAAT enhancer-binding protein (C/EBP) family of basic-leucine zipper (bZIP) transcription factors (29). Subsequent studies suggested a regulatory role of NF-IL6 for the genes encoding many acute-phase proteins and cytokines (30). The NF-IL6 motif on the rat cyclooxygenase-2 gene (5Ј-TTATGCAAT-3Ј located at Ϫ140 to Ϫ132 bp) was previously reported to contribute to the induction of cyclooxygenase-2 by forskolin, follicle-stimulating hormone, and luteinizing hormone in granulosa cells (4,8). Recently it was reported that the cyclooxygenase-2 induction was mediated by an activating transcription factor/CRE element (5Ј-CGTCACGTG-3Ј at Ϫ56 to Ϫ48 bp) on the mouse cyclooxygenase-2 gene promoter (7) or by CRE element (5Ј-TTCGTCA-3Ј at Ϫ59 to Ϫ53) in the human cyclooxygenase-2 gene (6). However, we could not demonstrate the involvement of the CRE motif in the TNF␣-dependent cyclooxygenase-2 induction in MC3T3-E1 cells. The AP2 element (5Ј-CCGCTGCGG-3Ј, Ϫ150 to Ϫ142 bp) close to the NF-IL6 was shown to be inactive (Figs. 3, 6, and 7). NFB was originally found as a binding protein to immunoglobulin light-chain enhancer (31), and it is now considered as  6 and 9) TNF␣. The nuclear extracts before the addition of TNF␣ were also incubated with g1 (lane 2). Anti-NFB p50 (lane 10) or anti-NFB p65 (lane 11) antibody was incubated together with the nuclear extracts from the cells stimulated with 20 ng/ml TNF␣. Purified NFB p50 protein was incubated with probe g1 (lane 7). Probe g2 with a mutation for NFB was incubated with the nuclear extracts from the cells stimulated with 20 ng/ml TNF␣ (lanes 13 and 14). Anti-NFB p50 antibody was also present (lane 14). Lanes 1, 8, and 12 were control runs without the nuclear extracts. Shifted bands indicated by ␣, ␤, and ␥ are described in the text. a ubiquitous transcription factor. It is constitutively present in the nuclei of B cells, certain T cell lines and monocytes, and has the structure of either homodimer or heterodimer (32)(33)(34)(35). The heterodimer of p50 and p65 subunits has higher affinity to the NFB site than the homodimer of either p50 or p65 (36). As examined by electrophoretic mobility shift assay (Fig. 5), the nuclear extracts of MC3T3-E1 cells gave the complex ␣ which is associated presumably with p50-p50 homodimer and p50-p65 heterodimer.
The double mutant of NFB and NF-IL6 showed a lower luciferase activity than a single mutant of either NFB or NF-IL6 (Fig. 7). Interaction of both NFB and NF-IL6 was suggested by a finding that band a was supershifted to band e by anti-NFB p50 antibody in electrophoretic mobility shift assay with a probe including the NF-IL6 motif (lane 9 in Fig. 6).
Individual binding sites for NF-IL6 and NFB are present in the promoter of the IL-6 gene, and the cooperation of these two factors plays an important role in transcription of IL-6 (37). A similar activation was reported for the IL-8 promoter, which also contains both NF-IL6 and NFB binding sites (37). The NFB p50 and NF-IL6 proteins directly interact in vitro, and the Rel homology domain and leucine-zipper motif, respectively, are important for this interaction (38).

FIG. 7. Mutations of NFB and NF-IL6 motifs.
A mixture of each luciferase plasmid which is wild or mutated (cross indicates point mutation of wild plasmid) (2.3 g) and ␤-galactosidase plasmid (1.2 g) was transfected into MC3T3-E1 cells (1.4 ϫ 10 5 cells/60-mm dish) by the lipofection method. The cells were cultured to confluency for 3 days, and incubated for 12 h with (closed column) or without (slashed column) 20 ng/ml TNF␣. The luciferase activity was assayed, and normalized with the ␤-galactosidase activity. Data are means Ϯ S.E. of triplicate determinations.