The transcription factor C/EBPbeta is essential for inducible expression of the cox-2 gene in macrophages but not in fibroblasts.

Cyclooxygenase-2 (COX-2) is the rate-limiting enzyme for the inducible synthesis of prostaglandins, and its up-regulated activity is thought to play a pathological role in diseases such as inflammatory bowel disease, rheumatoid arthritis, and cancer. Regulation of COX-2 expression is complex and appears to involve diversified mechanisms in different cell types and conditions. Here we make use of immortalized macrophages and fibroblasts that we have generated from C/EBPbeta-deficient mice to directly test and compare the specific role played by this factor in inducible COX-2 expression in these two cell types. We could demonstrate that COX-2 mRNA induction and promoter activity were profoundly impaired in C/EBPbeta(-/-) macrophages and could be rescued by expression of C/EBPbeta. The obligatory role of C/EBPbeta in COX-2 expression appeared to be mediated exclusively by the C/EBP element located at positions -138/-130 of the murine cox-2 promoter, and did not involve altered activity at the level of the other promoter elements described previously (the -402/-392 NF-kappaB site, the -59/-48 CRE/E box element, and a potential second C/EBP site located at positions -93/-85). In contrast, COX-2 induction was completely normal in C/EBPbeta-deficient fibroblasts, thus highlighting the diversity of cell-specific molecular mechanisms in determining inducible COX-2 expression and prostaglandins production.

Prostaglandins (PG) 1 play important roles in many cellular responses including cell growth, ovulation, and immune func-tions, and inhibition of their synthesis is the target of most known nonsteroidal anti-inflammatory drugs (1). PG production is controlled by two cyclooxygenases (COX-1 and COX-2) that mediate the initial conversion of arachidonic acid into PGH 2 , a precursor common to all prostanoids (2). Whereas COX-1 is expressed constitutively in most cells and its activity is regulated mainly by the availability of arachidonic acid, COX-2 is expressed at undetectable levels under normal conditions and rapidly and strongly induced by specific stimuli in different cell types (3). In particular, macrophages are prominent producers of prostaglandins during inflammatory processes in response to signals that trigger macrophage activation such as bacterial lipopolysaccharide (LPS). Altered COX-2 levels and consequent abnormally high PGs secretion are thought to be involved in diverse pathological processes, and COX-2-specific inhibitors hold high hopes for the treatment of cancer as well as chronic inflammatory diseases such as ulcerative colitis and rheumatoid arthritis (4,5).
The regulation of COX-2 synthesis occurs mainly at the transcriptional level, although mRNA stabilization is also involved in response to specific signals. The stimuli, signal transduction pathways, and transcription factors involved in the induction of cox-2 gene expression are extremely diversified and cell-specific. Thus, cox-2 transcription is activated by LPS and pro-inflammatory cytokines in macrophages and endothelial cells, by growth factors, serum and phorbol esters (phorbol 12-myristate 13-acetate (PMA)) in fibroblasts, by growth factors and PMA in epithelial cells, by serotonin and interleukin (IL)-1␤ in mesangial cells, and by follicle-stimulating and luteinizing hormones in ovarian granulosa cells (3). The promoter elements and transcription factors involved also vary according to the cell type, and their relative roles are often contradictory, probably reflecting the complexity of the pathways involved. Of the numerous cis-acting elements identified on the cox-2 mouse gene promoter, three have been proposed to play a fundamental role in most systems analyzed. The nuclear factor-B (NF-B) site at position Ϫ402/Ϫ392 has been implicated in cox-2 induction in different cell types (6 -11) but has recently been proposed not to be required in macrophages, osteoblasts, and fibroblasts (12)(13)(14). The element at position Ϫ59/Ϫ48 is made of two overlapping cyclic AMP-response element (CRE) and E box sites and is considered essential for both basal and induced transcription in most cellular systems (9, 12, 14 -21). Interestingly, whereas in macrophages and fibroblasts the CRE moiety is responsible for transcriptional activation by members of the CRE-binding protein (CREB) and activating protein-1 (AP-1) family (12,15), in rat granulosa and skin carcinoma cells the E box is predominantly involved through the interaction with the upstream stimulating factors 1 and 2 (16,17). Finally, the nuclear factor for IL-6/CCAAT enhancer-binding protein (NF-IL6/C/EBP) element at position Ϫ138/Ϫ130 was reported to play an important role in mediating signal-dependent transcriptional induction in macrophages, osteoblastic cells, pancreatic islet cells, skin carcinoma cells, and chondrocytes (6,9,10,12,13,17,18,20) but not in rat granulosa cells or in fibroblasts (15,16). Transcription factors belonging to the C/EBP family share a strong homology in their leucine zipper and DNA binding domains, and as a consequence are able to form both homo-and heterodimers and bind to the same DNA sequences (22). Different C/EBP family members can bind to the Ϫ138/Ϫ130 element, but their relative role in activating the cox-2 promoter in different cell types is unclear and sometimes contradictory (9,10,12,17,20,23). However, all these studies are based on transient transfection/overexpression experiments and do not take into account the endogenous and regulated ratio of C/EBP family members able to bind to the promoter under uninduced or induced conditions. Cells where specific transcription factors have been inactivated represent a precious tool to determine the physiological role played by a given factor in the transcriptional control of candidate target genes. Here we made use of immortalized cells that we have recently derived from C/EBP␤-deficient or wild type mice 2 to analyze the role played by this factor in regulating cox-2 expression. We found that COX-2 induction by LPS was profoundly defective in C/EBP␤ Ϫ/Ϫ macrophages, essentially due to impaired transcriptional induction via the Ϫ138/Ϫ130 C/EBP promoter element. In contrast, in fibroblasts C/EBP␤ was totally dispensable for cox-2 transcriptional induction in response to PMA treatment or v-Src transfection, thus unambiguously confirming the cell-specific nature of COX-2 regulation.

EXPERIMENTAL PROCEDURES
Cell Culture and Treatments-The generation of C/EBP␤ Ϫ/Ϫ -and C/EBP␤ ϩ/ϩ -immortalized macrophages is described elsewhere. 2 The cells were maintained at 37°C in a 5% CO 2 atmosphere in RPMI 1640 medium supplemented with 10% heat-inactivated fetal calf serum and containing 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin (standard medium). Cells at ϳ70% confluence were stimulated with 100 units/ml of interferon (IFN)␥ (kindly provided by G. Garotta, Ares-Serono, Geneva, Switzerland) for 16 h and with 100 ng/ml of LPS (Escherichia coli serotype 026:B6; Sigma) for the indicated times. For the inhibition of transcription, after 4 h of LPS treatment cells were incubated with 5 g/ml of actinomycin D (Sigma) for the indicated times.
Isolation of Bone Marrow-derived Macrophages-C/EBP␤ Ϫ/Ϫ and C/EBP␤ ϩ/ϩ mice were killed by asphyxiation with CO 2 , and bone marrow cells were collected as described previously (26). Briefly, bone marrow cells were mechanically isolated from femurs and cultured on 9-cm diameter bacteriological plates in RPMI 1640 standard medium supplemented with 30 -50% of L cell-conditioned medium as a source of macrophage colony-stimulating factor (26). Bone marrow macrophages were treated with 100 units/ml IFN␥ for 16 h, followed by 1 g/ml of LPS for 4 h prior to RNA extraction.
RNA Extraction and Northern and Slot Blot Analysis-Total RNA was prepared from untreated or stimulated cultured cells using the RNeasy Midi Kit (Qiagen Ltd., Crawley, UK) according to the manufacturer's instructions. 20 or 5 g of total RNA was analyzed by Northern blot or slot blot, respectively, as described previously (27). cDNA probes were labeled by random priming. The relative abundance of cox-2 mRNA was measured by PhosphorImager analysis and normal-ized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as an internal control.
For immunoblotting, 30 g of cell lysate were denatured in SDS, electrophoresed on a 10% SDS-polyacrylamide gel, transferred to nitrocellulose membrane, immunoblotted with an anti-COX-2 polyclonal goat antibody (Santa Cruz Biotechnology, Santa Cruz, CA), and detected by ECL reagent (Amersham Pharmacia Biotech).
Nuclear Extracts and EMSAs-Cells were harvested in phosphatebuffered saline, resuspended in 5 packed cell volumes of hypotonic buffer (10 mM Hepes, pH 7.9, 0.75 mM spermidine, 0.15 mM spermine, 0.1 mM EDTA, 10 mM KCl, 1 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride, and complete proteinase inhibitor mixture), and mechanically lysed in 2 packed cell volumes of hypotonic buffer. Sucrose was added to a final concentration of 6.75%, and samples were centrifuged at 13,000 ϫ g for 30 s at 4°C. Nuclear pellets were resuspended in Nuclear Resuspension buffer (20 mM Hepes, pH 7.9, 0.75 mM spermidine, 0.15 mM spermine, 0.2 mM EDTA, 2 mM EGTA, 25% glycerol, 2 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride, and complete proteinase inhibitor mixture), sonicated in a 4°C water bath, and rocked for 30 min at 4°C. Samples were then centrifuged at 58,000 rpm in a TLA 120.2 rotor for 90 min and supernatants snap-frozen and stored at Ϫ80°C. Protein concentration was determined by Bradford assay.
Electrophoretic mobility shift assay (EMSA) probes were made by annealing single-stranded oligonucleotides (MWG-Biotech) with 5Ј GATC overhangs. 1 picomole of probe was radiolabeled by filling in with [␣- 32  EMSAs were performed with 4 g of nuclear extract in 20 mM Hepes, pH 7.9, 1 mM EDTA, pH 8, and 2.5 mM DTT, containing 3 g of poly(dI-dC). The complexes were separated by electrophoresis on a 6% (C/EBP) or 5% (NF-B and CRE/E box) polyacrylamide, 0.25ϫ Tris borate/EDTA gel. For supershift experiments, 2 l of polyclonal purified antibody (Santa Cruz Biotechnology, Santa Cruz, CA) was incubated with nuclear extracts and poly(dI-dC) for 30 min on ice, prior to probe addition. Unlabeled double-stranded oligonucleotide competitors were preincubated at a 50-fold molar excess 10 min prior to probe addition. 1 g of a synthetic peptide carrying the C/EBP leucine zipper motif was used as described (28). PGE 2 Detection-Culture media of cells treated with IFN␥ and LPS were assayed with the Prostaglandin E 2 Enzyme Immunoassay System from Biotrak (Amersham Pharmacia Biotech) according to manufacturer's instructions. Values obtained (pg/well) were normalized to protein concentration of each sample.
Nuclear Run-on Assays-60 ϫ 10 6 cells per sample were harvested and lysed as for nuclear protein extraction (see above). Immediately after lysis, nuclei were prepared as follows: 1 ml of Sucrose buffer I (0.32 M sucrose, 3 mM CaCl 2 , 2 mM magnesium acetate, 0.1 mM EDTA, 10 mM Tris-HCl, pH 8, 1 mM DTT) and 5.6 ml of Sucrose buffer II (2 M sucrose, 5 mM magnesium acetate, 0.1 mM EDTA, 10 mM Tris-HCl, pH 8, 1 mM DTT) were added to the lysate. Nuclei were collected by ultracentrifugation over a cushion of Sucrose buffer II, resuspended in 225 l of Nuclear freezing buffer (50 mM Tris-HCl, pH 8.3, 40% v/v glycerol, 5 mM MgCl 2 , 0.1 mM EDTA), and stored at Ϫ80°C until use. Nuclear run-on was performed essentially as described (29). Equal amounts of radioactivity (about 5 ϫ 10 6 cpm/ml) were used to hybridize dot blots in 10 mM TES, pH 7.4, 10 mM EDTA, 0.2% SDS, 0.6 M NaCl at 65°C for at least 36 h. Dot blots contained 10 g of a linearized COX-2 or GAPDH cDNA plasmid or the empty Bluescript plasmid as a negative control. Membranes were washed as described (29) and treated with 10 g/ml RNase A for 30 min at 37°C. Blots were visualized by autoradiography and quantified using a PhosphorImager. Normalized values were obtained as COX-2:GAPDH intensity ratio.
Transient Transfections-The full-length cox-2 promoter (Ϫ963 base pairs) was generated by polymerase chain reaction using genomic DNA as template and the Pfu Turbo DNA Polymerase (Stratagene, La Jolla, CA). Primers used were Ϫ963, 5Ј-GGGCTAGCCCAACACAAACACAG-TAGGA-3Ј; COX2rev, 5Ј-GGCTCGAGGCAGTGCTGAGATTCTTCGT-3Ј, positioned at ϩ70 base pairs. The product was verified by sequence analysis and ligated into the firefly luciferase reporter plasmid pGL-3 Basic (Promega, Madison, WI). The other constructs were amplified from the Ϫ963/luc construct, sequenced, and cloned in the same vector. Primers used were COX2rev, indicated above, and one of the following: CMV-C/EBP␤ carried the rat cDNA (28) cloned into the pCEP4 plasmid (Invitrogen BV, Groningen, The Netherlands). The pSGT-Src-527 expression plasmid (30) was kindly provided by G. Superti-Furga (EMBL, Heidelberg, Germany). All plasmids were prepared using Endotoxin-free Plasmid Preparation Kits (Qiagen).
For transient transfection of immortalized macrophages, 10 7 cells were resuspended in 250 l of RPMI 1640 supplemented with 20% fetal calf serum, together with 5 g of the indicated firefly luciferase reporter plasmid and 1 g of the internal control pRL-TK, encoding Renilla luciferase (Promega). 2 g of CMV-C/EBP␤ plasmid was used for cotransfection experiments. Cells were electroporated in 0.4-cm cuvettes in a Bio-Rad Gene Pulser at 250 V and 950 microfarads, and each sample was seeded on two 3.5-cm diameter Petri dishes in RPMI 1640 standard medium and cultured for 18 h. One dish was treated with 1 g/ml of LPS for 4 h, and then cells were washed in ice-cold phosphatebuffered saline and scraped off the dish in 0.5 ml of Passive Lysis buffer (Promega).
Immortalized fibroblasts were plated at ϳ70% confluence onto 6-cm dishes 24 h prior to transfecting with 3 g of cox-2 reporter construct and 0.5 g of pRL-TK using the calcium phosphate method. 1.5 g of pSGT-Src-527 plasmid were used for co-transfection experiments. Cells were allowed to recover for 24 h and collected in Passive Lysis buffer.
Firefly and Renilla luciferase values were obtained by analyzing 20 l of cell extract using the Dual Luciferase kit (Promega), according to manufacturer's instructions, in a Lumat LB 9507 luminometer. Relative luciferase activity of cell extracts was typically represented as firefly luciferase value/Renilla luciferase value. Since C/EBP␤ activated the control pRL-TK plasmid, luciferase activity was normalized to protein content as measured by Bradford assay when C/EBP␤ was co-transfected.

RESULTS
cox-2 mRNA Induction Is Defective in Immortalized and Primary C/EBP␤ Ϫ/Ϫ Macrophages-The generation and characterization of immortalized macrophages from the spleen of C/EBP␤ Ϫ/Ϫ and C/EBP␤ ϩ/ϩ mice is described elsewhere. 2 To evaluate the expression of COX-2 mRNA in the absence of C/EBP␤, total RNA from several independent C/EBP␤ Ϫ/Ϫ (K1, K3, and K4) or C/EBP␤ ϩ/ϩ (W2, W3B, and W3E) macrophage cell lines either untreated or treated with IFN␥ and LPS was subjected to slot blot analysis using a COX-2 cDNA fragment as a probe. COX-2 mRNA was undetectable in both Ϫ/Ϫ and ϩ/ϩ untreated cells but was strongly induced in the wild type cell lines after 4 h of LPS treatment (Fig. 1, A and C, left panel). The induction was still appreciable at 8 h but had decreased considerably by 24 h after treatment (Fig. 1B). In contrast, very little induction could be detected in all three mutant cell lines analyzed. COX-2 expression was partially rescued in the revertant cell line r(K4) that was obtained by stably transfecting C/EBP␤ into the K4 Ϫ/Ϫ cells (Fig. 1C, left panel). Of note, both C/EBP␤ mRNA 2 and protein levels (data not shown) were about 3-4 times lower in the r(K4) cells than in wild type cell lines both before and after stimulation, in line with the partial rescue of COX-2 expression achieved. Taken together, these data establish a strong correlation between COX-2 mRNA induction and the presence of C/EBP␤. In agreement with this idea, COX-2 expression was also strongly defective in primary macrophages derived from the bone marrow of C/EBP␤-deficient mice (Fig. 1C, right panel).
Decreased COX-2 mRNA Correlates with Low Protein Expression and Impaired PGE 2 Secretion-We next analyzed COX-2 protein levels in the W2, K4, and r(K4) cell lines by Western blot. As shown in Fig. 2A, COX-2 was barely detectable in the mutant K4 cells both before and after stimulation, whereas it was strongly induced in the wild type W2 cells with levels comparable to the mouse macrophage cell line RAW264, used as a control. Expression of C/EBP␤ in the revertant cells allowed ϳ50% of COX-2 expression to be rescued. These differences were mirrored by the levels of prostaglandin E 2 secreted in the culture medium, which were negligible in the K4 cells, extremely high in the W2 cells, and intermediate in the revertant r(K4) cells (Fig. 2B).
COX-2 mRNA Expression Is Impaired at the Transcriptional Level-Nuclear run-on assays were performed using nuclei prepared from unstimulated or stimulated K4 and W2 cells. As shown in Fig. 3A, transcriptional rates were strongly induced by IFN␥/LPS treatment in nuclei from the wild type but not from the mutant cells, indicating that the low expression of COX-2 mRNA observed in the absence of C/EBP␤ was at least partly due to impaired transcription of its gene. To verify whether decreased mRNA stability could also play a role, cells were treated with actinomycin D after stimulation with IFN␥ and LPS, and total RNA was analyzed by Northern blot with a COX-2 cDNA probe. As expected, COX-2 mRNA was much less abundant in the mutant K4 cells. However, no RNA degradation was detected for up to 90 min after actinomycin D addition in either wild type or mutant samples (Fig. 3B), suggesting that COX-2 mRNA stability is not altered in the C/EBP␤ Ϫ/Ϫ cells.
DNA-Protein Interactions on the cox-2 Promoter Are Only Altered at the C/EBP-binding Site-Altered transcription of specific genes in transcription factor-deficient cells could well represent a secondary event, resulting from the altered expression or activity of distinct sets of transcriptional activators. We therefore decided to analyze the DNA-protein interactions taking place at the level of the cox-2 promoter sites previously shown to be involved in transcriptional induction of the gene (Fig. 4A) in order to assess whether the pattern of proteins binding to one or more of these sites was altered in the C/EBP␤ Ϫ/Ϫ cells. Previous studies on the mouse cox-2 promoter have variably identified an NF-B-binding site (NF-B, position Ϫ402/Ϫ392), a C/EBP-binding site (C/EBP1, position Ϫ138/Ϫ130), and an overlapping CRE/E box element (CRE/E box, position Ϫ59/Ϫ48), as important in LPS induction of the cox-2 promoter (Fig. 4A) (6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21). In addition, an element located at position Ϫ93/Ϫ85 (*C/EBP2), bearing some sequence similarity to a C/EBP-binding site, has recently been proposed to play a role in the LPS inducibility of the cox-2 promoter in Raw 264.7 cells (12).
We first analyzed the Ϫ138/Ϫ130 C/EBP1 element, a canonical C/EBP-binding site that has been shown to bind members of the C/EBP family (6,9,13,17,18,20,23). EMSA analysis using nuclear extracts from untreated wild type W2 cells revealed at least four distinct complexes (Fig. 4B, 1-4). Most bands were supershifted by antibodies directed against C/EBP␤ or C/EBP⑀, although complex 1 was not completely abolished by either antibody. Complexes 1 and 2 were dramatically increased in extracts from IFN␥/LPS-stimulated W2 cells. Importantly, all the newly induced activities could be supershifted by anti-C/EBP␤ antibodies. The pattern detected using extracts from untreated K4 mutant cells was similar to the one observed with untreated W2 cells, except that complex 2 appeared to be weaker and complexes 1 and 2 were partially reduced by anti-C/EBP␣ antibodies. Predictably, no supershift was obtained using anti-C/EBP␤ antibodies. In contrast to the strong increase in DNA binding observed upon IFN␥/LPS treatment in the W2 cells, no change was triggered in the C/EBP␤ Ϫ/Ϫ cells by stimulation. Moreover, no supershift was detected with any of the anti-C/EBP antibodies used. Competition with a cold C/EBP1 oligonucleotide abolished all binding in extracts from both untreated and treated K4 and W2 cells (Fig. 4C, self). Interestingly, however, competition with two distinct C/EBP-binding sites derived from the D site of the albumin promoter (31) or from the A site of the hemopexin promoter (32) could abolish all bands except the one corresponding to complex 3 that appeared therefore to be unique to the COX-2 C/EBP1 site. Since complex 3 was also not supershifted by any of the anti-C/EBP antibodies used (Fig. 4B), we performed a competition with a peptide carrying the leucine zipper domain from C/EBP␤ that we have shown previously (28) to interfere with the binding of all C/EBP members by competing for dimer formation. Similar to the AlbD and HpxA sites, the C/EBP␤-zipper peptide abolished the formation of all complexes with the exception of complex 3, suggesting that this complex may involve a non-C/EBP protein. Complex 3 was similar, however, in both treated or untreated K4 and W2 cells, and its identity was not investigated further. Likewise, the , and in vitro transcription was performed as described under "Experimental Procedures." Radiolabeled nascent mRNA transcripts were hybridized to membranes where plasmids either with no insert (Bluescript, pBS) or containing COX-2 or GAPDH cDNAs had been transferred by dot blot. The histograms show the values Ϯ S.E. from at least two experiments upon normalization to GAPDH. B, mRNA stability assay. Cells were either untreated or treated with IFN␥ for 16 h ϩ LPS for 4 h, followed by the addition of actinomycin D (ActD, 5 g/ml). Total RNA was isolated at the indicated times, and COX-2 levels were determined by Northern blot. The membrane was subsequently hybridized with GAPDH as an internal control.
identity of the protein(s) responsible for complex 1, not fully supershifted by any of the anti-C/EBP antibodies used but abolished by competition with either a C/EBP site or the C/EBP leucine zipper peptide, was not further explored since no difference could be detected between the mutant and the wild type cells.
We next examined the DNA-protein interactions occurring at the recently identified *C/EBP-2 site located at positions Ϫ93/ Ϫ85 (12). Binding to this sequence gave rise to three differentially migrating complexes (named A-C, Fig. 4D) in extracts from both K4 and W2 cells, with complexes B and C becoming similarly induced by IFN␥/LPS treatment in both cell types (Fig. 4D). However, these complexes were neither competed by the C/EBP1 site or by the C/EBP␤-zipper peptide nor supershifted by anti-C/EBP␣, -␤, -␦, or -⑀ antibodies (Fig. 4D and not  shown). These data strongly suggest that the *C/EBP2 site, although able to form complexes that are induced by IFN␥/LPS stimulation, does not directly bind C/EBP proteins, at least in macrophages.
Binding to the cox-2 NF-B site was similarly induced in both K4 and W2 cells (Fig. 5A), ruling out the idea that abnormal NF-B activation may be responsible for the impaired induction of COX-2 mRNA. Finally, we analyzed the DNAprotein interactions occurring at the overlapping CRE/E box elements from the proximal cox-2 promoter region. As shown in Fig. 5B, two closely migrating complexes could be detected using this site as a probe, and no difference was observed between either treated or untreated K4 and W2 cells. Both complexes were competed by the CRE/E box sequence itself and almost completely abolished by competition with a consensus E box sequence. In contrast, competition with a CRE consensus sequence only caused a slight decrease in binding. Some reports suggest that C/EBP factors can exert their activating role through interaction with the CRE/E box element (9,18). How- FIG. 4. DNA binding activity on the C/EBP sites of the murine cox-2 promoter. A, schematic representation of the proximal cox-2 promoter. The sites analyzed and their position relative to the transcription start site are indicated. B and C, the C/EBP site 1 from the murine cox-2 promoter was used as a probe. Arrows and numbers on the left indicate the different DNA-protein complexes detected. Nuclear extracts from the K4 and W2 cell lines either untreated (u/t) or treated with IFN␥ for 16 h ϩ LPS for 4 h (IϩL) were used. B, polyclonal antibodies directed against different C/EBP isoforms (C/EBP␣, -␤, -␦, or -⑀) were incubated together with the extracts where indicated. C, only extracts from IFN␥ ϩ LPS-treated K4 and W2 cell lines were used. Where indicated, one of the following competitors (comp) was included in the incubation mix: C/EBP1 unlabeled double-stranded oligonucleotide (self); C/EBP leucine-zipper peptide; C/EBP D site from the albumin promoter (AlbD); C/EBP site from the hemopexin promoter (HpxA). The asterisk indicates a complex that was not reproducibly obtained in all experiments. D, nuclear extract as in B were incubated with radiolabeled *C/EBP2 site from the cox-2 promoter. Either the *C/EBP2 (self) or the C/EBP1 unlabeled oligonucleotides were used as competitors. F, free radiolabeled probe. Data are representative of at least two independent experiments. ever, despite our attempts to identify such an interaction through either direct competition or supershift experiments, we have failed to confirm this observation (data not shown).
Defective Transcription Driven by the cox-2 Promoter in the Absence of C/EBP␤-In order to analyze directly the transcriptional activity of the cox-2 promoter in the presence or absence of C/EBP␤, we isolated a region of the mouse cox-2 promoter that was reported to fully support inducible transcription in several cell lines including macrophages, and we fused it to the firefly luciferase reporter gene (reporter COX-2-963/luc, Fig.  6A). This vector was then transiently transfected into K4 or W2 cells, in the presence or absence of an expression plasmid encoding C/EBP␤. As shown in Fig. 6B, transcriptional activity was minimal in the mutant K4 cells and about 3 times higher in the wild type W2 cells. C/EBP␤ co-transfection was able to induce cox-2 promoter transcription to similar levels in both cell types and was therefore sufficient to fully rescue the defective promoter activity observed in the mutant cells. LPS treatment did not achieve appreciable induction, perhaps because the transfection process itself appeared to activate the cells.
Serial deletions of the cox-2 promoter progressively eliminating different cis-acting elements were then generated and tested for transcriptional activity in the K4 and W2 cells (Fig.  6, A and C). Deletion of the NF-B Ϫ402/Ϫ392 site (reporter COX-2-203/luc) caused transcriptional activity to decrease by about 50% in both the K4 and W2 cells, although remaining considerably lower in the mutant as compared with the wild type cells. In contrast, deletion of the C/EBP1 site (reporter COX-2-119/luc) did not further affect transcriptional activity in the mutant K4 cells, although it considerably reduced it in the wild type W2 cells. The activity of this construct was comparable in the two cell types, suggesting that the effect of the absence of C/EBP␤ on cox-2 promoter activity is exerted at the level of the C/EBP1 site. Transcriptional activity further dropped similarly in both cell types upon deletion of the *C/EBP2 site (reporter COX-2-79/luc), to finally reach background levels upon deletion of the CRE/E box (reporter COX-2-45/luc).
The Obligatory Role of C/EBP␤ in cox-2 Transcription Is Cell Type-specific-Induction of COX-2 expression was found to be normal and even prolonged in granulosa cells of the ovary from C/EBP␤-deficient mice (33). Moreover, PMA and v-Src-mediated cox-2 induction in NIH 3T3 fibroblasts has been reported to depend solely on the CRE/E box element (15). In order to compare directly the role of C/EBP␤ in regulating COX-2 expression in macrophages and fibroblasts, we have generated immortalized fibroblasts from the C/EBP␤-deficient mice making use of the 3T3 protocol (not shown), and we analyzed COX-2 mRNA induction triggered by PMA in C/EBP␤ Ϫ/Ϫ and C/EBP␤ ϩ/ϩ cells. As shown in Fig. 7A, PMA-mediated COX-2 mRNA induction was equivalent in both cell types. Similar results were obtained when the fibroblasts were stimulated with serum, recombinant tumor necrosis factor-␣, or IL-1␤ (data not shown), supporting the idea that C/EBP␤ is not required for COX-2 expression in fibroblasts. Next, we transiently transfected the COX-2 Ϫ963/luc reporter into C/EBP␤ Ϫ/Ϫ and C/EBP␤ ϩ/ϩ fibroblasts in the presence or absence of a plasmid expressing the constitutively active v-Src. The transcriptional activity of this construct was equivalent in both cell types and was similarly induced by v-Src, in agreement with the idea that a C/EBP␤-independent pathway controls cox-2 expression in fibroblasts. DISCUSSION C/EBP␤-deficient mice developed an age-related lymphoproliferative disease associated with diffused plasmacytosis and mucosal inflammation, and displayed abnormal immune responses consistent with impaired macrophage functions. These included defective production of bioactive IL-12 and nitric oxide, impaired T helper 1 responses, and failure to kill intracellular bacteria and tumor cells (24,34). Indeed, the analysis of immortalized and primary macrophages derived from the mutant mice has recently allowed us to identify a number of genes whose induction upon cellular activation is variably compromised in the absence of C/EBP␤. 2 Importantly, we also found that several genes are either normally or even more efficiently induced in the mutant macrophages, thus suggesting that the responsiveness to IFN␥/LPS is not compromised. 2 The finding that COX-2 expression is also profoundly impaired in the mutant macrophages adds this gene to the growing list of C/EBP␤ target genes in the monocyte/macrophage lineage, improving our molecular understanding of the defective cellular functions detected in the mutant mice. Interestingly, COX-2 activity can be involved in both the initiation of the inflammatory response and in the resolution phase, when the synthesis of anti-inflammatory prostaglandins such as PGD 2 is prevalent (35). In the light of the lymphoproliferative and inflammatory phenotype of the mutant mice, it will be of interest to explore the specific contribution of impaired COX-2 synthesis in the different phases of inflammation.
The regulation of cox-2 gene transcription is complex and varies according to the cell type, and the stimulus applied and, probably as a consequence, the role attributed to the different promoter elements and transcription factors involved is some- times contradictory. The C/EBP site is considered important in regulating COX-2-inducible transcription in several different cellular systems (6,9,10,12,13,17,18,20). However, the specific role and relative contribution of different family members and in particular of C/EBP␤ and -␦, which are induced by most treatments that stimulate COX-2 expression, cannot be easily assessed in normal cells with the usual overexpression methods. Our finding that COX-2 induction by LPS is almost totally defective in C/EBP␤ Ϫ/Ϫ macrophages unambiguously demonstrates the non-redundant role of C/EBP␤ in cox-2 gene transcription in these cells. Although we cannot exclude that C/EBP␦ or other family members are also involved, our supershift experiments suggest that in our cells very little, if any, C/EBP␦, -␣, or -⑀ bind to the cox-2 C/EBP promoter element, at least in vitro. The protein levels of C/EBP␦ were very low both before and after LPS treatment in the C/EBP␤ ϩ/ϩ cells, 2 explaining why no binding could be detected. Although C/EBP␦ was in contrast appreciably induced by LPS in the C/EBP␤ Ϫ/Ϫ cells, 2 it was still apparently unable to bind the C/EBP1 site on the cox-2 promoter. A likely explanation could be the need for C/EBP␦ to bind to this site as a heterodimer with C/EBP␤, as suggested by results we have recently obtained in RAW 264 macrophages. 3 This would also explain why C/EBP␦ is unable to compensate for the absence of C/EBP␤ in the context of the cox-2 promoter. The conclusion that defective expression in the C/EBP␤ Ϫ/Ϫ macrophages is primarily due to the inability of the cox-2 promoter to undergo efficient transcription in the absence of this factor rests on several independent observations. Both COX-2 expression and PGE 2 secretion were partially rescued by the low level of C/EBP␤ expressed in the revertant cells, and COX-2 mRNA induction was also defective in primary macrophages. COX-2 mRNA induction in response to LPS, IL-1␤, and tumor necrosis factor-␣ in human peripheral blood monocytes has been shown to be partly or largely due, respectively, to increased mRNA stability (36,37). However, in agreement with the lack of a critical transcription factor, the dramatically lower COX-2 mRNA levels detected in the C/EBP␤ Ϫ/Ϫ macrophages upon LPS treatment appear to be due exclusively to defective transcriptional activation as assessed by run-on experiments, since COX-2 mRNA appears to be equivalently stabilized in both the mutant and the wild type cells. In addition, no difference was detected in the DNA binding activities interacting with the other two main cis-acting elements involved in cox-2 regulation. Indeed, NF-B activation was equivalent in the mutant and wild type cells. In agreement with previous work, the DNA binding pattern detected with the CRE/E box was unchanged by LPS treatment and was identical between the C/EBP␤ Ϫ/Ϫ and C/EBP␤ ϩ/ϩ cells. This last observation demonstrates that the nuclear proteins interacting with this element are expressed at a normal level in the absence of C/EBP␤ without formally ruling out the possibility that activation FIG. 6. Analysis of cox-2 promoter activity in C/EBP␤ ؊/؊ and C/EBP␤ ؉/؉ macrophages. A, schematic representation of the different cox-2 promoter-luciferase fusion plasmids used. The positions of the different deletions are indicated. B, the full-length reporter plasmid Ϫ963 was transiently transfected by electroporation into K4 or W2 cells, along with a plasmid expressing C/EBP␤ where indicated. Cells were divided into two dishes, and after 18 h one was treated with LPS for 4 h (solid bars) prior to cell lysis. Firefly luciferase activity values were normalized to protein concentration in each sample. C, the indicated reporter plasmids were electroporated into K4 or W2 cells, along with pRL-TK (Renilla luciferase) as a control for transfection efficiency. Cells were treated as described in B. Firefly luciferase activity values were normalized to the Renilla luciferase activity. Values are shown as mean Ϯ S.E. of at least six independent experiments. Open bars, untreated cells; solid bars, cells treated with LPS for 4 h. through phosphorylation may be altered in the C/EBP␤ mutant cells. However, we have found that LPS-induced phosphorylation of CREB and ATF-1, two of the factors potentially involved in activating the cox-2 promoter through interaction with the CRE/E box, was normal in the C/EBP␤ Ϫ/Ϫ macrophages. 3 Finally, transcription of the cox-2 promoter upon transient transfection was profoundly impaired in the C/EBP␤ Ϫ/Ϫ cells as compared with the wild type controls, but co-transfection with C/EBP␤ was sufficient to fully rescue transcriptional activity, again suggesting that the only missing player in promoting cox-2 transcription in the mutant cells is indeed C/EBP␤.
The role of NF-B in regulating cox-2 expression is ambiguous; although several studies (6 -11) strongly implicated NF-B activity and the Ϫ402/Ϫ392 NF-B site in cox-2 induction in many cell types including LPS-treated macrophages, other recent data (12)(13)(14) failed to confirm its importance. In our system, the Ϫ402/Ϫ392 NF-B site did contribute to promoter activity, since transcription levels dropped by about 50% in both the mutant and wild type cells following its deletion. Following deletion of the C/EBP-1 element, promoter activity further dropped in the wild type but remained the same in the mutant cells, thus suggesting that C/EBP␤ action mainly occurs through this site. In agreement with this idea, the activity of the promoter constructs became equivalent in both kinds of cells after deletion of the C/EBP-1 site. The activity of the COX-2-79/luc promoter construct, which maintains the CRE/E box element but lacks the Ϫ93/Ϫ85 site, was similarly decreased in both cell types, thus confirming that this element does play a role in mediating cox-2 transcription. Likewise, the complete loss of activity displayed by the Ϫ45/luc construct is in agreement with the essential role of the Ϫ59/Ϫ48 CRE/E box element previously described in several cell types (9, 12, 14 -21). In contrast with some previously published data (9,12,18), both these elements appeared to exert their function on cox-2 transcription independent of C/EBP␤, since no difference was detected between the mutant and wild type cells when deleting either of them. Moreover, we could not find any evidence for direct interaction of any C/EBP family member with either promoter element in our cells. However, we cannot rule out that C/EBP family members such as C/EBP␦ might be involved indirectly in promoting transcription from these sites.
Recently published work (38) has demonstrated a role for IRF-1 in promoting synergetic COX-2 induction by IFN␥ and LPS in macrophages. The promoter elements involved are, however, located upstream of the promoter region analyzed here, and therefore we could not assess whether the IRF-1-dependent regulation of COX-2 was in any way affected in the absence of C/EBP␤.
One of the unusual characteristics of COX-2 regulation is the extreme cell specificity of the mechanisms involved (3). The ability to generate and analyze both macrophages and fibroblasts from the C/EBP␤ Ϫ/Ϫ mice allowed us to compare directly the specific role played by this factor in the two cell types. The results were surprisingly clear-cut, with C/EBP␤ being absolutely required for COX-2 induction in macrophages but completely dispensable in fibroblasts, despite C/EBP␤ being expressed in both cell types. This difference may be due to a different activation threshold of the cox-2 promoter in macrophages and fibroblasts, as suggested by the observation that only the CRE/E box element is required for efficient promoter activity in fibroblasts (14). Alternatively, the involvement of different signals and pathways may lead in fibroblasts to the activation of distinct factors, making C/EBP␤ redundant. In light of these results, it will be of interest to determine the levels of COX-2 expression in a variety of different cell types in the C/EBP␤ Ϫ/Ϫ mice. Indeed, abnormally high COX-2 expression is thought to be involved in the development of pathological conditions, such as colon and skin cancer and rheumatoid arthritis (5,39). Elevated C/EBP␤ levels have been reported in epithelial cancers and rheumatoid arthritis (17,40), thus suggesting that altered C/EBP␤ activity might be directly responsible for COX-2 up-regulation in these conditions and that this factor could therefore represent a new potential therapeutic target. Indeed, in chondrocytes induction of the cox-2 and phospholipase A2 promoters by IL-1␤, the most abundant inflammatory cytokine in the arthritic joint, has been shown recently to be dependent on C/EBP␤ and -␦ (9,41). Moreover, C/EBP␤ Ϫ/Ϫ macrophages also display defective production of nitric oxide, 2 which is also implicated in the progressive destruction of the affected joints in rheumatoid arthritis. Direct determination of the specific role played by C/EBP␤ in inducing COX-2 expression in cells such as cartilage chondrocytes and skin or epithelial tumors will be instrumental in predicting the potential therapeutic use of inhibitors of C/EBP␤ activity in diseases involving these systems.