The Induction of Cyclooxygenase-2 mRNA in Macrophages Is Biphasic and Requires both CCAAT Enhancer-binding protein β (C/EBPβ) and C/EBPδ Transcription Factors

Prostaglandins are important mediators of activated macrophage functions, and their inducible synthesis is mediated by cyclooxygenase-2 (COX-2). Here, we make use of the murine macrophage cells RAW264 as well as of immortalized macrophages derived from mice deficient for the transcription factor CCAAT enhancer-binding protein β (C/EBPβ) to explore the molecular mechanisms regulating COX-2 induction in activated macrophages. We demonstrate that lipopolysaccharide-mediated COX-2 mRNA induction is biphasic. The initial phase is independent of de novo protein synthesis, correlates with cAMP-response element-binding protein (CREB) activation, is inhibited by treatments that abolish CREB phosphorylation and reduce NF-κB-mediated gene activation, and requires the presence of the transcription factor C/EBPβ. On the other hand, C/EBPδ appears to be essential in addition to C/EBPβ to effect the second phase of COX-2 gene transcription, which is important for maintaining the induced state and requires de novo protein synthesis. Indeed, both phases of COX-2 induction were defective in C/EBPβ−/− macrophages. Moreover, the synthesis of C/EBPδ was increased dramatically by treatment with lipopolysaccharide and, like COX-2 induction, repressed by combined inhibition of the MAPK and of the SAPK2/p38 cascades. Taken together, these data identify CREB, NF-κB, and both C/EBPβ and -δ as key factors in coordinately orchestrating transcription from the COX-2 promoter in activated macrophages.

Cyclooxygenase-2 (COX-2) 1 is a key enzyme catalyzing the rate-limiting step in the inducible production of prostaglandins (PG), and its synthesis can be readily induced in many different cell types in response to a variety of stimuli (1). Although the contrasting biological properties of the different prostanoids make it difficult to define their roles in physiological processes unambiguously, PG secretion by activated macrophages clearly represents an important step in the inflammatory process (2). Indeed, COX-2, as well as COX-1, the isoform responsible for the basal steady state production of PG, represents the main target for nonsteroidal anti-inflammatory drugs, and compounds that can specifically inhibit the inducible but not the basal production of PG (i.e. COX-2 but not COX-1 activity) are being tested for the treatment of chronic inflammatory diseases such as rheumatoid arthritis or ulcerative colitis (3)(4)(5).
Many studies have therefore recently focused on the mechanisms regulating inducible COX-2 expression in monocytic cells. Three main cis-acting elements have been identified on the murine COX-2 promoter that play a role in LPS-mediated induction of COX-2 transcription in macrophages. NF-B is a transcription factor involved in LPS-mediated induction of many cytokines and inflammatory products, and inhibition of NF-B activity has been reported to impair COX-2 mRNA induction (6 -11). The Ϫ138/Ϫ130 C/EBP element is generally believed to play an important role in COX-2 promoter induction in macrophages as well as in other cell types, mainly through interactions with the two C/EBP family members C/EBP␤ and -␦ (9 -16). Finally, the overlapping CRE/E-box recognition sequence located at positions Ϫ59/Ϫ48 appears to be the most generally required promoter element, being essential for both basal and induced COX-2 transcription in most cellular systems analyzed (10,12,13,(15)(16)(17)(18)(19)(20)(21).
We and others have previously shown that stimulation of macrophages with LPS elicits the activation of the classical mitogen-activated protein kinase (MAPK) cascade and the homologous stress-activated protein kinase 2 (SAPK2)/p38 pathway (7,22,23). Moreover, combined suppression of these pathways utilizing the small cell-permeant inhibitors PD 98059 (24,25) or U0126 (26), which specifically inhibit the activation of the MAPK kinase-1, together with SB 203580, a specific inhibitor of SAPK2/p38 activity (27), resulted in the coordinated inhibition of LPS-stimulated CREB/ATF1 phosphorylation and COX-2 mRNA and protein induction (28). However, the protein kinase A-mediated phosphorylation of CREB following cell treatment with forskolin did not trigger detectable COX-2 protein induction, suggesting that CREB activation, even if required, is not sufficient to activate COX-2 expression (28). Serendipitously, we have recently observed that COX-2 expression in response to LPS was profoundly impaired in macrophages derived from mice where the transcription factor C/EBP␤ was inactivated (29). COX-2 induction could be rescued by transient or stable re-expression of C/EBP␤, suggesting that this factor is required for efficient COX-2 gene tran-scription in macrophages. Here we explore the kinetics of COX-2 mRNA induction and how it correlates with the induced activities of CREB and C/EBP factors, making use of both the murine macrophage cell line RAW264 and immortalized C/EBP␤Ϫ/Ϫ or ϩ/ϩ macrophages. We demonstrate the existence of two waves of COX-2 induction. The first does not involve de novo protein synthesis, correlates with CREB and NF-B activation, and requires preexisting C/EBP␤, while the second involves newly synthesized C/EBP␦ and requires the DNA binding activity of C/EBP␤⅐C/EBP␦ heterodimers.
Cell Culture and Stimulation-RAW264 macrophages were maintained at 37°C in 5% CO 2 atmosphere in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) heat-inactivated fetal calf serum, 100 units/ml penicillin, 100 g/ml streptomycin. 2 h before stimulation, the medium was removed and replaced with 2 ml of Dulbecco's modified Eagle's medium. The cells were then stimulated with 100 ng/ml LPS or 20 M forskolin plus 10 M IBMX for the times indicated in the figure legends. Where indicated, SB 203580 (10 M) and/or U0126 (10 M) were added 1 h before stimulation.
The generation of C/EBP␤ Ϫ/Ϫ and 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 containing 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. Cells at ϳ70% confluence were stimulated with 100 units/ml of interferon-␥ for 16 h and with 100 ng/ml LPS for the indicated times.
Immunoblotting Analysis-Proteins were denatured in SDS, electrophoresed on a 4 -12% SDS-polyacrylamide gel, and transferred to nitrocellulose membranes. Ponceau S staining was performed in order to ensure equivalent gel loading. Membranes were then incubated with the antibodies described below, which were detected using the enhanced chemiluminescence reagent (ECL). For immunoblotting of C/EBP␤ or C/EBP␦, 50 g of nuclear cell extracts were electrophoresed and immunoblotted using a monoclonal C/EBP␤ antibody or a polyclonal C/EBP␦ antibody, respectively. For immunoblotting of IB␣, 30 g of total proteins were electrophoresed and immunoblotted using a polyclonal anti-IB␣ antibody. For immunoblotting CREB, cell lysates (30 g of protein) were electrophoresed and immunoblotted using, respectively, an anti-CREB antibody or an anti-phosphospecific CREB antibody recognizing CREB phosphorylated at Ser 133 and ATF-1 phosphorylated at Ser 63 .
EMSAs were performed with 4 g of nuclear extract in 20 mM Hepes (pH 7.9), 1 mM EDTA, and 2.5 mM dithiothreitol, containing 3 g of poly(dI-dC). The complexes were separated by electrophoresis on a 6% (for C/EBP) or 5% (for NF-B and CRE/E-box) polyacrylamide-0.25ϫ Tris borate-EDTA gel. For supershift experiments, 2 l of polyclonal purified antibody was incubated with nuclear extracts and poly(dI-dC) for 30 min on ice prior to the probe addition. Unlabeled double-stranded oligonucleotide competitors were preincubated at a 50-fold molar excess 10 min prior to the probe addition.
RT-PCR-Total RNA was prepared from LPS-stimulated or control RAW264 cells using the RNeasy Mini Kit according to the manufacturer's protocol. Total RNA was measured, and 50 ng was reverse transcribed using Promega avian myeloblastosis virus reverse transcriptase (5 units/ml) with the oligonucleotides GTTGGATACAGGCCAGACTT-TGTTG and GAGGGTAGGCTGGCCTATAGGCT (for amplification of the "housekeeping gene," hypoxanthine guanine phosphoribosyltransferase (HPRT)), together with those for the COX-2 gene (CAGC-AAATCCTTGCTGTTCC and TGGGCAAAGAATGCAAACATC) or for the C/EBP␦ gene (CGGCACAGTCCGAGAAAAGG and TTGAAGAAC-TGCCGGAGGCC). Conditions for PCR amplification of the resulting first-strand DNA template were 94°C denaturing for 30 s, 60°C annealing for 1 min, 68°C extension for 1 min, 30 cycles using thermostable Tfl DNA polymerase (5 units/ml), and 1 mM MgSO 4 . The PCR products showed a band of 515 bp for COX-2, a band of 233 bp for C/EBP␦, and a band of 352 bp for HPRT.
Statistical Analysis-Results obtained after densitometric quantifications were analyzed using the two-tailed t test. A p value of Ͻ0.05 was considered statistically significant.

COX-2 mRNA Induction Is
Biphasic-Phosphorylation of CREB obtained by treatment of RAW264 macrophages with forskolin and IBMX did not trigger detectable accumulation of COX-2 protein after 4 h, suggesting that a distinct factor(s) induced by LPS but not by forskolin is required for COX-2 induction (28). Since, however, forskolin induces a faster and more transient phosphorylation of CREB compared with LPS stimulation, we decided to also analyze COX-2 mRNA induction at shorter time points. Indeed, forskolin was able to trigger an increase of the COX-2 mRNA (1.8-fold) already after 30 min of treatment (Fig. 1A, compare lanes 1 and 2), corresponding to maximal CREB phosphorylation (28). This induction peaked (3.3-fold) 1 h after the treatment (Fig. 1A, lane 3) and rapidly decreased, having almost returned to unstimulated levels (1.14-fold) by 2 h (Fig. 1A, lanes 4 -6).
Taken together, these results suggest a biphasic activation of the COX-2 gene. The first phase, corresponding to the initial activation, correlates with the kinetics of CREB phosphorylation and does not involve de novo protein synthesis, while the second phase, involved in the maintenance of the induced state, requires the action of some newly synthesized factor(s).

The LPS-mediated Induction of C/EBP␦, but Not of C/EBP␤, Is Suppressed by Treatment with U0126 plus SB 203580 -
Members of the C/EBP family, and particularly C/EBP␤ and -␦, may well be involved in the second phase of COX-2 transcriptional induction, since their synthesis is increased by LPS in a number of cell types (36). We have therefore analyzed their induction following LPS treatment in the presence or absence of the protein kinase inhibitors that abolish COX-2 expression. C/EBP␤ is already present in untreated macrophages, but it is increased by 2-3-fold following LPS treatment, peaking at 4 h (not shown). We have shown previously that treatment with PD 98059 and/or SB 203580 did not affect the induction of C/EBP␤ triggered by LPS (28). Indeed, U0126 and SB 203580, alone or in combination, were unable to modify the 3-fold LPSinduced increase of all three C/EBP␤ isoforms, the full-length protein (FL), the liver-activating protein (LAP), and the liverinhibitory protein (LIP) (Fig. 2A, upper panel, compare lane 1  with lanes 2-5).
The levels of C/EBP␦ were in contrast almost undetectable in untreated RAW264 cells, but they were increased dramatically following LPS treatment ( Fig. 2A, lower panel, compare lanes 1  and 2). Strikingly, the LPS-mediated induction of C/EBP␦ was almost totally abolished by combined treatment with U0126 plus SB 203580 ( Fig. 2A, lower panel, compare lanes 2 and 5) and only slightly decreased by treatment with either compound alone (lanes 3 and 4). This is in marked contrast to what was observed with C/EBP␤, and suggests that C/EBP␦ may represent the factor, or one of the factors, whose synthesis is required to maintain COX-2 transcriptional induction. Treatment with U0126 plus SB203580 Abolishes the Induction of C/EBP␤⅐C/EBP␦ DNA Binding Activities-Next, we analyzed by EMSA the DNA-protein interactions occurring at the level of the Ϫ138/Ϫ130 C/EBP site from the murine COX-2 promoter. Four differentially migrating complexes could be detected using nuclear extracts from untreated RAW264 cells (Fig. 2B, lanes 1, 2, and 11). Complexes 1 and 2 were induced by 5-fold upon LPS treatment (Fig. 2B,  . The inhibition of DNA binding appeared to be specific to the newly induced activities, since the inhibitors did not affect the formation of complexes using extracts from untreated cells (Fig. 2B, compare lanes 11  and 12).
All DNA-protein complexes detected at the level of the COX-2 C/EBP site could be abolished by an excess of unlabeled double-stranded oligonucleotide carrying either the same sequence (self) or the sequence of a known C/EBP binding site from the hemopexin promoter (HpxA) (Fig. 2C, lanes 6 and 7,  13 and 14, and 20 and 21), but not by an unrelated oligonucleotide (not shown). To assess if C/EBP proteins were involved in the formation of the different complexes detected and particularly of induced complexes 1 and 2, we performed supershift experiments using polyclonal antibodies against C/EBP␣, -␤, -␦, or -⑀ and nuclear extracts from RAW264 cells either untreated or treated with LPS (Fig. 2C). In extracts from untreated cells, all complexes could be supershifted by anti-C/ EBP␤ antibodies (Fig. 2C, lane 3) and did not contain any of the other tested C/EBP family members, as confirmed by densitometric analysis of the retarded bands (Fig. 2C, lanes 2, 4, and 5), although the same antibodies could readily supershift complexes obtained with different extracts and probes (data not shown). Upon LPS treatment, still no binding of either C/EBP␣ or C/EBP⑀ could be detected (Fig. 2C, lanes 9 and 12), while C/EBP␤ was involved in the formation of all four complexes, since all, including those induced by LPS, were abolished by specific antibodies raised against this protein (Fig. 2C, lane 10). C/EBP␦, which was absent from the complexes formed using extracts from untreated cells, was in contrast detected as part of the LPS-induced complexes 1 and 2, 40% of which could be supershifted by anti-C/EBP␦ antibodies (Fig. 2C, lane 11). Of note, the amount of complexes supershifted could not be increased using more anti-C/EBP␦ antibodies (data not shown). Taken together, these data suggest that complexes 1 and 2 only contain C/EBP␤ homodimers in untreated RAW264 macrophages, while they are formed partly of C/EBP␤ homodimers and partly of C/EBP␤⅐C/EBP␦ heterodimers in LPS-treated cells.
We next asked whether binding of C/EBP␤, C/EBP␦, or both was affected by U0126 and SB 203580. As expected, all residual DNA-protein complexes obtained upon LPS treatment in the presence of U0126 and SB 203580 could still be supershifted by FIG. 1. COX-2 mRNA induction in RAW264 macrophages as determined by semiquantitative RT-PCR. A, cells were left untreated or stimulated for the times indicated with 20 M forskolin plus 10 M IBMX, and total RNA was extracted and subjected to RT-PCR using specific primers for COX-2 and HPRT (internal control). Densitometric analysis of three independent experiments was carried out. COX-2 values were normalized against HPRT values, and statistical analysis was performed using a two-tailed t test. The differences between lanes 1 and 2 or lanes 1 and 3 were found to be statistically significant (p Յ 0.0002). B, cells were incubated for 30 min in the presence or absence of 50 g/ml CHX and then stimulated for 1.5 or 4 h with or without LPS (100 ng/ml) in the continuous presence or absence of the inhibitor. Total RNA was extracted and analyzed as in A. Densitometric and statistical analysis was performed as in A. The differences between lanes 1 and 2 or lanes 5 and 6 (p Ͻ 0.0005) and between lanes 6 and 8 (p Ͻ 0.0001) were statistically significant.
anti-C/EBP␤ antibodies (Fig. 2C, lane 17). Interestingly, as confirmed by densitometric analysis, no supershift with anti-C/EBP␦ antibodies could be detected anymore after treatment with the inhibitors (Fig. 2C, lane 18), suggesting that the DNA binding activity involving C/EBP␤⅐C/EBP␦ heterodimers may represent the main target for U0126 and SB 203580 action.
DNA Binding to the CRE/E-box and NF-B Elements of the COX-2 Promoter Is Not Impaired following U0126 and SB 203580 Treatment-To verify if the inhibition of DNA binding by treatment with U0126 and SB 203580 is specifically limited to the C/EBP proteins among the factors playing a role in COX-2 gene transcription, we have examined the pattern of DNA-protein complexes forming at the level of the two other main cis-acting elements involved in the LPS-inducible activation of the COX-2 promoter. EMSA experiments were therefore performed using the CRE/E-box element located at positions Ϫ59/Ϫ48 and the Ϫ402/Ϫ392 NF-B binding site as probes and nuclear extracts from RAW264 cells either untreated or treated with LPS in the presence or absence of a combination of U0126 and SB 203580 (Fig. 3).
As expected from knowledge that the proteins involved can bind constitutively to DNA, binding to the CRE/E-box element was already detected in extracts from untreated cells (Fig. 3A, lane 1) and increased by 1.9-or 1.3-fold, respectively, upon 1.5 or 4 h of LPS treatment (Fig. 3A, compare lanes 1, 3, and 5). Treatment with the inhibitors enhanced by 2-fold the binding prior to LPS treatment (compare lanes 1 and 2), but this phenomenon was not investigated further. In contrast, binding to the NF-B element, which was not detected in extracts from untreated cells (Fig. 3B, lanes 1 and 2) was strongly increased by treatment with LPS for 1.5 h (Fig. 3B, lane 3) and decreased by 50% after 4 h (Fig. 3B, lane 5). Importantly, binding to

FIG. 2. Effect of different protein kinase inhibitors on the LPS-mediated induction of C/EBP␤ and ␦ synthesis and C/EBP DNA binding activity in RAW264 macrophages.
A, cells were incubated for 1 h in the presence or absence of SB 203580 (SB) and/or U0126 (U) and then stimulated for 4 h with or without 100 ng/ml LPS in the continuous presence or absence of the inhibitors. Nuclear extracts were prepared and immunoblotted with an anti-C/EBP␤ antibody (upper panel) that recognizes all three isoforms of the protein, full length (FL), liver-activating protein (LAP), and liver-inhibitory protein (LIP) (indicated by the arrows) or an anti-C/EBP␦ antibody (lower panel). Equivalent gel loading was assessed by Ponceau S staining of each filter prior to immunostaining. Densitometric and statistical analysis of three independent experiments was performed. The differences between lanes 1 and 2 of the upper and lower panels (p Ͻ 0.003) and between lanes 2 and 5 of the lower panel (p ϭ 0.001) were found to be statistically significant. B, nuclear extracts prepared as in A were analyzed by EMSA using the Ϫ198/Ϫ130 C/EBP site from the murine COX-2 promoter as a probe. The arrows and numbers on the left indicate the different DNA-protein complexes detected. Densitometric and statistical analysis of complexes 1 and 2 from three independent experiments was performed. The differences between lanes 1 and 2 and lanes 3 and 4 (p ϭ 0.04) and between lanes 3 and 4 and lanes 9 and 10 (p ϭ 0.03) were found to be statistically significant. C, nuclear extracts were prepared and analyzed as in B. Where indicated, polyclonal antibodies directed against different C/EBP isoforms (C/EBP␣, -␤, -␦, or -⑀) were also included in the incubation mix. For competition experiments, a 50-fold molar excess of one of the following unlabeled oligonucleotides was used: Ϫ138/130 COX-2 C/EBP site (self) or C/EBP site from the hemopexin promoter (HpxA). Densitometric and statistical analysis of complexes 1 and 2 from three independent experiments was performed. The differences between lanes 1 and 3 (p ϭ 0.0005), between lanes 8 and 10 (p ϭ 0.0001), between lanes 8 and 11 (p ϭ 0.02), and between lanes 15 and 17 (p ϭ 0.004) were found to be statistically significant.
neither element was inhibited by U0126 and SB 203580 treatment.
Intriguingly, LPS-induced nuclear NF-B DNA binding activity was both increased and prolonged by treatment with the inhibitors as shown in Fig. 3B by comparing lane 3 with lane 4 (4-versus 8-fold induction) and lane 5 with lane 6 (2-versus 3.5-fold induction). Since increased NF-B activation may be due to impaired IB␣ resynthesis after LPS-induced degradation, we have assessed IB␣ levels in RAW264 cells both before and after 1 h of LPS treatment, a time when NF-B-triggered IB␣ resynthesis should be completed (37), in the presence or absence of U0126 and/or SB 203580 (Fig. 3C). IB␣ levels were indeed reduced about 4-fold in the presence of both inhibitors (Fig. 3C, compare lanes 2 and 5).
Neither C/EBP␤ nor C/EBP␦ Is Induced by Forskolin Treatment-If C/EBP␦ and/or C/EBP␤ are indeed the factors whose synthesis is required to effect the second phase of COX-2 transcriptional induction, they should not be induced by forskolin treatment, since this only triggers transient COX-2 mRNA induction that is extinguished before the newly synthesized factors could start accumulating. Indeed, the levels of C/EBP␤ as detected by Western blot were only increased slightly (2.1fold) following forskolin treatment in comparison with the much stronger 5.5-fold induction obtained with LPS (Fig. 4A,  upper panel). Even more strikingly, forskolin treatment completely failed to induce C/EBP␦ in contrast to the dramatic 22.3-fold induction triggered by LPS treatment (Fig. 4A, lower  panel). In agreement with these findings, forskolin could only trigger a weak (1.4-fold) increase of C/EBP DNA binding activ-  Fig. 2A were analyzed by EMSA using the Ϫ59/Ϫ48 CRE/E-box (A) or the Ϫ402/Ϫ392 NF-B site (B) from the murine COX-2 promoter as a probe. F, free probe. Densitometric and statistical analysis of two independent experiments was performed. In B, the differences between lanes 3 and 4 or lanes 5 and 6 (p Յ 0.001) were statistically significant. C, RAW264 macrophages were incubated for 1 h in the presence or absence of SB 203580 (SB) and/or U0126 (U) and then stimulated for 1 h with or without LPS in the continuous presence or absence of the inhibitors. Cell lysates were prepared, and 30 g of total protein were analyzed by immunoblotting using an anti-IB␣ antibody. Equivalent gel loading was assessed by Ponceau S staining of each filter prior to immunostaining. Densitometric and statistical analysis of four independent experiments was performed. The difference between lanes 2 and 5 (p ϭ 0.005) was statistically significant.

FIG. 4. Comparison of LPS-or forskolin-mediated induction of C/EBP␤ and C/EBP␦ synthesis and DNA binding activity.
A, RAW264 macrophages were left untreated or stimulated for 4 h with LPS or forskolin (F) plus IBMX. Nuclear extracts were prepared and analyzed by Western blotting using anti-C/EBP␤ or anti-C/EBP␦ antibody. The full-length (FL) and LAP isoforms of C/EBP␤ are indicated. Equivalent gel loading was assessed by Ponceau S staining of each filter prior to immunostaining. Densitometric and statistical analysis of three independent experiments was performed. The differences between lanes 1 and 2 of the upper panel (p ϭ 0.005) and of the lower panel (p ϭ 0.004), but not between lanes 1 and 3 of both panels, were found to be statistically significant. B, nuclear extracts prepared as in A were analyzed by EMSA using the Ϫ198/Ϫ130 C/EBP site from the murine COX-2 promoter as a probe. Densitometric and statistical analysis of complexes 1 and 2 from three independent experiments was performed. The difference between lanes 1 and 2 and lanes 3 and 4 (p ϭ 0.02) was statistically significant. C, nuclear extracts obtained from forskolin (F) plus IBMX treatment were analyzed as in B and, where indicated, were preincubated with polyclonal antibodies directed against different C/EBP isoforms (C/EBP␣, -␤, -␦, or -⑀). The arrows and numbers on the left indicate the different DNA-protein complexes detected. Densitometric and statistical analysis of three independent experiments was performed. The difference between lanes 1 and 3 (p ϭ 0.005) was statistically significant.
ity that contained C/EBP␤ but not C/EBP␦ as assessed by EMSA and confirmed by densitometric analysis of supershift experiments (Fig. 4, B and C).
Both the Early and the Late Phase of LPS-mediated COX-2 Induction Are Defective in the Absence of C/EBP␤-We have reported that COX-2 expression in macrophages is defective in the absence of C/EBP␤ following 4 h of LPS treatment (29). To extend these data to earlier time points, we have analyzed COX-2 mRNA levels at various times after LPS treatment in C/EBP␤Ϫ/Ϫ and ϩ/ϩ macrophages. As shown in Fig. 5A, COX-2 mRNA slowly accumulated in the mutant cells, but its levels remained dramatically lower than in the wild type cells at all time points analyzed, suggesting that the basal levels of C/EBP␤ present in untreated cells may already be required to initiate COX-2 expression in conjunction with the U0126 and SB 203580-sensitive activation of CREB, NF-B, and possibly other transcription factor(s). Importantly, in the C/EBP␤Ϫ/Ϫ cells, LPS-induced CREB phosphorylation was normal (Fig.  5B), as was the LPS-mediated NF-B DNA binding induction (29). These data indicate that defective COX-2 expression in the C/EBP␤Ϫ/Ϫ cells is probably a direct consequence of the absence of C/EBP␤ and is not due to defective activation of the signaling pathways leading to CREB or NF-B activation.
The Induction of C/EBP␦ mRNA Requires de Novo Protein Synthesis-Little is known about the mechanisms regulating the LPS-mediated induction of C/EBP␦ in RAW264 cells. Our observation that C/EBP␦ induction, like COX-2 induction, is abolished by treatment with U0126 plus SB 203580 suggests the possibility that perhaps the same mechanisms involved in COX-2 transcriptional activation may also be responsible for the activation of the C/EBP␦ gene. In order to test this idea, we analyzed by RT-PCR the levels of C/EBP␦ mRNA in RAW264 macrophages after forskolin treatment or after treatment with LPS in the presence or absence of CHX (Fig. 6). C/EBP␦ mRNA could not be induced by forskolin treatment at any of the time points analyzed (Fig. 6A), and its 3-fold induction by LPS was abolished by CHX treatment after both 1.5 and 4 h as indicated by densitometric analysis (Fig. 6B, compare lane 2 with lane 4, and compare lane 6 with lane 8). This was in marked contrast to COX-2 mRNA induction, suggesting that distinct additional factor(s) are probably required to activate the C/EBP␦ gene.

DISCUSSION
The regulation of the COX-2 promoter in macrophages is complex and involves different promoter elements and transcription factors, but their relative roles are not completely understood. We demonstrate here that induction of the COX-2 mRNA following LPS treatment is biphasic, presumably effected by preexisting transcription factors that become posttranslationally activated in the first phase while requiring the synthesis of a new factor(s) for the second phase.
The initial phase correlates with the activation of the transcription factor CREB, both upon LPS and forskolin treatment. Moreover, both CREB activation and the first phase of COX-2 induction by LPS are insensitive to inhibition of protein synthesis. Notwithstanding the strong correlations between CREB activation and COX-2 induction, recent work proposing that CREB cannot activate the COX-2 promoter in transient transfection assays imposes caution (16), and further work is needed to find out whether CREB is required to initiate COX-2 transcription and which other factors are involved.
NF-B is a good candidate to be one of these factors. Inhibition of the p38/SAPK pathway (38) and of the MAPK pathway (39), which impairs COX-2 induction, is known to cooperatively repress NF-B-dependent gene expression, and we show that in RAW264 macrophages treatment with SB 203580 plus U0126 impairs IB␣ induction, known to be NF-B-dependent (37). Taken together, these observations suggest that in our system as well NF-B transactivating capacity may be decreased by treatment with the inhibitors. This decrease might therefore at least in part account for the inhibition of LPSmediated COX-2 induction triggered by treatment with SB 203580 plus U0126. Activating protein-1 factors and, in particular, c-Jun have been proposed to play an important role in COX-2 promoter activation both by co-transfection assays and by inhibition studies with a dominant negative form of c-Jun N-terminal kinase (16). Indeed, it has been proposed that it is c-Jun and not CREB that binds to the CRE in the COX-2 promoter. However, the finding that the LPS-induced COX-2 induction is suppressed by a combination of SB 203580 plus PD 98059 cannot be explained by the inhibition of c-Jun N-terminal kinase activation, because the LPS-induced phosphorylation of c-Jun is unaffected by these compounds. 3 Likewise, the induction of COX-2 by forskolin cannot be explained by activation of c-Jun N-terminal kinase, since this kinase is not activated by forskolin (40). However, our results do not exclude the possi-bility that c-Jun N-terminal kinase and c-Jun activity may play a role in COX-2 gene transcription.
Importantly, neither CREB nor NF-B can be sufficient to both initiate and maintain COX-2 transcription that is still intensely active 3 h after LPS treatment (29), when CREB phosphorylation has long since subsided (28) and NF-B nuclear localization is already strongly decreased (37). Members of the C/EBP family of transcription factors, and in particular C/EBP␤ and -␦, are good candidates to represent the transcription factors whose synthesis is required for the second, CHXsensitive phase of COX-2 induction. Indeed, their synthesis is induced by LPS in RAW264 macrophages, although C/EBP␤ is already present at appreciable levels in unstimulated cells. The finding that forskolin does not significantly increase either C/EBP␤ or C/EBP␦ levels and that treatment with U0126 plus SB 203580, which abolishes COX-2 induction, also inhibits the 3 S. Morton and P. Cohen, unpublished results. LPS-mediated induction of C/EBP␦, hence abolishing the binding of C/EBP␤⅐C/EBP␦ heterodimers to the COX-2 promoter, are consistent with the idea that C/EBP␤ and C/EBP␦ are essential to mediate the second phase of LPS-mediated COX-2 induction in macrophages.
While C/EBP␦ levels are very low in unstimulated cells, C/EBP␤ is already present at appreciable levels before LPS treatment and appears to play an obligatory role both in the initiation and in the maintenance of COX-2 gene activation, since both phases are profoundly impaired in C/EBP␤-deficient macrophages. Although C/EBP␤ transcriptional activity can be increased by phosphorylation (41,42), no data are available describing specific phosphorylation events occurring in LPStreated macrophages and, in preliminary experiments, we could not detect phosphorylation taking place following LPS treatment. 4 We therefore favor the idea that C/EBP␤ is already active in unstimulated cells but is either unable to bind to the promoter prior to CREB and/or NF-B activation or is not sufficient on its own to initiate transcription.
The model depicted in Fig. 7 describes the potential interplay of activated and inducible transcription factors taking place at the level of the COX-2 promoter in macrophages before and after LPS treatment. Although no data concerning promoter occupancy in vivo are available, at least in vitro both the CRE/E-box and the C/EBP elements can be already occupied under unstimulated conditions by preexisting members of the CREB/ATF and C/EBP families, respectively (Fig. 7A). The activation of the MAPK and SAPK pathways by LPS results, among other events, in phosphorylation of CREB and in improved NF-B activity, and the activation of the IB kinase complex by MAPK/extracellular signal-regulated kinase kinase kinase-1 leads to phosphorylation and degradation of IB and hence to the migration of NF-B to the nucleus. Both phosphorylated CREB and NF-B are known to recruit histone acetylases such as CBP/p300, which in turn contribute to making the promoter more accessible to transcription factors and help in bridging the transcription factor-CBP complexes to components of the basal transcription machinery (43,44). These changes are likely to allow preexisting C/EBP␤ to bind more stably to its recognition site, adjacent to the CRE. This step is required to initiate transcription (Fig. 7B), perhaps because C/EBP␤ can also bind to CBP/p300, thus stabilizing its interaction with the promoter (45). At the same time, LPS also triggers transcriptional induction of the C/EBP␤ and -␦ genes through still uncharacterized mechanisms (see below). During later phases of induction, CREB phosphorylation subsides, and NF-B is sequestered back into the cytoplasm by newly synthesized IB, but at this stage, more C/EBP␤ and newly made C/EBP␦ are present and capable of interacting with the promoter (Fig. 7C). Either their increased abundance or, more likely, the availability of C/EBP␤⅐C/EBP␦ heterodimers in addition to C/EBP␤ homodimers is able to overcome the need for additional factors binding to the CRE element, perhaps even through direct or indirect interactions with the CRE site itself as proposed previously (10,15). In addition, other transcription factors able to bind to the CRE site, such as the upstream stimulating factor-1 or -2 and members of the activating protein-1 family, may also come into play either in the initial phases or once the promoter has been activated (13,18,21).
The induction of C/EBP␤ and -␦ appears to be regulated differentially in RAW264 macrophages. Both genes can be induced by a variety of stimuli in different cell types and particularly by proinflammatory cytokines and LPS (36). In hepatocytes, the transcription factor STAT3 is thought to be involved in activating both genes in response to interleukin-6 (46 -48), and C/EBP␤ gene transcription has been proposed to be regulated by CREB/ATF factors in hepatocytes as well as in the promonocytic cells U937 (49,50). In RAW264 cells, C/EBP␤ induction appears not to require CREB, since it is not affected by the treatments that abolish CREB activation (28). In contrast, CREB may well be involved in the induction of the C/EBP␦ gene in LPS-treated cells, since the same treatments that inhibit CREB phosphorylation also abolish C/EBP␦ induction. However, the induction of C/EBP␦ also requires distinct as yet unidentified newly synthesized factors, since it is abolished by CHX.
Interestingly, although the relative abundance of C/EBP␤ and -␦ appeared to be at least equivalent after LPS induction, C/EBP␦ binding to the COX-2 promoter was only observed as part of a heterodimer with C/EBP␤. This suggests that C/EBP␤ homodimers and/or C/EBP␤⅐C/EBP␦ heterodimers may display a higher affinity for this site. Interestingly, no binding of C/EBP␦ to this site was detected in C/EBP␤Ϫ/Ϫ macrophages despite appreciable levels of protein being present, supporting the idea that C/EBP␦ may be unable to bind as a homodimer to the COX-2 C/EBP site (29). This might also explain why C/EBP␦ cannot compensate for the absence of C/EBP␤ in the mutant cells. Since C/EBP␦ induction is abolished by the same treatments that inhibit CREB activation, we could not establish whether C/EBP␦, in conjunction with C/EBP␤, would be sufficient to bypass the need for CREB and/or NF-B activation. However, recent data suggest that this might be the case, since co-transfection of C/EBP␦, but not of C/EBP␤, could activate transcription of a COX-2 reporter in the absence of LPS treatment (16).