Tumor Necrosis Factor-α Inversely Regulates Prostaglandin D2 and Prostaglandin E2 Production in Murine Macrophages

Increased synthesis of insulin-like growth factor-1 is induced in murine macrophages by prostaglandin E2 (PGE2) and tumor necrosis factor-α (TNFα). Accordingly, we have investigated mechanisms regulating synthesis of PGE2 that might contribute to autocrine/paracrine effects on insulin-like growth factor-1 production. In response to zymosan, TNFα specifically induced a 5-fold increase in PGE2 synthesis, at the same time decreasing PGD2 production in a reciprocal fashion. Activators of cyclic AMP-dependent protein kinase (PKA), such as PGE2 itself or dibutyryl cyclic AMP, did not modify PGE2 production by themselves but potentiated the TNFα-induced increase in PGE2; this effect required both RNA and protein synthesis. No significant change in arachidonate release or production of other eicosanoids was observed. The inducible form of cyclooxygenase-2 (COX2) but not of the constitutive form COX1 was implicated in the generation of both PGE2 and PGD2 in these cells by use of specific inhibitors and effects of dexamethasone. Neither COX1 nor COX2 protein levels were affected by TNFα or PKA activators used alone, whereas in association, marked up-regulation of COX2 mRNA and protein was observed. Incubations of cells carried out with PGH2demonstrated that PGE2 synthase activity was increased after a TNFα pretreatment. Taken together, our results suggest that TNFα induced a switch from the PGD2 to PGE2synthesis pathway by regulating PGE2 synthase expression and/or activity and that activators of PKA markedly potentiated the TNFα-induced increase in PGE2 through up-regulation of COX2 gene expression.

Increased synthesis of insulin-like growth factor-1 is induced in murine macrophages by prostaglandin E 2 (PGE 2 ) and tumor necrosis factor-␣ (TNF␣). Accordingly, we have investigated mechanisms regulating synthesis of PGE 2 that might contribute to autocrine/paracrine effects on insulin-like growth factor-1 production. In response to zymosan, TNF␣ specifically induced a 5-fold increase in PGE 2 synthesis, at the same time decreasing PGD 2 production in a reciprocal fashion. Activators of cyclic AMP-dependent protein kinase (PKA), such as PGE 2 itself or dibutyryl cyclic AMP, did not modify PGE 2 production by themselves but potentiated the TNF␣-induced increase in PGE 2 ; this effect required both RNA and protein synthesis. No significant change in arachidonate release or production of other eicosanoids was observed. The inducible form of cyclooxygenase-2 (COX2) but not of the constitutive form COX1 was implicated in the generation of both PGE 2 and PGD 2 in these cells by use of specific inhibitors and effects of dexamethasone. Neither COX1 nor COX2 protein levels were affected by TNF␣ or PKA activators used alone, whereas in association, marked up-regulation of COX2 mRNA and protein was observed. Incubations of cells carried out with PGH 2 demonstrated that PGE 2 synthase activity was increased after a TNF␣ pretreatment. Taken together, our results suggest that TNF␣ induced a switch from the PGD 2 to PGE 2 synthesis pathway by regulating PGE 2 synthase expression and/or activity and that activators of PKA markedly potentiated the TNF␣-induced increase in PGE 2 through up-regulation of COX2 gene expression.
Macrophages are known to generate prostaglandins (PGs) 1 in response to various stimuli such as endotoxin (1,2), phorbol myristate acetate (3), or phagocytic particles (4). In addition to playing important roles in such biologic processes as cell proliferation (5), inflammatory and immune responses (6 -8), and the production of extracellular matrix proteins (9), prostaglandins may also act in an autocrine/paracrine fashion to modulate the responses of the macrophages themselves (10 -12). We have previously shown that in murine macrophages, prostaglandin E 2 (PGE 2 ) increased the synthesis of insulin-like growth factor-1 (IGF-1), a growth factor for fibroblasts, by a TNF␣-independent signaling pathway. The effect of TNF␣ on this PGE 2induced process was additive (13). Since macrophages are themselves a potent source of PGE 2 , autocrine up-regulation of IGF-1 was a distinct possibility. Furthermore, TNF␣ might itself alter the IGF-1 response by, in part, enhancing production of PGE 2 . Accordingly, in this study, we addressed the effects of TNF␣ and PGE 2 on PGE 2 production in murine bone marrow-derived macrophages. Since PGE 2 acts by receptormediated generation of cyclic AMP and activation of protein kinase A (PKA), the effect of direct addition of dibutyryl cyclic AMP (dbcAMP) was also examined.
The first enzymatic step in eicosanoid synthesis is the release of free arachidonic acid (AA) from membrane phospholipid. Depending on the cell type and stimulus, different forms of phospholipase A 2 (PLA 2 ), including an 85-kDa cytosolic PLA 2 (cPLA 2 ) (14, 15), 14-kDa secreted group II PLA 2 (sPLA 2 ) (16 -18), or a calcium-independent cytosolic PLA 2 (19,20), have each been described to play a role in mediating AA release and subsequent prostaglandin synthesis. Agents that stimulate AA release in macrophages, including zymosan, phorbol 12-myristate 13-acetate, calcium ionophore A23187, and okadaic acid, can activate cPLA 2 by enhancing serine phosphorylation of the enzyme (21). More recently, a role for mitogen-activated protein kinase has been described in mediation of agonist-induced activation of cPLA 2 (22,23). Cytokine-induced changes in gene expression of sPLA 2 have been reported in rat mesangial cells (24) and human synovial cells (25), but the activity of these enzymes is presumably also regulated by the extent and location of their secretion.
Prostaglandin endoperoxide-synthase or cyclooxygenase (COX) catalyzes the conversion of AA to PGH 2 , which is then metabolized by one or more terminal synthases to a variety of biologically active prostanoids (26). Cyclooxygenase is a key enzyme in prostanoid synthesis and possesses both fatty acid cyclooxygenase activity (producing PGG 2 from AA) and PG hydroperoxidase activity (converting PGG 2 to PGH 2 ). COX2 is a recently described form of cyclooxygenase that is induced in a number of cells by proinflammatory stimuli, which contrasts to the lack of induction seen with the previously characterized constitutive enzyme, COX1 (27). Thus, COX2 is thought to contribute to the generation of prostanoids at sites of inflammation (28), (see Ref. 29 for review).
In the current study, the ability of murine macrophages to synthesize PGE 2 in response to a phagocytic particle (zymosan) was selectively induced by TNF␣. In contrast, a concomitant decrease in PGD 2 production was observed after TNF␣ priming, suggesting a striking switch from production of PGD 2 to PGE 2 and regulation of PG synthase activity. In addition, we show herein that activators of PKA markedly potentiated the TNF␣-induced increase in PGE 2 through up-regulation of COX2 gene expression. Our results suggest that (i) PGE 2 might act as an autocrine mediator to stimulate and to maintain the differentiation of uncommitted macrophages into IGF-1 producing cells and (ii) this differentiated cell type may participate in an exocrine fashion to regulate the inflammatory response by its ability to synthesize and release PGE 2 .

EXPERIMENTAL PROCEDURES
Animals-C3H/HeJ mice were bred in the Biological Resource Center at the National Jewish Center. The C3H/HeJ strain is lipopolysaccharide hyporesponsive and was chosen to minimize the effects of trace lipopolysaccharide contamination (30). Approximately 8-week-old mice were killed by CO 2 narcosis.
Materials-Dulbecco's modified Eagle's medium was obtained from Bio-Whittaker (Walkersville, MD). Fetal bovine serum (Hybri-Max) as well as all chemical compounds were purchased from Sigma except for valeryl salicylate and NS-398, which were purchased from Cayman Chemical (Ann Arbor, MI). Nonidet P-40 was purchased from Boehringer Mannheim, and nitrocellulose membranes were purchased from Bio-Rad. Recombinant murine TNF␣ was generously provided by Genentech, Inc. Zileuton (A64077) was obtained from Abbott Laboratories (North Chicago, IL). Zymosan was purchased from Sigma and opsonized using a human serum pool as described (3). The antibodies used included generous gifts of antibody to PGE 2 and PGD 2 (Dr. J. Maclouf, INSERM, Paris), 6-keto-PGF 1␣ (Dr. K. Allen, Colorado State University, Fort Collins, CO) and TXB 2 (Dr. Frank Fitzpatrick, University of Colorado Health Science Center, Denver). Antibodies for LTB 4 and LTC 4 were purchased from Advanced Magnetics (Boston, MA). RNAzol B was purchased from Tel-Test, Inc. (Friendswood, TX), and BCA protein assay was purchased from Pierce. Murine COX1 cDNA probe was purchased from Cayman Chemicals (Ann Arbor, MI), and murine COX2 mRNA analysis was performed using a 1.8-kilobase pair cDNA insert kindly provided by Dr Voelkel (UCHSC, Denver). Human glyceraldehyde 3-phosphodehydrogenase (GAPDH) cDNA was provided by J. Shannon (National Jewish Center, Denver). Both monoclonal antibodies were kindly provided by Dr. J. Maclouf (INSERM, Paris).
Isolation and Culture of Bone Marrow-derived Macrophages-Murine bone marrow-derived macrophages were obtained using a technique previously described in detail (31). Dulbecco's modified Eagle's medium containing 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin, 0.37% (w/v) NaHCO 3 , 10% (v/v) heat-inactivated fetal bovine serum, and 10% (v/v) L929 cell-conditioned medium as a source of colony-stimulating factor-1 was used for the isolation, culture, and stimulation of murine bone marrow-derived macrophages. Bone marrow cells were flushed asceptically from the dissected pelvises, femurs, and tibias of C3H/HeJ mice with a jet of complete medium directed through a 25-gauge needle to form a single cell suspension. The bone marrow cells were dispensed at a concentration of 0.45 ϫ 10 6 cells/well in 24-well tissue culture plates for the eicosanoid study, 11 ϫ 10 6 cells/100-mm diameter culture dishes for protein immunoblotting, and at a concentration of 1.8 ϫ 10 6 cells/well in 6-well tissue culture plates for RNA analysis. The cells were matured at 37°C under a 10% (v/v) CO 2 atmosphere for 5 days before use.
Quantification of Eicosanoids-After incubation of macrophages for the indicated period of time with various stimuli, the cell monolayers were washed twice in Ca 2ϩ , Mg 2ϩ -free phosphate-buffered saline, pH 7.4; then, cells were incubated for 1 h with the phagocytic stimulus (50 particles of opsonized zymosan per cell) or with 10 M exogenous AA in 2% BSA-Dulbecco's modified Eagle's medium. The culture medium was centrifuged to remove nonadherent cells and assayed for AA metabolites using enzyme immunoassay methods previously described (32,33). LTB 4 , LTC 4 , TXB 2 , and 6-keto-PGF 1␣ (the stable metabolite of PGI 2 ) were assayed directly from fresh supernatants while PGD 2 and PGE 2 were measured as their methyl oximes after derivatization with methoxamine as described by Kelly et al. (34). Macrophage monolayers were scraped in 500 l of 0.5 N sodium hydroxide, and total protein was determined using a BCA protein assay. Eicosanoid levels were all expressed as picograms per microgram of macrophage protein. Total content of free AA in both cell lysates and supernatants was measured using a quantitative mass spectrometric assay (35). Briefly, macrophage monolayers maintained in culture medium (0.45 ϫ 10 5 cells/ml) were lysed by adding 1 ml of MeOH directly in the well, and lysates (2 ml) were acidified to pH 3 with 100 l of 1 N HCl. Free AA was extracted twice with one volume of hexane and dried under vacuum. The pentafluorobenzyl ester was synthesized by adding a 10% solution of pentafluorobenzyl bromide in acetonitrile (50 l) and an equal volume of a 10% solution of diisopropylamine in acetonitrile and then allowed to stand at room temperature for 10 min. After vacuum evaporation of reagents, the derivative was dissolved in 100 l of hexane for electron capture ionization mass spectrometric analysis (36). Free arachidonic acid levels were expressed as nanograms per well using a calibration curve. PGE 2 and PGD 2 Synthase Activities-Mouse macrophages were isolated, cultured, and stimulated as described above. Prostaglandin synthase activity was performed according to Kelner and Uglik (37). After a 16-h incubation period, cells (2 ϫ 10 6 ) were collected by scraping and resuspended in 100 l of cold buffer (50 mM TRIS, pH 7.5, 0.1 M sodium chloride, 0.1 M EDTA). The cells were lysed by freezing/thawing them twice in a acetone-dry ice bath and centrifuged at 14,000 rpm in a microcentrifuge. Supernatants were warmed to 37°C in a water bath for 1 min, and reduced glutathione (6 g in 10 l) was added for 1 min. Then, 150 ng of PGH 2 was added for 1 min. The reaction was terminated by adding 10 l of a 25 mM FeCl 2 solution. Spontaneous conversion was evaluated by incubating PGH 2 in buffer alone with no cell. The concentrations of PGE 2 and PGD 2 at 0 and 1 min were determined by enzyme immunoassay.
Quantification of mRNA Transcripts-The expression of COX1, COX2, and GAPDH mRNA transcripts was determined by Northern analysis. Total cellular RNA was extracted from macrophage monolayers with 1 ml of RNAzol B for 1.8 ϫ 10 6 cells (38). 10 g of total RNA was electrophoresed under denaturing conditions through a 1.2% agaroseformaldehyde gel and then transferred to NYTRAN hybridization filters as described (39). Total RNA was covalently linked to the membrane by UV cross-linking (0.12 joules) using a Stratalinker 1800 (Stratagene). The blots were hybridized with 10 6 dpm/ml of 32 P-labeled cDNA probes for 18 h and washed to a final stringency of 0.2 ϫ SSC at 42°C, and autoradiograms were prepared by exposure to Kodak X-OMAT AR film at Ϫ70°C. To ensure that the differences in transcript expression were specific to COX1 and COX2 mRNA and not due to differential loading of RNA, the blots were stripped in 2% glycerol for 3 min at 100°C and reprobed with GAPDH. The density of the autoradiographic signals was quantified by scanning densitometry combined with integration using the "Image 1.35" software run on a Macintosh microcomputer. The results were expressed as a ratio of COX to GAPDH expression as indicated.
Data Analysis-Results are presented as the mean Ϯ S.E. for at least three separate experiments. Comparisons between groups were made using the student's paired t test.

Effect of TNF␣, PGE 2 , and dbcAMP on PGE 2 Synthesis and
Release-Murine bone marrow-derived macrophages were cultured for 5 days in the presence of colony-stimulating factor-1, followed by priming with TNF␣, PGE 2 , or dbcAMP alone or in combination for a period of 16 h. After the culture medium was removed and cell monolayers washed, macrophages were stimulated for 1 h with 50 particles per cell of opsonized zymosan in fresh culture medium to promote production and release of arachidonic acid metabolites. Previous analysis had shown this amount and time of exposure to zymosan to be optimal (data not shown). As shown in Fig. 1, 10 ng/ml TNF␣ led to a 5-fold increase in PGE 2 production, whereas no change was observed in response to either 0.1 mM dbcAMP or 1 M PGE 2 treatment alone. However, when cells were incubated with the combination TNF␣ϩdbcAMP or TNF␣ϩPGE 2 , synergistic responses were observed corresponding to a 4-fold increase and a 2-fold increase compared with TNF␣ alone for dbcAMP and PGE 2 , respectively.
Release of Free Arachidonic Acid-To determine the point at which the TNF␣ and PGE 2 were exerting their effect, the release of free AA was analyzed to explore the possibility that these effectors may affect the balance of deacylation and reacylation. To block metabolism of free AA released by COX and 5-lipoxygenase (5-LO), cells were incubated in the presence of 5 M indomethacin to block the COX activity (40) and 5 M zileuton to inhibit 5-lipoxygenase (41) for the 1-h stimulation period with zymosan. In Fig. 2, intracellular plus extracellular AA content is shown, as determined by mass spectrometry and expressed as nanogram/0.45 ϫ 10 6 cells. Independently of the presence of the inhibitors, no significant variations in AA release were observed in response to any of the stimuli. Differences observed in AA release between the presence and absence of the COX and 5-lipoxygenase inhibitors are likely due to utilization of AA as substrate. Similar results were obtained when AA release was analyzed after the cells had been labeled with [ 3 H]AA (data not shown). These results suggest that the effect of TNF␣ on PGE 2 levels with or without dbcAMP is not due to enhanced phospholipase activity.
Effect of TNF␣, PGE 2 , and dbcAMP on Synthesis and Release of Other Eicosanoids-To determine whether the observed induced increase in PGE 2 production was exclusive to PGE 2 , the effect on other AA metabolites known to be produced by macrophages was examined. The production of two cyclooxygenase metabolites, TXB 2 and 6-keto-PGF 1␣ (the stable PGI 2 metabolite), as well as two 5-lipoxygenase products (LTC 4 and LTB 4 ), were quantified in the supernatant of cells primed for 18 h with the various stimuli and then stimulated with zymosan. Table I shows that untreated cells released high levels of TXB 2 (12.95 pg/g), whereas release of the other eicosanoids was low. In contrast to PGE 2 (Fig. 1), TNF␣ alone or in combination with dbcAMP did not significantly increase the production of any of these four AA metabolites studied. Interestingly, a slight but significant increase in TXB 2 and LTB 4 release was observed in dbcAMP-treated cells.
TNF␣ Inversely Regulates PGD 2 and PGE 2 Production-In contrast to the effect of the cytokine on thromboxane, PGI 2 , or leukotrienes, TNF␣ had a dramatic inhibitory effect on the production of PGD 2 . Fig. 3A depicts the dose-dependent variations in PGE 2 and PGD 2 synthesis and release in response to increasing concentration of TNF␣. Whereas unprimed macrophages, when stimulated with zymosan, synthesize and release high amounts of PGD 2 (22.89 pg Ϯ 4.99 pg of PGD 2 /g of total protein; n ϭ 6), the release of PGE 2 by these same cells remained at a low level (1.75 pg Ϯ 0.34 pg/g). However, when cells were incubated for a 16-h period with increasing concentration of TNF␣ (0.01 to 100 ng/ml), a simultaneous decrease in PGD 2 release and increase in PGE 2 production was observed that dropped to 7.5 Ϯ 0.88 pg/g and increased to 10.33 Ϯ 0.3 pg/g, respectively (at 10 ng/ml TNF␣). Interestingly, the overall amount of released prostaglandins (PGE 2 ϩPGD 2 ) remained unchanged (15.2 pg/g for no TNF␣ versus 15.5 with 10 ng/ml TNF␣ stimulation; Fig. 3A). The same reciprocal pattern of prostaglandin production was seen when cells were incubated with 10 ng/ml TNF␣ for 16 h and then stimulated for 1 h with 10 M exogenous AA instead of zymosan (Fig. 3B), suggesting that the role of zymosan is to stimulate PLA 2 , thus generating the AA, and that the effects of TNF␣ are downstream of this step. All together, the results shown in Fig. 3 suggest that TNF␣ might act by switching the synthesis of prostaglandins from the PGD 2 to the PGE 2 pathway.
Requirement of Protein and RNA Synthesis in PGE 2 and PGD 2 Production-To further investigate the molecular basis involved in the regulation of PGE 2 and PGD 2 synthesis in murine macrophages, cells were treated with 10 ng/ml TNF␣ alone or in combination with 0.1 mM dbcAMP in the presence or absence of 5 g/ml cycloheximide or 0.5 g/ml actinomycin D. As illustrated in Fig. 4, the observed increase in PGE 2 production after stimulation of cells with TNF␣ alone or in combination with dbcAMP was completely abolished by either cycloheximide or actinomycin D, suggesting an absolute requirement for both RNA and protein synthesis. In contrast, neither cycloheximide nor actinomycin D prevented the inhibition of PGD 2 production by TNF␣. Interestingly, only cycloheximide but not actinomycin D inhibited basal PGD 2 levels. These results suggest that, unlike the TNF␣ϩdbcAMP-mediated induction of PGE 2 synthesis, basal production of PGD 2 does not require new RNA synthesis. The observed decrease in PGD 2 synthesis following cycloheximide treatment might reflect the turnover of constitutive enzyme(s) involved in the basal production of this prostaglandin.
TNF␣ and dbcAMP or PGE 2 Synergistically Up-regulate COX2 Gene Expression and Protein Synthesis-Since protein and RNA synthesis were involved in the enhanced production of PGE 2 by the combination of TNF␣ϩdbcAMP and because basal production of PGD 2 was dependent of protein synthesis, we next examined steady state mRNA levels of the constitutive (COX1) and the inducible form (COX2) of cyclooxygenase. Messenger RNA levels were analyzed by Northern blot after a 12-h stimulation period, and hybridization with a GAPDH cDNA probe was performed to ensure for equal loading of total RNA. As shown in Fig. 5, COX1 mRNA was constitutively expressed in control macrophages while COX2 mRNA level remained undetectable. Neither 1 M PGE 2 nor 0.1 mM dbcAMP affected the expression of either COX1 or COX2, whereas 10 ng/ml TNF␣ inhibited the basal expression of COX1 mRNA and seemed to slightly increase COX2 mRNA levels. When cells were incubated with the combination TNF␣ϩPGE 2 or dbcAMP, elevation in COX2 mRNA levels was observed. The TNF␣induced inhibition of COX1 mRNA expression was not significantly altered by the presence of PGE 2 or dbcAMP. As seen in Fig. 6, Western blot analysis of macrophage whole cell lysates  4 , and LTB 4 production in murine macrophages Release of arachidonic acid metabolites was quantified in the supernatant as described in Fig. 1 and under "Experimental Procedures." Results are expressed as picograms per microgram protein, and values represent the mean Ϯ S.E. of 5 (TXB 2 ), 4 (LTB 4 ), and 3 (LTC 4 ) different cultures or represent the mean Ϯ variation of two independent experiments (6-keto-PGF 1␣ ). *, p Ͻ 0.05 compared with values of untreated.  using specific monoclonal antibodies for COX1 and COX2 showed that the COX2 protein levels paralleled those of the transcript. The low content of COX2 protein in untreated cells was not affected by any of the stimuli used alone but was enhanced when cells were incubated with the combination of TNF␣ϩdbcAMP. In contrast, the basal expression of COX1 protein remained unchanged in response to any of the treatments. Although the TNF␣-induced down-regulation of COX1 mRNA levels was observed at 12 h, no significant change in the corresponding protein was seen after 18 h. This latter observation suggests that the constitutive form of COX may be relatively stable so that protein levels were not yet affected at 18 h despite the absence of transcript at 12 h.
Involvement of COX2 Activity in Both PGE 2 and PGD 2 Productions-To further characterize which COX isozyme was involved in the up-regulated synthesis of PGE 2 , inhibitors of COX1 and/or COX2 activities were used. As shown in Fig. 7A, the COX inhibitor indomethacin totally prevented the TNF␣ϩdbcAMP-induced increase in PGE 2 synthesis. The COX2 inhibitor NS-398 (40) duplicated the indomethacin results. However, a concentration 10 times higher (10 M versus 1 M) was required. In contrast, neither sulindac sulfide (42) nor valeryl salicylic acid (43), which are 10 times more potent inhibitors of COX1 activity, decreased PGE 2 levels even when used at 10 M. Similar patterns of inhibition were obtained when cells were primed with TNF␣ alone (data not shown). To determine whether the basal levels of PGD 2 were produced as a consequence of COX1 or COX2 activity, PGD 2 release was analyzed in untreated macrophages incubated with increasing concentration of the inhibitors used above. Results shown in Fig. 7B suggest that COX2 and not COX1 is involved in PGD 2 synthesis despite the low basal amounts of both COX2 mRNA and protein.
Glucocorticoids have been reported to prevent the induction of COX2 (44,45). To further examine the possibility that only COX2 was involved in production of PGD 2 (before priming) as well as the induced synthesis of PGE 2 , the effect of steroids on COX1 and COX2 mRNA levels as well as PGE 2 and PGD 2 production was examined. At 12 h, total RNA was extracted, and the same membrane was consecutively hybridized with COX2, COX1, and GAPDH cDNA probes (Fig. 8). After scanning of the autoradiographs, COX1 and COX2 values were  5. Northern blot analysis of COX1, COX2, and GAPDH mRNA from control and treated macrophages. Total cellular RNA was extracted at 12 h, and 15 g of RNA hybridized with a 32 P-labeled COX2 cDNA probe as described under "Experimental Procedures." After stripping, the same blot was reprobed with 32 P-labeled COX1 cDNA and finally with GAPDH cDNA to demonstrate equal loading of mRNA. The Northern blot shown here is representative of three independent experiments.
FIG. 6. Anti-COX2 and anti-COX1 immunoblots of macrophage lysates. Macrophage whole cell lysates after a 16-h stimulation period were probed with monoclonal antibodies for COX1 and COX2. Comparison to a protein ladder and to the migration of pure protein (standard) identify immunoreactive COX1 and COX2. As described by others (58), murine COX2 migrates as two bands, the lower band resulting from inefficient N-glycosylation that has no effects on enzyme activity. This Western blot is representative of three separate experiments.
normalized to the density of the GAPDH band. Dexamethasone prevented the TNF␣ϩdbcAMP-induced increase in COX2/ GAPDH mRNA level by 70%, whereas slight increase in steady state levels of COX1/GAPDH mRNA was observed in control.
To analyze the effect of glucocorticoids on PGE 2 and PGD 2 production, cells were primed with stimuli for 16 h as before, but because glucocorticoids inhibit PLA 2 activity (46), cells were incubated in the presence of 10 M exogenous arachidonic acid rather than zymosan to bypass the requirement for PLA 2 . As depicted in Fig. 9, the presence of 1 M dexamethasone inhibited both constitutive production of PGD 2 (Fig. 9B) and induced increase in PGE 2 synthesis (Fig. 9A), suggesting that expression of COX2 is necessary for both the synthesis and release of these two prostaglandins.
TNF␣ Increases PGE 2 Synthase Activity-To determine whether PGE 2 and/or PGD 2 synthase activities were modified following TNF␣ treatment, conversion of exogenously supplied PGH 2 to both PGE 2 and PGD 2 was examined. As shown in Fig.  10, a 16-h preincubation period with 10 ng/ml TNF␣ induces a 5.2-fold increase in PGE 2 synthase activity after incubation of cell lysate for 1 min with 150 ng of exogenous PGH 2 . This increase in PGE 2 synthesis is similar to the one observed when TNF␣-primed cells were incubated with zymozan ( Fig. 1). When PGD 2 synthase activity was examined, no difference between unstimulated and TNF␣-treated macrophages was observed, suggesting that TNF␣ increases PGE 2 synthase activity without affecting PGD 2 synthase activity. We next examined the effect of RNA and protein synthesis inhibitors on the activity of the PGE 2 synthase (Fig. 11). TNF␣-induced increase in PGE 2 synthase activity was abolished using either the protein synthesis inhibitor cycloheximide (CHX) or the RNA synthesis inhibitors actinomycin D (AD) or dichlororibofuranosylbenzimidazole. These results suggest that up-regulation of PGE 2 synthase activity by TNF␣ is dependent of both protein and RNA synthesis.

DISCUSSION
As shown herein, stimulation of murine macrophages by TNF␣ induced a switch in prostaglandin synthesis from PGD 2 to the PGE 2 without affecting production of other eicosanoids. We also demonstrated that PGE 2 itself as well as the membrane-permeable PKA activator dbcAMP markedly and selectively potentiated the TNF␣-induced increase in PGE 2 synthesis. This enhancement could not be explained by increasing activity of PLA 2 because no change in AA release was observed. In contrast, we suggest that the enhancing effect of cAMP (PGE 2 ) is mediated through up-regulation of COX2 gene expression. PGE 2 inhibits macrophage cytocidal activity and can act as an autocrine-negative feedback stimulus to limit this function (11,12). In the other direction, we have previously shown that in murine macrophages, PGE 2 increased the synthesis of IGF-1 and thus might contribute an autocrine signal to enhance the role of macrophages in wound healing and fibrosis (13). Herein, we show that (i) TNF␣ selectively induced the production of PGE 2 by increasing PGE 2 synthase activity and (ii) PGE 2 itself up-regulated its own synthesis by TNF␣primed macrophages. These observations help refine the complex, and mutually exclusive, regulating pathways for synthesis of IGF-1 and for proteins involved in cytocidal macrophage function, respectively (47,48).
The bone marrow-derived macrophages generated PGD 2 as their predominant prostanoid when stimulated to release arachidonate by uptake of zymosan (or upon addition of free arachidonate). In contrast, when incubated with TNF␣, the cells switched their production to PGE 2 without any significant changes in the release of other eicosanoids from either the 5-lipoxygenase or the cyclooxygenase pathways. The total level of PGs released (PGE 2 plus PGD 2 ) was similar in untreated and in TNF␣-treated cells. Our results suggest that TNF␣ induced a switch in PGD 2 and PGE 2 synthesis from their common precursor PGH 2 , i.e. that the cytokine was acting at the level of PGD 2 and/or PGE 2 synthases rather than at the level of active PLA 2 or cyclooxygenase. Such a switch has been reported by others in rat Kupffer cells where in vitro activation of protein kinase C resulted in reduced PGE 2 and enhanced PGD 2 synthase activity (49).
In our study, TNF␣ did not change the amounts of arachidonate released by the uptake of zymosan. Furthermore, the switch was also evident when arachidonate was added directly without any phagocytic stimulus, i.e. in the absence of obvious PLA 2 activation. Our results also indicate that the TNF␣induced switch to PGE 2 was not mediated through alteration of COX isotype usage. TNF␣ did not modify either COX1 or COX2 protein levels at 18 h (the time of zymosan addition and pros-taglandin measurement) despite a slight increase in COX2 and a decrease in COX1 mRNA steady state levels observed 12 h after TNF␣ treatment. In addition, use of inhibitors of COX1 and COX2 suggested that COX2 but not COX1 was involved in both the production of PGD 2 before TNF␣ treatment as well as the PGE 2 formation afterward.
To further investigate the possibility that TNF␣ might act directly by either increasing PGE 2 synthase activity or by decreasing PGD 2 synthase activity, we incubated cell lysates from control or TNF␣-treated cultures with PGH 2 as a substrate before the measurement of prostaglandins. Our results provided evidence that TNF␣ priming leads to preferential enzymatic conversion of PGH 2 into PGE 2 while PGD 2 generation remains unchanged (Fig. 10). Thus, the observed switch from PGD 2 to PGE 2 does not seem to be mediated by depletion of glutathione following TNF␣-induced oxidant stress since both PGE 2 and PGD 2 synthases are glutathione-dependent enzymes (50,51). We next examined whether the observed TNF␣-induced increase in PGE 2 synthase activity was dependent on protein or RNA synthesis. Concomitant incubation of macrophages with TNF␣ and either cycloheximide or actinomycin D demonstrated that both inhibitors abolished the TNF␣-induced effect (Fig. 11). The observation that cycloheximide or actinomycin D prevented the TNF␣-induced increase in PGE 2 synthesis suggests that TNF␣ acts through up-regulation of PGE 2 synthase itself or of a regulatory protein (Fig. 4). In addition, TXB 2 levels did not change, suggesting that TNF␣ acts selectively by inducing PGE 2 activity.
Taken together, our results demonstrate that PGE 2 synthase activity is increased in TNF␣-primed macrophages compared with unstimulated cells. The unchanged PGD 2 synthase activity suggests that TNF␣ does not have any effect at this level. The TNF␣-induced switch from PGD 2 to PGE 2 observed in the absence of exogenously added PGH 2 may be the result of preferential use of the endogenous substrate PGH 2 by the PGE 2 synthase pathway.
In previous studies, we showed that in murine macrophages, PKA activators such as PGE 2 and dbcAMP stimulated the synthesis of IGF-1 by a process that was independent of, and additive with, that induced by TNF␣ (13). As described herein, PKA activators had only a slight enhancing effect by themselves on eicosanoid synthesis in response to zymosan uptake. On the other hand, they did synergistically potentiate the TNF␣-induced increase in PGE 2 production. Synergism between cAMP elevating agents and TNF␣ or interleukin-1␤ have also been reported in up-regulation of group II PLA 2 mRNA in rat mesangial cells (52)(53)(54). Agents that increase cellular levels of cAMP used alone have additionally been shown to increase secretion of PLA 2 from vascular smooth muscle cells (55) and in mesangial cells (54). However, when we examined free AA release by mass spectrometry, no significant change was observed in these macrophages after treatment with PKA activators used alone or in association with TNF␣.
However, the combination of TNF␣ and dbcAMP together did markedly up-regulate COX2 protein and mRNA. Neither stimulus was effective by itself, although, as noted above, TNF␣ decreased the steady state levels of mRNA for COX1. Incubation of cells with stimulus and cycloheximide or actinomycin D demonstrated that both protein and mRNA synthesis were necessary for the increase in PGE 2 production induced by TNF␣ plus dbcAMP. Glucocorticoids such as dexamethasone inhibit the induction of COX2 in cultured cells with little effect on COX1 mRNA and protein levels (44,45). In the current studies, dexamethasone inhibited TNF␣ plus dbcAMP-induced COX2 mRNA levels as well as PGE 2 generation. The implication that COX2 was required for the PGE 2 synthesis was con-FIG. 10. PGE 2 and PGD 2 synthase activities. Macrophages (2 million) were incubated with or without 10 ng/ml TNF␣ for 16 h. Exogenous PGH 2 (150 ng) was added to cell lysates, and the tubes were incubated at 37°C for 1 min in the presence of reduced glutathione. The reaction was stopped with FeCl 2 , and PGE 2 and PGD 2 levels were quantified by enzyme immunoassay. Spontaneous conversion of PGH 2 into PGE 2 and PGD 2 obtained in the absence of cells was subtracted from each value. Results are expressed as ng/ml and represent the mean Ϯ S.E. of five different cultures. *, p Ͻ 0.05: TNF versus control at 1 min.
FIG. 11. TNF␣-induced increase in PGE 2 synthase activity is dependent of protein and RNA synthesis. Macrophages were incubated with or without 10 ng/ml TNF␣ for 16 h in the presence or absence of 5 g/ml CHX, 0.5 g/ml AD or 50 g/ml dichlororibofuranosylbenzimidazole (DRB). PGE 2 synthase activity was assayed as described in Fig. 10. Results expressed as ng/ml represent the mean Ϯ variation of two separate experiments. firmed by use of COX2-selective inhibitors. Importantly, inhibitors of COX2 prevented PGE 2 production after TNF␣ stimulation whether or not the process was further enhanced by dbcAMP. This suggested that COX2 was the critical enzyme involved even without its up-regulation by the combined stimuli. In fact, even the PGD 2 produced after zymosan stimulation of macrophages that had not been exposed to either TNF␣ or PKA activators was prevented by COX2 inhibitors and not by agents more selective toward COX1. All together, our results suggest that TNF␣ and dbcAMP synergistically up-regulate COX2 gene expression and, as a consequence, provide more substrate (PGH 2 ) for PGE 2 production in TNF␣primed macrophages. The synergistic effect of the combination TNF␣ϩdbcAMP on COX2 expression might be explained by the up-regulation of TNF␣ receptors by agents that increase cAMP levels as it has been described in vitro in human histiocytic lymphoma cell line U-937 (56). Interestingly, we showed herein that TNF␣ alone leads to a slight increase in COX2 mRNA steady state levels (Fig. 5).
The observed switch from PGD 2 to PGE 2 synthesis after activation of murine macrophages is biologically relevant because, in these cells, PGD 2 did not increase PKA activation or intracellular levels of cAMP (data not shown), although others have reported that it can induce adenylate cyclase activity and PGE 2 production in endothelial cells (57). In contrast and as expected, PGE 2 was a potent activator of PKA and cAMP generation (13). Thus, PGE 2 might be considered to provide a positive feedback stimulus in TNF␣-primed macrophages for its own synthesis. Since TNF␣ by itself induced the switch to PGE 2 production by increasing PGE 2 synthase activity, the PGE 2 so produced would be expected to further enhance its production via up-regulation of COX2 with subsequent positive (e.g. IGF-1 synthesis) or negative (e.g. cytocidal activity) autocrine/paracrine effects on the macrophages.