Nitric oxide increases tumor necrosis factor production in differentiated U937 cells by decreasing cyclic AMP.

Nitric oxide (NO) increases tumor necrosis factor (TNF) synthesis in human peripheral blood mononuclear cells by a cGMP-independent mechanism. NO has been shown to inhibit adenylate cyclase in cell membranes. Since cAMP down-regulates TNF transcription, we examined the possibility that NO enhances TNF synthesis by decreasing cAMP. U937 cells were induced to differentiate using phorbol myristate acetate (100 nM for 48 h) and then were incubated for 24 h with sodium nitroprusside (SNP) or S-nitroso-N-acetylpenicillamine (SNAP). These NO donors increased TNF production (7.0- and 15.6-fold, respectively, at 500 μM) in a dose-dependent manner (p = 0.002). However, SNP and SNAP did not elevate cGMP levels in U937 cell cultures, and the cGMP analog, 8-bromo-cGMP, had no effect on TNF production. In contrast, SNP (p = 0.001) and SNAP (p = 0.009) decreased intracellular cAMP levels by up to 51.5% over 24 h and, in the presence of a phosphodiesterase inhibitor, blunted isoproterenol-stimulated increases in cAMP by 21.8% (p = 0.004) and 27.6% (p = 0.008), respectively. H89, an inhibitor of cAMP-dependent protein kinase, dose dependently increased TNF production in phorbol myristate acetate-differentiated U937 cells in the absence (6.5-fold at 30 μM; p = 0.035), but not in the presence (p = 0.77) of SNAP. Conversely, the cAMP analog dibutyryl cAMP (Bt2cAMP) blocked SNAP-induced TNF production (p = 0.001). SNP and SNAP (500 μM) increased relative TNF mRNA levels by 57.5% (p = 0.045) and 66.2% (p = 0.001), respectively. This effect was prevented by Bt2cAMP. These results indicate that NO up-regulates TNF production by decreasing intracellular cAMP.

First, NO is a free radical with the ability to react with a variety of enzymes besides soluble guanylate cyclase. NO has been shown to catalyze the covalent binding of NAD to glyceraldehyde-3-phosphate dehydrogenase (20), oxidize iron-containing proteins such as aconitase or ribonucleotide reductase (21)(22)(23), and nitrosylate tyrosine and cysteine residues in a variety of proteins (24 -26). Second, some effects of NO cannot be reproduced with cell permeable cGMP analogs. For example, the synthesis of tumor necrosis factor ␣ (TNF␣), a proinflammatory cytokine implicated in tissue injury and shock (27), is increased in human peripheral blood mononuclear cells (28) and lipopolysaccharide-stimulated neutrophil preparations (29) by exogenous NO. Although NO increases cGMP concentrations in these cells, cGMP analogs have no effect on TNF␣ production (28,29). Collectively, these investigations suggest that NO might use cGMP-independent signaling pathways for some of its cellular functions.
Recently, adenylate cyclase has been added to the list of enzymes that can be modified by NO (30). Treatment of cell membranes with NO decreases cAMP production by inhibiting calmodulin activation of type I adenylate cyclase, presumably through thiol nitrosylation at the calmodulin-binding site (30,31). Notably, increases in cAMP in leukocytes activate cAMPdependent protein kinase (PKA). This kinase phosphorylates transcription factors that bind to the cAMP-response element on the TNF␣ promoter, thereby inhibiting TNF␣ mRNA transcription (32)(33)(34)(35). The effect of NO on type I adenylate cyclase suggests that NO might up-regulate TNF␣ synthesis in human monocytes by decreasing cAMP concentrations.
We investigated this question using U937 cells, a human monocytic cell line that differentiates into monocyte-macrophage-like cells and produces TNF␣ when exposed to phorbol myristate acetate (PMA) (36 -38). The specific objectives were as follows: 1) to demonstrate that NO up-regulates TNF␣ production in PMA-differentiated U937 cells and test the cGMPdependence of this effect; 2) to determine whether NO alters resting or stimulated cAMP concentrations in intact cells; 3) to investigate the effect of inhibitors or activators of PKA on NO-stimulated TNF␣ production in this system; and 4) to determine if NO-induced changes in TNF␣ mRNA levels were consistent with a cAMP mechanism.

EXPERIMENTAL PROCEDURES
Reagents and Cells-PMA, S-nitroso-N-acetylpenicillamine (SNAP), 8-bromo-cGMP, dibutyryl cGMP (Bt 2 cGMP), dibutyryl cAMP * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Human neutrophils were isolated from the venous blood of normal volunteers by dextran sedimentation and Ficoll-Hypaque density centrifugation as described previously (39). U937 cells obtained from ATCC (Rockville, MD) were cultured in RPMI 1640 supplemented with 25 mM HEPES, 10% heat-inactivated fetal calf serum, penicillin (100 units/ ml), streptomycin (100 g/ml), and 2% (w/v) L-glutamine (all from Biofluids, Inc., Rockville, MD). The cells were grown at 37°C in a humidified atmosphere containing 5% CO 2 . Differentiation was induced by incubating U937 cells with PMA (100 nM) for 48 h. The cells were then washed three times with Hank's balanced salt solution without Ca 2ϩ and Mg 2ϩ (HBSSϪ) to remove residual PMA.
Measurement of TNF␣ Production-Differentiated U937 cells (5 ϫ 10 5 cells/ml) were incubated with each of the following reagents for 24 h: increasing concentrations of SNP or SNAP (0 -500 M); 8-bromo-cGMP or Bt 2 cGMP (0 -1000 M); or Bt 2 cAMP (0 -100 M) in the presence of varying doses of SNAP (0 -500 M). In another experiment, differentiated U937 cells were pretreated for 6 h with varying concentrations of H89 (0 -30 M), a cell-permeable PKA inhibitor. The cells (5 ϫ 10 5 cells/ml) were then washed three times with HBSSϪ and incubated for an additional 24 h in the absence or presence of SNAP (500 M). TNF␣ release into the medium was measured using an enzyme-linked immunosorbent assay (R&D Systems, Minneapolis, MN). All reagents were dissolved into RPMI 1640, and tested negative for endotoxin using a limulus amoebocyte lysate assay (BioWhittaker, Inc., Walkersville, MD).
Determination of Total cGMP-At a concentration of 4 ϫ 10 6 cells/ml, differentiated U937 cells or freshly isolated human neutrophils were preincubated for 15 min in a shaking water bath at 37°C with Hank's balanced salt solution with Ca 2ϩ and Mg 2ϩ (HBSSϩ) containing 1 mM IBMX. After addition of SNP or SNAP (0 -1000 M), these cells were incubated for 0 -120 min followed by the addition of iced ethanol (65% (v/v) final concentration). The amount of cGMP present in the samples was quantitated by an enzyme immunoassay according to the manufacturer's protocol (Amersham).
Determination of cAMP-Differentiated U937 cells (5 ϫ 10 5 cells/ml) were incubated in RPMI 1640 containing increasing doses of SNP or SNAP (0 -500 M) for 24 h. The cells were then scraped and spun for collection of cell pellets. Intracellular cAMP was extracted by lysing the cell pellets in 65% iced ethanol and centrifuging at 2000 ϫ g for 15 min. The resultant supernatants were dried and cAMP was quantitated using an enzyme immunoassay according to the manufacturer's protocol (Amersham).
To investigate the effect of NO on agonist-stimulated cAMP re-sponses, differentiated U937 cells (5 ϫ 10 5 cells/ml) were pretreated for 15 min with HBSSϪ containing 500 M SNP or SNAP in the presence of IBMX (1 mM). Adenylate cyclase agonists, isoproterenol (1 M), or PGE 2 (10 M) were then added and total cAMP was extracted at several time points (0 -15 min) and quantitated as described above. Ribonuclease Protection Assay (RPA)-Differentiated U937 cells (5 ϫ 10 5 cells/ml) were incubated under the following conditions for 3.5 h: RPMI 1640 alone or with 8-bromo-cGMP (100 M), Bt 2 cAMP (100 M), SNP (500 M), SNAP (500 M), or Bt 2 cAMP (100 M) and SNAP (500 M). RNA was isolated using TRI REAGENT-LS. The entire open reading frame of human TNF␣ cDNA, in the PAW711 plasmid, was kindly provided by Dr. Alice M. Wang (40). A 253-base pair TNF␣ gene fragment, obtained by digestion with the enzymes AvaI and HincII (Stratagene, La Jolla, CA), was subcloned into the pGEM3Z vector (Promega, Madison, WI). The recombined pGEM3Z plasmid was subsequently sequenced to confirm the presence of the TNF␣ fragment insert. Single-strand antisense mRNA probe labeled with 35 S was prepared by in vitro transcription using SP6 polymerase. RPA was performed with a RPA II kit (Ambion Inc., Austin, TX) using 50 g of total RNA according to the manufacturer's instructions. Results obtained were expressed as percentages of the concurrently run ␤-actin controls.
Statistics-All data are expressed as mean Ϯ S.E. All p values are two-sided. To analyze dose-response curves in each experiment, the data was first summarized using the nonparametric Sen-Theil estimate of regression slope (41). From these dose-response slope estimates, one sample t tests (against the hypothesis of H 0 : slope ϭ 0) were done to determine whether the observed dose-response was significant. To determine if the TNF␣ response to doses of SNAP was monotonically decreasing as the Bt 2 cAMP dose increased, in addition to using the same technique as described above, we also used Page's nonparametric test (42) for ordered alternatives in a 2-way ANOVA layout (the first factor was the increasing doses of Bt 2 cAMP, the second factor was experiment). To evaluate the effects of NO on the isoproterenol-induced cAMP response in PMA-differentiated U937 cells, the area under the curve was determined for each condition/experiment. Then, nominal p values were computed for each of the two comparisons of interest: isoproterenol versus isoproterenol/SNAP, and isoproterenol versus isoproterenol/SNP. Last, these nominal p values were adjusted by multiplying each by 2, to take into account the multiple comparisons (Bonferroni adjustment) (43). When comparing relative TNF␣ mRNA levels, we used paired t tests.

RESULTS
Effect of NO on TNF␣ Production by PMA-differentiated U937 Cells-First, we confirmed previous reports (37, 38) that PMA-differentiated U937 cells produce TNF␣ (Fig. 1). Next, we demonstrated that exogenous NO donors, SNP or SNAP, increased TNF␣ release from PMA-differentiated U937 cells over a 24-h incubation period in a dose-dependent manner (p ϭ 0.002 for both). At the highest concentrations examined (500 M), SNP and SNAP increased TNF␣ production 7.0-and 15.6fold, respectively (Fig. 1). exogenous NO donors was assessed using human neutrophils as a positive control (Fig. 2). After 2 h incubation in the presence of IBMX (1 mM), neither SNP nor SNAP elevated total cGMP in U937 cells (p Ͼ 0.1). In contrast, and as expected (28), either SNP or SNAP increased total cGMP production by human neutrophils in a dose-dependent manner (p ϭ 0.012 and p ϭ 0.004, respectively).

Effect of NO on Total cGMP in U937 Cell
To exclude the possibility that cGMP was quickly degraded despite the presence of a phosphodiesterase (PDE) inhibitor, thereby masking an increase in cGMP production, total cGMP was also measured at earlier time points (0, 1, 3, 5, 10, and 15 min) in the presence of IBMX (1 mM) after exposure to SNAP (1 mM). SNAP increased total cGMP in neutrophils which reached a maximum at 15 min. However, this NO donor had no effect on cGMP production in PMA-differentiated U937 cells (Fig. 2,  inset). Furthermore, in separate experiments, SNP or SNAP had no effect on total cGMP in naive U937 cells, or in U937 cells differentiated with retinoic acid with or without 1␣,25-dihydroxyvitamin D 3 (data not shown).
Effect of NO on cAMP Levels in PMA-differentiated U937 Cells-To determine if NO altered cAMP levels, intact PMAdifferentiated U937 cells were incubated in the presence of increasing concentrations of SNP or SNAP for 24 h. Either SNP or SNAP decreased intracellular cAMP levels ( Fig. 4) in PMAdifferentiated U937 cells in a dose-dependent manner (p ϭ 0.001 and p ϭ 0.009, respectively). Compared with culture medium alone, SNP or SNAP at the highest concentrations examined (500 M) decreased intracellular cAMP by 37.8 and 51.5%, respectively.
To confirm that NO decreases cAMP levels in PMA-differentiated U937 cells, the effect of NO on agonist-stimulated cAMP responses was examined in the presence of IBMX (1 mM). As shown in Fig. 5A, SNP or SNAP (500 M) blunted isoproterenolstimulated increases in total cAMP by 21.8% (p ϭ 0.004) and 27.6% (p ϭ 0.008), respectively. Furthermore, the inhibitory effects of SNP (p ϭ 0.045) or SNAP (p ϭ 0.011) were also demonstrated using PGE 2 (10 M) instead of isoproterenol to activate adenylate cyclase (Fig. 5B). This experiment was performed in HBSSϪ to avoid activation of Ca 2ϩ /calmodulin-dependent PDE I (44,45).
Effect of Inhibitors or Activators of PKA on TNF␣ Production by PMA-differentiated U937 Cells-Finding that NO but not cGMP increased TNF␣ production, and that NO decreased intracellular cAMP levels, we next examined the effect of cellpermeable agents that either inhibit (H89) or activate (Bt 2 cAMP) PKA on TNF␣ production. Preincubation of PMAdifferentiated U937 cells with H89 increased TNF␣ production (6.5-fold at 30 M) in a dose-dependent manner (Fig. 6, p ϭ  0.035). The addition of SNAP elevated TNF␣ production and eliminated the dose effect of H89 (p ϭ 0.77). Conversely, Bt 2 cAMP blocked SNAP-induced TNF␣ production (Fig. 7). With increasing concentrations of Bt 2 cAMP (0 -100 M), the dose-dependent effect of SNAP on TNF␣ production was abolished (p ϭ 0.001). Cell viability by trypan blue exclusion was not decreased at the concentrations of Bt 2 cAMP employed (0 -100 M, data not shown).
Effect of NO and Cyclic Nucleotide Analogs on TNF␣ mRNA Levels in PMA-differentiated U937 Cells-Relative TNF␣ mRNA levels were measured using a RPA to investigate the effects of NO, cGMP, and cAMP on TNF␣ mRNA transcription (Fig. 8). SNP or SNAP increased TNF␣ mRNA levels by 57.3% (p ϭ 0.045) and 66.2% (p ϭ 0.001), respectively. An analog of cAMP, Bt 2 cAMP, decreased TNF␣ mRNA levels (p ϭ 0.002), and prevented the effect of SNAP (Bt 2 cAMP versus Bt 2 cAMP and SNAP: p ϭ 0.93). In contrast to the NO donors, 8-bromo-cGMP had no effect on TNF␣ mRNA levels (p ϭ 0.52). DISCUSSION We demonstrated that NO increased TNF␣ production in PMA-differentiated U937 cells by decreasing intracellular cAMP levels, indicating that NO uses cAMP, rather than cGMP as a second messenger for some of its cellular effects. This conclusion is based on these findings: 1) two structurally dissimilar NO donors increased TNF␣ production in a dose-dependent manner; 2) both SNP and SNAP increased cGMP concentrations in human neutrophil cultures, but had no effect on cGMP concentrations in PMA-differentiated U937 cell cultures; 3) cell-permeable analogs of cGMP, 8-bromo-cGMP and Bt 2 cGMP, did not alter TNF␣ production by PMA-differentiated U937 cells; 4) SNP or SNAP not only decreased intracellular cAMP in a dose-dependent manner, but also blunted isoproterenol-and PGE 2 -stimulated cAMP responses in PMAdifferentiated U937 cells; 5) an inhibitor of PKA, H89, increased TNF␣ release in the absence but not in the presence of SNAP; 6) conversely, an activator of PKA, Bt 2 cAMP, abolished the effect of SNAP on TNF␣ production; and 7) finally, NO donors and Bt 2 cAMP but not 8-bromo-cGMP caused changes in relative TNF␣ mRNA levels that were consistent with a cAMP mechanism for the observed effects of NO. Collectively, these experiments demonstrate that NO-induced up-regulation of TNF␣ production in this human cell line uses cAMP, not cGMP, as its second messenger.
Many of the known effects of NO have been attributed to its ability to generate cGMP through its action on soluble guanylate cyclase (15)(16)(17)(18). However, we were unable to demonstrate that NO donors increase cGMP in either naive or differentiated U937 cells. It seems unlikely that our inability to detect NO-stimulated increases in cGMP was due to degradation of cGMP. U937 cells have extremely low cGMP hydrolytic activity and do not contain the cGMP-specific PDE isoenzyme (PDE V) (45,46). Furthermore, our experiments were conducted in the presence of a potent, nonselective PDE inhibitor. These data indicate that U937 cells lack NO-sensitive soluble guanylate cyclase. Moreover, membrane permeable cGMP analogs, 8-bromo-cGMP and Bt 2 cGMP, were unable to mimic the effect of NO on TNF␣ production, a finding that has also been reported in human peripheral blood mononuclear cells and neutrophil preparations (28,29). These results further suggest that NO regulates TNF␣ production in PMA-differentiated U937 cells by a cGMP-independent mechanism.
The ability of NO to decrease intracellular cAMP levels and blunt isoproterenol-and PGE 2 -stimulated cAMP responses provide direct evidence for our speculation that NO increases TNF␣ production in PMA-differentiated U937 cells by decreasing cAMP levels. Decreases in intracellular cAMP can result either from its reduced synthesis by adenylate cyclase or from increased catabolism due to increased PDE activity (47). The result that NO decreased cAMP concentrations in the presence of IBMX, a nonspecific PDE inhibitor, support the hypothesis that changes in cAMP levels were due to decreased synthesis, rather than increased catabolism by PDE. Interestingly, cGMP can either increase cAMP hydrolysis by activating PDE II (45,46) or decrease cAMP hydrolysis by inhibiting PDE III (48,49). Increased or decreased cAMP hydrolysis mediated by cGMP is unlikely in our experiments since U937 cells lack PDE II ac-  6. Effect of H89, a PKA inhibitor, on TNF␣ production. PMA-differentiated U937 cells were pretreated for 6 h with H89, washed with HBSSϪ and then incubated for an additional 24 h with or without SNAP (500 M). TNF␣ released into the medium is presented as the mean Ϯ S.E. of three independent experiments, each run in duplicate.
tivity (45,46), a PDE inhibitor was used, and as shown here, U937 cells do not produce cGMP in response to a NO signal. However, in other cell types that contain NO-sensitive soluble guanylate cyclase, the ability of NO to decrease cAMP production may be masked by decreased cAMP hydrolysis via cGMPmediated inhibition of PDE III activity (48 -51).
As already noted, NO has been shown to inhibit calmodulindependent adenylate cyclase activity in isolated cell membranes by oxidizing cysteine residues at the calmodulin-binding site (30). Other studies have shown that calcium ionophores potentiate cAMP responses in human peripheral blood mononuclear cells and neutrophils stimulated with isoproterenol and PGE 2 , and this potentiation was inhibited by calmodulin inhibitors (52)(53)(54). These observations indicate that the calmodulindependent adenylate cyclase subtype that is inhibited by NO is present in human leukocytes. Although this suggests a possible mechanism for NO modulation of intracellular cAMP levels, other possibilities exist. Substitution of cysteine residues for other amino acids in the ␤ 2 -adrenergic receptor markedly shifts the dose-response curve to the right for isoproterenol-stimulated increases in intracellular cAMP concentrations (55). This suggests that NO could reduce agonist-stimulated cAMP responses by decreasing receptor affinity through the nitrosylation of key cysteine-containing domains.
Inhibitors and activators of PKA were used to further explore the possibility that NO was using a cAMP-dependent signaling pathway. H89, a specific cell-permeable inhibitor of PKA (56), dose-dependently increased TNF␣ release. This finding demonstrates in PMA-differentiated U937 cells that basal cAMP levels and the resulting degree of PKA activation are inhibitory of TNF␣ synthesis. Interestingly, H89, at the doses tested, did not further up-regulate SNAP-induced TNF␣ production, suggesting that PKA was maximally inactivated by the decrease in cAMP levels caused by NO. In contrast, Bt 2 cAMP only slightly suppressed basal TNF␣ release, but completely abolished SNAP-induced increases in TNF␣ production. This cAMP analog can permeate cell membranes and is resistant to hydrolysis by PDE (57), enabling it to persist in cell cultures and mimic prolonged elevations of intracellular cAMP. Toxicity caused by the butyrate moiety of the Bt 2 cAMP molecule was unlikely to be responsible for this effect, since the concentrations of Bt 2 cAMP used were relatively low and Bt 2 cGMP (up to 1 mM), which also contains a butyrate moiety did not alter TNF␣ production. Furthermore, cell viability by trypan blue exclusion was not decreased by Bt 2 cAMP. Together, these re-sults reinforced our conclusion that NO increases TNF␣ production in PMA-differentiated U937 cells by decreasing cAMP levels.
Similarly, NO donors were also found to increase relative TNF␣ mRNA levels and Bt 2 cAMP completely prevented this effect. An analog of cGMP, 8-bromo-cGMP, had no effect. These results are consistent with a cAMP mechanism acting at the level of transcription for the observed effects of NO on TNF␣ production in PMA-differentiated U937 cells. Evidence is also available that cAMP can down-regulate TNF␣ expression at a post-transcriptional level in monocytes and macrophages (58 -60).
Previously, NO was demonstrated to increase TNF␣ mRNA levels in HL-60 cells (61), but we found no change in TNF␣ mRNA levels in human neutrophil preparations (28). These inconsistent findings may be ascribed to the different methods employed to measure mRNA levels. In our present study, we measured TNF␣ mRNA levels using a RPA, which may be quantitatively more reliable than the reverse transcription polymerase chain reaction assay used in our previous experiments with neutrophils (28). Furthermore, differentiated U937 cells may contain more copies of TNF␣ mRNA than neutrophils. Besides the cAMP mechanism, our data do not exclude the possibility of additional mechanisms for the up-regulation of TNF␣ production by NO. For example, NO could activate or induce other transcription factors, such as NF-B (29).
In conclusion, the present study indicates that NO increase TNF␣ production in PMA-differentiated U937 cells by decreasing intracellular cAMP. To our knowledge, this is the first demonstration in intact cells that NO signal transduction can use cAMP rather than cGMP to regulate cell function.