Inducible Nitric Oxide Synthase Binds, S-Nitrosylates, and Activates Cyclooxygenase-2

See allHide authors and affiliations

Science  23 Dec 2005:
Vol. 310, Issue 5756, pp. 1966-1970
DOI: 10.1126/science.1119407


Cyclooxygenase-2 (COX-2) and inducible nitric oxide synthase (iNOS) are two major inflammatory mediators. Here we show that iNOS specifically binds to COX-2 and S-nitrosylates it, enhancing COX-2 catalytic activity. Selectively disrupting iNOS–COX-2 binding prevented NO-mediated activation of COX-2. This synergistic molecular interaction between two inflammatory systems may inform the development of anti-inflammatory drugs.

Inflammatory processes are mediated by multiple molecular mechanisms. Two of the most prominent are the production of nitric oxide (NO) by inducible NO synthase (iNOS) and the formation of prostaglandins by cyclooxygenase-2 (COX-2; prostaglandin H2 synthase) (1, 2). COX-2 inhibitors have attained widespread use as anti-inflammatory agents, although they elicit potentially adverse side effects (1, 3, 4), whereas iNOS inhibitors are not presently employed therapeutically. Inflammatory stimuli elicit the synthesis of iNOS and COX-2 proteins with similar time courses, which suggests that the two systems may interact (5, 6). Stimulants of iNOS such as bradykinin (7) and lipopolysaccharide (LPS) plus interferon-γ (IFN-γ), two components of endotoxin, enhance prostaglandin formation (8). NOS inhibitors prevent the formation of prostaglandins (9).

To determine whether iNOS and COX-2 interact, we used a murine macrophage cell line (RAW264.7) in which LPS and IFN-γ massively activate both iNOS and COX-2. INOS immunoprecipitated with COX-2–specific antibodies from lysates of cells treated with LPS–IFN-γ (Fig. 1A). This was also observed in transfected human embryonic kidney cells (HEK293T) overexpressing both proteins (fig. S1A). The two enzymes also coimmunoprecipitated from peritoneal macrophages obtained from mice injected with thioglycollate, an inflammatory stimulus that induces peritonitis or pleuritis (fig. S1B). To determine whether catalytic activity of the enzymes influences their interactions, cells that were induced by LPS–IFN-γ were also treated with the iNOS-selective inhibitor 1400W (Fig. 1B) or the COX-2–selective inhibitor SC58125 (Fig. 1C). Coimmunoprecipitation of iNOS and COX-2 by antibodies specific to either protein was unaffected by either inhibitor. The binding of iNOS and COX-2 was selective, because COX-1 did not immunoprecipitate with iNOS. To map the binding sites on both proteins, we generated selective deletions of iNOS (Fig. 1, D to F) and COX-2 (Fig. 1, G and H) sequences. The amino acid segment 1 to 144 of iNOS, which is within the oxygenase domain, is required, whereas the C terminus of COX-2 mediates binding and includes amino acids 484 to 604, which do not exist in COX-1.

Fig. 1.

COX-2 and iNOS bind selectively in vitro and in intact cells. (A) RAW264.7 cells were treated with LPS (2 μg/ml) and IFN-γ (100 U/ml). COX-2 was immunoprecipitated by COX-2–specific antibody and analyzed by Western blot with antibodies against COX-2 and iNOS. Control indicates untreated cells. (B and C) RAW264.7 cells were treated with LPS–IFN-γ with or without an iNOS inhibitor 1400W (100 μM) or COX-2 inhibitor SC58125 (100 μM). Cell lysates were subjected to immunoprecipitation (IP) and Western blot analysis with antibodies against COX-2 and iNOS. (D) The fragments of iNOS denoted in red bind to full-length COX-2, whereas fragments labeled purple do not, as determined by coimmunoprecipitation of full-length COX-2 by iNOS fragment fused to glutathione S-transferase (GST). The numbers represent the number of the amino acid sequence. (E) Transfected HEK293T cells expressing COX-2 and iNOS fragments expressed as fusion proteins with GST were precipitated with glutathione-conjugated beads. Proteins were detected by Western blot with antibodies against GST or COX-2. (F) Transfected HEK293T cells expressing COX-2 and epitope-tagged (Myc) iNOS fragments were immunoprecipitated with Myc-specific antibody and then analyzed by Western blot. (mock: The cells were treated with transfection reagent without the plasmid.) (G) Generated fragments of COX-2 that bind to full-length iNOS are labeled in red; those that do not bind are labeled in yellow. (MBD, membrane-binding domain). (H) Transfected HEK293T cells expressing iNOS and Myc-tagged COX-2 fragments were immunoprecipitated with Myc-specific antibody and analyzed by Western blot. (mock: The cells were treated with transfection reagent without the plasmid.)

The two major mechanisms whereby NO influences its intracellular targets are stimulation of guanylyl cyclase by direct binding of NO to iron in heme at the active site of guanylyl cyclase (10) or S-nitrosylation of protein targets on appropriate cysteines (11, 12). Because COX-2 has heme at its active site (13), this would be a potential target. However, NO binding to heme in COX-1 does not alter its activity (14). COX-2 also contains 13 cysteines whose roles are not fully understood (15). To explore the possibility of S-nitrosylation of COX-2 by NO, we examined multiple NO donors including nitroso-S-glutathione (GSNO) (Fig. 2A), sodium nitroprusside (SNP), spermine-NO, and (Z)-1-[N-(2-aminoethyl)-N-(2-ammonioethyl) amino]diazen-1-ium-1,2-diolate (DETA-NONOate) (fig. S3A). Using the biotin switch method in which all the S-nitrosylated cysteines are selectively biotinylated (16), we observed that all four NO donors elicited S-nitrosylation of COX-2 in transfected HEK293T cells expressing COX-2–Myc (Fig. 2A). S-Nitrosylation of COX-2 was also observed in RAW264.7 cells treated with LPS–IFN-γ. This was prevented when cells were treated with iNOS inhibitor 1400W (Fig. 2B and fig. S3B). The biotin switch method was specific, as H2O2 did not elicit S-nitrosylation (fig. S4). We also ruled out the possibility that sulfenic acid modification was detected by the biotin switch assay by demonstrating that arsenite, which reverses sulfenic acid modifications but not S-nitrosylation, failed to provide the biotin switch signal afforded by ascorbate using GSNO with purified COX-2 or LPS–IFN-γ treatment of RAW 264.7 cells (fig. S4B). In some instances there may be no need to deliver NO directly to targets, as some actions of NO are prevented by hemoglobin, which sequesters freely diffusible NO (17). We examined the effects of hemoglobin on S-nitrosylation of COX-2 under varying conditions. In transfected HEK293T cells expressing COX-2, hemoglobin prevented the S-nitrosylation elicited by GSNO (fig. S5A), whereas it failed to alter S-nitrosylation of COX-2 in RAW264.7 cells activated by LPS–IFN-γ (fig. S5B). Thus, in the more physiologic macrophage cell line, the S-nitrosylation of COX-2 induced by an inflammatory stimulus does not appear to be elicited by freely diffusible NO.

Fig. 2.

S-Nitrosylation of COX-2 enhances enzyme activity. (A) COX-2 expressed in transfected HEK293T cells is S-nitrosylated in the presence of GSNO (100 μM) or glutathione (reduced form) (100 μM) as determined by biotin-switch assay. All the S-nitrosylated proteins were precipitated and COX-2 was detected by Western blot with COX-2–specific antibody. COX-2 was selectively S-nitrosylated by GSNO. (B) LPS–IFN-γ treatment of RAW264.7 cells elicits S-nitrosylation of COX-2, which is prevented by the iNOS inhibitor 1400W (100 μM). COX-2 was selectively S-nitrosylated by endogenously generated NO. (C) COX-2enzyme activity was measured from the cell lysate of transfected HEK293T cells expressing COX-2–Myc in the presence or absence of SNP and ascorbate. Bars represent the mean ± SEM of three independent cell cultures performed in triplicate (*statistically significant by Student's t test). (D) COX-2–Myc expressed in transfected HEK293T cells is S-nitrosylated by various concentrations of GSNO. The dose-dependence of GSNO-mediated activation of PGE2 was measured. Data were pooled from at least three independent determinations, each in triplicate. (E) COX-2–Myc expressed in transfected HEK293T cells is S-nitrosylated in the presence of SNP and reversed by the addition of ASC. All the S-nitrosylated proteins were precipitated, and COX-2 was detected by Western blot with COX-2–specific antibody. (F) Recombinant human COX-2 was treated with SNP, and COX-2 activity was measured (n = 3, control: Vmax = 81.3 ± 4.8 nmol/min per mg, Km = 16.2 ± 2.2 μM; SNP: Vmax = 132 ± 6.5 nmol/min per mg, Km = 17.0 ± 2.0 μM). (G) Recombinant human COX-2 was treated with SNP, and its turnover rate (kcat) was measured in the presence of various concentrations of sucrose. Data were expressed as kcat-control over kcat in each viscosity versus viscosity ratio.

To determine whether S-nitrosylation of COX-2 alters enzyme activity, we examined transfected HEK293T cells expressing COX-2–Myc. The NO donor SNP, added to cell lysates, elicited a twofold increase in COX-2 activity, reflecting S-nitrosylation. Ascorbic acid reversed S-nitrosylation (16, 18) and prevented the increase (Fig. 2, C and D). The reversal by ascorbate of COX-2 activation by NO donors is not merely a reflection of ascorbate influences on enzyme substrates or intermediate products, as ascorbate failed to affect COX-2 activity in preparations not treated with SNP. Further evidence that S-nitrosylation and COX-2 activation are related is the closely similar concentration-response relation between the effects of the NO donor GSNO on S-nitrosylation and on COX-2 activity (Fig. 2E).

NO activates COX-2 by increasing its apparent Vmax without changing its Km (Fig. 2F). The higher concentration of SNP required to activate COX-2 in vitro compared with intact cells accords with earlier studies showing greater potency of NO donors in intact cells (19). To ascertain the kinetic basis for NO activation of COX-2, we conducted enzyme assays with increasing concentrations of sucrose to augment viscosity and slow down enzyme kinetics (Fig. 2G). As expected, with increasing viscosity, the ratio of control enzyme activity to the activity in more viscous solutions increased. This increase was diminished in SNP samples, consistent with SNP's accelerating the release of product from the enzyme.

To determine which of the 13 cysteines of COX-2 are critical for the augmentation of COX-2 activity elicited by S-nitrosylation, RAW 264.7 cells were transfected to express the N-terminal 483 amino acids or the C-terminal 120 amino acids of COX-2. LPS–IFN-γ treatment induced S-nitrosylation of the C-terminal fragment (which contains three cysteines) but not the N-terminal fragment (fig. S6). To ascertain which of these three cysteines is responsible for augmented COX-2 activity, each was mutated to serine. The mutation in which Ser is substituted for Cys526 (C526S) prevented activation of COX-2 by the NO donor SNP, whereas the C561S mutation did not (fig. S6). The C555S mutation abolished enzyme activity, so the effects of NO stimulation could not be assessed. Individual mutation of the 13 cysteines in COX-2 did not detectably diminish total S-nitrosylation of the enzyme, which suggests that multiple cysteines can be S-nitrosylated, but only C526 is responsible for enzyme activation by NO.

To clarify the influence of NO on prostaglandin E2 (PGE2), a prostaglandin synthesized by COX-2 formation in a physiologic context, we examined RAW264.7 cells. The formation of PGE2 in response to LPS–IFN-γ was inhibited by the iNOS inhibitor 1400W, with 50% reduction of PGE2 formation at drug concentrations that provide 50% inhibition of iNOS activity (Fig. 3A). Specificity of the NO association was evident by inhibition of PGE2 formation with the active l-isomer of the NOS inhibitor N-nitro-l-arginine methyl ester (l-NAME) but not by d-NAME; the effects of l-NAME were reversed by added l-arginine (Fig. 3B). Thus, about 50% of induced COX-2 activity is determined by S-nitrosylation.

Fig. 3.

Endogenously generated NO enhances COX-2 activity. (A) RAW264.7 cells were activated by LPS–IFN-γ and treated with various concentrations of iNOS inhibitor 1400W for 18 hours. The dose dependence of 1400W-mediated suppression of PGE2 and nitrite was then measured. Data were pooled from at least three independent determinations, each in triplicate (*statistically significant by Student's t test). (B) Combinations of l-NAME (500 μM), l-NAME + l-Arg (1 mM) or d-Arg (1 mM), and d-NAME (500 μM) were added to RAW264.7 cells treated with LPS–IFN-γ. PGE2 was measured. The data were pooled from three independent experiments performed, each in triplicate. (C) PGE2 and nitrite were measured from primary peritoneal macrophages isolated from wild-type (WT) or iNOS knockout (KO) mice. Macrophages were treated with LPS–IFN-γ or untreated (*statistically significant by Student's t test). (D) S-Nitrosylation of COX-2 of WT primary peritoneal macrophages treated with LPS–IFN-γ is abolished in iNOS KO macrophages. All the S-nitrosylated proteins were precipitated, and COX-2 was detected by Western blot with COX-2–specific antibody.

As RAW264.7 cells are a continuous macrophage cell line that may not behave the same as macrophages in intact organisms, we used peritoneal macrophages from mice lacking iNOS. PGE2 formation from macrophages of mice treated with LPS–IFN-γ was reduced in the iNOS knockout mice by ∼70%, in parallel with a similar reduction in nitrite formation by the macrophages (Fig. 3C) and a decrease in S-nitrosylated COX-2 (Fig. 3D). These observations concur with findings of decreased urinary PGE2 in iNOS knockout mice (20).

We hypothesized that the increase in PGE2 formation by iNOS activation reflects binding of iNOS to COX-2 to deliver NO in appropriate proximity for S-nitrosylation. To explore this possibility we blocked iNOS–COX-2 binding with the fragment of COX-2 (amino acids 484 to 604), which binds iNOS (Fig. 4A). Expression of COX-2(484–604) in transfected RAW264.7 cells abolished the coprecipitation of iNOS and COX-2. Instead, COX-2(484–604) associated with iNOS (Fig. 4A). Moreover, this interference of binding between COX-2 and iNOS by COX-2(484–604) decreased S-nitrosylation of COX-2 in RAW264.7 cells (Fig. 4B). The dominant-negative effect of COX-2(484–604) reduced PGE2 formation by more than 50%, whereas expression of a COX-2 fragment of amino acids 1 to 483, which does not bind iNOS, failed to influence PGE2 formation (Fig. 4, C and D).

Fig. 4.

COX-2–Myc fragment attenuates iNOS binding to COX-2– and NO-mediated activation of PGE2 production. Transfected RAW264.7 cells expressing COX-2–Myc fragments 1 to 483 or 484 to 604 were treated with LPS–IFN-γ. (A) Cell lysates were immunoprecipitated with rabbit iNOS-specific antibody and analyzed by Western blot with antibodies against mouse iNOS, goat COX-2, and mouse Myc. (B) COX-2–Myc fragment (484 to 604) decreases S-nitrosylation of COX-2 in RAW264.7 cells. All the S-nitrosylated proteins were precipitated and COX-2 was detected by Western blot with COX-2–specific antibody. (C) Transfected RAW264.7 cells expressing the indicated COX-2 fragments were treated with LPS–IFN-γ. PGE2 levels were measured and the data were pooled from three independent experiments performed, each in triplicate (*statistically significant by Student's t test). (D) PGE2 and the indicated COX-2 fragments were visualized with confocal microscopy using antibodies against mouse Myc and rabbit PGE2. Images of COX-2 (red) and PGE2 (green) were superimposed to show colocalization. Nuclei were visualized with Hoechst staining (blue). In D1, arrows point to two RAW264.7 cells, only one of which is expressing the COX-2 fragment 484 to 604 (red). In D2, the same two cells are analyzed for presence of endogenous PGE2 after activation of RAW264.7 cells by LPS–IFN-γ treatment. Immunofluorescent staining shows a reduction in the PGE2 expression in cells expressing COX-2(484–604) compared with the nontransfected cell (D2). This observation contrasts with D4, where the arrows point to a nontransfected cell and a transfected cell expressing COX-2(1–483). D5 does not show a reduction of PGE2 in the transfected cell as compared with the nontransfected cell.

In summary, our study establishes a physiologic binding interaction of iNOS and COX-2 bringing NO in proximity to COX-2, facilitating its S-nitrosylation and activation, and fitting with earlier findings that NOS inhibition decreases prostaglandin formation (9, 21). Our findings accord with recent evidence that many physiologic actions of NO require its delivery to molecular targets (12, 22, 23). Whereas scaffolding proteins such as CAPON (22) or PSD95 (23) link neuronal NOS, respectively, to Dexras1 (22) and N-methyl-d-aspartate receptors (23), iNOS and COX-2 bind directly. The molecular synergism between iNOS and COX-2 may represent a major mechanism of inflammatory responses. Drugs that block the iNOS–COX-2 interaction may be anti-inflammatory, synergizing with COX-2 inhibitors and permitting lower doses. As the binding site on iNOS is in the catalytic domain, derivatives of iNOS inhibitors that also prevent binding to COX-2 may decrease both NO and prostaglandin formation.

Supporting Online Material

Materials and Methods

Figs. S1 to S6


References and Notes

View Abstract

Navigate This Article