S-nitrosylation/activation of COX-2 mediates NMDA neurotoxicity

Abstract

Glutamate/N-methyl-d-aspartate (NMDA) receptor-mediated neurotoxicity involves cyclooxygenase (COX)-2. We demonstrate that this neurotoxicity reflects activation of COX-2 by S-nitrosylation after selective binding of neuronal nitric oxide synthase (nNOS) to COX-2. nNOS, via its PDZ domain, binds COX-2 with the generated NO S-nitrosylating and activating the enzyme. Selective disruption of nNOS—COX-2 binding prevents NMDA neurotoxicity.

Cyclooxygenase (COX) is a key regulatory enzyme in the biosynthesis of the prostaglandins (1). Two principal forms of the enzyme exist (2). COX-1 is constitutive and in the stomach forms the cytoprotective prostaglandins whose depletion in response to nonsteroidal antiinflammatory drugs (NSAIDs) elicits gastric irritation. By contrast, COX-2 is induced by inflammatory stimuli so that COX-2 inhibitors are antiinflammatory with lesser gastric disturbance than nonspecific COX inhibitors (1, 3, 4). Because central neurons are not mediators of inflammation, it was not anticipated that they should possess COX-2. In a screen for immediate-early genes induced by brain stimulation, Worley and associates (5) discovered abundant inducible COX-2 in the brain localized exclusively in neurons (6). Unlike peripheral COX-2, the brain enzyme is also constitutive.

Neuronal COX-2 is activated predominantly by glutamate/ N-methyl-d-aspartate (NMDA) receptor stimulation under basal conditions, being diminished by NMDA antagonists (5). This regulation, together with the localization of COX-2 to synaptic spines (6), implies a role for COX-2 in physiologic neurotransmission and neuronal plasticity. COX-2 is also activated during glutamate-dependent neurotoxicity. Selective COX-2 inhibitors (7–10) and targeted deletion of COX-2 genes (11) decrease neurotoxicity and stroke damage, whereas COX-2 overexpression exacerbates focal ischemic damage (12, 13).

Besides the prostaglandins, nitric oxide (NO) is a major small-molecule mediator of inflammation formed by the inducible form of NO synthase (iNOS) (14, 15). Recently we discovered synergistic interactions between the NO and prostaglandin systems. We showed that iNOS binds selectively to COX-2 with the generated NO S-nitrosylating COX-2 and activating prostaglandin formation (16). Selective disruption of iNOS–COX-2 binding presents NO-mediated activation of prostaglandin formation. The iNOS–COX-2 inflammatory system presumably is most relevant to inflammation in the periphery because the brain contains only low levels of iNOS. We wondered whether neuronal NOS (nNOS) might interact with the COX-2 system in the brain. In the present work we demonstrate that nNOS binds COX-2. NO generated by nNOS directly S-nitrosylates and activates COX-2. Moreover, selective disruption of nNOS–COX-2 binding substantially reduces NMDA neurotoxicity.

Results

nNOS and COX-2 display similar regional distributions in mouse brain, with highest levels in the cerebellum (Fig. 1a). COX-2 occurs in the luminal portion of the endoplasmic reticulum, but multiple other intracellular localizations have been reported (17, 18). Immunohistochemical studies reveal COX-2 concentrated in perinuclear regions but also distributed throughout the cytoplasm with multiple colocalizations with nNOS (Fig. 1b). Even though COX-2 is constitutively expressed in neurons, the basal level is extremely low, and hence we augmented endogenous COX-2 expression levels by eliciting seizures. Endogenous nNOS and COX-2 coprecipitate from postseizure brain tissue (Fig. 1c). To map their binding sites, we first showed that nNOS and COX-2, overexpressed in HEK293T cells, coprecipitate (Fig. 1d). We mapped the binding domain of COX-2 to its C-terminal 10 amino acids (Fig. 1e). nNOS binding is mediated by its PDZ domain (Fig. 1f), whereas a domain near the catalytic region mediates iNOS binding to COX-2 (16).

Fig. 1.

COX-2 interacts with nNOS. (a) Western blot analysis of basal level COX-2 expression in cortex, hippocampus, hypothalamus, and cerebellum. (b) Colocalization of nNOS and COX-2 in cerebellar granular cell cultures. The localizations of COX-2 and nNOS were determined by immunohistochemistry. Cerebellar granular cells were fixed, permeabilized, blocked, and then incubated with COX-2 and nNOS antibodies. Immunostaining was visualized by confocal microscopy. Images of COX-2 (green) and nNOS (red) are superimposed to show colocalization. Nuclei were visualized with Hoechst staining (blue). (c) Coimmunoprecipitation of COX-2 and nNOS in mouse brain cerebellum after seizure. Mice were subjected to maximal electrical shock-induced seizures. Cerebellum lysates were immunoprecipitated (IP) with COX-2 antibody and analyzed with antibodies against mouse COX-2 and rabbit nNOS. (d) Coimmunoprecipitation of COX-2 and nNOS in HEK293T cells. Cell lysates were immunoprecipitated with COX-2 antibody and analyzed by Western blotting with antibodies against mouse COX-2 and rabbit nNOS. (e) Mapping COX-2 sites mediating binding to nNOS. HEK293T cells were transfected with Myc-COX-2 fragments and full-length nNOS. Cell lysates were immunoprecipitated with myc antibody and analyzed by Western blotting with anti-myc antibody and anti-nNOS antibody. (f) Mapping nNOS sites mediating binding to COX-2. HEK293T cells were transfected with full-length COX-2 and myc-nNOS fragments. Cell lysates were immunoprecipitated with COX-2 antibody and analyzed by Western blotting with anti-COX-2 antibody and anti-myc antibody. All experiments were repeated at least five times, and representative figures are shown.

Using the biotin switch method (19), we established that overexpression of nNOS elicits S-nitrosylation of COX-2, which is inhibited by the NOS inhibitor N(G)-nitro-l-arginine methyl ester (l-NAME) (Fig. 2a). COX-2 is physiologically S-nitrosylated under basal conditions in cerebellar granular preparations, with S-nitrosylation abolished in cultures from nNOS-deleted mice (Fig. 2b). S-nitrosylation of COX-2 is stimulated by NMDA receptor activation and by ionomycin, a calcium ionophore (Fig. 2c). S-nitrosylation of COX-2 reflects nNOS–COX-2 binding because the PDZ domain of nNOS, used as a dominant-negative peptide, blocks nNOS–COX-2 binding and abolishes S-nitrosylation of COX-2 (Fig. 2d).

Fig. 2.

nNOS-derived NO stimulates COX-2 activity in vivo and in vitro by S-nitrosylation. (a) S-nitrosylation of COX-2 in HEK293T cells with or without NOS inhibitor l-NAME, determined by the biotin switch assay. HEK293T cells were transfected with nNOS and/or COX-2 in the presence and absence of 1 mM l-NAME. S-nitrosylated proteins were precipitated and analyzed with Western blotting using COX-2 antibody. (b) S-nitrosylation of constitutive COX-2 in brain cerebellum. Homogenates from wild-type and nNOS-knockout mice were subjected to the biotin switch assay and Western blot analysis. (c) NMDA and ionomycin stimulate the S-nitrosylation of COX-2 in cerebellar granular neurons. Cerebellar granular neurons were treated with either 300 μM NMDA or 1 μM ionomycin. S-nitrosylated COX-2 was determined by biotin switch assay and Western blotting with COX-2 antibody. (d) Dominant-negative peptide (nNOS-PDZ domain) blocks nNOS–COX-2 binding and S-nitrosylation of COX-2. HEK293 cells stably expressing nNOS were transfected with COX-2 alone or with combinations of COX-2 and different amounts of myc-nNOS-PDZ plasmids. After transfection, cell lysates were immunoprecipitated with nNOS antibody, and bound proteins were detected with COX-2 or myc antibodies. Quantification of immunoprecipitated blot is shown. S-nitrosylation of COX-2 was determined by biotin switch assay and Western blotting with COX-2 antibody. (e) NMDA activates COX-2 activity in the presence and absence of 1 μM cycloheximide in cerebellar granular neurons. COX-2 activity was assayed in the cell lysates of primary cerebellar granular neurons treated with 300 μM NMDA and 5 μM glycine for 5 min. Bars represent the mean ± SEM of three independent cell cultures performed in triplicate. Asterisks indicate statistically significant differences by Student's t test (*, P < 0.05). (f) Inhibition of nNOS and COX-2 prevents NMDA neurotoxicity. Cell viability was assessed by MTT assay. All experiments were repeated at least five times, and representative figures are shown. Bars represent the mean ± SEM of three independent cell cultures performed in triplicate. Asterisks indicate statistically significant differences by Student's t test (*, P < 0.05).

Binding of nNOS to COX-2 directly elicits its activation (Table 1). Incubation of purified nNOS with COX-2 increases COX-2 activity more than 2-fold with the activation abolished by the selective nNOS inhibitor vinyl-l-NIO, deletion of the substrate arginine, or deletion of the cofactor NADPH. The dominant-negative PDZ domain of nNOS also markedly reduces COX-2 activation, establishing that nNOS must bind to COX-2 to provide the close proximity that enables NO to activates COX-2.

Table 1.

nNOS activates COX-2 in vitro

Conditions	COX-2 activity, μmol/min/mg
nNOS + COX-2	1.65 ± 0.15*
COX-2	0.75 ± 0.06
nNOS + COX-2 + vinyl-l-NIO	0.79 ± 0.07
nNOS + COX-2 + arginine	0.85 ± 0.09
nNOS + COX-2 + NADPH	0.89 ± 0.08
nNOS + COX-2 + nNOS-PDZ	0.96 ± 0.11

*Significantly higher than all other conditions (P < 0.05).

COX-2 mRNA levels in brain cultures are stimulated by NMDA (5), and we demonstrate augmentation of COX-2 protein after NMDA treatment with blockade by the NOS inhibitor l-NAME [supporting information (SI) Fig. S1]. To discriminate newly synthesized COX-2 from activation of preexisting enzyme mediated by S-nitrosylation, we treated cerebellar granular cultures with 1 μM cycloheximide, which inhibits protein synthesis >90% and prevents NMDA-mediated increases in COX-2 protein (Fig. 2e). NMDA markedly enhances COX-2 activity in cycloheximide-treated cultures, although less than in untreated cultures, with the activation abolished in cultures of nNOS-deleted mice. Thus, NMDA activates already synthesized COX-2.

Selective COX-2 inhibitors block NMDA neurotoxicity (7, 9). In cerebellar granular cultures, the selective COX-2 inhibitor NS-398 reduces NMDA neurotoxicity, which is also diminished in cultures from nNOS-knockout mice (Fig. 2f). NMDA neurotoxicity in cerebellar granular cell cultures is decreased by ≈70–80% after transfection of the cultures with dominant-negative constructs, the PDZ domain of nNOS, or the C-terminal amino acid peptide of COX-2 (Table 2). This finding establishes that nNOS–COX-2 binding, with attendant S-nitrosylation and activation of COX-2, mediates NMDA neurotoxicity.

Table 2.

Dominant-negative nNOS and COX-2 constructs block NMDA neurotoxicity

Constructs	No. of dead cells with NMDA treatment/no. of dead cells without NMDA treatment, -fold
GFP	3.75 ± 0.35
GFP-nNOS-PDZ	1.67 ± 0.18*
GFP-COX-2 (484–604)	1.56 ± 0.25

*Significantly higher than all other conditions (P < 0.05).

Discussion

In the present work we have established that nNOS physiologically binds COX-2, that this binding “delivers” NO to COX-2 to elicit S-nitrosylation and activation, that the stimulation of COX-2 activity by NMDA reflects the binding and S-nitrosylation, and that NMDA neurotoxicity also depends on nNOS–COX-2 binding. The nNOS–COX-2 interactions differ somewhat from the iNOS–COX-2 we reported (16). Mapping studies established that an area of iNOS including the catalytic domain is critical for the binding of iNOS to COX-2 (16). By contrast, the PDZ domain of nNOS mediates its binding to COX-2 so that the PDZ domain can function as a dominant-negative to disrupt the binding of the two enzymes. In our earlier work we demonstrated stimulation of COX-2 by NO donors. In the present work we provide more compelling evidence for COX-2 regulation by NO derived from nNOS. Thus, under in vitro conditions nNOS activates COX-2, and blockade of nNOS–COX-2 binding by a dominant-negative nNOS construct prevents such activation. The earlier study showed that iNOS–COX-2 interactions influence prostaglandin formation but did not elucidate physiological consequences. Here, we show that NMDA neurotoxicity depends on nNOS–COX-2 binding because dominant-negative constructs of nNOS and of COX-2 block neurotoxicity.

In functional experiments reported here, we have focused on neurotoxicity. It is likely that nNOS–COX-2 interactions mediate neuronal formation of prostaglandins under basal conditions. Thus, it is well established that physiological NDMA neurotransmission augments calcium entry into neurons, which binds to calmodulin and activates nNOS (20). Basal “constitutive” levels of COX-2 are diminished by treatment with NMDA antagonists (6, 16), which presumably reflects loss of the NMDA/nNOS/COX-2 S-nitrosylation cascade.

Our findings may have therapeutic relevance. Drugs that block nNOS–COX-2 binding might reduce cerebral levels of prostaglandins to diminish stroke damage and possibly other neurodegenerative dysfunctions. For iNOS–COX-2 binding, the catalytic domain of iNOS mediates binding, affording the possibility of dual inhibitors. Arginine derivatives are well established to inhibit NOS activity. Conceivably, modifications of such agents could also block iNOS–COX-2 binding so that the resultant drugs would decrease levels of both NO and prostaglandins. We do not know whether such a dual inhibitor for nNOS–COX-2 is feasible. The PDZ domain of nNOS is not critical for catalytic activity because its deletion does not diminish enzyme function. The nNOS PDZ domain is at the extreme N-terminal portion of the molecule immediately adjacent to the catalytic domain, suggesting that a drug acting at the intersection of these two domains might inhibit enzyme activity and COX-2 binding. Such an agent would be uniquely beneficial in the treatment of stroke and neurodegenerative conditions.

Materials and Methods

Cell Culture.

HEK293T cells were from the American Type Culture Collection and were maintained in Dulbecco's modified Eagle's medium with 10% FBS, 2 mM l-glutamine, 100 units/ml penicillin, and 100 μg/ml streptomycin in a humidified incubator at 37°C with 5% CO₂. Cerebellar granule neurons were prepared from wild-type and nNOS-knockout mice (14-day gestation) and grown in Neurobasal medium (Invitrogen) with 2% B27 supplement, 10% FBS, 25 mM KCl, and 2 mM l-glutamine.

Reagents.

Chemicals were purchased from Sigma unless otherwise noted. Rabbit anti-nNOS antibody was generated in our laboratory. Mouse anti-COX-2 antibody and anti-GAPDH antibody were from BD Biosciences.

Plasmid Constructions.

Full-length COX-2 was made as described in ref. 16. COX-2 and nNOS plasmids were all cloned into pCMV-Myc (Clontech) with primers harboring SalI/NotI restriction sites. GST-nNOS-PDZ was cloned into pCMV-GST (21) with SalI/NotI sites. GFP-nNOS-PDZ and GFP-COX-2 (483–604) were cloned into pEGFP-C1 vector (GeneBank) with restriction sites at EcoRI and SalI.

Protein Purification.

Recombinant protein GST-nNOS PDZ domain was purified through GST-Sepharose and eluted with 50 mM glutathione. Eluate was dialyzed with buffer containing 50 mM Tris·HCl (pH 7.4), 1 mM EDTA, and complete proteinase inhibitor, and the GST tag was cleaved with 1 mM thrombin.

Immunoprecipitation.

Cells were lysed in lysis buffer containing 100 mM Tris·HCl (pH 7.4), 150 mM NaCl, 10% vol/vol glycerol, 0.5% Triton X-100, and protease inhibitors. Cell lysates were centrifuged and supernatants obtained. Protein A–Sepharose beads were added to the supernatant to preclear nonspecific binding. Then COX-2 antibody or Myc antibody was added and incubated with precleared lysates at 4°C. After overnight incubation, protein A–Sepharose beads were added for 1 h. The pellets were washed four times with lysis buffer and eluted with SDS/PAGE sample buffer, which was subjected to SDS/PAGE and analyzed by Western blotting.

S-nitrosylation biotin switch assay.

S-nitrosylated protein was detected as described in ref. 19. In brief, cells were lysed in HEN [250 mM Hepes-NaOH (pH 7.7), 1 mM EDTA, and 0.1 mM neocuproine] buffer, and reduced cysteines were blocked with blocking buffer containing 4 mM methyl methanethionsulfonate at 50°C for 20 min. S-nitrosylated cysteines were reduced with 1 mM ascorbate and biotinylated with 4 mM biotin-[N-(6-(biotinamido)hexyl)-3′-(2-pyridyldithio)-propionamide] (Pierce). The biotinylated proteins were precipitated by streptavidin–agarose beads, eluted with SDS/PAGE sample buffer, and analyzed by Western blotting.

COX-2 enzymatic assay.

COX-2 activity was monitored by using a COX assay kit from Cayman Chemical according to the manufacturer's instructions. For activation of COX-2 by nNOS in vitro, recombinant nNOS and COX-2 were preincubated with reaction mix containing 50 mM Tris·HCl (pH 7.4), 0.1 μM calmodulin, 100 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 0.1 mM EGTA, 2 mM CaCl₂, 5 mM glucose, and 100 mM l-arginine for 1 h. Reaction products were desalted by using Micro Bio-Spin P6 prepacked columns from Bio-Rad. The eluate from these columns was collected, and COX-2 was assayed as described.

Immunohistochemistry.

Cerebellar granular cells were fixed in 4% paraformaldehyde in PBS for 5 min and permeabilized in 0.1% Triton X-100 for 30 s. Cells were blocked in 10% goat serum at room temperature for 1 h and incubated with COX-2 antibody and nNOS antibody overnight at 4°C. Rhodamine (rabbit) or fluorescein (mouse)-conjugated secondary antibodies were then added at 1:10,000 for 1 h at room temperature, and localizations were assessed with a fluorescence microscope (Zeiss).

Cell death assay.

Cytotoxicity of nNOS-knockout and wild-type neurons cells was monitored by using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assay. NMDA neurotoxicity was induced by treating the 8- to 10-day-old culture with Earle's balanced salt solution (EBSS) containing 300 μM NMDA and 5 μM glycine for 5 min. After 20–24 h, MTT was added for ≈3 h and eluted with DMSO. Cell viability was monitored by measuring the absorbance of 1 ml of solution at 580 nm at 630 nm to normalize for cell debris/background.

The neuroprotective effect afforded by transfected dominant negative peptides [GFP-nNOS-PDZ domain and GFP-COX-2 (483–604)] was monitored by propidium iodide (PI) staining and Hoechst double staining as described in ref. 22 with some modifications. Neurons were transfected at DIV 6 with 1 μg of GFP dominant-negative constructs and GFP plasmid as control using Lipofectamine 2000. Transfected neurons were treated with NMDA for 5 min. Toxicity was assayed 20–24 h after NMDA exposure by counting the dead and GFP-positive cells. Cells were exposed to 40 μg/ml PI dissolved in BSS for 10 min to identify cells with disrupted membranes. After washing with BSS, the cells were fixed with 4% paraformaldehyde for 10 min followed by washing with BSS. The cells were then treated with 10 μg/ml Hoechst 33258 (Sigma) in BSS for 4 min in the dark and washed with BSS. The cells were stored in PBS for later assessment of nuclear morphology. All staining procedures were performed at room temperature. Cells were examined under a fluorescence microscope (Zeiss). Toxicity was assayed by counting the dead cells with PI staining and the total GFP-positive cells by microscopic examination with computer-assisted cell counting. Cell death was determined as the ratio of dead to total cells and quantified by counting 500–1,000 cells.