Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Cyclooxygenase (COX) is the rate-limiting enzyme in the production of prostaglandins and, as such, is a key target for many anti-inflammatory drugs. There are two known isoforms, COX-1 and COX-2, which have quite distinct expression patterns and biological activities. COX-1 is a constitutively expressed protein found in most tissues, whereas COX-2 expression can be induced by a variety of mitogens, including cytokines, hormones, and phorbol esters (Kujubu et al., 1991; O'Banion et al., 1992). Inflammatory stimuli have been found to have little effect on COX-1 expression but are well known to lead to a rapid rise in COX-2 mRNA, suggesting COX-2 plays a role in the process of inflammation (O'Banion et al., 1992; Cao et al., 1995).

Recent evidence indicates that COX-2 may also play an important role in neurodegeneration, in particular, excitotoxic neuron death (O'Banion, 1999; Hewett et al., 2000). We previously found a selective elevation in COX-2, but not COX-1, mRNA in apoptotic brain cells during response to kainic acid (KA)-induced excitotoxicity (Tocco et al., 1997), and suggest that COX-2 may be involved in pathways leading to neuronal death. To further explore the function of neuronal COX-2 in excitotoxicity, we prepared transgenic mice with neuron-specific overexpression of human (h)COX-2 and found that COX-2 potentiated the intensity and lethality of KA seizures in vivo, as well as glutamate neurotoxicity in vitro (Kelley et al., 1999). The mechanism by which neuronal COX-2 may influence excitotoxicity is unknown.

In this study, to identify the molecular mechanism involved in neuronal COX-2-mediated responses in brain that may influence excitotoxic neurodegeneration, we used complementary DNA expression arrays. We found that the expression of the endogenous cell cycle-dependent kinase (CDK) inhibitor-inhibitor kinase (INK) p18^INK4, a specific inhibitor of CDK 4,6 (Zindy et al., 1997), is a downstream target of neuronal hCOX-2 expression in the brain. The functional relevance of this evidence was further confirmed in in vitro studies showing that the phosphorylation of the retinoblastoma (Rb) tumor suppressor protein, an index of CDK activity (Weinberg, 1995), is induced in neurons overexpressing hCOX-2 during response to glutamate-mediated excitotoxic apoptotic damage. Most importantly, we found that the CDK inhibitor, flavoperidol, attenuated the hCOX-2-mediated potentiation of apoptotic neuron damage. The study suggests that a mechanism by which COX-2 may influence excitotoxicity is through promotion of cell cycle activity.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Transgenic Mice. Homozygous transgenic mice [C57BL/6J × C3H (B6C3) 9-month-old females) with overexpression of hCOX-2 in neurons and nontransgenic control littermates were previously described from our laboratory (Kelley et al., 1999). Transgenic mice were identified by dot blot hybridization of tail skin DNA samples with a random-primed, 930-bp EcoRI fragment that contains the entire simian virus 40 sequence present in the hCOX-2 transgene (Kelley et al., 1999). In this study, mice were sacrificed by cervical dislocation and the brains were stored at 70°C.

Drug Treatment. The preferential COX-2 inhibitor nimesulide was provided by Helsinn Healthcare (Lugano, Switzerland) (Warner et al., 1999). Nimesulide was mixed directly with powdered rodent feed in a mixing drum into a homogeneous preparation, which was then formulated into rodent half-inch feed pellets. All ingredients of the diet were from Zeigler Bros., Inc. (Garners, PA), and formulated diets were stored at 4°C. Mice (four per cage) had access to food and water ad libitum, and diets were replenished regularly every 3 days. Based on recent evidence indicating that 1500 mg celecoxib/kg in feed is the maximum tolerable dose in rodents for 50 weeks of treatment (Pentland et al., 1999), and that the relative potency of COX-2 inhibition by celecoxib and nimesulide is equivalent (Warner et al., 1999), we treated our mice with an equivalent dose of nimesulide (1500 mg/kg of diet) for 3 months. At sacrifice, necropsied mice did not reveal any detectable gross alteration of renal or intestinal abnormalities in the nimesulide-treated group. In previous studies we found that treatment with 1500 mg/kg nimesulide in the diet for 3 months is well tolerated in B6C3 mice without detectable adverse effect based on weight loss criteria and overall lack of gastrointestinal pathology.

Complementary DNA Expression Arrays. Brain (cerebral cortex) poly(A⁺) RNA was purified from four mice per group using the Oligotex mRNA Mini Kit (Qiagen, Valencia, CA). Fluorescent-labeled antisense cDNA probes were generated by reverse transcription, following procedures recommended by Incyte Genomics, Inc. (Palo Alto, CA). The reference probe (control mRNA, pool of n = 4) was labeled with a green-fluorescent dye, Cy3 (P1), whereas the test probes (hCOX-2 mRNA, pool of n = 4) were labeled with the red-fluorescent dye Cy5 (P2). Individual mouse UniGem microarray chips (Incyte Genomics, Inc.) containing 8832 independent genes, including known and expressed sequence tag sequences, were simultaneously hybridized with both probes. Hybridization of Cy3- and Cy5-labeled probes for each of the elements on the microarray chip was simultaneously recorded and quantified using a two-head laser scanner. The linear fitting curves of the P1 and P2 fluorescence signal were then analyzed by linear regression scatter plot analysis that revealed a very tight distribution pattern clustered in an almost 45^o diagonal line. This evidence suggested that probe preparation, hybridization condition, data collection, and microarray preparation had negligible influences in the microarray studies. In our studies, signals covering less than 40% of the surface area and/or signals having fluorescence intensity less than 600 units (range from 0 to 2200 units) were considered to be background hybridization and were excluded from analysis. The cDNA microarray study was conducted in duplicate, and genes that were consistently up- or down-regulated in hCOX-2 transgenics by 1.8-fold relative to wild-type (WT) were further analyzed. Genes, including p18^INK4 mRNA were grouped into clusters defined by cellular and/or biochemical functions using the GEM Tool software package (Incyte Genomics, Inc.). The 1.8 cut-off level was derived from the analysis of independent UniGem complementary DNA microarrays (n = 4, cDNA microarray) designed to identify consistency and reproducibility in mouse cDNA chip hybridization. In this study the expression for each gene represented in the cDNA chip expressed in the mouse brain was normalized to a mean equal to 1. The standard deviations of each gene were then calculated. The 95th percentile of the empirical distribution of the standard deviations (0.236) was used to build the model for mouse gene expression variability. Our results suggested that for 95% of the genes, gene expression levels have to be outside the 1± 0.5-fold range for P < 0.05 significance level. Although the differentially regulated gene products that were considered for further investigation in this study by cDNA array had to be outside the 1± 0.5-fold range, we found that the altered expression of p18^INK4 mRNA by RNase protection was in the range of a 30% change (see Results). Thus, the cDNA microarray evidence overestimated the differential expression of p18^INK4 mRNA in the brain relative to the RNase protection assay evidence (Materials and Methods).

Western Analysis. Tissue culture lysates were homogenized and resolved by SDS-polyacrylamide gel electrophoresis (10% acrylamide). Proteins were transferred to a nylon membrane (Transblot Membrane; Bio-Rad, Hercules, CA), blocked overnight (4°C) (Superblock; Pierce Chemical, Rockford, IL) and immunoreacted for 3 h in a 1:1000 dilution of a rabbit polyclonal anti-phospho-Rb (Ser⁷⁹⁵) antibody (PhosphoPlus; New England Biolabs, Beverly, MA). A C-terminal control antibody, which detects phosphorylated-independent levels of Rb (1:1000; PhosphoPlus; New England Biolabs) was also used in parallel studies to control for specificity of phosphorylated Ser⁷⁹⁵ Rb, and gave negative results (not shown). Immunoreactivities were visualized autoradiographically using a chemiluminescence detection kit (SuperSignal; Pierce) for horseradish peroxidase-labeled goat antirabbit IgG (Vector Laboratories, Burlingame, CA) according to the manufacturer's instruction. -Actin immunoreactivity (anti--actin, 1:1000; Sigma-Aldrich, St Louis, MO) controlled for selectivity of changes. Immunoreactivities were quantified by Western blot imaging analysis using the Bio-Rad Gel Doc 2000 gel documentation system.

RNase Protection Assay. Total RNA was assayed with the RiboQuant Multiprobe RNase Protection Assay System (BD PharMingen, San Diego, CA). A custom probe containing cDNA template for the p18^INK4 was used. The probe set included the housekeeping genes L32 and glyceraldehyde phosphate dehydrogenase (GAPDH) for normalization of assay conditions. Details on the generation of ³²P-labeled antisense RNA probes and conditions of the RNase protection assay were provided by the manufacturer's instructions as previously described (Luterman et al., 2000). The radioactively labeled RNase protection fragments were quantified using a Storm 860 Phosphor Screen Scanner with the ImageQuant software package (Molecular Dynamics, Sunnyvale, CA). Each RNase protection assay analysis was conducted with 10 µg total RNA, according to A₂₆₀ values. Data were expressed as a ratio of p18^INK4 mRNA normalized to the constitutively expressed GAPDH mRNA. Normalization of p18^INK4 mRNA signals to L32 did not change the outcome of the results (not shown).

Primary Neuronal Cultures and Glutamate-Mediated Neurotoxicity. Embryonic (E14-E16) cortico-hippocampal primary neuronal cultures derived from hCOX-2 transgenic mice were prepared as previously described (Kelley et al., 1999). Briefly, after brain dissection, mechanical trituration, and centrifugation, neurons were seeded onto poly-D-lysine-coated 96-well plates at a density of 2 × 10⁵ cells per well. The absence of astrocytes (<1-2%) was confirmed by the lack of glial-fibrillary acidic protein immunostaining (not shown). Glutamate and/or flavoperidol (L86-8275, ()-cis-5,7-dihydroxy-2-(2-clorophenyl)-8[4-(3-hydroxy-1-methyl)-piperydinyl]-4H-benzopyran-4-one]) (donated by David Park) were added to 7-day-old cultures for the appropriate time as discussed in the text. Detection of neuronal apoptotic nuclear morphology in vitro was assessed by incubation of cultures with 1 µg/ml of bisbenzimide (Hoechst 33258) (Sigma-Aldrich) for 15 min at room temperature and then coverslipped as previously described (Didier et al., 1996). The number of neurons with apoptotic nuclei (condensed chromatin) were digitized, counted, and expressed as percentage of the total number of neurons in six to eight random fields of each chamber (n = 4-6 chambers per group, per culture).

Prostaglandin Assay. PGE₂ concentration was measured by an enzyme-linked immunoassay kit (Cayman Chemical, Ann Arbor, MI) according to the manufacturer's instructions. Briefly, pulverized brain tissue stored in liquid N₂ was homogenized in 0.1 M phosphate-buffered saline (containing 1 mM EDTA and 10 µM indomethacin), mixed with an equal volume of ethanol, and centrifuged. The supernatant was diluted with 50 mM acetic buffer and purified through an affinity column (Cayman). After the column was equilibrated with column buffer (0.1 M phosphate-buffered saline, 7.7 mM NaN₃, 0.5 M NaCl₂) followed by UltraPure water, the supernatant was eluted from the 4-ml column by adding the elution solution and allowing it to pass through the packing material. The eluate was then evaporated and redissolved in enzyme-linked immunoassay buffer, applied to a 96-well plate precoated with goat antimouse IgG, and incubated with PGE₂ monoclonal antibody and recovery tracer for 18 h at 4^oC. After incubation with the PGE₂ monoclonal, the plate was rinsed fives times with washing buffer and developed using Ellman's reagent for 1 h at room temperature. The PGE₂ concentration was determined spectrophotometrically and calculated by plotting the standard % B/B0 (percentage of sample or standard bound/maximum bound) versus PGE₂ concentration (in picograms per milliliter).

Statistical Analysis. Differences between two groups were analyzed by a two-tailed t test. Analysis of variance was used to compare three or more groups, and Bonferroni's multiple comparisons test was used to detect differences across multiple groups.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

p18^INK4 mRNA Expression Is Decreased in the Cerebral Cortex of hCOX-2 Transgenic Mice. Brains from 9-month-old female hCOX-2 transgenic mice overexpressing hCOX-2 under the regulation of the neuron-specific enolase promoter were collected and used for high-throughput mRNA screening with complementary DNA array microarray chips containing probes for 8832 mouse sequences; brain mRNA from age- and gender-matched WT mice were used as reference control. Genes showing consistent differential expression at the steady-state mRNA level or turnover in the brain of hCOX-2 transgenics were considered for further analysis by RNase protection assay. Among others, mRNA expression of CDK 4,6 p18^INK4 was found to be consistently decreased in the cerebral cortex of hCOX-2 transgenics and further considered for its potential role in COX-2-mediated responses in the brain.

View larger version (28K): [in this window] [in a new window]   Fig. 1.   hCOX-2-mediated decrease of p18INK4 mRNA expression in the brain of hCOX-2 transgenics coincides with elevation of PGE2 content. The p18INK4 mRNA expression (A) and the PGE2 content (B) were assessed in the cerebral cortex of 9-month-old female homozygous hCOX-2 transgenics or wild-type control littermates. Bar graphs represent mean ± S.E.M. (, P < 0.05 versus wild type; n = 4-6 per group). NMS, nimesulide. Inset, representative RNase protection assay showing the hybridized (protected) fragments coding part of the p18INK4 (273 bp) and GAPDH (98 bp) open-reading frame.

Overexpression of Exogenous hCOX-2 Potentiates Glutamate-Mediated Apoptotic Neuron Damage in Vitro and Is Prevented by the CDK Inhibitor, Flavoperidol. Initial experiments were performed to characterize glutamate-mediated apoptotic damage using primary cortico-hippocampal neuron cultures derived from WT (B6C3) embryos. In these studies, neuronal death was evaluated by condensed pyknotic nuclear morphology using a bisbenzimide assay. We found that concentrations of glutamate ranging from 10 to 100 µM caused a dose-dependent increase of neurons with evident apoptotic damage (2- to 3-fold induction), 12 to 24 h after treatment (not shown). Based on this evidence, subsequent studies were performed using a dose of 50 µM glutamate for 24 h, which achieved a significant amount of apoptotic neuron damage within the linear range of increasing glutamate toxicity assessed by bisbenzimide in our culture conditions (not shown). In control studies we also found that the treatment of cortico-hippocampal neurons with the noncompetitive NMDA receptor antagonist MK801 (10 µM) blocked glutamate-mediated apoptotic damage (50 µM glutamate, 24 h treatment) (not show). Using primary cortico-hippocampal neuron cultures derived from transgenic embryos with neuronal overexpression of hCOX-2 or from WT littermate control embryos, we then tested the hypothesis that COX-2 in neurons could influence glutamate (50 µM glutamate, 24 h treatment)-mediated apoptotic damage.

View larger version (13K): [in this window] [in a new window]   Fig. 2.   hCOX-2 overexpression in cortico-hippocampal neuron cultures potentiates glutamate apoptotic damage. A, cortico-hippocampal neuron cultures were treated with glutamate (GLU, 50 µM) and/or flavoperidol (FLAV, 1 µM) 24 h before apoptotic neuron damage was assessed by bisbenzimide assay. Values represent means ± S.E.M. of determinations made in three separate cultures. , P < 0.01 versus control untreated WT group; , P < 0.01 versus WT glutamate-treated group. B, glutamate-mediated regulation of Ser795 Rb phosphorylation (24-h time point) in pools of cortico-hippocampal neuron cultures derived from hCOX-2 transgenic or WT control embryos. The consistency of potentiation of glutamate-mediated induction of Ser795 Rb phosphorylation in hCOX-2 cultures was confirmed in a replica Western blot; replica values are shown as percentage of their own untreated control: WT cultures, 202%; hCOX-2 cultures, 372%. In A, the number of neurons with apoptotic nuclei (condensed chromatin) were counted from six to eight random fields within premarked reticules of 1 mm2, n = 4-6 culture chambers per group, per culture.

View larger version (112K): [in this window] [in a new window]   Fig. 3.   Potentiation of glutamate apoptotic neuron damage in cortico-hippocampal neuron cultures derived by hCOX-2 transgenics and protection by flavoperidol treatment. Representative micrographs of vehicle-treated (PBS) (a-c) and glutamate-treated (24 h, 50 µM; d-f) cortico-hippocampal neuron cultures derived from WT (a and d) or hCOX-2 (b and e) transgenic embryos, as indicated. In panel f, cortico-hippocampal neuron cultures derived from hCOX-2 embryos treated with 50 µM glutamate (as in panel e) were cotreated with 1 µM flavoperidol (flav). In panel c, control hCOX-2 neuron cultures were treated with 1 µM flavoperidol alone. In panels a to f, arrows point toward bisbenzimide fluorescence-positive neurons characterized by condensed pyknotic nuclear morphology.

The COX-2 Preferential Inhibitor Nimesulide Neuroprotects While Preventing Glutamate-Mediated Induction of Ser⁷⁹⁵ pRb Phosphorylation in WT Cortico-hippocampal Neurons. Using primary cortico-hippocampal neuron cultures derived from WT control embryos (B6C3), we then tested the hypothesis that preferential COX-2 inhibitors may neuroprotect against glutamate-mediated apoptotic damage, possibly through the modulation of Ser⁷⁹⁵ pRb phosphorylation.

View larger version (25K): [in this window] [in a new window]   Fig. 4.   The preferential COX-2 inhibitor nimesulide neuroprotects against glutamate-mediated apoptotic damage coincidental to inhibition of Ser795 pRb phosphorylation. In A, cortico-hippocampal neuron cultures derived from wild-type (B6C3) embryos were treated with glutamate (GLU, 50 µM) and/or nimesulide (NMS, 1 µM) 24 h before apoptotic neuron damage was assessed. Cultures treated with 0.01% DMSO did not differ from treated controls (not shown). , P < 0.001 versus untreated group. In B, parallel cortico-hippocampal neuron cultures, derived from WT control embryos as shown in A, demonstrated that the elevation of Ser795 Rb phosphorylation by glutamate treatment (24 h, 50 µM) is prevented by cotreatment with NMS (1 µM). Values represent means ± S.E.M. of determinations made in five separate cultures for glutamate alone and three separate cultures for the other groups; , P < 0.01 versus untreated control cultures. In B inset, representative Western blot immunoreactivity of Ser795 Rb. In A and B, apoptotic neurons were counted from six to eight random fields within premarked reticules of 1 mm2 (n = 4-6 culture chambers per group, per culture).

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, using complementary DNA microarray technique, we found that the mRNA expression of the INK p18^INK4, an inhibitor of CDK 4,6 that controls the activation of the Rb tumor suppressor protein phosphorylation, is decreased in the brain of adult hCOX-2 homozygous transgenics. This complementary DNA microarray finding was confirmed by RNase protection assay. Further assays indicated that chronic treatment of the transgenic mice with the preferential COX-2 inhibitor nimesulide (Warner et al., 1999) in the diet prevents the hCOX-2-mediated decrease of the CKI p18^INK4 mRNA expression. This evidence in vivo was extended to in vitro studies revealing that hCOX-2 overexpression in primary cortico-hippocampal neuron cultures potentiates excitotoxic-mediated apoptotic damage, which coincided with elevation of Ser⁷⁹⁵ pRb, an index of CDK 4,6 activation (Connell-Crowley et al., 1997). This finding, coupled with the observation that the CDK inhibitor flavoperidol and the preferential COX-2 inhibitor nimesulide may prevent apoptotic neuron damage, suggests that a mechanism by which neuronal COX-2 may influence excitotoxicity is through activation of cell cycle activity. Consistent with our evidence, recent studies also suggested that the loss of expression of the endogenous CDK inhibitor p16^INK4 in neurons and the activation of cell cycle machinery may be responsible for delayed neuronal cell death in a model of neuronal ischemic damage (Katchanov et al., 2001).

Two classes of CDK inhibitors, the INK4 (p16, p18, p21) and the cycle-inhibitory protein (CIP) (p21, p27, p57) families, have been defined based on sequence similarity (Elledge et al., 1996) and found to differ in specificity and mechanisms of inhibition (Zindy et al., 1997). Unlike the CIP family, which effectively inhibits multiple classes of cell cycle G₁ kinases including CDK2, CDK3, and CDK4 and CDK6, the INK4 family is specific for CDK4 and CDK6 (reviewed in Elledge et al., 1996). The link between CDK activity and cell cycle is provided by phosphorylation of the Rb, a phosphoprotein that regulates growth in the G₁ phase of the cell cycle (Weinberg, 1995), in part by binding to and inhibiting critical regulatory proteins that include members of the E2F family of transcription factors (Connell-Crowley et al., 1997). We also note that, physiologically, Rb is differentially regulated by CDK complexes; for example, there is evidence that CDK 4,6 phosphorylation steps appear to activate pRb, with subsequent phosphorylation steps being inactivating (Ezhevsky et al., 2001). pRb is phosphorylated on a defined number of serine and threonine residues by activation of CDK 4,6 during G₁(Zarkowska and Mittnacht, 1997). These events appear to play an important role in neurodegeneration, especially because pRb phosphorylation occurs in dying cells (Gill et al., 1998), and it appears that overexpression of E2F promotes cell death (Qin et al., 1994).

Our in vitro studies show that hCOX-2 overexpression in primary cortico-hippocampal neurons derived from homozygous hCOX-2 transgenics promotes glutamate-mediated apoptotic damage. This is consistent with our previous finding, showing that neuronal hCOX-2 potentiates the intensity and lethality of KA excitotoxicity (Kelley et al., 1999). We have found that a potential mechanism by which hCOX-2 may promote excitotoxicity in vitro is by influencing Ser⁷⁹⁵ pRb phosphorylation. Conversely, this was found to be prevented by treatment of neurons with the COX-2 preferential inhibitor, nimesulide (Warner et al., 1999). This evidence is of high interest, particularly in view of the finding that cell cycle abnormalities occur in several neurodegenerative diseases (reviewed in Raina et al., 2001), including Alzheimer's disease (AD) (Raina et al., 2000) and models of AD (Giovanni et al., 1999, 2000; Xiang et al., 2001). Also interesting, and consistent with our findings, is the recent observation that the CDK inhibitor flavoperidol may neuroprotect against excitotoxicity (Park et al., 2000), thus further underscoring the implication of cell cycle activity involvement in excitotoxic neurodegeneration.

Ongoing studies on cancer continue to show that the loss of endogenous CDK inhibitors such as INKs and CIP-KIP is sufficient to precipitate abnormal cell proliferation, coincidental with pRb activation (Elledge et al., 1996; Connell-Crowley et al., 1997; Schreiber et al., 1999). This is in agreement with other studies on brain tumors showing that the loss of endogenous CDK inhibitors is sufficient to precipitate uncontrolled proliferation (Nishikawa et al., 1995; Ueki et al., 1996). However, apoptotic neuron death appears to be closely tied together with abnormal cell cycle in neurodegenerative diseases. This has been also demonstrated in AD (Raina et al., 2001), where abnormal cell cycle activities in the brain were correlated with neuronal death (Herrup and Busser, 1995). The general interpretation of these findings is that if neurons committed to a permanent cessation of cell division, then, once they are forced to reenter the cell cycle, they die (Raina et al., 2000). Thus, the attempt to reenter the cell cycle may be a significant factor that accompanies neurodegeneration and AD. Our study brings a new perspective to this hypothesis by suggesting that neuronal COX-2, which is also elevated in the AD brain (Ho et al., 2001), may influence the excitotoxic neuronal cell in postmitotic neurons by promoting an unsuccessful attempt to reenter the mitotic cycle as schematized in Fig. 5.

View larger version (23K): [in this window] [in a new window]   Fig. 5.   Hypothetical mechanisms by which neuronal COX-2 may promote glutamate excitotoxicity through cell cycle activity.

The apparent protective effect of COX inhibitors in AD (reviewed in Pasinetti, 2001) suggests that COX may be involved in neurodegeneration. The mechanism of action for the beneficial role of nonsteroidal anti-inflammatory drugs in AD is unclear; it is generally assumed that their effects are mediated by competitive inhibition of COX catalytic activity, thus reducing the production of inflammatory prostaglandins from membrane-derived arachidonate (Warner et al., 1999). However, the role of COX-2 inhibitors in neurodegeneration may be far more complex. For example, although there is evidence that COX-2-specific inhibitors neuroprotect in vitro against glutamate toxicity (Hewett et al., 2000), recent evidence suggests that inhibition of COX-2 may also worsen responses to traumatic brain injury (Dash et al., 2000) and KA seizures (Baik et al., 1999). The demonstration in the present study that neuronal hCOX-2 overexpression in the brain may influence cell cycle activities and increase susceptibility to excitotoxicity suggests a novel, noninflammatory role for COX-2 in the brain, and has important implications for understanding its role in neurodegeneration and, possibly, the development of therapeutic strategies for neurodegenerative diseases.