The purpose of this study was to determine whether the selective cyclooxygenase-2 (COX-2) inhibitor rofecoxib [4-[4-(methylsulfonyl)phenyl]-3-phenyl-2(5H)-furanone] could effectively prevent hippocampal neuronal injury in an animal model of excitotoxic neurodegeneration. COX-2 protein levels increased between 3 and 6 h, peaked at 12 h, and declined to near baseline levels 24 h after injection of N-methyl-d-aspartate (NMDA; 18 nmol) into the CA1 region of the left hippocampus. Mice that were fed ad libitum a control rodent diet for 4 days before and 3 days after injection of NMDA demonstrated marked neuronal loss in the primary cell layers of the ipsilateral CA1, CA3, and dentate gyrus (50, 30, and 20% cell loss, respectively). This injury was potently and dose-dependently reduced by feeding animals a diet standardized to deliver 15 or 30 mg/kg rofecoxib per day. Neurodegeneration in the CA1 region was reduced by 30.1 ± 5.6 and 51.5 ± 9.0%, respectively; in the CA3 by 64.6 ± 12.4 and 69.0 ± 14.1%, respectively; and in the dentate gyrus by 47.8 ± 15.2 and 58.0 ± 18.2%, respectively. Moreover, rofecoxib chow slightly but significantly reduced injury-induced brain edema. These findings demonstrate that rofecoxib can ameliorate excitotoxic neuronal injury in vivo and, as such, may be a particularly promising pharmaceutical for the treatment of neurological diseases associated with overactivation of NMDA receptors.
The generation of arachidonic acid and eicosanoids is an important component of the inflammatory response in most cells. Free arachidonic acid is metabolized into prostaglandins and thromboxanes via the enzyme cyclooxygenase (COX) (for review, see Smith et al., 1991). There are two unique COX genes, COX-1 and COX-2. Both are remarkably similar in structure and have similar enzyme kinetics, yet important differences distinguish the two isozymes. The COX-1 gene promoter lacks a TATA box motif (Kraemer et al., 1992; Wang et al., 1993) and is constitutively active in most cells (Xu et al., 1997; Ye and Liu, 2002). In contrast, the COX-2 promoter is not basally active in most cell types but can be strongly and rapidly induced by growth factors and proinflammatory mediators (Kujubu et al., 1991; Fletcher et al., 1992; O'Banion et al., 1996). Neurons are an important exception. COX-2 is constitutively present in the perinuclear, dendritic, and axonal areas of glutamatergic neurons, particularly in the cortex, hippocampus, and amygdala, of animals, including humans (Yamagata et al., 1993; Breder et al., 1995; Yasojima et al., 1999; Hewett et al., 2000). Within the hippocampus, basal COX-2 expression is observed in dentate granule cells and in CA1-CA3 pyramidal neurons (Yamagata et al., 1993; Breder et al., 1995; Adams et al., 1996; Sandhya et al., 1998). Although there is heterogeneity in the level of expression within neuronal populations, COX-2 expression is uniformly up-regulated in the cortex and hippocampus following generalized excitatory stimulation, such as seizures (Yamagata et al., 1993; Adams et al., 1996). This up-regulation is blocked by N-methyl-d-aspartate (NMDA) receptor antagonists indicating that neuronal COX-2 expression is linked to excitatory neuronal activity (Yamagata et al., 1993; Adams et al., 1996). A direct link between NMDA receptor-induced neuronal COX-2 induction, activity, and cell death was first established in cortical cell cultures in vitro (Hewett et al., 2000) and subsequently confirmed in the cortex in vivo (Iadecola et al., 2001). As for the hippocampus, recent work from our laboratory demonstrated that a nonselective cyclooxygenase inhibitor, naproxen (Laneuville et al., 1994), ameliorated hippocampal parenchymal cell death and edema formation mediated by excessive activation of neuronal NMDA receptors in vivo (Silakova et al., 2004). Whether this protection occurred via a COX-2-dependent mechanism is addressed herein.
Materials and Methods
NMDA, chloral hydrate, paraformaldehyde, high-performance liquid chromatography grade 2-methylbutane, and thionin were all obtained from Sigma Chemical Co. (St. Louis, MO). Pentobarbital sodium was purchased from Abbott Laboratories (North Chicago, IL). Control and Rofecoxib [4-[4-(methylsulfonyl)phenyl]-3-phenyl-2(5H)-furanone] Purina Laboratory Rodent Diets were supplied by Merck Research Laboratories (West Point, PA). The rabbit polyclonal anti-COX-2 antibody was purchased from Cayman Chemical (Ann Arbor, MI). The WesternBreeze chemiluminescent immunodetection kit was purchased from Invitrogen (Carlsbad, CA). X-Ray film was from GE Healthcare UK, Ltd. (Little Chalfont, Buckinghamshire, UK).
Intrahippocampal Injection of NMDA. Male mice (ND-4; 28–30 g; Harlan, Indianapolis, IN) were anesthetized with 400 mg/kg chloral hydrate i.p. and placed in a Kopf 900 Small Animal Stereotaxic Frame with mounted model 926 Mouse Adaptor (Kopf Instruments, Tujunga, CA). NMDA (18 nmol in 0.1 M phosphate-buffered saline, pH 7.4) was injected (350 nl) using a Kopf 5000 microinjection unit equipped with a 5-μl Hamilton syringe and a 26-gauge blunt-tipped needle at stereotaxic coordinates 1.9 mm posterior to bregma, 1.2 mm lateral from the midline, and 1.6 mm ventral from the skull surface (Franklin and Paxinos, 1997). Injections were made in 12 min, and the needle was left in place for an additional 10 min to minimize backflow. Control animals received intrahippocampal injections of PBS (0.35 μl). After the procedure, the skin wound was sutured, and animals were returned to their cages under ad libitum conditions. Before, during, and directly after the procedure (up to 1 h), animal temperature was maintained at 35.8–36.2°C using a Harvard homeothermic blanket control unit for small animals. This protocol was in accordance with the National Institutes of Health guidelines for the use of live animals and was approved by the Institutional Animal Care and Use Committee of The University of Connecticut Health Center.
Western Blot Analysis. Animals were left alone (naive) or microinjected into the left hippocampus with either PBS or NMDA and then sacrificed 3, 6, 12 or 24 h later. Hippocampi were dissected and frozen, and either a Nonidet P-40 (NP-40) or radioimmunoprecipitation assay lysis buffer containing protease inhibitors (see below) was added directly to the tubes. Thirty micrograms of total cellular protein from each animal was separated by SDS-8% polyacrylamide gel electrophoresis under reducing conditions and then electrophoretically transferred onto nitrocellulose. Membranes were placed in PBS (>70°C) for 2 min and then immunoblotted for COX-2 (rabbit polyclonal, 1/5000) following the procedure outlined in the Western-Breeze chemiluminescent immunodetection kit. Results were recorded on X-ray film, and digitized images were analyzed by computer-assisted densitometry (Gel-Pro Analyzer; MediaCybernetics, Silver Springs, MD). The constituents of the NP-40 lysis buffer were 1% NP-40, 50 mM Tris, pH 7.5, 150 mM NaCl, 5 mM EDTA, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 5 mM iodoacetamide, and 1 mM phenylmethylsulfonyl fluoride. The radioimmunoprecipitation assay buffer contained identical ingredients plus the following: 0.5% deoxycholate, 0.1% SDS, and 5 mM EGTA.
Rofecoxib (Vioxx) Treatment. Twenty-four animals were randomly divided into three groups (eight animals/group) and fed a control diet or a diet containing rofecoxib standardized to deliver an approximate dose of 15 or 30 mg/kg/day for 4 days before intrahippocampal NMDA microinjection. The chow was administered to the animals in a coded fashion to ensure that the experiments were performed and analyzed blindly. Animals were maintained on their respective diets until sacrifice 72 h later. The 4-day pretreatment paradigm was initiated to ensure that steady-state drug levels were reached before experimental manipulation (Depre et al., 2000). This regimen was well tolerated and yielded 500 nM (for 15 mg/kg/day) and 1 μM (for 30 mg/kg/day) plasma drug concentrations as determined by Pauline Luk at Merck Frosst Laboratories (Pointe Claire, QC, Canada) via high-performance liquid chromatography with UV detection.
Assessment of Neuronal Injury. Seventy-two hours after microinjection, mice were sacrificed under general anesthesia (pentobarbital; 100 mg/kg i.p.) by transcardiac perfusion with ice-cold 0.9% NaCl followed by 4% PBS-buffered paraformaldehyde. Brains were removed and postfixed for an additional hour and then placed step-wise into 15% sucrose/H2O and finally 30% sucrose/H2O solution for storage (2–3 days; 4°C). 2-Methylbutane-frozen (–80°C) brains were cut serially (1.2 to 2.7 mm posterior to bregma) into 30-μm coronal sections using a CM1900 cryostat (Leica Microsystems, Inc., Deerfield, IL),. Every 5th section was mounted and stained with 0.1% thionin, a standard Nissl stain used for the histological verification of lesions (Powers and Clark, 1955). The extent of hippocampal injury was assessed via image analysis (NIH Image; Scion Software; Scion Corporation, Frederick, MD) as described previously in detail (Silakova et al., 2004). The mean-integrated density on the ipsilateral (I) and contralateral (C) sides of eight to 10 analyzed serial images were calculated and compared. The counts in the contralateral hippocampus served as an internal control for each animal, as injury was restricted to the ipsilateral hippocampus. Data are expressed as the percentage difference in mean-integrated density [100 × (C – I)/C].
Measurement of Brain Edema. Twenty-six mice were randomly divided into three groups (8–9 animals/group) and fed either control diet or diets standardized to deliver 15 or 30 mg/kg/day rofecoxib for 4 days before intrahippocampal microinjection of NMDA. After injection, animals were returned to their cages and were maintained on their respective diets ad libitum until sacrifice 24 h later. Brain water content in the ipsilateral and contralateral hippocampi (derived from a two mm thick coronal section centered on the mark of injection) was measured using the wet-dry method (Dempsey et al., 2000).
Statistical Analysis. Percentage data are by nature non-normally distributed. Thus, the arcsine square root of the percentage data (Steel and Torrie, 1980) was analyzed via one- or two-way ANOVA followed by the appropriate post hoc test as described in each figure legend. Values of zero or less were set at 1 × 10–20 before transformation. Significance was assessed at p < 0.05.
COX-2 Protein Expression Is Increased in Hippocampus following NMDA Microinjection. Western blot was used to study COX-2 protein expression. The modest expression of COX-2 protein expressed in naive control brain was unaffected by microinjection of PBS (350 nl) into the left hippocampus (Fig. 1). However, COX-2 protein expression increased between 3 and 6 h after NMDA injection, appeared to peak at 12 h, and declined to near baseline levels 24 h later (Fig. 1).
Treatment with Rofecoxib Decreases Neuronal Death in the Hippocampal Formation after NMDA Injection. This same unilateral injection of NMDA resulted in degeneration of areas of CA1 (56.1 ± 1.5%), CA3 (28.5 ± 4.3%), and dentate gyrus (21.6 ± 4.3%) that extended ≈1mm (rostral to caudal) around the site of injection (Figs. 2 and 3). As demonstrated previously (Silakova et al., 2004), neuronal degeneration was restricted to the injected side with no neuronal loss detected in the contralateral hippocampus (Fig. 4). Systemic administration of the selective COX-2 inhibitor rofecoxib partially but potently prevented the injury in a dose-dependent manner in all three areas of the hippocampus under study (Figs. 2, 3, and 4). Injury in the CA1, CA3, and dentate gyrus was reduced by 30.1 ± 5.6, 64.6 ± 12.4, and 47.84 ± 15.1%, respectively, after administration of chow standardized to deliver 15 mg/kg rofecoxib per day (Fig. 3). Animals that received rofecoxib chow standardized to deliver 30 mg/kg/day enjoyed even greater protection with a 51.5 ± 9.0, 69.0 ± 14.1, and 58.0 ± 18.2% reduction in excitotoxic neurodegeneration observed in the CA1, CA3, and dentate gyrus, respectively (Fig. 3).
Brain Edema Resulting from NMDA Injection Is Reduced by Systemic Rofecoxib Treatment. Microinjection of NMDA into mouse hippocampus results in significant brain edema on both the ipsilateral (injected) and contralateral (uninjected) side 24 h postinjury (Silakova et al., 2004) (Fig. 5). Animals that were administered the rofecoxib diets demonstrated a reduction in edema on both sides, although brain water content of the ipsilateral side never reached levels measured in naive controls (designated by horizontal line) (Fig. 5).
Excitotoxicity, primarily due to overactivation of the NMDA subtype of glutamate receptor, contributes to neuronal loss in both acute neurological injuries and chronic neurodegenerative diseases (Choi, 1988; Meldrum and Garthwaite, 1990; Coyle and Puttfarcken, 1993). With the exception of memantine (Reisberg et al., 2003), human clinical trials using NMDA receptor antagonists have proven to be disappointing (Doble, 1999; Lee et al., 1999). The reasons for these failures may be diverse and could depend on whether the initial injury develops slowly over time or is rapidly initiated. In the latter case, it is likely that compounds that prevent excitotoxic neuronal injury after initial receptor binding of glutamate has occurred may actually be more clinically practical. Recent work from our laboratory demonstrates that systemic treatment of a commercially available and clinically useful nonselective cyclooxygenase inhibitor, naproxen (Laneuville et al., 1994), ameliorates hippocampal and parenchymal cell death and edema formation mediated by excessive activation of neuronal NMDA receptors in vivo (Silakova et al., 2004). Results from the present study, demonstrating similar and significant neuroprotection by rofecoxib p.o. in the same in vivo excitotoxicity model, support the idea that the cell death occurs predominantly via a COX-2-dependent mechanism. Rofecoxib (Vioxx) was chosen because it is a potent, orally active COX-2 inhibitor (Chan et al., 1999; Van Hecken et al., 2000) that, at the beginning of this study, was commercially available, although now it has been voluntarily withdrawn from the market. It shows a greater than 1000-fold selectivity for the inhibition of COX-2 over COX-1 (Chan et al., 1999), has a long half-life (≈17 h) (Depre et al., 2000), readily crosses the blood-brain barrier (Halpin et al., 2000; Dembo et al., 2005), and effectively decreases prostaglandin biosynthesis in rodent brain (Candelario-Jalil et al., 2003b; Teismann et al., 2003).
The idea that COX-2 contributes to excitotoxic neuronal injury is certainly not new. Animals overexpressing COX-2, as well as cells derived from said animals, are more susceptible to injury induced by the excitotoxins, kainate, and glutamate (Kelley et al., 1999; Mirjany et al., 2002). Selective inhibition of COX-2 effectively ameliorates cortical brain damage caused via direct intracortical injection of NMDA (Iadecola et al., 2001) and hippocampal oxidative damage following intraperitoneal injection of kainate (Candelario-Jalil et al., 2000). Glutamate receptor-mediated injury to cortical, hippocampal, and cerebellar granule cell neurons in vitro is also reduced when COX-2 is pharmacologically inhibited (Hewett et al., 2000; Strauss and Marini, 2002; Carlson, 2003; McCullough et al., 2004). In addition, COX-2 expression is increased in brains from animals subjected to experimental manipulations mimicking neurological diseases that have a known excitotoxic component (Collaco-Moraes et al., 1996; Nogawa et al., 1997; Sanz et al., 1997; Nakayama et al., 1998). Pharmacological inhibition of COX-2 or use of COX-2 null mutant animals in these same models has, in most cases, proven beneficial (Nakayama et al., 1998; Iadecola et al., 2001; Cernak et al., 2002; Drachman et al., 2002; Candelario-Jalil et al., 2003b; Gopez et al., 2005). Importantly, up-regulation of COX-2 is reported to occur in neurons and non-neuronal cells in human brains following a lethal cerebral ischemic insult (Sairanen et al., 1998; Iadecola et al., 1999) in Alzheimer's diseased brains (Pasinetti and Aisen, 1998; Hoozemans et al., 2001), in post mortem Parkinson's disease specimens (Teismann et al., 2003), and in spinal cord (Yasojima et al., 1999), cortex, and hippocampus of amyotrophic lateral sclerosis patients (Yokota et al., 2004), indicating that these experimental observations may have direct relevance to human pathology.
Studies specifically using rofecoxib are rare but, in general, support the findings presented herein. In rat, rofecoxib treatment, at concentrations similar to what was used in the present study (10 mg/kg/day), selectively diminishes hippocampal neuronal cell loss induced by kainate administration (Kunz and Oliw, 2001). Loss of cholinergic neurons in the nucleus basalis following quisqualate injection is ameliorated in rats receiving rofecoxib (3 mg/kg/day) (Scali et al., 2003), and treatment with rofecoxib (5–20 mg/kg/day) increases survival of CA1 hippocampal neurons in gerbils subjected to global ischemia (Candelario-Jalil et al., 2003b). Interestingly, transgenic mice engineered to express humanmutated SOD-1, a murine model for amyotrophic lateral sclerosis, experienced a delay in locomotor changes after administration of rofecoxib intraperitoneally (10 mg/kg; three times weekly) (Azari et al., 2005) and a 20% increase in their average lifespan compared with control animals when three times that dose was employed (Klivenyi et al., 2004). Finally, treatment with rofecoxib preserved nigral neurons and tyrosine hydroxylase-positive fibers following MPTP treatment, an experimental model of Parkinson's disease (Teismann et al., 2003).
Along with direct cytoprotection, reduction in postinjury brain edema remains a potential therapeutic target because elevation of intracranial pressure can lead to microvasculature compression, resulting in impairment of cerebral blood flow. In addition to histological preservation of neurons, our data demonstrate that the animals fed rofecoxib chow had less injury-induced brain edema than animals receiving control chow. However, it seems unlikely that improvement of hemodynamics could account for the neuroprotection afforded by rofecoxib, because significant water elevation in the injured cortex remained. Interestingly, this increase in water content was nearly completely prevented by systemic administration of the nonselective COX inhibitor naproxen in this same model (Silakova et al., 2004), suggesting that products derived from COX-1 metabolism also contribute.
Two prevailing hypotheses for how COX-2 activation results in neuronal injury exist: generation of toxic reactive oxygen species and/or generation of prostaglandins. Relevant experimental data support both theories, indicating that either or both can, under certain circumstances, contribute to the injury process (Kontos, 1985; Kukreja et al., 1986; Candelario-Jalil et al., 2003a; Carlson, 2003; Teismann et al., 2003; Jiang et al., 2004; Manabe et al., 2004; Kawano et al., 2006). However, numerous studies also describe a neuroprotective function for prostaglandins (Akaike et al., 1994; Cazevieille et al., 1994; McCullough et al., 2004). Thus, whether prostanoids are injurious or protective may relate to the type of prostaglandin produced, the signaling receptor(s) activated, and the immediate physical environment in which they act. Unfortunately, we were unable to explore the mechanism of neuroprotection vis à vis prostaglandins or reactive oxygen species in this study, because rofecoxib became unavailable to us after its withdrawal.
In summary, the present study demonstrates significant neuroprotection by rofecoxib against excitotoxicity in vivo, lending further support to the general idea that cell death associated with overactivation of NMDA receptors occurs via a COX-2-dependent mechanism. More specifically, this study demonstrates directly the effectiveness of rofecoxib as a potential therapeutic agent against neurological disorders associated with overactivation of NMDA receptors should it return to the market. Vioxx was voluntary withdrawn because of concerns over increased risk of cardiovascular events after prolonged dosing. (Catella-Lawson et al., 1999; McAdam et al., 1999) (for review, see FitzGerald, 2003). In the meantime, it will be important to determine whether compounds marketed with COX-2 inhibitory properties, such as celecoxib (Celebrex), meloxicam (Mobic), or lumiracoxib (Prexige), can prevent excitotoxic neurodegeneration in vivo, with similar efficacy and over an extended therapeutic time window. Although it is not yet known whether the side effects associated with the prolonged use of rofecoxib represent a true class effect, these concerns would necessarily exclude the use of these drugs as prophylactics for neurodegenerative disease. However, the possibility that short-term benefits could be gained in individuals afflicted with acute neurological disorders associated with NMDA receptor overactivation should not be overlooked.
We thank Tracy F. Uliasz for technical assistance, Merck Research Laboratories for providing us with control and rofecoxib diets, and Pauline Luk (Merck Frosst Laboratories) for the plasma rofecoxib measurements.
This work was supported by the NINDS, National Institutes of Health Grant NS36812 and the Merck Medical School Grant Program. S.J.H. is an Established Investigator of the American Heart Association.
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
ABBREVIATIONS: COX, cyclooxygenase; NMDA, N-methyl-d-aspartate; rofecoxib, [4-[4-(methylsulfonyl)phenyl]-3-phenyl-2(5H)-furanone]; PBS, phosphate-buffered saline; NP-40, Nonidet P-40.
- Received June 23, 2006.
- Accepted September 6, 2006.
- The American Society for Pharmacology and Experimental Therapeutics