Contributions of cyclooxygenase-2 to neuroplasticity and neuropathology of the central nervous system

Abstract

Cyclooxygenase (COX) enzymes, or prostaglandin-endoperoxide synthases (PTGS), are heme-containing bis-oxygenases that catalyze the first committed reaction in metabolism of arachidonic acid (AA) to the potent lipid mediators, prostanoids and thromboxanes. Two isozymes of COX enzymes (COX-1 and COX-2) have been identified to date. This review will focus specifically on the neurobiological and neuropathological consequences of AA metabolism via the COX-2 pathway and discuss the potential therapeutic benefit of COX-2 inhibition in the setting of neurological disease. However, given the controversy surrounding the use of COX-2 selective inhibitors with respect to cardiovascular health, it will be important to move beyond COX to identify which down-stream effectors are responsible for the deleterious and/or potentially protective effects of COX-2 activation in the setting of neurological disease. Important advances toward this goal are highlighted herein. Identification of unique effectors in AA metabolism could direct the development of new therapeutics holding significant promise for the prevention and treatment of neurological disorders.

Introduction

Arachidonic acid (AA) is a C₂₀ polyunsaturated fatty acid (5,8,11,14-eicosatetraenoic acid) produced by the elongation and desaturation of linoleic acid, an essential ω-6 fatty acid. In the central nervous system (CNS), endothelial cells and astrocytes are among the cells that synthesize AA from linoleic acid. Interestingly, although neuronal membranes are rich in AA, neurons appear to be incapable of desaturating elongated precursor fatty acids (Moore et al., 1991). Thus, neurons of the CNS must acquire AA from their environment; neighboring astrocytes are a likely source (Moore et al., 1991).

A general overview of AA metabolism is shown in Fig. 1. AA is enriched on the cytoplasmic face of the lipid bilayer, where it is esterified at the sn-2 position of membrane phospholipids, including inositol phospholipids. The rate of release of esterified AA from membranes is controlled by the phospholipase A₂ (PLA₂) family of phospholipases (Kudo & Murakami, 2002), or alternatively, by diacylglycerol lipase (Meiri et al., 1998). In neurons, release of AA has been linked to excitatory neuronal activity and appears to be specifically coupled to N-methyl-d-aspartate (NMDA) receptor activity (Dumuis et al., 1988, Lazarewicz et al., 1988, Sanfeliu et al., 1990, Tapia-Arancibia et al., 1992, Taylor & Hewett, 2002). Free AA is a substrate for cyclooxygenases (COXs) and lipoxygenases (LOXs), which catalzye production of prostaglandin H₂ (PGH₂) or hydroperoxyeicosatetraenoic acids (HPETE), respectively. PGH₂ and HPETEs are further metabolized to potent bioactive lipid mediators by specific synthases for prostaglandins (PG), thromboxanes (TXs), and leukotrienes (LTs). These eicosanoids affect cell function by binding to specific cell surface 7 transmembrane G-protein coupled receptors (Narumiya et al., 1999, Kobayashi & Narumiya, 2002, Tsuboi et al., 2002, Brink et al., 2003) and, in the case of cyclopentenone PG, nuclear receptors (Negishi & Katoh, 2002, Ide et al., 2003).

AA and its metabolites affect a wide spectrum of activities in the nervous system. This review will focus specifically on the neurobiological and neuropathological consequences of AA metabolism via the COX pathway. The mechanism of AA release and the function of the LOX pathway in the CNS have been reviewed recently by others (Kim et al., 1999, Manev et al., 2000, Sapirstein & Bonventre, 2000, Sun et al., 2004).

COX enzymes, or prostaglandin-endoperoxide synthases (PTGS), are heme-containing bis-oxygenases that catalyze the first committed reaction in metabolism of AA to prostanoids and thromboxanes (Fig. 1). COXs possess both a COX and a peroxidase function (Smith & Song, 2002). The COX reaction catalyzes the incorporation of 2 molecules of O₂ into AA to form PGG₂. The peroxidase reaction catalyzes the 2 electron reduction of this endoperoxide intermediate to form PGH₂ (Smith et al., 1996). Production of PGH₂ by COX enzymes is self-limited by suicide inactivation, an innate kinetic property of both the COX and peroxidase reactions (Smith & Song, 2002). Thus, subsequent production of PGH₂ necessitates synthesis of new COX protein.

Two unique COX genes, COX-1 and COX-2, have been identified and extensively characterized. The COX-1 gene product was cloned first (DeWitt & Smith, 1988, Merlie et al., 1988, Yokoyama et al., 1988), followed soon thereafter by COX-2 (Kujubu et al., 1991, Hla & Neilson, 1992, O'Banion et al., 1992). Both genes code for 70 kDa polypeptides that exhibit > 60% identity. Moreover, both isozymes assemble as homodimers, reside as integral membrane proteins within the lumen of the endoplasmic reticulum and nuclear envelope, and exhibit only subtle differences in enzyme kinetics (Kulmacz et al., 1994, Swinney et al., 1997). A putative third isoform, COX-3, was recently reported. However, it was found to be a splice variant of the COX-1 transcript (Chandrasekharan et al., 2002, Qin et al., 2005, Snipes et al., 2005). As such, it should be considered a variant of the COX-1 gene and not a new COX isozyme. A detailed discussion on the existence of alternative splice variants of both COX isozymes and their putative function, if any, are reviewed elsewhere (Davies et al., 2004, Simmons et al., 2004, Kis et al., 2005).

Although COX-1 and COX-2 are remarkably similar in many respects, important differences distinguish the 2 isozymes. Perhaps foremost among these is their striking difference in gene expression (Tanabe & Tohnai, 2002). 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 in an SP1-dependent manner (Xu et al., 1997, Ye & Liu, 2002). In contrast to COX-1, 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. The COX-2 gene promoter contains a TATA box motif and a number of cis-acting elements, including CREB, C/EBP, NF-IL6, AP-1, SP1, and NF-κB consensus sequences, which control transcriptional responsiveness under various conditions (Fletcher et al., 1992, Appleby et al., 1994). Moreover, the 3′ untranslated region of COX-2 mRNA contains sequences, including an A/U-rich element consisting of overlapping AUUUA motifs, which control mRNA stability and/or translation (Dixon et al., 2000, Cok & Morrison, 2001). Optimal COX-2 protein expression likely requires coordination between transcriptional and post-transcription mechanisms.

In the CNS, COX-1 protein is constitutely expressed in both glia and neurons. Interestingly, unlike most tissues, COX-2 protein is also constitutitively expressed in the CNS, where it is localized primarily within neurons. It is not detected in glia under physiologic conditions, with the exception of radial glia of the spinal cord (Ghilardi et al., 2004). However, astrocytes and microglia can express COX-2 after exposure to proinflammatory mediators in vitro or following CNS injury in vivo (Minghetti & Levi, 1995, Busija et al., 1996, O'Banion et al., 1996, Bauer et al., 1997, Hewett, 1999, Maslinska et al., 1999). It is worthy of note that, unlike neurons of the CNS, peripheral dorsal root ganglion neurons appear to constitutively express COX-1 exclusively and lack detectable COX-2 expression, either under basal conditions or during peripheral inflammatory states (Dou et al., 2004).

Constitutive CNS COX-2 protein is detected in the perinuclear, dendritic and axonal domains of neurons, particularly in cortex, hippocampus, amygdala and dorsal horn of the spinal cord of both rodent and human CNS (Yamagata et al., 1993, Breder et al., 1995, Adams et al., 1996, Beiche et al., 1996, Kaufmann et al., 1996, Sandhya et al., 1998, Samad et al., 2001). Within the hippocampus, COX-2 protein is observed in dentate gyrus and CA1–CA3 pyramidal layers, where it appears to colocalize selectively with glutamatergic neurons (Kaufmann et al., 1996). Importantly, the level of neuronal COX-2 expression within the CNS appears to be coupled to excitatory neuronal activity. For example, COX-2 protein expression in the brain and spinal cord is upregulated by seizure activity and peripheral inflammation, respectively (Yamagata et al., 1993, Beiche et al., 1996, Samad et al., 2001). This activity-dependent up-regulation of COX-2 expression is dependent on NMDA receptor activity (Yamagata et al., 1993, Adams et al., 1996). Moreover, some studies suggest that this may result in part from an increase in NF-κB-dependent COX-2 gene transcription (Lee et al., 2004, Tegeder et al., 2004). However, additional studies are required to unequivocally determine the molecular determinants of constitutive and inducible COX-2 expression in neurons.

In addition to differences in expression, COX isozymes exhibit subtle structural differences that influence protein stability. In this regard, COX-1 protein is more susceptible to proteolytic inactivation than COX-2 (Guo et al., 1997) and this may contribute to the exceptionally short half-life of COX-1 relative to COX-2 (Fagan & Goldberg, 1986, Shao et al., 2000). Structural differences also appear to account for differences in substrate utilization. Notably, COX-2 but not COX-1 can metabolize the endocannabinoids, anandamide and 2-arachidonoylglycerol (Yu et al., 1997, Kozak et al., 2000), generating ethanolamine and glycerol conjugated prostanoids, respectively (Kozak et al., 2002). Depletion of endocannabinoids by COX-2 appears to be a functionally important metabolic pathway in neurons, restricting endocannabinoid action in certain regions of the CNS (Kim & Alger, 2004, Slanina et al., 2005, Slanina & Schweitzer, 2005). To what extent, if any, the conjugated prostanoid products perform novel functions within the CNS remains to be determined.

Acetyl salicylic acid (aspirin) had been employed therapeutically for its analgesic and anti-inflammatory properties long before its mechanism of action was known. It was not until the 1970s that it was found to inhibit PG production (Ferreira et al., 1971, Smith & Willis, 1971, Vane, 1971) by acetylating COX (Roth et al., 1975, Hemler & Lands, 1976, Roth et al., 1977). It subsequently became clear that most of the therapeutic properties, as well as adverse actions, of an entire class of important drugs termed non-steroidal anti-inflammatory drugs (NSAIDs) could be ascribed to inhibition of COX (Moncada & Vane, 1979). NSAIDs inhibit PGH₂ production by preventing access of AA to the catalytic tyrosine residue of the COX, while having little effect on the activity of the peroxidase function (Rome & Lands, 1975). Many NSAIDs inhibit both COX-1 and COX-2 with little specificity (Meade et al., 1993, Mitchell et al., 1993). In mammals, PGs protect the gastrointestinal tract from irritation by decreasing the amount of gastric acid secretion; by inducing vasodilatation of the blood vessels that feed the gastric mucosa; and by triggering the production of gastric mucus and fluid. Because COX-1 is the enzyme that is primarily responsible for the biosynthesis of PG that mediate these effects, administration of NSAIDs that inhibit COX-1, such as aspirin and ibuprofen, cause unfortunate side effects such as stomach irritation (Whittle & Vane, 1984, Andrews et al., 1994, Kargman et al., 1996). Since COX-2 is preferentially induced in migratory cells by inflammatory stimuli, many pharmaceutical companies designed novel COX-2 selective inhibitors with the idea that the beneficial anti-inflammatory actions of NSAIDs would benefit the patient without causing as much discomfort and side effects (Futaki et al., 1993, Masferrer et al., 1994b, Vane et al., 1998).

The subtle structural differences between COX-1 and -2 have allowed the development of therapeutically relevant small molecule inhibitors that demonstrate a high degree of isoform selectivity (Smith & DeWitt, 1995, Bhattacharyya et al., 1996, Kurumbail et al., 1996). To wit, the amino acid residues that encompass the COX inhibitor-binding site can be classified into 2 groups. The first, termed the inner shell residues, mediates direct contact with the inhibitors. The second group, termed the second shell, also encloses the inhibitor-binding site but does not make direct contact with the inhibitor. Among all of the amino acids that comprise the first shell of both COX-1 and COX-2, only amino acid 523 differs between the 2 enzymes. This position is occupied by valine in COX-2 and isoleucine in COX-1 (Gierse et al., 1996). Because valine is a smaller amino acid than isoleucine, it contributes to the formation of a second pocket that forms an extension off of the inhibitor-binding site. As a result, the size of the binding site in COX-2 is 25% larger than that of COX-1 (Wong et al., 1997). This difference allows inhibitor molecules to gain access to amino acid residue 513 in COX-2 (arginine) that provides an excellent hydrogen-binding site for an inhibitor that is only specific for COX-2 (Kurumbail et al., 1996, Luong et al., 1996). Differences also exist in the amino acids that comprise the second shells of these two enzymes. Phenylalanine 503 in COX-1 directly contacts a conserved inner shell residue (leucine 384) and creates a smaller inhibitor-binding site. In COX-2, this amino acid residue at position 503 is replaced by leucine that is considerably smaller than phenylalanine, resulting in a larger binding site (Gierse et al., 1996, Guo et al., 1996, Wong et al., 1997, Schneider et al., 2002).

These differences in the NSAID binding sites of COX-1 and COX-2 have been exploited for the development of a new generation of COX-2-selective inhibitors, primarily for the treatment of inflammatory conditions. Of these Meloxicam (Mobic™; Boehringer Ingelheim), Nimesulide and Etodolac (Lodine™; Wyeth-Ayerst) were first identified as potent anti-inflammatory drugs with less ulcerogenic activity than traditional NSAID. These preferentially inhibit COX-2 with a COX-2/COX-1 ratio between 0.1 and 0.01 depending on the test system used (Engelhardt et al., 1996). Needelman and his group at Monsanto/Searle made inhibitors with greater selectivity for COX-2 and one of those, Celecoxib (Celebrex™; Pfizer), is currently the only COX-2 selective drug still marketed in the US with FDA approval for the treatment of familial adenomatous polyposis, osteo- and rheumatoid arthritis and acute pain. A second drug, rofecoxib (Vioxx™), developed by Merck-Frost, was being marketed for osteo- and rheumatoid arthritis, acute pain and migraine headache until being voluntarily withdrawn from the market in September 2004 due to safety concerns (see Section 4). Pfizer, at the behest of the FDA, discontinued marketing a second-generation COX-2 selective inhibitor, valdecoxib (Bextra™), in April 2005. Other specific COX-2 inhibitors that are in development include Merck's etoricoxib (Arcoxia™; approved in Europe but not in the US), Novartis' lumiracoxib (Prexige™; on sale in Britain and Australia but not yet approved elsewhere in the European Union or the US), and GlaxoSmithKline's GW406381X (Beswick et al., 2004, Bingham et al., 2005). The latter is a more potent and selective COX-2 inhibitor and exhibits higher brain penetration than either rofecoxib or celecoxib.