Selective COX-2 inhibition prevents progressive dopamine neuron degeneration in a rat model of Parkinson's disease

The role of microglia in the pathogenesis of neurodegenerative disorders is not clear [1]. Increasing evidence suggests that an inflammatory reaction accompanies the pathological processes seen in many neurodegenerative disorders, including Parkinson's disease (PD) [2‐4]. Glial activation is part of a defense mechanism to remove debris and pathogens and promote tissue repair. However, inflammatory activation of microglial cells may contribute to the neurodegenerative process through structural invasion and the release of pro-inflammatory cytokines, reactive oxygen species (ROS), nitric oxide (NO) and excitatory amino acids at synapses and cell bodies. In cell culture and animal models, inflammation contributes to neuronal damage, and anti-inflammatory drugs have been shown to provide some neuroprotection in different paradigms [5‐7] including PD models [8, 9]. Reactive microglia inhibit neuronal cell respiration via NO and cause neuronal cell death in vitro [10] and in vivo [11]. Interestingly, microglial cell activation by chronic infusion of lipopolysaccharide (LPS) appears to be capable of inducing a selective degeneration of nigral dopamine (DA) neurons [11]. Intranigral injection of LPS, but not of cytokines, induces DA degeneration [11, 12]. LPS induces NO production and release from microglia and also release of pro-inflammatory cytokines such as IL-1β and TNF-α, which may also participate in cytotoxicity [13].

In PD there is evidence of an increase in oxidative and inflammatory nigral environment [2, 14‐16]that includes the presence of cyclooxygenase (COX)-immunoreactive activated microglial cells in the substantia nigra (SN) [17], elevated levels of TNF-α and other pro-inflammatory cytokines in the cerebrospinal fluid (CSF) [18, 19]. DA neurons in the SN express TNF-α receptor 1 [3] which may contribute to the selective susceptibility of DA neurons to microglial toxicity. Supporting a role of inflammation in DA degeneration, mice deficient in TNF-α receptors are resistant to selective DA toxins [20]. In PD patients, a polymorphism in the TNF-α gene, leading to high production of TNF-α, was found to be more frequent than in matched healthy controls and to be related to earlier onset of the disease [21]. In addition, the results of a recent epidemiological study suggest that nonsteroidal anti-inflammatory drugs (NSAID) might delay or prevent onset of PD [22]. NSAIDs target p38 mitogen-activated protein kinase (MAPK) in addition to their main target COX [23] and inhibition of p38MAPK phosphorylation blocks NO release from activated microglial cells [24].

Selective COX-2 inhibitors lack the adverse effects of conventional NSAIDs, which inhibit both isoforms of COX (constitutive and inducible). COX-2 is induced by pro-inflammatory stimuli and cytokines [25]. Inhibition of the inducible form (COX-2) accounts largely for the therapeutic (anti-inflammatory) actions of NSAIDs whereas inhibition of the constitutively expressed form (COX-1) is responsible for the gastrointestinal side effects [25].

In this study we used the intrastriatal administration of 6-OHDA in the rat to evaluate the protective effect of selective COX-2 inhibition by celecoxib. Like other toxic and genetic models, this model has limitations, but it provides a time window to test neuroprotective strategies, as DA neurons die retrogradely over the course of several weeks [26‐29]. We have previously shown that, in this model, DA cell death is accompanied by microglial cell activation [28].

Following the 6-OHDA intrastriatal infusion, rats received either celecoxib at 20 mg/kg (COXIB group, N = 12) or vehicle (N = 13) by oral gavage daily until the time of sacrifice at 12 (n = 4/5) or 21(n = 8/8) days post surgery. First, we examined in vivo the effect of celecoxib compared to vehicle using micro PET and ¹¹C PK11195, a peripheral benzodiazepine receptor ligand that binds to microglia [28] and ¹¹C CFT, a cocaine analog that binds to the DAT. At 12 days we observed a decrease in CFT binding (~60% of contralateral BP) in the 6-OHDA lesioned striatum of both experimental groups (Fig. 1A). No striatal ¹¹C PK11195 binding was present in the COXIB group (n = 3) while in the vehicle group (n = 2) there was binding ipsilateral to the 6-OHDA lesion, as described previously. No differences between COXIB (n = 3) and vehicle (n = 3) were observed at 21 days at the striatal level (Fig. 1A).

Next we examined the effect of celecoxib on the DA system. We analyzed the striatal volume of TH+ fibers (Fig. 1B,1C,1D) and quantified the number of TH+ cell bodies in the SN (Fig. 1E). At 12 days there was a severe (~85%) decrease in TH immunoreactive fibers in the striatum and less marked loss of TH+ cells (~40%) in the SN, and there were no differences between groups in either of these measures (Fig 1B,1D). However, at 21 days animals in the COXIB group displayed significantly larger volumes of DA terminal fibers in the striatal areas (> 50%, t = 7.8, p = 0.01) and corresponding larger number of TH+ cell bodies in the SN (t = 5, p < 0.05, Fig. 1B,1C,1D). In the vehicle group there was a significant progression of TH+ cell loss from 12 to 21 days (t = 3.5, p < 0.01) (Fig. 1E), which is consistent with previous studies [26‐29]. This progression did not occur in the COXIB group (p = 0.7). In addition to the treatment effect, there was a significant recovery of striatal fiber density at 21 days in both groups. At this time point, the striatal TH fiber volume was directly correlated with the number of remaining DA cells in the SN (p < 0.01).

Using a selective antibody directed against CD11b (C3R), we studied activated microglia by morphological analysis and performed stereological quantification in the striatum (Fig. 2) and ventral midbrain (Fig. 3). In the COXIB group, a significant reduction in the number of activated microglia was seen in both striata (treatment effect F_1,32 = 7, p = 0.01). This effect of selective COX-2 inhibition was more pronounced in the striatum at the 12-day time point after the toxin injection than at 21 days after 6-OHDA (Fig 2A,2B,2C). In addition to a significant increase in cell density, in the vehicle group the predominant morphology of microglial cells was amoeboid (activated) as opposed to ramified (resting) (Fig 2B and 2D). In the ventral midbrain microglial cell density was significantly higher ipsilateral to the 6-OHDA injection in the vehicle group at 21 days. ANOVA revealed a significant effect of time (F_1,24 = 809, p < 0.001), lesion side (F_1,24 = 17. p < 0.001) and treatment (F_1,24 = 11.6, p < 0.01) on microglial density (Fig 3E,3F). Astrocytes immuno-labeled with GFAP were also quantified in the striatum. There was a significant astrogliosis ipsilateral to the 6-OHDA injection (F_1,28 = 28, p < 0.001, Fig 4). This lesion effect decreased but was still noticeable at 21 days in the vehicle group. Interestingly, no effect of treatment (p = 0.9) was observed for astrocyte density (GFAP+ cells/mm³) (Fig 4E,4F). These results show that celecoxib produced a selective reduction of local microglial reaction in response to the neurotoxin.

Regular intake of nonaspirin NSAIDs (and high dose of aspirin) has been reported to be associated to a 45% lower risk of PD in 2 large cohorts [22]. In this study we found that COX-2 inhibition by celecoxib decreased microglial activation and was associated with a prevention of the progressive degeneration seen in the 6-OHDA retrograde lesion model of PD. We used the striatal administration of 6-OHDA because neuronal damage and death, characteristically progress over weeks. This provides a close (although accelerated) model for the cascade of degenerative events that occurs in PD. Between weeks 2 and 3, DA cell loss progressed from 40 to 65 % in vehicle treated animals, as previously described for this model [26, 27, 29, 38]. In contrast, we did not observe such a progression of DA neuronal cell loss in the COXIB group. It is worth noting that TH striatal fiber density was not correlated (more extensive) with DA cell numbers at 12 days, likely reflecting TH down-regulation in the acute stage of degeneration. Consistent with this explanation, there was a significant recovery of TH fibers in both groups at 21 days (Fig 1B,1C,1D) to levels that matched and corresponded to the number of DA neurons remaining in the substantia nigra. Based on this temporal pattern we propose that COX-2 inhibition protects a nigral neuronal cell population with reversible damage [15, 39, 40]. Such neurons are damaged and have reduced axonal TH expression at 2 weeks. Left to the natural evolution of the progressive degeneration half of these DA neurons will eventually die [26‐29]. Our results show that COX-2 inhibition resulted in a complete protection of these damaged DA neurons. This effect can be dependent on specific intraneuronal effects of celecoxib and/or related to a classical anti-inflammatory mechanism, through microglial cell inhibition. Interestingly, the reduction of microglia activation by celecoxib was stronger at 12 days (Fig. 2E), while at this time point there were no differences in DA markers between groups. The reduction in microglia was not accompanied by changes in astroglial reaction to the striatal injury (Fig. 4).

With infectious or tissue injury stimuli, including inflammatory or selective DA terminal lesions of the striatum, microglia can both proliferate and transform morphologically into reactive forms [28, 41]. The reactive microglia's amoeboid movement and activities in injured neural tissue include macrophage activity and presumed synaptic stripping along dendrites [42]. In the current PD model, microglial invasion and continued presence in the lesioned striatum and substantia nigra could contribute to long-term synaptic disconnection of the damaged DA terminal afferents. Such loss of normal neuron target interactions and trophic support can lead to DA neuronal vulnerability, atrophy or death [43, 44]. In experimental in vivo PD models with delayed DA neuronal death, various exogenous trophic factor support of DA neurons [29, 43‐46] or intracellular signalling related to neuroimmunophilins [30] can prevent long-term progressive DA degeneration to a similar degree to that seen by COX-2 inhibition in the present study. In chronic degenerative situations involving the striatum and midbrain and during marked fluctuations in neuronal ionic, metabolic and functional status, astroglia are thought to play a more homeostatic, trophic and protective role for DA neurons and terminals than microglia [47, 48]. Our evidence clearly demonstrate that selective COX-2 inhibition did not reduce the typical astroglial response to injury in the striatum, while to a large extent preventing expression of the morphologically activated microglial phenotype. In fact, COX-2 inhibition had by 3 weeks of treatment (compared to vehicle) caused a mild elevation of the number of reactive astrocytes in the striatum contralateral to the lesion (Fig. 4). In the period of progressive DA neuronal death between the 2^nd and 3^rd week after DA terminal injury by 6-OHDA, our data indicate that nigro-striatal DA terminals were restored in striatum, and a significant population of DA neurons where spared from cell death in the substantia nigra by COX-2 inhibition. These results point to an altered reactive astroglial to reactive microglial cell ratio by COX-2 inhibition that may provide insights for how to create favorable conditions for prevention of progressive neurodegenerative cascades during and after neuronal injury similar to that seen in PD.

Microglial cells can also produce and release pro-inflammatory cytokines, in particular TNFα, and cytotoxic molecules including ROS and NO [34] although such responses are non-specific to lesion type [41]. After 6-OHDA intrastriatal infusion, there is an acute increase in TNF-α in the striatum [49]. Pro-inflammatory cytokines IL-1β and TNF-α activate the p38MAPK cascade and NFkB translocation to the nucleus, resulting in transcriptional upregulation of COX-2. TNF-α activates COX-2 via the JNK pathway [50] and induction of NFkB [51]. Importantly p38 MAPK stabilizes the mRNA of COX-2 and other pro-inflammatory factors [52, 53]. Activated microglia cells release NO [54, 55] and superoxide free radical [11]. DA neurons are particularly vulnerable to this type of inflammation induced oxidative stress as DA metabolism and DA autoxidation generate ROS [56]. Celecoxib (at low dose) [57, 58] and other NSAIDs (and minocycline) inhibit p38MAPK leading to a decrease in COX-2 production, decreased mRNA stability and decreased PGE₂ release. It is possible that celecoxib, by blocking COX-2 enzymatic activity and by inhibition of the p38MAPK pathway, constrained the inflammatory response induced by striatal 6-OHDA thus limiting, to a certain extent, the progressive DA neuronal death. Similar protective effects of selective COX-2 inhibitors have been reported in excitotoxicity and ischemia models [7, 59‐61].

The benefit we observed can also stem directly from neuronal inhibition of COX-2, which is one of several enzymes capable of oxidizing DA to reactive DA quinone [62]. DA quinones can deplete the cells of antioxidants, inactivate enzymes and increase α-synuclein protofibrils [56, 63]. Induction of COX-2 results in an inflammatory cascade accompanied by formation of ROS. Recent work using the mouse MPTP model of PD suggests that an intraneuronal mechanism can be sufficient to achieve neuroprotection in that specific acute paradigm [64, 65]. Therefore the reported effects could be attributed to a direct decrease of inflammatory mediators inside the neuron or to inhibition of release of proinflammatory and toxic factors from microglia [24, 40]. The temporal relationship between glial activation and neurodegeneration suggests that microglial activation plays a key role in amplifying the toxic effect and thereby exacerbating DA cell loss, although it cannot be determined by these experiments whether microglia inhibition is absolutely required to achieve neuroprotection. Massive and prolonged microglial cell activation has been observed in aged mice exposed to MPTP, associated with a progressive loss of TH+ neurons [66]. Sugama et al. and our data strongly suggest that microglia activation prolongs an oxidative environment after the initial toxic insult, leading to the subsequent loss of neurons that have a reversible damage [39, 40]. Specifically, in the retrograde 6-OHDA lesion paradigm that we used here, ~50% of the DA neurons with reversible damage will die between weeks 2 and 3 after the initial injury. In this study celecoxib treatment rescued this population. Therefore inhibition of COX-2 by celecoxib appears to be able, directly, and through inhibition of microglia activation to result in a reduction of DA cell degeneration.

The presence of activated microglia in the brain of PD patients [2] and after MPTP exposure both in humans [67]and monkeys [4] supports the existence of an ongoing inflammatory process that can contribute to the progression of the disease. The origin of neuroinflammation is unknown and is probably different for different individuals, being a common response to a variety of pathogenic insults. If indeed chronic neuroinflammation contributes to the progression of the degenerative process [15], anti-inflammatory drugs could prevent or slow down the disease, independently of the causative factors.