Introduction
Prostaglandin endoperoxide synthases or cyclooxygenases (COX-1 and COX-2) play a central role in the inflammatory cascade by converting arachidonic acid (AA), released from membrane phospholipids by a phospholipase A
2 (PLA
2), into prostaglandin endoperoxide H
2, which in turn is converted to bioactive prostanoids by specific terminal synthases. The two COX isoforms share 60% homology in their amino acids sequence and have comparable kinetics; however they also show individual differences. COX-1 is normally constitutively expressed in most tissues and thought to be involved in homeostasis, whereas COX-2 is inducible upon inflammatory and other stimuli [
1]. However, in the central nervous system (CNS), COX-1 and COX-2 are both constitutively expressed and COX-2 is mainly 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 [
2‐
4]. In the CNS, COX-2 has been implicated in important physiological functions such as synaptic transmission, neurotransmitter release, blood flow regulation, and sleep/wake cycle [
5‐
9].
Both COX-1 and COX-2 have been shown to play important roles in an inflammatory response, their contribution being different depending on the type of insult, the time after insult, and the tissue examined [
6,
10]. Because COX-2 is highly inducible by inflammatory stimuli it has been traditionally considered as the most appropriate target for anti-inflammatory drugs [
2,
11]. However, the exact role of each COX isoform in neuroinflammation is unclear. While we have recently reported that genetic deletion or pharmacological inhibition of COX-1 significantly ameliorate the neuroinflammatory response and brain injury following lipopolysaccharide (LPS) treatment [
12], the role of COX-2 in the neuroinflammatory process remains controversial. For instance, COX-2 deficient (COX-2
-/-) mice have been reported to be resistant to the febrile response induced by peripheral injection of LPS [
13]. On the other hand, selective pharmacological inhibition of COX-2, but not of COX-1, increases the expression of several pro-inflammatory genes in the vascular associated cells and the parenchymal microglia after systemic injection of LPS [
14].
In this study we examined the neuroinflammatory response of COX-2
-/- and wild type (COX-2
+/+) mice to intracerebroventricular (icv) injection of LPS, which is a model of direct activation of brain innate immunity [
15‐
19].
LPS, a component of the outer cell wall of gram-negative bacteria, mediates its effect through the CD14 receptor, a glycosylphosphatidylinositol-linked membrane protein that is present on microglial cells. The LPS-CD14 complex, together with other adaptor proteins, binds to the toll-like receptor 4 (TLR4), which is present on microglia, but not on astrocytes, oligodendrocytes or cortical neurons [
20]. This initiates a bifurcated signal transduction cascade that leads to the transcription of inflammatory and immune response genes, primarily
via nuclear factor- κB (NF-κB) activation but also through c-Fos/c-Jun and Janus kinase (JAK)/signal transducer and activator of transcription 3 (STAT3)-dependent pathways [
21]. The signaling events ultimately lead to the production of free radicals generated by NADPH oxidase, myeloperoxidase and inducible nitric oxide synthase (iNOS) in combination with cytokines and chemokines [
22,
23], which are mediators of the LPS-induced injury [
15,
16,
24]. In this regard, previous data suggest that interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF-α) can contribute to neuronal death in models of acute CNS injury as well as in chronic neurodegenerative disease [
25].
In this study, we demonstrate that COX-2-/- mice are more susceptible than COX-2+/+ mice to LPS-induced neuronal injury and exhibit an increase in microglia and astrocyte activation, and increases in the expression of genes and proteins for inflammatory cytokines, chemokines, reactive oxygen species-generating enzymes, such as iNOS and NADPH oxidase, and in the expression of STAT3 and suppressor of cytokine signaling 3 (SOCS3) signaling molecules. COX-2+/+ mice chronically treated with celecoxib, a COX-2 selective inhibitor, also exhibit an increased neuroinflammatory response compared to untreated wild-type mice.
Discussion
In this study we demonstrate for the first time that genetic deletion of COX-2 enhanced the neuroinflammatory response and increased the susceptibility to neuronal damage induced by centrally injected LPS. We also showed that chronic treatment with a selective COX-2 inhibitor, celecoxib, also increases LPS-induced protein levels of IL-1β, a major proinflammatory cytokine, of phosphorylated STAT3, a transcription factor involved in the progression of the inflammatory cascade, and of NADPH oxidase subunit p67
phox, a marker of oxidative stress. We have previously demonstrated that this chronic dosing paradigm of celecoxib (6000 ppm for 6 weeks) leads to a plasma concentration of 18.2 ± 5.8 μg/ml [
28]. Assuming 98% binding of celecoxib to plasma proteins and that only free celecoxib can cross the blood brain barrier [
38], brain concentration of celecoxib is approximately 640 nM, well above the IC
50 (39 nM) of celecoxib for COX-2 [
28]. These plasma concentrations are within the same order of magnitude of steady state concentrations (2–3 μg/ml) observed in humans after acute administration of 400–800 mg of celecoxib, doses clinically used for the treatment of rheumatoid arthritis and familial adenomatous polyposis [
39].
In this study, FJB-positive neurons were only observed in the COX-2
-/- mice, suggesting that COX-2 deletion increases the susceptibility to LPS-induced neurodegeneration. There is a conflicting view about the role of COX-2 in neurodegeneration and neurotoxicity [
40]. For instance, COX-2 inhibition is believed to be neuroprotective in models such as MPTP (1-methyl 4-phenyl 1,2,3,6-tetrahydropyridine), quisquallic acid induced damage [
41,
42], and centrally injected NMDA-induced neurotoxicity [
43,
44]. However, in these studies, where the toxins directly damage neurons, COX-2-mediated cytotoxicity does not appear to be linked to the inflammatory response [
41]. On the other hand, pre-treatment with COX-2 inhibitors or genetic deletion of COX-2 has been shown to increase seizure activity and neuronal damage in response to kainate [
28,
45], and to exacerbate endotoxin-induced ocular inflammation [
46] and tissue damage in ConA- and acetaminophen-induced hepatotoxicity [
47,
48]. Another study reported, in support of our observations, that selective pharmacological inhibition of COX-2 with NS-398 increases the transcription of inflammatory genes (mPGES-1, TLR2, CD14, MCP-1) in vascular associated brain cells and parenchymal microglia after systemic injection of LPS [
14]. While these conflicting data highlight the importance of investigating the distinct roles of COX-1 and COX-2 in physiology and pathology, our findings suggest that COX-2-derived products selectively mediate a protective effect in the development and/or the resolution of inflammation in the brain after endotoxin activation of the innate immune system. In this regard, a recent review emphasizes that COX-2 mediates neuroprotection
via specific anti-inflammatory lipid mediators [
49]. Furthermore, Gilroy and colleagues demonstrated that selective COX-2 inhibitors, by blocking the production of PGE
2 and PGD
2, disturbed the resolution phase of inflammation, leading to delay in return to homeostasis [
50].
COX-1 protein levels were not significantly changed by LPS in either COX-2
-/- or celecoxib-treated mice compared to COX-2
+/+ mice, indicating that the increased neuroinflammatory response was not due to an increased compensatory expression of COX-1 in response to LPS when COX-2 is either genetically abrogated or pharmacologically inhibited. Increases in microglial activation and in the induction of cytokines and chemokines in the COX-2
-/- mice could contribute to the susceptibility to LPS-induced damage. Overexpression of chemokines, small pleiotropic chemoattractant cytokines that promote leukocytes activation and migration, has been recently implicated in many neurological disorders including multiple sclerosis, and Alzheimer's disease [
51,
52]. The overexpression of chemokines observed in the COX-2
-/- mice after LPS may increase the leukocytes and monocytes recruitment in the inflamed brain and cause neuronal damages, in the absence of a "switch off" mechanism. The increased expression of cytokines could be due to the incapacity of the tissue to resolve the inflammation, leading to a persistent activation of the inflammatory cascade. One possibility is that COX-2 deletion or inhibition leads to a reduction in anti-inflammatory mediators or neurotrophic factors, which would impair the brain ability to resolve the inflammation.
iNOS and NADPH oxidase may also contribute to microglia-mediated LPS induced neurotoxicity by increasing the production of extracellular reactive oxygen and nitrogen species, which, in turn, stimulate the microglial release of pro-inflammatory mediators that, like radical oxygen species, are toxic to neurons [
53,
54]. In this regard, NADPH inhibitors suppress LPS-induced expression of iNOS, IL-6, IL-1β and TNF-α in glial cells
in vitro [
55] and NADPH oxidase has been shown to regulate COX-2 mediated PGE
2 production in cultured microglia [
56].
The JAK/STAT pathway is a key player in the intracellular response to cytokines. SOCS3 is a potent inhibitor of the JAK/STAT signaling cascade, negatively regulating signal transduction pathways mediated by a variety of cytokines. SOCS3 has been suggested to play a critical role in integrating the neuroimmunoendocrine circuits [
57]. Although NF-κB p65 expression were similar in COX-2
+/+ and COX-2
-/- mice after LPS, we found that the mRNA expression of STAT3 and the levels of phosphorylated STAT-3 were significantly higher in the COX-2
-/- mice compared to wild type mice. SOCS3 was also upregulated in the COX-2
-/- mice compared to COX-2
+/+ mice after LPS. SOCS3 is a negative modulator of inflammatory cytokine signaling [
58] and can be induced by inflammatory stimuli such as LPS, TNF-α and IL-6 [
58,
59]. SOCS3 mRNA up-regulation in the COX-2 deficient mice can thus be viewed as the consequence of the higher cytokine production in these mice after LPS. In this regard, SOCS3 overexpression has been shown to lead to neuroblastoma cell death [
60]. Overall, our data indicate a dysregulation of the cytokine signaling pathway in the COX-2
-/- mice, which may mediate the increased neuroinflammatory response.
While independent epidemiological studies indicate that non steroidal anti-inflammatory drugs (NSAIDs) administration prevents or delays the onset and risk of developing Alzheimer's disease [
61‐
63], clinical trials using COX-2 selective inhibitors in patients with mild to severe cognitive impairment, have been unsuccessful to date [
64‐
67], with the exception of a small double blind, placebo-controlled study with indomethacin, a preferential COX-1 inhibitor [
68]. We have recently demonstrated that genetic deletion or pharmacological inhibition of COX-1 significantly attenuates glial cells activation and the neuroinflammatory response, oxidative stress and neuronal damage in response to icv injected LPS [
12]. Our results show that while COX-1 selective inhibition may be beneficial, selective inhibition of COX-2 appears not to be beneficial in neurodegenerative diseases with a marked inflammatory component and may explain the failure of selective COX-2 inhibitors to protect AD patients from cognitive decline in clinical trials [
64‐
67].
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SA performed the experiments, contributed to study design and data analysis and wrote the manuscript. RL provided the COX-2+/+ and COX-2-/-mice and reviewed the manuscript. FB directed the work, contributed to study design, reviewed the data, and wrote the manuscript. All authors read and approved the final manuscript.