Cyclooxygenase-2 contributes to oxidopamine-mediated neuronal inflammation and injury via the prostaglandin E2 receptor EP2 subtype

Cyclooxygenase-2 (COX-2) triggers pro-inflammatory processes that can aggravate neuronal degeneration and functional impairments in many neurological conditions, mainly via producing prostaglandin E2 (PGE2) that activates four membrane receptors, EP1-EP4. However, which EP receptor is the culprit of COX-2/PGE2-mediated neuronal inflammation and degeneration remains largely unclear and presumably depends on the insult types and responding components. Herein, we demonstrated that COX-2 was induced and showed nuclear translocation in two neuronal cell lines – mouse Neuro-2a and human SH-SY5Y – after treatment with neurotoxin 6-hydroxydopamine (6-OHDA), leading to the biosynthesis of PGE2 and upregulation of pro-inflammatory cytokine interleukin-1β. Inhibiting COX-2 or microsomal prostaglandin E synthase-1 suppressed the 6-OHDA-triggered PGE2 production in these cells. Treatment with PGE2 or EP2 selective agonist butaprost, but not EP4 agonist CAY10598, increased cAMP response in both cell lines. PGE2-initiated cAMP production in these cells was blocked by our recently developed novel selective EP2 antagonists – TG4-155 and TG6-10-1, but not by EP4 selective antagonist GW627368X. The 6-OHDA-promoted cytotoxicity was largely blocked by TG4-155, TG6-10-1 or COX-2 selective inhibitor celecoxib, but not by GW627368X. Our results suggest that PGE2 receptor EP2 is a key mediator of COX-2 activity-initiated cAMP signaling in Neuro-2a and SH-SY5Y cells following 6-OHDA treatment, and contributes to oxidopamine-mediated neurotoxicity.


Results
Neuro-2a and SH-SY5Y cells are TH positive and susceptible to 6-OHDA-mediated cytotoxicity. Neuro-2a is a mouse neuroblastoma cell line derived from neural crest with many features of neurons, including neurofilaments 28 ; whereas SH-SY5Y is a human originated cell line that was initially isolated from a bone marrow biopsy removed from a four-year-old girl with neuroblastoma 29 . Because of their neuronal background and neuron-like properties, these two cell lines have been widely used as in vitro models to study neuronal function and differentiation, axonal growth, neuronal signaling, neurotoxicity, and neurodegeneration, particularly in Parkinson's disease (PD) [30][31][32] . We purchased both cell lines directly from the American Type Culture Collection (ATCC), and first examined their neuronal background by immunochemistry. As shown in Fig. 1, both cultured Neuro-2a and SH-SY5Y cells widely expressed NeuN -a canonical neuronal biomarker, and tyrosine hydroxylase (TH) -the enzyme responsible for the first step of the dopamine synthesis in the SNpc neurons. In fact, the vast majority of cells -82.1% Neuro-2a and 73.1% SH-SY5Y cells -were identified both NeuN and TH positive.
As a synthetic oxidopamine, 6-OHDA is a widely used neurotoxin to bleach dopaminergic neurons and destruct the nigrostriatal pathway in rodent PD models [21][22][23] . Consistent with previous findings 25, 31 , 6-OHDA was shown here to decrease the viability of both Neuro-2a cells ( Fig. 2A) and SH-SY5Y cells (Fig. 2B) in a concentration-dependent manner with similar EC 50 values (~110 µM), measured by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) reduction assay 24 hr after treatment. It appears that incubation for an additional day did not further increase the 6-OHDA-mediated cytotoxicity in these two cell lines ( Fig. 2A and B); therefore, we used up to 24 hr treatment for 6-OHDA throughout the rest of this study.

6-OHDA induces COX-2 expression and nuclear translocation.
The molecular mechanisms whereby 6-OHDA destroys dopaminergic neurons remain elusive, but it has been proposed that the neurotoxin enters the cells via dopamine transporter where it promotes chronic inflammatory processes and oxidative stress 21,33,34 . As a primary pro-inflammatory mediator, COX-2 is well known for its pathogenic roles in many chronic inflammation-associated conditions. To investigate the effect of 6-OHDA on COX-2 expression, we next measured COX-2 mRNA levels in the neurotoxin-treated cells using qPCR. We found that COX-2 was quickly and robustly induced in a time-dependent manner by 6-OHDA at both a low concentration (75 µM, ~EC 25 ) and a high concentration (150 µM, ~EC 75 ) in Neuro-2a cells (P < 0.05 and P < 0.01 for 24-hr incubation with 6-OHDA at 75 µM and 150 µM, respectively, Fig. 3A) and SH-SY5Y cells (P < 0.001 and P < 0.05 for 24-hr incubation with 75 µM and150 µM 6-OHDA, respectively, Fig. 3B). Immunochemistry further revealed that COX-2 was present at basal levels mainly in the cytosol of both Neuro-2a cells (Fig. 4A) and SH-SY5Y cells (Fig. 4B), where it synthesizes homeostatic prostanoids that are essential for many normal physiological functions. With 6-OHDA stimulation [75 µM for Neuro-2a cells and 150 µM for SH-SY5Y cells; concentrations were chosen based on their higher COX-2 induction ( Fig. 3A and B)], COX-2 was remarkably induced in both the nucleus and cytosol of Neuro-2a and SH-SY5Y cells (P < 0.05, Fig. 4C). Interestingly, 6-OHDA-treated cells showed higher percentages of COX-2 expression in the nuclei when compared to vehicle-treated cells (P < 0.001, Fig. 4D), suggestive of a nuclear translocation of COX-2 triggered by 6-OHDA stimulation. Taken together, these results demonstrate that 6-OHDA treatment induces COX-2 activation characterized by expression induction and nuclear translocation in both Neuro-2a and SH-SY5Y cells.

6-OHDA treatment causes PGE 2 biosynthesis and pro-inflammatory cytokine production.
To investigate the involvement of COX-2 in 6-OHDA-mediated inflammation and neurotoxicity, we next quantified the prostaglandins secreted into the culture medium from the neurotoxin-treated cells using ELISA. We found that the 6-OHDA stimulation substantially increased PGE 2 in the culture medium by nearly 5-fold in Neuro-2a cells (P < 0.01 with 75 µM 6-OHDA incubation for 24 hr, Fig. 5A left) and 3-fold in SH-SY5Y cells (P < 0.01 with 150 µM 6-OHDA incubation for 24 hr, Fig. 5A right). Moreover, 6-OHDA treatment significantly upregulated the pro-inflammatory cytokine -interleukin-1β (IL-1β) within Neuro-2a cells (P < 0.05 with 75 µM 6-OHDA incubation for 6 hr; P < 0.001 with 150 µM 6-OHDA incubation for 24 hr, Fig. 5B left) and SH-SY5Y cells (P < 0.05 with 150 µM 6-OHDA incubation for 6 hr, Fig. 5B right), measured by qPCR for its mRNA levels. Immunostaining was performed to show the expression of NeuN (red) and TH (green) in Neuro-2a cells (A) and SH-SY5Y cells (B). The cell nuclei were visualized by DAPI staining (blue) and cells were counted in 5 random fields at a magnification of 200x. Data were reported as the percentage of cells/field. Note that the majority of cells are both NeuN and TH positive. The size of SH-SY5Y cells on average is smaller than that of Neuro-2a cells, which together with their slower growth rate might explain the lower cell density of SH-SY5Y cells. Scale bar = 100 µm.

PGE 2 mediates cAMP signaling via EP2 receptor in Neuro-2a and SH-SY5Y cells. PGE 2 has four
receptors EP1-EP4, among which EP2 and EP4 are Gα s -coupled and mediate cAMP-dependent pathways that are involved in PGE 2 -mediated chronic inflammation and neurodegeneration 1, 2, 12 . To investigate the PGE 2 /cAMP signaling in Neuro-2a cells, we treated the cells with increasing concentrations of PGE 2 , EP2 selective agonist butaprost, or EP4 selective agonist CAY10598. Forskolin, a direct activator of the adenylyl cyclase, was used to show the maximal capability of the cells to synthesize cAMP. The cytosol cAMP levels were measured by a time-resolved fluorescence energy transfer (TR-FRET) assay. As shown in the Fig. 6A, both PGE 2 and Butaprost induced cAMP accumulation in Neuro-2a cells in a concentration-dependent manner (P < 0.01 for 10 µM PGE 2 , 1 or 10 µM butaprost). Particularly, 10 µM butaprost nearly peaked the cAMP levels in these cells when compared to positive treatment by 100 µM forskolin (P < 0.01). On the contrary, EP4 selective agonist CAY10598 merely slightly increased the cAMP levels in these cells at 10 µM (Fig. 6A), a concentration that should be able to fully activate the EP4 receptor 37 .
Using high-throughput screening (HTS) and subsequent medicinal chemistry for hit optimization, we developed a series of novel small-molecule compounds that are among the first-generation EP2 selective antagonists [38][39][40][41][42][43][44] . Compound TG6-10-1 was created by introducing a trifluoromethyl group in the methylindole ring of an HTS hit compound TG4-155, aiming to improve its pharmacodynamic and pharmacokinetic properties (Fig. 6B) 38 . Development of these novel compounds provides unprecedented opportunities to study this receptor through an inhibitory strategy 45 . Both compounds TG4-155 and TG6-10-1 showed robust inhibition on EP2 receptor in butaprost-treated Neuro-2a cells, demonstrated by an up to 10-fold rightward shift of butaprost dose-response curve when either compound (10 µM) was present (Fig. 6C).
It appears that 10 µM TG4-155 showed remarkable inhibition on EP2-mediated cAMP production in these cells (Figs 6C and 7B), but merely moderately reduced the 6-OHDA-induced cytotoxicity ( Fig. 8A and B). It should be noted that the cell-based cAMP assay was carried out in HBSS system requiring TG4-155 incubation for less than 1 hr, whereas the MTT cytotoxicity assay was performed 24 hr after the compound was added into the culture medium containing 10% FBS. It is likely that TG4-155 was more stable in HBSS than in culture medium, and thus showed more robust effect in the cAMP assay. Furthermore, TG6-10-1 showed slightly higher neuroprotection and inhibition on PGE 2 synthesis than TG4-155 in both 6-OHDA-treated Neuro-2a and SH-SY5Y cells (Fig. 8A-C), although its potency of inhibiting EP2 receptor was considerably lower than that of TG4-155 (Figs 6C and 7B). This incongruity might also be explained by the structural stability of these two compounds. TG4-155 showed instant half-life (<10 min) in mouse or human liver microsomes, suggestive of its metabolic instability in vitro 38 . In addition, TG4-155 had a relatively short plasma half-life (~0.6 hr) after systemic administration in mice 38 . With the introduction of fluorine (Fig. 6B), TG6-10-1 showed improved in vitro metabolic stability 38 , as well as longer plasma half-life in mice (~1.6-1.8 hr) 39,41,42 . Owing to its more stable SCIentIfIC RepoRts | 7: 9459 | DOI:10.1038/s41598-017-09528-z chemical structure, TG6-10-1 might act on EP2 receptor longer than TG4-155 in the culture systems, thereby causing more protection of the 6-OHDA-treated cells and better inhibition on their PGE 2 production, although both compounds were added into the culture medium 24 hr prior to the MTT assay. Nevertheless, these data together suggest that PGE 2 signaling via EP2 receptor, but not EP4, contributes to 6-OHDA-induced cytotoxicity and PGE 2 induction in Neuro-2a and SH-SY5Y cells.

Discussion
In the present study, we demonstrated that stimulation with 6-OHDA rapidly caused COX-2 activation -expression upregulation and nuclear translocation -in both mouse Neuro-2a and human SH-SY5Y cells, leading to pro-inflammatory reaction characterized by PGE 2 synthesis and cytokine (IL-1β) induction. We also found that PGE 2 mediated cAMP-dependent pathways in these cells mainly through EP2 receptor subtype. The finding that 6-OHDA-induced cytotoxicity in these NeuN-and TH-positive cells was largely blocked by pharmacological inhibition on EP2 receptor, but not EP4 receptor, sheds light on the COX-2 downstream signaling pathway that is involved in the neurotoxin-mediated neuronal inflammation and injury.
6-OHDA is a commonly used neurotoxin to induce the neuronal loss in animal PD models; however, the molecular mechanisms underlying the progressive death of dopaminergic cells in the SNpc are not fully understood. Here we presented evidence that COX-2 is induced by 6-OHDA in both mouse and human neuronal cell lines that express TH (Fig. 3A and B), leading to PGE 2 synthesis ( Fig. 5A and C) and IL-1β secretion (Fig. 5B). Therefore, 6-OHDA-provoked inflammatory processes in these cells are -at least partially -mediated by COX-2 and prostaglandin cascade. Upregulated COX-2 has been reported in many neurological disorders including stroke 6 , epilepsy 39,47,48 , and neurodegenerative diseases 49,50 , and has been proposed to essentially contribute to the neuronal injury via initiating pro-inflammatory processes and imposing oxidative stress on the neurons 51,52 . COX-2 is selectively induced in SNpc dopaminergic neurons in 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-administered mice 53 , and in 6-OHDA-treated rats 54 . The genetic ablation or pharmacological inhibition of COX-2 is protective for the TH neurons and relieves the oxidative stress in these animals 53,55 . Epidemiological studies over the past decades also suggest that the chronic use of non-steroidal anti-inflammatory drugs (NSAIDs) might be associated with reduced incidence of PD [56][57][58] . These preclinical and clinical data suggest an essential role for COX-2 in the neurodegenerative process of PD. However, the conception of COX-2 as a therapeutic target has been dampened over the past decade due to the growing recognition of adverse effects of several COX-2 inhibitor drugs on the cardiovascular and cerebrovascular systems 59 , inspiring us that the downstream prostaglandin signaling pathways might provide alternative targets for therapeutics with more specificity.
In response to external stimuli such as injury, infection and inflammation, COX-2 converts the cell membrane-released arachidonic acids to PGH 2 in the cytoplasm where the unstable intermediate prostaglandins are further catalyzed to prostanoids by tissue-specific isomerases 1 . Unexpectedly, we found translocation of COX-2 from the cytoplasm to nucleus in the 6-OHDA-treated Neuro-2a and SH-SY5Y cells (Fig. 4A-D). The translocation of COX-2 between the nucleus and cytosol has been reported in IL-1β-treated vascular endothelial cells 60 , retinal Müller cells after hypoxia 61 , human breast carcinoma 62 , and bladder cancer cells 63 , but not in brain neurons prior to this study. COX-2 activity has been linked to several nuclear receptors and transcription factors such as peroxisome proliferator-activated receptors (PPARs), octamer-binding transcription factor 4 (Oct-4), and nuclear factor-κB (NF-κB) [63][64][65][66] . Whether the nuclear translocation of COX-2 in these two neuroblastoma cell lines would indicate a novel function of COX-2 in the transcriptional regulation of genes that are associated with 6-OHDA-promoted oxidative stress and apoptosis needs further investigation.
EP2 and EP4 are the two currently known PGE 2 receptor subtypes that are coupled to Gα s and mediate cAMP-dependent signaling pathways in both brain neurons and glia 2, 12 . PGE 2 signaling via the EP2 receptor leads to α-synuclein aggregate-mediated neurotoxicity in mouse MPTP model of PD 18 . Moreover, EP2 receptor activation facilitates microglial and astrocytic inflammatory responses to MPTP and causes loss of dopaminergic neurons in the SNpc 19 , suggesting an exacerbating role of EP2 receptor in the pathogenesis of PD. Conversely, pharmacological activation of EP4 receptor has been reported to prevent the MPTP-induced loss of SNpc neurons, whereas the genetic ablation of the receptor aggravated MPTP-associated pro-inflammatory processes 20 . These previously findings together suggest that EP2 and EP4 might function oppositely during the degenerative progression in PD or other neurodegenerative diseases 12, 67-69 . Consistently, we showed that the inhibition on EP2 receptor -but not EP4 -significantly reduced 6-OHDA-triggered neuronal death (Fig. 8A and B). In fact, EP4 receptor inhibition by GW627368X showed a trend of exacerbating the 6-OHDA-mediated cytotoxicity (Fig. 8A).
COX-2 inhibition by celecoxib completely blocked 6-OHDA-induced PGE 2 synthesis in both Neuro-2a and SH-SY5Y cells (Fig. 5C). However, highly COX-2-selective celecoxib -at a relatively high concentration (up to 20 µM) which would completely shut down the enzymatic activity of COX-2 -was only able to block up to 35% cytotoxicity in these 6-OHDA-treated cells (Fig. 8A), suggesting that some COX-2/PGE 2 -independent mechanisms might also be involved in 6-OHDA-mediated neurotoxicity. Nevertheless, the EP2 antagonists afforded a similar level of neuroprotection to that celecoxib provided following 6-OHDA treatment (Fig. 8A), insinuating that the COX-2-mediated injury in these cells can mostly be attributed to EP2 receptor activation by elevated PGE 2 . However, the contribution of Gα q -coupled EP1 and Gα i -coupled EP3 receptors to the 6-OHDA-induced neurotoxicity cannot be excluded in this study, as PGE 2 presumably also activates these two EP receptors following 6-OHDA-triggered COX-2 induction. Also interestingly, EP2 receptor activation by butaprost has been reported to be neuroprotective in rat midbrain cultures containing approximately 5% dopaminergic neurons 70 . These seemingly differing results [see also in N-Methyl-D-aspartate (NMDA)-induced excitotoxicity in rat primary cortical cultures [71][72][73][74] are not unexpected, given that EP2 is expressed in both glial cells and neurons 71,75 , and whether its activation is beneficial or detrimental might be determined by its cell type-specific expression 2 .
SCIentIfIC RepoRts | 7: 9459 | DOI:10.1038/s41598-017-09528-z Therefore, whether these selective EP2 antagonists can provide sufficient beneficial effects on SNpc neurons following MPTP or 6-OHDA treatment in vivo remains an important topic for the future study.

Materials and Methods
Cell culture. To  Technology, #2792), or COX-2 (abcam, #ab15191) overnight, and followed by staining with Alexa Fluor ® 488 or 546-conjugated secondary antibodies (Invitrogen, Carlsbad, CA, USA) for 2 hr and DAPI (10 µg/ml in PBS, Invitrogen) for 10 min. Slides were mounted using DPX mountant (Electron Microscopy Sciences, Hatfield, PA, USA). Images were obtained using EVOS FL Auto Cell Imaging System (Invitrogen). The fluorescence intensity was quantified using ImageJ software developed at the National Institutes of Health (NIH).
Cell viability assay. Cell viability was measured using the Vybrant MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-d iphenyltetrazolium bromide] reduction assay kit (Invitrogen) as previously described 45 . In brief, cells were cultured in 96-well plates (5,000 cells/well). After treatment, MTT was added into each well at a final concentration of 0.5 mg/ml and the cells were incubated at 37 °C for 4 hr. Living cells converted MTT to insoluble formazan, which was dissolved in DMSO. Absorbance of the formazan was measured by a microplate reader (Molecular Devices, Sunnyvale, CA, USA) at 540 nm with a reference wavelength at 630 nm. The dose-response curves were generated and EC 50 values were calculated using OriginPro software (OriginLab, Northampton, MA, USA).
Quantitative PCR. mRNA levels of interested genes were quantified by quantitative PCR (qPCR) as described previously 76,77 . In brief, total RNA was isolated using Trizol (Invitrogen) with the PureLink RNA mini kit (Qiagen, Hilden, Germany) from cultured cells. RNA concentration and purity were measured by a NanoDrop spectrophotometer (Thermo Fisher Scientific). First-strand complementary DNA (cDNA) synthesis was performed with 1 µg of total RNA, using SuperScript III One-Step RT-PCR System (Invitrogen) according to manufacturer's guidelines. 1.5 µl of cDNA was used to set up each 20-µl qPCR reaction mixture following SYBR Green PCR Master Mix manual (Thermo Fisher Scientific). Real-time PCR was performed using the StepOnePlus Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). Cycling conditions were as follows: 95 °C for 2 min followed by 40 cycles of 95 °C for 15 s and then 60 °C for 1 min. Melting curve analysis was used to verify single-species PCR product. Fluorescent data were acquired at the 60 °C step. The cycle thresholds for GAPDH was used as an internal control for relative quantification. Samples without cDNA template served as the negative controls. Primers used for qPCR were listed in Table 1. The mRNA level of an interested gene was normalized to the control in each experiment to limit unwanted sources of variation. PGE 2 measurement. The PGE 2 levels in culture medium were measured by ELISA kit from Arbor Assays (Ann Arbor, MI, USA). In brief, cells were cultured in 24-well plates with 90,000 cells/well for Neuro-2a and
SCIentIfIC RepoRts | 7: 9459 | DOI:10.1038/s41598-017-09528-z 150,000 cells/well for SH-SY5Y cells. After each treatment, 50 µl medium was taken for PGE 2 measurement according to the manufacturer's protocol. The optical density generated from each well was measured in a microplate reader (Molecular Devices) at 450 nm. A standard curve for PGE 2 was run within each experiment.

Cell-based TR-FRET cAMP assay. Cytosol cAMP was measured with a cell-based homogeneous
time-resolved fluorescence resonance energy transfer (TR-FRET) method (Cisbio Bioassays, Codolet, France) as described previously 45,71,75 . The assay is based on generation of a strong FRET signal upon the interaction of two molecules: an anti-cAMP antibody coupled to a FRET donor (cryptate) and cAMP coupled to a FRET acceptor (d2). Endogenous cAMP produced by cells competes with labeled cAMP for binding to the cAMP antibody and thus reduces the FRET signal. Cells were seeded into 384-well plates in 30 µl complete medium (4,000 cells/well) and grown overnight. The medium was carefully withdrawn and 10 µl Hank's Buffered Salt Solution (HBSS) (Hyclone, Logan, Utah, USA) plus 20 µM rolipram was added into the wells to block phosphodiesterase. The cells were incubated at room temperature for 30 min, and then treated with vehicle or tested compound for 5-10 min before addition of EP agonists. The cells were further incubated at room temperature for 40 min, then lysed in 10 µl lysis buffer containing the FRET acceptor cAMP-d2 and 1 min later another 10 µl lysis buffer with anti-cAMP-cryptate was added. After 60-90 min incubation at room temperature, the FRET signal was measured by an Envision 2103 Multilabel Plate Reader (PerkinElmer, Waltham, MA, USA) with excitation at 340 nm and dual emissions at 665 nm and 590 nm for d2 and cryptate (100 µs delay), respectively. The FRET signal was expressed as: F665/F590 × 10 4 and normalized to the controls to indicate the cAMP levels. The dose-response curves were generated and EC 50 values were calculated using OriginPro software (OriginLab).
Statistical analysis. Statistical analyses were performed using Prism (GraphPad Software, La Jolla, CA, USA) by one-way ANOVA with post-hoc Bonferroni/Dunnett's test for multiple comparisons or Student's t-test as appropriate. P < 0.05 was considered to be statistically significant. All data are presented as mean + or ± SEM.