Cyclooxygenase-2 inhibitor prevents radiation-enhanced infiltration of F98 glioma cells in brain of Fischer rat.

Abstract Purpose: Radiation induces a neuro-inflammation that is characterized by the expression of genes known to increase the invasion of cancer cells. In Fischer rats, brain irradiation increases the infiltration of cancer cells and reduced the median survival of the animals. In this study, we have determined whether these adverse effects of radiation can be prevented with the cyclooxygenase-2 (COX-2) inhibitor meloxicam. Materials and methods: Brain of Fischer rats treated or not with meloxicam were irradiated (15 Gy) and then implanted with the F98 glioma cells. The median survival of the animals, the infiltration of F98 cells, and the expression of inflammatory cytokines and pro-migration molecules were measured. Results: Meloxicam reduced by 75% the production of prostaglandin E2 (bioproduct of COX-2) in irradiated brains validating its anti-inflammatory effect. Median survival was increased to control levels by the treatment of meloxicam following brain irradiation. This protective effect was associated with a reduction of the infiltration of F98 cells in the brain, a complete inhibition of radiation-enhancement of matrix metalloproteinase-2, and a significant reduction of tumor necrosis factor α (TNF-α) and tumor growth factor β1 (TGF-β1) expression. Using invasion chambers, interleukin-1β (IL-1β) stimulated by 5-fold the invasiveness of F98 cells, but this stimulation was completely inhibited by meloxicam. This suggests that a cooperation between IL-1β and COX-2 are involved in radiation-enhancement of F98 cell invasion. Conclusions: Our results indicate the importance of reducing the inflammatory response of normal brain tissue following irradiation in an effort to extend median survival in F98 tumor-bearing rats.


Introduction
Glioblastoma multiforme (GBM) is the most aggressive primary brain neoplasm, taking the lives of patients within a median of 12 -15 months after diagnosis and standard treatment (Stupp et al. 2005). Because of their infi ltrative nature, complete resection of GBM tumors is a virtually unreachable goal (Brandes et al. 2008). Th erefore, the objective of the surgery is to maximize the extent of resection, thereby alleviating neurological defi cits caused by the local mass effect, and allowing a decrease in steroids doses (Berger and Hadjipanayis 2007). Cancer cells left behind are targeted using radiotherapy and chemotherapy.
Ionizing radiation produces a therapeutic eff ect by eradicating tumor cells, but also causes damage to the surrounding healthy tissues. Th erefore, acute and late eff ects of radiotherapy on the brain are common and represent a signifi cant source of morbidity (Lawrence et al. 2010). Th e acute side-eff ects of radiotherapy include nausea, vomiting, headache, vertigo and seizures, while late eff ects such as cognitive disturbance can be seen (Anand et al. 2012). Th is narrow therapeutic index leads to an optimization of radiation dose based on the overall tolerance of healthy tissues instead of giving a high therapeutic dose that would eliminate most cancer cells (Chakravarti and Palanichamy 2008).
Th e short life expectancy after treatment is associated with a rapid recurrence of the glioma, typically within 2 -3 cm from the resection cavity (Burger et al. 1983, Gaspar et al. 1992. Although this rapid infi ltration and proliferation of glioma cells after treatment has been known for decades, the molecular mechanisms involved are still largely unknown (Mangiola et al. 2010). Th erefore, to extend the life expectancy of GBM patients, it is paramount to identify what stimulates the infi ltration of glioma cells in brain and then assess new therapeutic modality to prevent it.
In an animal model developed to mimic the stimulation of glioma cell infi ltration in brain after irradiation, it was shown that a sublethal in vitro irradiation of rat 9L glioma cells resulted in a greater number of tumor satellites after their injection into the striatum of rat brain (Wild-Bode et al. 2001). Another study using the rat glioma F98 cells showed that irradiation of these cancer cells has only a marginal eff ect. Stimulation of the F98 cell infi ltration was mainly induced by the irradiated healthy brain and resulted in a reduction of median survival of rats bearing this tumor (Desmarais et al. 2012). Th is stimulation was associated with pro-infi ltration mediators released from irradiated normal brain, such as cyclooxygenase-2 (COX-2), interleukin-1 β (IL-1 β ) and matrix metalloproteinase-2 (MMP-2) (Desmarais et al. 2012). Th is suggests that the rapid recurrence of GBM could be partially attributed to an infl ammatory response induced by radiation in the brain. Th is adverse eff ect of radiation was also observed in other cancers. For example, irradiation of the mouse mammary gland stimulated the migration of breast cancer cells, and increased the number of circulating tumor cells and the number of lung metastases (Bouchard et al. 2013).
In the present study, we assessed the ability of the COX-2 inhibitor meloxicam to prevent the stimulation of F98 cell infi ltration that is induced by irradiated brain of Fischer rats, and determined whether this inhibition of COX-2 led to an increase of the median survival of the animals. To do so, the F98 cells were implanted after irradiation of the brain. Th is protocol eliminated the possibility that a reduction of cancer cell infi ltration would be caused in part by toxic eff ects of radiation on these cells. Th us, the specifi c eff ect of meloxicam on the irradiated brain was better addressed. Th e eff ect of meloxicam on the expression of infl ammatory cytokines IL-1 β , interleukin-6 (IL-6), tumor growth factor β 1 (TGF-β 1) and tumor necrosis factor α (TNF-α ) in irradiated animals was also assessed.

Cell culture
Th e murine cell line F98 was obtained from American Type Culture Collection (ATCC, Manassas, VA, USA) and tested negative for the mouse antibody production (MAP) assay by Charles River Laboratories (Wilmington, MA, USA). Cells were grown in monolayer using Dulbecco ' s Minimal Essential Medium supplemented with 10% foetal bovine serum (Gibco, Burlington, ON, Canada), 26.2 mM of sodium bicarbonate, 2 mM L-glutamine and a mix of penicillin (100 μ l/ml) and streptomycin (100 μ g/ml). Cells were incubated at 37 ° C in a humidifi ed environment with 5% CO 2 and propagated upon confl uence every 3 days.

Animals and groups studied
Th e experimental protocol was approved by the institutional ethical committee and complied with regulations of the Canadian Council on Animal Care. Adult male Fischer rats weighting 225 -260 g were bought from Charles River laboratories (Montreal, QC, Canada). Animals were kept into our specialized facility and fed ab libitum prior to surgery and irradiation. Animals were randomly assigned to one of the following groups ( n ϭ 8 -10 per group): Group 1, tumor implantation without irradiation; group 2, brain irradiation (15 Gy) and tumor implantation 1 day later; group 3, tumor implantation without irradiation ϩ meloxicam; and group 4, brain irradiation (15 Gy) and tumor implantation 1 day later ϩ meloxicam. Meloxicam was administered daily by intraperitoneal injection starting 48 h prior irradiation until the animals were sacrifi ced on day 15. No toxicity was observed in animals treated with Meloxicam. Half of each group was sacrifi ced 15 days post tumor implantation to measure the infi ltration of F98 cells in brain. Th e other half was used to assess the median survival time. Th e protocol for animal experiments is show in Figure 1.

Tumor implantation
Anaesthesia was induced by an intraperitoneal injection of ketamine:xylazine (87:13 mg/kg). Animals were then mounted onto a stereotaxic frame. A midline incision was performed and a burr hole was punctured using a 16-gauge needle. Th e burr hole coordinates were, using the bregma as a reference point, 1 mm anterior, and 3 mm lateral to the right. Th en 1 ϫ 10 4 cells suspended in a volume of 5 μ l of non-supplemented Dulbecco ' s Minimal Essential Medium were injected into the brain with a 26-gauge needle at a depth of 6 mm from the surface of the skull. Th e suspension of cells was infused at a constant rate of 1 μ l/min using a micro-infusion pump (WPI model UMP3, Sarasota, FL, USA). Th e needle was then slowly withdrawn to minimize the risks subdural and extracranial seeding. Animals were observed daily for apparition of signs of neurological defi cits (hemiparesis, ataxia) and increased intracranial pressure (lethargy and cachexia).

Brain irradiation
Rats were anesthetized (ketamine:xylazine, 87:13 mg/kg) and positioned on a stereotactic frame adapted for use in the Leksell Gamma Knife (model 4C) (Elekta AB, Norcross, GA, USA) (Charest et al. 2009). Th e 14-mm collimators were used to deliver to the whole brain a single dose of 15 Gy. Rats were then brought back to the animal facility for their recovery from anaesthesia and subsequent follow-up. In human, GBM are frequently treated with 60 Gy administered in 30 fractions of 2 Gy. Daily irradiation in animals can be performed using 3% isofl urane as an anesthetic, a procedure that we have already done in mice (Bouchard et al. 2013). However, this set-up was not available to irradiate rats, and 30 daily injections of ketamine:xylazine would be fatal. Th e Biological Equivalent Dose (BED) was calculated to compare the fractionated protocol of 60 Gy in 30 fractions with a single dose of 15 Gy. Using α / β ratio of 3, the BED calculated for a single dose of 15 Gy was 90 Gy. Th is is comparable to the clinical standard of 30 fractions of 2 Gy (BED of 100 Gy).

Brain processing
Euthanasia was carried out by exsanguination and intracardiac perfusion of 30 ml formaldehyde 4% for histological analysis, or phosphate buff ered saline (PBS) for molecular quantifi cations. Brain specimens were removed and kept in formaldehyde for 48 h prior coronal plane sectioning using a brain matrix (taking the implantation needle mark on the cortex as a reference point for the slicing), and fi nally embedded into paraffi n. Th e blocks were cut into 5 μ m thick slides and stained with haematoxylin and eosin (H & E). For biomarker quantifi cations, brains were snap frozen into liquid nitrogen and pulverised into a fi ne powder and kept at Ϫ 80 ° C for further quantifi cations.

Morphological analysis of brain tumors
Brain sample slices, including the F98 tumor, were scanned for each animal using a Nikon 9000 fi lm scanner to get a view of the whole brain coronal section. Th en tumor morphology was registered and analyzed using the Image Pro software (Media Cybernetics, Bethesda, MD, USA). Briefl y, areas of interest were highlighted prior to analysis. Tumors were divided into two distinct areas: Primary tumor and invasive tumor. Areas corresponding to the primary tumor were defi ned as the central region with a well-defi ned edge, while the invasive areas corresponded to clusters of neoplastic cells which have migrated from the primary tumor and did not display direct physical contact with the primary nodule. Th e following parameters were measured: Surface of the primary tumor, overall surface of the invasive tumor, and distance of infi ltration of neoplastic cell clusters from the edge of the primary tumor.

Prostaglandins E2 quantifi cation by liquid chromatography/tandem mass spectrometry
Brains were snap frozen with liquid nitrogen and pulverised brain tissues ( ∼ 100 mg) were homogenized with a dounce homogenizer in 3 ml of acetone-saline solution (2:1) containing 10 ng of prostaglandin E2-d4 (PGE2-d4) which contains four deuterium atoms at the 3, 3 ′ , 4, and 4 ′ positions (internal standard, Cayman Chemical, Ann Arbor, MI, USA) and 0.05% butylated hydroxytoluene to prevent the oxidation of prostanoids. Th e homogenate was transferred to a screw-top tube, vortexed for 1 min, and centrifuged (10 min, 2000 g , 4 ° C). Th e supernatant was transferred to another tube and mixed with 2 ml hexane by vortexing for 1 min. After centrifugation (10 min, 2000 g , 4 ° C), the upper phase containing lipids was discarded. Th e lower phase was acidifi ed with 30 μ l 2 M formic acid and then 2 ml chloroform containing 0.05% butylated hydroxytoluene were added. Th e mixture was vortexed and again centrifuged (10 min, 2000 g , 4 ° C) to separate the two phases. Th e lower phase containing chloroform was transferred to a conical centrifuge tube for evaporation with a SpeedVac Concentrator (Sarant, Nepean, ON, Canada). Samples were reconstituted in 200 μ l methanol:10 mM ammonium acetate buff er, pH 8.5 (70:30) prior to liquid chromatography/tandem mass spectrometry (LC/MS/MS) analyses. All extraction procedures were performed under low light and low temperature conditions to minimize potential photooxidation or thermal degradation of eicosanoid metabolites.
Prostaglandin E2 (PGE2) was quantifi ed by LC/MS/MS using an API 3000 mass spectrometer (Applied Biosystem, Streetsville, ON, Canada) equipped with a Sciex turbo ion spray (AB Sciex, Concord, ON, Canada) and a Shimadzu pump and controller (Columbia, MD, USA). Prostaglandins were chromatographically resolved using a Kromasil column 100-3.5C18 150 ϫ 2.1 mm (Eka Chemicals, Valleyfi eld, QC, Canada). A linear acetonitrile gradient from 45 -90% during 12 min at a fl ow rate of 200 μ l/min was used. Th e mobile phase consisted of water buff ered with 0.05% of acetic acid (A) and acetonitrile 90% with acetic acid 0.05% (B). Injection volume was 5 μ l per sample which were kept at 4 ° C during analysis. Individual products were detected using negative ionization and the monitoring of the transition m/z 351 → 271 for PGE2 and 355 → 275 for PGE2d 4 at collision energy of Ϫ 25 V. Area under the curves for specifi c ions were used for quantifi cations.

Matrix metalloproteinase-2 quantifi cation by zymography gel
Brain tissues pulverized into a fi ne powder were homogenized with a RIPA buff er (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 % Triton X-100, 0.5% sodium deoxycholate, 0.1% sodium dodecyl-sulfate (SDS) and protease inhibitor cocktail (BD, Mississauga, ON, Canada). Th e homogenates were then centrifuged at (10,000 g , 4 ° C) for 10 min. Supernatants were collected and stored at Ϫ 20 ° C for further analyses. Protein quantifi cation was made with the DC protein assay (Bio-Rad Laboratories, Mississauga, ON, Canada). Zymography analysis was performed as previously reported (Lemay Glioma cell infi ltration stimulated by radiation 627 the addition of IL-1 β . Cells that crossed the layer of Matrigel and the porous membrane 24 h later were fi xed, stained and counted under the microscope. Each experimental condition was performed in triplicate and repeated 2 times.

Statistical analysis
Data are expressed as the mean Ϯ standard deviation mean. Statistical analyses for the median survival time were performed using non-parametric Mann-Whitney test for mean comparisons and Chi-square for distribution comparisons, while the other results was analysed using a t -test. A value of p Յ 0.05 was considered as signifi cant. Th e non-irradiated groups were compared to the irradiated ones, and those treated with meloxicam were compared to the irradiated groups.

Optimizing the dose of meloxicam
Th e COX-2 inhibitor meloxicam was injected daily at 0.1, 0.5 or 1 mg/kg in an attempt to prevent the stimulation of PGE2 production induced by irradiating the brain of Fischer rats. As we previously reported (Desmarais et al. 2012), an irradiation of 15 Gy has increased by 9.7-fold the production of PGE2 measured 4 h post-irradiation, and by 7.2-fold ( p Ͻ 0.05) at day 15 ( Figure 2). Treatment with meloxicam at all doses tested resulted in an important and similar reduction by 2 -4 fold of PGE2. However, even with the highest concentration of meloxicam, the level of PGE2 in irradiated brain remained higher than in non-irradiated brain. Th e dose of 0.1 mg/kg was elected for use in all other experiments.

Meloxicam prevented the reduction of median survival induced by radiation
Brain irradiation before implantation of F98 cells reduced the median survival of rats by 28% (Table I and Figure 3, control: 25.0 days vs. brain 15 Gy ϭ 18.2 days, p ϭ 0.0005). Treatment with meloxicam completely prevented this adverse eff ect of et al. 2011). Th is gel allows the distinction between matrix metalloproteinases (MMP) and proMMP which migrate according to their molecular weights. After the electrophoresis, proMMP is artifi cially activated without losing its propeptide. Gels are then incubated in activation buff er which allows proMMP and MMP to cleave locally the gelatin resulting in clear bands after staining with Coomassie blue. Briefl y, samples were applied on a 12% polyacrylamide-SDS gel containing 0.1% gelatine and electrophoresed at 150 V during 3 h at 4 ° C. After removal of SDS from the gel by incubating in 2.5% Triton X-100 (30 min, 4 times), the gel was incubated at 37 ° C for 18 h in 40 mM Tris-HCl, pH 7.5, containing 10 mM CaCl 2 , 1 μ M ZnCl 2 , 200 mM NaCl, and stained with Coomassie blue R-250. Gels were scanned and the bands analyzed using Image J 1.34n (public domain Java image, NIH).

mRNA levels measured by quantitative polymerase chain reaction
Total RNA extractions were performed on brain tissues pulverized into a fi ne powder using the Absolutely RNA Microprep Kit (Stratagene La Jolla, CA, USA) as previously described (Desmarais et al. 2012), and RNA integrity was assessed with an Agilent 2100 Bioanalyzer (Agilent, Montreal, QC, Canada). Reverse transcription was performed with 2 μ g total RNA in a total volume of 20 μ l containing Transcriptor reverse transcriptase, random hexamers, deoxyribonucleotides (dNTP) (Roche Diagnostics, Laval, QC, Canada), and 10 units of RNAseOUT (Invitrogen, Burlington, ON, Canada) following the manufacturer ' s protocol. All forward and reverse primers were individually resuspended in stock solutions (20 -100 μ M) containing Tris-ethylenediaminetetraacetic acid buff er and subsequently diluted as a primer pair down to 1 μ M in RNase DNase-free water (IDT, Coralville, Iowa, USA). Quantitative PCR (qPCR) reactions were performed in 10 μ l in 96-well plates on a Realplex2 thermocycler (Eppendorf, Mississauga, ON, Canada) with 5 μ l of 2X Fast-Start Universal SYBR Green Master mix (Roche Diagnostics, Laval, QC, Canada), 10 ng (3 μ l) cDNA, and 200 nM fi nal (2 μ l) primer pair solutions. Th e following cycling conditions were used: 10 min at 95 ° C; 50 cycles: 15 sec at 95 ° C, 30 sec at 60 ° C, 30 sec at 72 ° C. Relative expression levels were calculated using the qBASE framework (Hellemans et al. 2007) and the housekeeping genes anti-tubulin β 5 ( TUBB5 ), pumilio homolog 1 ( PUM1 ) and ribosomal protein L-19 ( RPL19 ) for rat cDNA. In every qPCR run, a no template control was performed for each primer pair and these were consistently negative. Primer sequences are listed in the Supplementary  radiation. Th e median survival of the treated animals was similar to the one measured in the non-irradiated control group (control vs. brain 15 Gy ϩ meloxicam, p ϭ 0.44; brain 15 Gy vs. brain 15 Gy ϩ meloxicam, p ϭ 0.009). Only a small but signifi cant increase in median survival was observed in non-irradiated animals treated with meloxicam (control: 25.0 days vs. brain non-irradiated ϩ meloxicam ϭ 27 days, p ϭ 0.03).

Meloxicam reduced the radiation-enhancement of cancer cell infi ltration
As we previously reported (Desmarais et al. 2012), the distance of cancer cell infi ltration from the edge of the tumor was multiplied by 2.7 when the brain was irradiated before implantation of the F98 cells ( p ϭ 0.005, Figure 4A), compared to non-irradiated control. Consequently, the brain surface infi ltrated by cancer cells was 2.5 times larger, which is reported as the ratio surface infi ltrated/primary tumor ( p Ͻ 0.05, Figure 4B). In this study, we found that the increased infi ltration induced by irradiating brain was partially reversed with the use of the COX-2 inhibitor meloxicam. Indeed, the infi ltration distance was still 1.97-fold longer in irradiated brain of animals that were also treated with meloxicam compared to non-irradiated brains ( p Ͻ 0.05, Figure 4A). Th is eff ect of meloxicam was not associated to a signifi cantly modifi cation of the surface of the primary tumor (data not shown).
Supporting the role of COX-2 and its bioproduct PGE2 in cancer cell invasion, addition of this prostaglandin to the F98 cells stimulated their invasiveness by 1.5-fold ( p ϭ 0.007), as assessed in vitro with invasion chambers ( Figure 4C).
Migration of cancer cells requires the cleavage of components of extracellular matrix by proteases, such the matrix metalloproteinase-2 (MMP-2) (Park et al. 2006). In irradiated brain, activity of MMP-2 was enhanced by 1.52-fold ( p ϭ 0.002) 4 h post-irradiation, and 2.52-fold at day 15 ( p ϭ 0.0003), as assessed with a gel zymography. Th is stimulation was completely prevented after treatment with the COX-2 inhibitor meloxicam ( Figure 4D).

Expression of genes involved in the production of PGE2
An important increase of PGE2 level was measured 4 h post-irradiation and was still signifi cantly elevated on day 15 ( Figure 2). As PGE2 is produced by COX-1 and COX-2, we determined whether the expression of these enzymes was stimulated in irradiated brain. A low but signifi cant expression of COX-1 was measured 4 h post-irradiation, which decreased on day 15 but still remained elevated compared to control ( Figure 5A). No increase of the mRNA of COX-2 was measured 4 h post-irradiation, but a signifi cant stimulation was observed on day 15 ( p ϭ 0.004) ( Figure 5B). Treatment with meloxicam did not signifi cantly modify the mRNA level of COX-1, and COX-2 measured 4 h post-irradiation. A signifi cant reduction was measured for COX-2 only at day 15 post-irradiation (day 15, p ϭ 0.03) ( Figure 5A and B).
Arachidonic acid is released from phospholipid membrane by the phospholipase A2 (PLA2) and then metabolized to various prostaglandins and leukotrienes via COX-1, COX-2, or 5-lipoxygenenase, depending on the type of cells (Kuwata et al. 2014). Four hours after brain irradiation, the mRNA level of the isoform sPLA2 was signifi cantly increased by 1.9-fold ( p ϭ 0.015). Th is stimulation was even more important at day 15 (2.7-fold, p ϭ 0.037), and was largely prevented by treating the animals with meloxicam ( Figure 5C). On the other hand, expression of the cytosolic form of PLA2 (cPLA2) was not signifi cantly modifi ed after the irradiation ( Figure 5D; 4 h post-irradiation, p ϭ 0.069).

Eff ect of meloxicam on the expression of proinfl ammatory cytokines
As COX-2 plays a central role in the infl ammatory response, we investigated whether treatment with the COX-2 inhibitor meloxicam could indirectly aff ect expression of the proinfl ammatory cytokines IL-1 β , IL-6, TGF-β 1 and TNF-α that are known to stimulate the invasion of cancer cells (Huang et al. 2009, Ding et al. 2010. For the cytokines IL-1 β , IL-6, and TNF-α , a signifi cant increase of their mRNA was observed 4 h post-irradiation ( p Ͻ 0.05), but not for TGF-β 1 ( p ϭ 0.21). Th ese stimulations were temporary as their expression at day 15 were similar or smaller to the level measured in non-irradiated brains ( Figure 6A -D). Treatment with meloxicam led to a small reduction that was signifi cant only at 4 h post-irradiation for TGF-β 1 ( p ϭ 0.006), and TNF-α ( p ϭ 0.024).
As the highest stimulation of mRNA expression was measured with IL-1 β , its ability to stimulate the in vitro invasion of F98 cells was assessed with invasion chambers. Th e addition of IL-1 β (10 ng/ml) produced an increase in the number of F98 cells that crossed the layer of Matrigel by 5-fold ( p ϭ 0.015, Figure 6E). Th is stimulation of F98 invasion was completely inhibited by adding meloxicam (IL-1 β vs. infi ltration. Our results showed that treatment with meloxicam largely prevented the stimulation of F98 cell infi ltration in the brain; more importantly, the median survival time was similar to non-irradiated animals. Th e benefi t of meloxicam treatment was also specifi c to irradiated animals as the COX-2 inhibitor only slightly improved the median survival of non-irradiated animals implanted with the F98 cells. Nevertheless, in some non-irradiated animals, their median survival seems to be increased by meloxicam. Since growth of the tumor may also induce some infl ammation in brain, treatment with meloxicam could reduce this stimulation of F98 cell infi ltration resulting in a longer median survival. Overall, our study supports the importance of inhibiting the infl ammatory reaction induced by radiation in brain tissue surrounding the tumor cells to prolong the median survival of the animals. MMP degrade extracellular matrix proteins, thereby opening routes to the infi ltration of glioma cells (Nakada et al. 2003, Larkins et al. 2006, Lemay et al. 2011. We showed that MMP-2 levels were higher than normal into the irradiated brains, and this upregulation persisted at day 15 postirradiation. Consequently, remodelling of ECM through an enhancement of MMP-2 activity could be an important feature explaining the enhancement of the infi ltrative properties of cancer cells observed into irradiated brains. Th is stimulation of MMP-2 was completely blocked by meloxicam treatment. Supporting the role of meloxicam in the reduction of MMP-2 expression, it was previously reported that the IL-1 β ϩ meloxicam, p ϭ 0.019). Th is supports that the contribution of IL-1 β to the enhancement of F98 invasion can be effi ciently inhibited by a COX-2 inhibitor.

Discussion
A neuro-infl ammatory process occurs after irradiation of the brain that is associated with a stimulation of the expression of pro-infl ammatory genes, some of which are known for their ability to increase cancer cell invasion (Kyrkanides et al. 2002, Moore et al. 2005, Moravan et al. 2011. Th e potential impact of this infl ammation on the median survival of irradiated animals bearing a brain tumor was only recently appraised (Desmarais et al. 2012). Irradiation of Fischer rat brain followed by the implantation of F98 cells favored the infi ltration of F98 cells, thereby resulting in a reduction in median survival time of the animals. Similar fi ndings were observed with a sub-curative dose of radiation delivery to the F98 tumor cells implanted in Fischer rat brain with no prior radiation exposure. Th is adverse eff ect of radiation was associated with pro-infi ltration mediators released from irradiated brain parenchyma, such as IL-1 β , MMP-2, and PGE2, the bioactive product of COX-2 (Wild- Bode et al. 2001, Desmarais et al. 2012.
As COX-2 seems to play an important role in radiationinduced infl ammation in brain and cancer cell infi ltration, we investigated whether treatment with the COX-2 inhibitor meloxicam could prevent the stimulation of F98 cell  rons (Tanaka et al. 2009), and it is also preferentially localised in neocortex, hippocampus, amygdala, and limbic cortices (Yamagata et al. 1993, Kaufmann et al. 1996. Th is distribution of the two COX isoforms in the brain may suggest that the relative importance of COX-1 and COX-2 in the stimulation of glioma cell infi ltration might vary according to the tumor location in the brain. In our study, the F98 cells were implanted in the caudate nucleus. Administration of a COX-2 inhibitor increased the median survival of irradiated rats to the same level that was measured in the nonirradiated group. Th is protective eff ect was obtained even if the increased infi ltration of cancer cells has not been completely blocked. Th erefore, we cannot exclude that a COX-1 inhibitor would further inhibit radiation-enhancement of cancer cell infi ltration. It also remains to be determined whether inhibition of COX-1 and/or COX-2 would be required for tumors implanted in other brain areas. COX-1 is usually considered as a housekeeping gene whose activity is dependent solely on availability of its substrate, the arachidonic acid. However, a recent study reported that COX-1 mRNA was rapidly induced in rat cerebral cortex in a model of reversible ischemia suggesting that expression COX-1 in brain might be stimulated bioactive product of COX-2, PGE2, increased the production of MMP-2 resulting in an enhancement of cancer cell invasion . It is noteworthy that radiationinduced F98 cell infi ltration was not completely inhibited by meloxicam. Th is might suggest that other proteases than MMP-2 also contribute to the infi ltration of F98 in irradiated brains. Th e benefi cial eff ect of meloxicam was obtained at a dose that did not completely inhibit the stimulation of PGE2 production. Although the level of PGE2 was reduced by approximately 75%, it was still between 2.4 and 3.6 times higher than that measured in non-irradiated brains. A similar observation was made in the brain of C3HeN mice where an increase of COX-2 expression was measured 4 h after an irradiation of 35 Gy (Moore et al. 2005). Specifi c COX-2 inhibition with NS-398 lowered brain PGE2 levels by about 60%, while a treatment with the COX-1 inhibitor SC560 was required to completely inhibit the elevation of PGE2 (Moore et al. 2005).
COX-1 and COX-2 are preferentially expressed by diff erent cell types and brain areas. COX-1 is abundantly expressed in microglia, even under resting conditions (Yermakova et al. 1999), while COX-2 is predominantly observed in brain neu-

Glioma cell infi ltration stimulated by radiation 631
COX-2 is constitutively expressed (Yamagata et al. 1993). COX-2 expression is tightly regulated at the transcription/ translation level (Masferrer et al. 1994, Smith andDewitt 1996) and it can be induced by infl ammatory cytokines, such as IL-1 β , Kuwata et al. 2014. In our rat model, a signifi cant increase of IL-1 β mRNA was induced in the brain 4 h post-irradiation and was followed by a under specifi c conditions (Holtz et al. 1996). Our results showed that an exposure to 15 Gy was not suffi cient to induce the expression of COX-1 in Fischer rat brain in the fi rst 15 days after irradiation, as measured by level of mRNA.
Regarding COX-2, few tissues express this cyclooxygenase under normal circumstances, except in the brain where  results support that a COX-2 inhibitor could be a good candidate as adjuvant to radiotherapy in patients treated for a GBM; however, the dosage suitable for clinical use should be determined to ensure that no toxicity would be induced. Since IL-1 β has been linked to the stimulation of the infi ltration of cancer cells, an inhibitor targeting this cytokine could also be investigated in clinical trials. stimulation of COX-2 detected at day 15. Th erefore, the signifi cant elevation of PGE2 measured 4 h post-irradiation is not likely directly related to an increase in COX-2 expression.
An increase in the level of PGE2 can also be induced by enhancing the liberation of arachidonic acid from membrane glycerophospholipids regulated by PLA2 enzymes (Kuwata et al. 2014). Th e expression of two isoforms of PLA2 (cPLA2 and sPLA2) can be linked to infl ammation as their expression can be stimulated by IL-1 β (Xin et al. 2007, Lee et al. 2010. Indeed in rat fi broblasts, IL-1 β induced the release of arachidonic acid that peaked at 6 h after treatment and then gradually declined , Kuwata et al. 2014. A stimulation of the production of PGE2 was then observed that was completely inhibited by the COX-2 inhibitor NS-398 . In our study, large increase of IL-1 β in irradiated brain was associated with a stimulation of sPLA2 expression, but not with cPLA2. Th is suggests that the elevation of PGE2 observed 4 h post-irradiation might be related to a stimulation of sPLA2 expression. Th e role of IL-1 β in neuro-infl ammation may be important since this cytokine increased the invasion of F98 cells, as measured with the in vitro invasion chamber assay. Th e IL-1 β pathway was also related to COX-2 since the stimulation of F98 cell invasion induced by this cytokine was completely inhibited by meloxicam. Since the animals were treated daily with meloxicam, this COX-2 inhibitor could inhibit the stimulation of F98 cell infi ltration induced by IL-1 β in irradiated brain. Similar results were previously reported with the breast cancer cell MDA-MB-231, where a stimulation of their invasion by IL-1 β was prevented by a COX-2 inhibitor . In C3H/HeN mouse brain following irradiation, a COX-2 selective inhibitor signifi cantly attenuated the level of induction of IL-1 β , further supporting the link between IL-1 β and COX-2 (Moore et al. 2005). Th is suggests that the cooperation between IL-1 β and COX-2 for the stimulation of cancer cell invasion is not specifi c to rat glioma F98 cells.
Th e stimulation of glioma cell infi ltration induced by radiation may involve other infl ammatory cytokines. Th is hypothesis is supported by in vitro invasion assays where IL-6 , TNF-α (Huang et al. 2009) and TGF-β 1 (Merzak et al. 1994) increased the invasion of glioma cells. In irradiated Fischer rat brains, a small but signifi cant increase was measured for IL-6 and TNF-α . However, unlike in C3H/ HeN mouse brain (Kyrkanides et al. 2002), inhibition of COX-2 by meloxicam did not signifi cantly attenuate the levels of these infl ammatory mediators in Fischer rat brain. Although we cannot rule out that these cytokines might stimulate the infi ltration of glioma cells in other models, our results support that they did not play a key role in the F98 Fischer rat glioma model.
In conclusion, a rapid infi ltration and proliferation of glioma cells after radiation treatment are known for decades and seems to involved pro-infl ammatory cytokines. In this study, we shown that inhibition of COX-2 prevents the stimulation of F98 cell infi ltration in irradiated brain resulting in a longer median survival time of the animals. Th ese