Induction of COX-2 Enzyme and Down-regulation of COX-1 Expression by Lipopolysaccharide (LPS) Control Prostaglandin E2 Production in Astrocytes*

Background: The relative contribution of COX-2 and COX-1 to prostanoid formation under neuroinflammation is complex. Results: LPS induced COX-2 and mPGES1 but down-regulated COX-1 and TS in astroglia. These effects accounted for the high production of PGE2. Conclusion: PGE2 after LPS results from the coordinated COX-2 up-regulation and COX-1 down-regulation in astrocytes. Significance: Changes in COX-2 and COX-1 expression mediate astroglial PGE2 generation in neuroinflammation. Pathological conditions and pro-inflammatory stimuli in the brain induce cyclooxygenase-2 (COX-2), a key enzyme in arachidonic acid metabolism mediating the production of prostanoids that, among other actions, have strong vasoactive properties. Although low basal cerebral COX-2 expression has been reported, COX-2 is strongly induced by pro-inflammatory challenges, whereas COX-1 is constitutively expressed. However, the contribution of these enzymes in prostanoid formation varies depending on the stimuli and cell type. Astrocyte feet surround cerebral microvessels and release molecules that can trigger vascular responses. Here, we investigate the regulation of COX-2 induction and its role in prostanoid generation after a pro-inflammatory challenge with the bacterial lipopolysaccharide (LPS) in astroglia. Intracerebral administration of LPS in rodents induced strong COX-2 expression mainly in astroglia and microglia, whereas COX-1 expression was predominant in microglia and did not increase. In cultured astrocytes, LPS strongly induced COX-2 and microsomal prostaglandin-E2 (PGE2) synthase-1, mediated by the MyD88-dependent NFκB pathway and influenced by mitogen-activated protein kinase pathways. Studies in COX-deficient cells and using COX inhibitors demonstrated that COX-2 mediated the high production of PGE2 and, to a lesser extent, other prostanoids after LPS. In contrast, LPS down-regulated COX-1 in an MyD88-dependent fashion, and COX-1 deficiency increased PGE2 production after LPS. The results show that astrocytes respond to LPS by a COX-2-dependent production of prostanoids, mainly vasoactive PGE2, and suggest that the coordinated down-regulation of COX-1 facilitates PGE2 production after TLR-4 activation. These effects might induce cerebral blood flow responses to brain inflammation.

Cyclooxygenases (COX), 3 also known as prostaglandin G/H synthases, play a crucial role in inflammation and are targets of widely used nonsteroidal anti-inflammatory drugs. There are two main COX enzymes, COX-1 and COX-2, that participate in the metabolism of arachidonic acid generating the unstable product prostaglandin (PG) G 2 that is reduced to PGH 2 . PGH 2 is the substrate of prostaglandin isomerases that give rise to a family of vasoactive compounds called prostanoids, including molecules such as prostaglandins, thromboxane, and prostacyclin. There is cellular specificity for the production of certain prostanoids, and they exert different actions depending on the type of molecule produced and on the specific receptors that become activated (1). Although COX-1 is constitutively expressed in most tissues, COX-2 is an inducible enzyme that responds to pro-inflammatory stimuli.
COX-2 is induced in brain cells under pathological conditions, but the role of the COX isoforms in brain diseases is not clearly established. Cox-2-deficient mice are protected against brain ischemia (2), and inhibition of COX-2 provides beneficial effects against ischemic damage and neuronal death (3)(4)(5)(6), suggesting a detrimental effect of COX-2 in stroke. In contrast, in neurodegenerative diseases, COX-2 inhibitors are not protective in mouse models of Alzheimer disease (7) and did not show benefits in clinical trials in Alzheimer disease patients (8) or in patients with mild cognitive impairment (9). Furthermore, COX-2 inhibitors increase the risk of cardiovascular and cerebrovascular pathology (10), and Cox-2-deficient mice show exacerbated brain inflammation, leukocyte infiltration, and blood-brain barrier damage after exposure to the bacterial lipopolysaccharide (LPS) (11)(12)(13)(14)(15), suggesting some beneficial action of COX-2 in inflammation. Furthermore, COX-2 might contribute to neurovascular coupling because COX-2 inhibitors abrogate the increases in cerebral blood flow (CBF) induced by neuronal activation in rats (16). Exposure to LPS has been reported to induce vasodilation (17) and increase CBF (18) through a mechanism involving inducible NOS and the NOX2 subunit of the superoxide-producing enzyme NADPH oxidase. Because LPS induces strong expression of COX-2 in the brain, it is feasible that vasoactive COX-2 products might also be involved in CBF regulation.
In this study we examined the effect of intracerebral administration of LPS on the cellular expression of COX-2 and found strong up-regulation in microglia and astrocytes. Because astrocytes are recognized as important players in CBF regulation under physiological and pathological conditions (19), we then investigated the prostanoids induced by LPS and the COX isoforms involved in prostanoid generation in purified astrocyte cultures. The results show that the LPS challenge strongly induced COX-2 in astrocytes through a MyD88/NFB-dependent mechanism, show the crucial role of COX-2 in prostanoid production after LPS, and show that PGE 2 is the major product of arachidonic acid metabolism under these experimental conditions. Furthermore, we found that LPS down-regulates Cox-1 gene expression and that Cox-1-deficient cells produce more PGE 2 than the WT, indicating some negative effect of COX-1 on the COX-2-dependent production of PGE 2 in astrocytes after LPS.

EXPERIMENTAL PROCEDURES
Animals-Animal work was authorized by the Ethical Committee of the University of Barcelona, and it was performed in agreement with the local regulations and in compliance with the Directives of the European Community. Four-month-old male Sprague-Dawley rats were obtained from Charles River (Lyon, France). MyD88 knock-out (KO) mice in a C57Bl/6 background were obtained from Oriental Bioservices, Inc. (Kyoto, Japan). MyD88 KO mice (Ϫ/Ϫ) were crossed with wild type (WT) (ϩ/ϩ) C57Bl/6 mice (Charles River), and a colony of MyD88 heterozygous mice (ϩ/Ϫ) was kept in the animal house of the University of Barcelona School of Medicine. Each individual animal born from the heterozygous progenitors was genotyped, and the MyD88 KO and the MyD88 WT animals were selected for the studies described below. COX-1 and COX-2 heterozygous mice were from Taconics Inc. (Hudson, NY). COX-2 or COX-1 heterozygous females (ϩ/Ϫ) were crossed with homozygous (Ϫ/Ϫ) males; all had a mixed B6;129P2 background. We genotyped each animal of the offspring, and the KO animals were selected for the studies, although the corresponding WT animals were used as controls.
LPS Administration to Rodents-Rats and mice were anesthetized with isofluorane and placed in a stereotaxic apparatus for injection of LPS or the vehicle (phosphate-buffered saline (PBS)) in the right striatum. For rats, LPS (5 l of 1 g/ml) or the same volume of vehicle was injected at the following coordinates according to the atlas of Paxinos and Watson (22) in relation to Bregma as follows: 0.5 mm antero-posterior, 3 mm lateral, and 5 mm ventral. For mice, LPS (0.7 l of 1 g/ml) or vehicle was injected at the following coordinates: 0.5 mm antero-posterior, 2 mm lateral, and 3 mm ventral. After 8 h, animals were anesthetized with isofluorane and perfused through the heart with saline to remove blood from the brain vessels, and brain tissue was obtained after dissection of the ipsilateral and contralateral striatum and was immediately frozen and kept at Ϫ80°C until further use. A different set of animals was processed for immunohistochemistry.
Cell Cultures-Glial cell cultures enriched in astrocytes were prepared from the cerebral cortex of 1-2-day-old rats or mice as described previously (21,23), with minor modifications. In brief, cells were maintained at 37°C in a humidified atmosphere of 5% CO 2 , 95% air in culture medium (DMEM (Invitrogen) for rat astrocytes, and DMEM/F-12 nutrient (1:1) (Invitrogen) for mouse astrocytes). Media were supplemented with 10% fetal bovine serum (FBS; Invitrogen) and 4 ml/liter of a mixture of penicillin/streptomycin 10,000 units, 10,000 g/ml (Invitrogen). Cells were subcultured to obtain purified astroglia cultures, as follows. At confluence after 8 -10 days in vitro, cells were treated with 4 M antimitotic cytosine arabinoside (Sigma) for 5 days to eliminate dividing cells, i.e. mostly microglia and progenitors. Flasks were shaken overnight, and the remaining astrocyte adherent monolayer was detached with trypsin 0.0125%, 0.2 mM EDTA and seeded at 10 ϫ 10 4 cells/ml with incubation medium (as above). Purified astrocytes were treated when cells reached confluence at 4 days after subculturing. FBS was reduced to 1% 16 h prior to treatments.
Rat astrocyte cultures contained only 2.01 Ϯ 1.68% of contaminating microglia cells, as reported (21). Purified mouse astrocyte cultures also contained very little microglia, as estimated by immunofluorescence and by examining the expression of CD11b mRNA (see below for description of these methods). After immunofluorescence with an antibody against a microglial marker (Iba-1) and an antibody against glial fibrillary acidic protein (GFAP) to label astrocytes (supplemental Fig. 2), we counted (n ϭ 24 fields for two cultures, using ϫ20 magnification) the percentage of Iba-1 immunoreactive cells and estimated that the % of contaminating microglia in the astrocyte cultures was 0.77 Ϯ 0.49%. Independently, we calculated the percentage of CD11b expression per culture by real time RT-PCR as a marker of microglia and used purified microglia cultures (obtained as reported previously (24)) as a reference for 100% CD11b expression. According to this procedure, CD11b expression (mean Ϯ S.D., n ϭ 5) in astroglia cultures was 1.41 Ϯ 1.22%, supporting that contaminating microglia cells were very scarce in the purified astroglia cultures.
For experiments with MyD88 KO, Cox-1 KO, and Cox-2 KO cells, individual astrocyte cultures were obtained from each newborn animal and, after genotyping, the Ϫ/Ϫ (KO) and ϩ/ϩ (WT) cultures were selected for use in further experiments. Experiments in KO and WT cells were carried out in parallel.
Immunocytochemistry-Astrocytes were seeded on polylysine-coated coverslips. Cells were washed in PBS and fixed in 4% paraformaldehyde for 30 min. Cells were permeabilized with 0.2% Triton X-100 (Sigma) in PBS for 10 min, blocked with 10% goat or horse serum in PBS for 1 h, and incubated overnight at 4°C with one of the following primary antibodies: a rabbit polyclonal antibody against Iba-1 (019-19741, Wako Chemicals GmbH, Neuss, Germany) diluted 1:1,000 or a monoclonal antibody against GFAP (G3893, Sigma) diluted 1,1000. The next day, cells were washed and incubated with green fluorescence Alexa Fluor 488 dye-labeled goat anti-rabbit IgG antibody and Alexa Fluor 546 goat anti-mouse IgG diluted 1,1000 (Molecular Probes) for 1 h at room temperature. Thereafter, astrocytes were stained with Hoechst to visualize the nuclei. The coverslips were mounted onto microscope slides using Mowiol mounting medium (Calbiochem, Merck). Observations were performed with an Olympus IX70 fluorescence microscope.
For immunohistochemistry in brain tissue, animals were perfused through the heart with saline followed by paraformaldehyde ( 4%) in phosphate buffer, pH 7.4. The brain was removed, post-fixed with paraformaldehyde overnight, and then kept in phosphate buffer before slicing it in a vibratome to obtain 30-m-thick coronal sections. Brain sections were cryoprotected in a solution containing glycerol and were kept frozen at Ϫ20°C. Immunohistochemistry was performed free-floating with vibratome sections, as reported previously (25). Endogenous peroxidases were blocked with 3% hydrogen peroxide and 10% methanol in PBS for 25 min. Sections were incubated for 2 h in 3% normal horse or goat serum for mouse monoclonal or rabbit polyclonal antibodies, respectively, to block unspecific binding sites, washed in T-PBS (PBS containing 0.5% Triton X-100), and incubated overnight at 4°C with either mouse monoclonal antibody against COX-1 (160110, Cayman Chemical) diluted 1:100, or rabbit polyclonal anti-COX-2 antibody (160126, Cayman Chemical) diluted 1:500. Thereafter, the sections were rinsed in T-PBS and incubated for 1 h with a biotinylated secondary antibody (1:200, Vector Laboratories), followed by incubation with 1% avidin-biotin-peroxidase complex (ABC kit, Vector Laboratories). The reaction was visualized with 0.05% diaminobenzidine in 0.03% hydrogen peroxide in PBS. Double immunohistochemistry was carried out following the first immunoreaction with COX-1 or COX-2. The second primary antibodies used were as follows: a rabbit polyclonal antibody against GFAP (Z0334, DakoCytomation) diluted 1:500, which labels astroglia, or a rabbit polyclonal antibody against Iba-1 (as above) diluted 1:500 to detect microglia. Sections were then incubated with the avidin-biotin complex, washed with 0.01 M sodium phosphate buffer, pH 6, and preincubated for 10 min with 0.01% benzidine dihydrochloride and 0.025% sodium nitroferricyanide in 0.01 M sodium phosphate buffer, pH 6. The reaction was developed with this solution containing 0.005% H 2 O 2 . Immunoreaction controls included omission of the first or second primary antibodies. The induction of COX-2 in microglia and in astrocytes was assessed by counting the number of GFAP ϩ and of Iba-1 ϩ cells expressing COX-2 in the ipsilateral striatum. Microscopic photographs (ϫ20 objective) of three areas surrounding the injection site were taken per brain section in three brain sections per animal. The proportion of microglia (Iba-1 ϩ ) and of astroglia (GFAP ϩ ) expressing COX-2 was calculated in each photograph, and the average value of all the photographs per animal was calculated. Values are expressed as the mean of four animals treated with LPS and three animals treated with PBS.
Real Time RT-PCR-Total RNA was extracted using the PureLink RNA kit (Invitrogen). RNA quantity and purity were determined using ND-1000 microspectrophotometer (Nano-Drop Technologies, Wilmington, DE). One g of total RNA was reverse-transcribed using a mixture of random primers (High Capacity cDNA reverse transcription kit, Applied Biosystems). Real time quantitative RT-PCR analysis was carried out by SYBR Green I dye detection (11761500, Invitrogen) using the iCycler iQTM Multicolor real time detection system (Bio-Rad). PCR primers were designed with the aid of Primer3 software to bridge the exon-intron boundaries within the gene of interest to exclude amplification of contaminating genomic DNA. Several genes were assayed as loading controls (Hprt1, Sdha, Ywhaz, and Rpl14). Rpl14 was the control gene showing the best stability after LPS treatment and was chosen for normalization. Primers (see list in Table 1) were purchased from IDT (Conda, Spain). Optimized thermal cycling conditions were as follows: 1 min at 50°C, 8 min, and 30 s at 95°C and 40 cycles of 15 s at 95°C and 30 s at 60°C in which an optical acquirement were performed. Data were collected after each cycle and were graphically displayed (iCycler iQTM real time detection system software, version 3.1, Bio-Rad). Melt curves were performed upon completion of the cycles to ensure absence of nonspecific products. Quantification was performed by normalizing cycle threshold (Ct) values with the Rpl14 control gene Ct, and analysis was carried out with the 2Ϫ⌬⌬CT method (26).
Statistical Analyses-One-way analysis of variance was used for comparisons between multiple groups, after testing for normality, followed by the post hoc Bonferroni test. Comparison between two groups was carried out with the t test after verifying normal distribution; otherwise, the nonparametric Mann Whitney test was used. Linear or nonlinear regression analyses

Intracerebral Administration of LPS Induces Cox-2 in
Microglia and Astroglia-Intracerebral administration of LPS to rats induced mRNA expression of Tnf-␣ (Fig. 1A) in the ipsilateral hemisphere at 8 h. LPS also increased the expression of Cox-2 mRNA (Fig. 1B) and COX-2 protein (Fig. 1, C and D). LPS induces TLR-4 activation and recruitment of the MyD88 adaptor protein that mediates activation of the transcription factor NFB and induction of target genes, such as the pro-inflammatory cytokine TNF-␣ (21). In agreement with this, MyD88 KO mice did not show induction of Tnf-␣ (Fig. 1E) or Cox-2 (Fig.  1F) mRNA in the ipsilateral hemisphere after LPS, indicating that induction of both genes was MyD88-dependent.
LPS up-regulated the expression of Iba-1 mRNA in the ipsilateral hemisphere at 8 h suggesting microglial activation, whereas expression of Gfap mRNA was not modified at this time point (supplemental Fig. 3). However, by immunohistochemistry (Fig. 2), we detected a strong induction of COX-2, not only in microglia (Fig. 2, A-D) but also in astrocytes (Fig. 2, E-G) of the ipsilateral hemisphere 8 h after LPS. Quantification of the immunohistochemistry showed increased numbers of COX-2 immunoreactive microglia (Iba-1) and astrocytes (GFAP) after LPS (Fig. 2, P and Q). COX-1 was expressed under basal conditions preferentially in microglia (Fig. 2, H-K) and, with a lower intensity, in astroglia (Fig. 3, L-O), and it was not up-regulated by LPS (Fig. 2, H-O).
We then undertook an in vitro study in purified cultures of astroglia treated with LPS to unravel the mechanisms underlying COX-2 induction and prostanoid release induced by TLR-4 activation, and the effects of deficiency or inhibition of either COX-1 or COX-2.
Regulation of COX-2 Expression in Astrocytes Challenged with LPS-LPS induced Tnf-␣ (Fig. 3A) and Cox-2 (Fig. 3B) mRNA and protein expression (Fig. 3, C and D) in cultured astrocytes, as it did in vivo (Fig. 1, A-D). The transcription factor NFB was involved in COX-2 induction because silencing the p65 subunit of NFB with siRNA attenuated Cox-2 mRNA induction (Fig. 3E). However, this effect was not observed by silencing other genes, such as the gp91 subunit of

Cox-2 Induction and Cox-1 Down-regulation by LPS
NADPH oxidase complex (Fig. 3E) that was reported to mediate COX-2 induction after LPS in microglia (27). The involvement of NFB in mediating the induction of COX-2 after LPS was further substantiated by the use of the inhibitor PDTC, which attenuated the effect of LPS (Fig. 3F). We then used astrocytes from mice deficient in MyD88 or corresponding WT mice to explore whether COX-2 induction was dependent on the MyD88 pathway in these cells, as observed previously in vivo (Fig. 1, E and F). LPS failed to induce Tnf-␣ mRNA in MyD88-deficient astrocytes (Fig. 3G), which did not express Cox-2 mRNA (Fig. 3H) or protein (Fig. 3I). Therefore, COX-2 induction after LPS is dependent on MyD88 and NFB.
MAPKs participate in LPS signaling (28) and can mediate COX-2 induction (29). We used specific inhibitors of MAPK pathways to unravel their contribution in COX-2 up-regulation after LPS in astrocytes. Cox-2 mRNA induction was severely reduced by the p38 inhibitor SB23906 and by the JNK inhibitor SP600125 but not by the MEK inhibitor U0126 (Fig. 4A). Likewise, COX-2 protein expression 8 h after LPS was very sensitive to SB239063 (from 1 M) (Fig. 4B) and SP600125 (from 10 M) (Fig. 4C), whereas U0126 (1-25 M) had a negligible effect (Fig.  4D). This finding was validated with another MEK inhibitor, PD98059 (from 10 to 40 M) (Fig. 4D). The same result was found at 4 h (Fig. 4E). In agreement with this, the production of  FEBRUARY 24, 2012 • VOLUME 287 • NUMBER 9 PGE 2 , as assessed by ELISA 8 h after LPS, was reduced by p38 and JNK inhibitors but not after MEK inhibition (Fig. 4F).

Cox-2 Induction and Cox-1 Down-regulation by LPS
We then silenced the expression of MAPK1 (ERK2), MAPK10 (JNK3), and MAPK14 (p38) using siRNA. Western blotting (Fig. 4G) showed that siRNAs reduced the corresponding protein expression by 65-70%. Silencing p38 and JNK3 attenuated the expression of Cox-2 mRNA and protein, but no significant effects were observed after silencing MAPK1 (Fig. 4,  H-J). Therefore, we can conclude that induction of COX-2 expression after LPS is strongly dependent on the MyD88 pathway NFB and on p38 and JNK pathways.
Regulation of COX-1 Expression after LPS-We also examined whether the expression of constitutive COX-1 was affected by LPS. The expression of Cox-1 mRNA was significantly reduced 8 h after LPS in WT astrocytes, and the same effect was observed in Cox-2 KO cells (Fig. 5A), which we verified did not express Cox-2 mRNA (Fig. 5B) or protein (Fig. 5, C and D). Although the reduction of Cox-1 mRNA by LPS was already seen at 4 h (Fig. 5, G and I), COX-1 protein expression was unaltered at this time point, but a slight reduction was seen at 8 and 24 h (Fig. 5, E and F). The delay in the reduction of the amount of COX-1 protein after the decreased expression of Cox-1 mRNA might be due to the presence of the constitutive protein that needs to follow its turnover before reductions in mRNA can be translated into protein decreases.
LPS-induced reduction of COX-1 was not prevented by MAPK inhibition (Fig. 5, G and H) or by silencing MAPK expression with siRNA (Fig. 5, I and J). However, LPS-induced down-regulation of Cox-1 mRNA was dependent on the MyD88 pathway because LPS did not reduce it in MyD88 KO mice (Fig. 5K). To better substantiate this finding, we examined whether down-regulation of COX-1 also occurred in vivo in the mouse brain after intracerebral LPS administration. Expression of Cox-1 mRNA was significantly reduced 8 h after injection of LPS but not after injection of PBS (Fig. 5L). This effect was strongly attenuated in MyD88-deficient mice (Fig. 5L), supporting that it was MyD88-dependent. PDTC attenuated the reduction of Cox-1 mRNA induced by LPS in cultured cells, suggesting that NFB was involved (Fig. 5M).
Because COX-1 and COX-2 expression responded in an opposite way to the LPS challenge, we examined whether Cox-

Cox-2 Induction and Cox-1 Down-regulation by LPS
1-deficient mice (Fig. 5N) showed up-regulation of Cox-2 mRNA (Fig. 5O) and protein (Fig. 4O) after LPS, which they did. These results support that although LPS strongly induces COX-2, it represses the expression of COX-1, and both responses are dependent on the MyD88 pathway, whereas p38 and JNK MAPK are involved in up-regulating COX-2 but not in down-regulating COX-1.

LPS Modifies the Expression of Prostaglandin Isomerases-
The types of prostanoids that are produced after COX activation depend on the action of specific prostanoid isomerases, i.e. the enzymes responsible for the production of prostanoids from COX-derived PGH 2 . LPS induced strong mRNA expression of one of the isoforms of PGE 2 synthase, the microsomal PGE synthase-1 (mPges-1) (Fig. 6A), in WT and Cox-1-or Cox-2-deficient cells. Like for COX-2, induction of mPGES-1 after LPS was dependent on the MyD88 pathway because MyD88-deficient cells showed no increase of mPges-1 mRNA expression (Fig. 6B). These findings show that LPS up-regulates the expression of mPges-1 through the MyD88 pathway, in a manner coordinated with the induc-  FEBRUARY 24, 2012 • VOLUME 287 • NUMBER 9

Cox-2 Induction and Cox-1 Down-regulation by LPS
6462 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 287 • NUMBER 9 • FEBRUARY 24, 2012 tion of COX-2, to strongly generate PGE 2 . In contrast to the above findings, LPS down-regulated the expression of prostacyclin synthase (Pgis) mRNA (Fig. 6C) and, to a greater extent, thromboxane synthase (Ts) mRNA (Fig. 6D). Accordingly, although after LPS the expression of mPGES-1 protein significantly increased (Fig. 6E), the expression of TS protein tended to be progressively lower than in controls (Fig. 6F). The latter effects on TS paralleled the down-regulation of COX-1 expression after LPS (Fig. 5, A and E), suggesting common regulatory pathways.
LPS Exposure Induces Prostanoid Production in Astrocytes-COX enzymes metabolize arachidonic acid to prostaglandins PGG 2 and PGH 2 that are rapidly converted by cell-specific prostaglandin isomerases into different prostanoids, including  . LPS induces expression of mPges-1 mRNA but not of the enzymes that synthesize other prostanoids. Astrocytes of Cox-1 or Cox-2 KO mice (Ϫ/Ϫ) and their respective wild type (ϩ/ϩ) astrocytes were treated with LPS (10 ng/ml), and mRNA was extracted at 8 h. A, LPS strongly induces microsomal PGE 2 synthase-1 (mPGES1) mRNA in the different genotypes. c, control. B, induction of mPges1 mRNA after LPS is dependent on the MyD88 pathway, as shown by lack of mPges-1 mRNA up-regulation in MyD88-deficient (MyD88 Ϫ/Ϫ ) cells after LPS. C and D, expression of prostacyclin synthase (PGIS) mRNA (C) and that of TS mRNA (D) is reduced after LPS in all genotypes. E and F, accordingly, in astrocytes from C57 WT mice, LPS significantly up-regulates the expression of mPGES1 protein (E), although TS protein shows a nonsignificant tendency to progressively decrease with time versus controls (F). n ϭ 3 for each genotype. One symbol, p Ͻ 0.05; two symbols, p Ͻ 0.01; three symbols, p Ͻ 0.001. * indicates comparison versus control; & indicates comparison versus LPS in WT. PGE 2 , PGF 2␣ , prostacyclin (PGI 2 ), and thromboxane A 2 (TxA 2 ), among others. We examined the profile of several prostanoids induced by LPS in the culture medium of rat astrocytes by ELISAs. TxA 2 has a half-life of only a few seconds (30), and its production is typically assessed by measuring TxB 2 , which is a stable metabolite. PGI 2 has a half-life of 60 min in plasma, but it is stable for only a few minutes in buffer (30), and its production is typically monitored by measurement of 6-keto-prostaglandin F 1␣ (PGF1␣). LPS caused a very strong accumulation of PGE 2 in the cell culture medium of rat astrocytes from 2 to 24 h (Fig. 7, A and B), and to a lesser extent it increased the concentration of TxB 2 (Fig. 7, C and D) and of PGF1␣ (Fig. 7, E and F) FIGURE 7. LPS induces secretion of prostanoids to the culture medium. Purified cultures of rat astrocytes were exposed to LPS (10 ng/ml) for different times, and the medium was collected and studied by ELISA. ELISAs were carried out in five independent experiments, and curves from representative experiments are shown.

Cox-2 Induction and Cox-1 Down-regulation by LPS
at 8 and 24 h. The cPLA 2 inhibitor arachidonyltrifluoromethyl ketone fully prevented the production of prostanoids (Fig. 7, G-I), suggesting the involvement of cPLA 2 in arachidonic acid mobilization after LPS.
Prostanoid Production after LPS Treatment Is Prevented by COX-2 Inhibitors-COX-2 inhibitor NS-398 strongly blocked LPS-induced PGE 2 , TxA 2 , and PGI 2 production, whereas COX-1 inhibition with SC-560 only partly attenuated the generation of prostanoids after LPS in rat astrocytes (Fig. 7, B, D,  and F). These results suggested that COX-2 was the main mediator of prostanoid production after LPS and pointed to a small contribution of COX-1 due to a weak inhibitory effect of SC-560. Although SC-560 is widely used to inhibit COX-1 and specific inhibition of this enzyme has been shown in cellfree systems, cell studies suggest that this compound may also exert some nonspecific inhibitory effects on COX-2 (31). This possible inhibition of COX-2 might explain why we found some partial inhibitory effects of SC-560 on prostanoid production in our system. To further investigate if SC-560 has COX-1-independent effects, we used astrocytes from COX-1-deficient mice. SC-560 significantly (p Ͻ 0.001) reduced the production of PGE 2 , PGF1␣, and TxB 2 (supplemental Fig. 4) induced by LPS in Cox-1 KO cells, thus indicating that this compound may have COX-1-independent effects.
Prostanoid Production after LPS Treatment Is Dependent on COX-2-We showed above that the induction of COX-2 by LPS was strongly inhibited in MyD88-deficient cells (Fig. 3, H and I).
For this reason, we then examined whether LPS-induced prostanoid production was abrogated in these cells. Compared with the previous findings of prostanoid release after LPS in rat astrocytes, we noticed that mouse astrocytes produced less thromboxane and more prostaglandin than rat astrocytes, although in both species PGE 2 was the prostanoid more abundantly generated in response to LPS. Lack of MyD88 prevented the production of prostanoids after LPS (Fig. 8, A-C), thus further supporting that COX-2 was the main mediator of prostanoid production after this challenge. Because COX-2 was not induced in MyD88 KO cells, the slightly higher production of TxB 2 after LPS than in control MyD88 KO cells might be attributable to COX-1 activity and related to the finding that the basal COX-1 expression was not down-regulated after LPS in MyD88 KO cells (Fig. 4I).
We then used astrocytes obtained from mice deficient in Cox-1 or Cox-2 and their corresponding WT controls to validate the above findings, excluding possible interferences because of nonspecific effects of the drug inhibitors. Astrocytes lacking Cox-2 did not produce PGE 2 (Fig. 8D), PGI 2 (Fig. 8E), or TxA 2 (Fig. 8F) in response to LPS, thus confirming that COX-2 was the main enzyme involved in the production of prostanoids induced by LPS in astrocytes. Under basal nonstimulated conditions, the concentration of TxB 2 was not reduced in cells lacking Cox-2 compared with the WT, although they showed very low levels of PGE 2 and PGF1␣, suggesting that COX-1 is involved in the low basal production of TxA 2 in astrocytes. This is in agreement with the previous observation in MyD88-deficient cells, where LPS did not induce COX-2 but did not downregulate COX-1 either. These cells showed an increase in the production of TxB 2 after LPS (Fig. 8C) that is attributed to the basal activity of COX-1 metabolizing the arachidonic acid newly generated after LPS-induced cPLA 2 activation.
Cox-1-deficient cells produced PGE 2 after LPS to a greater extent than the corresponding WT astrocytes (Fig. 8G), showing that COX-2 is the enzyme responsible for PGE 2 production and suggesting some negative regulatory effect of COX-1 on PGE 2 production after LPS. Also, LPS increased the production of PGI 2 , as assessed by measuring PGF1␣ (Fig. 8H), and TxA 2 , as assessed by measuring TxB 2 (Fig. 8I), in Cox-1-deficient cells suggesting the involvement of COX-2. To add further support to these findings, we silenced Cox-1 with siRNA (Fig. 8J). Under these conditions, a small but significant increase in the production of PGE 2 after LPS was observed (Fig. 8K), whereas the production of PGI 2 (Fig. 8L) and TxA 2 (Fig. 8M) was not altered. Altogether, these results show that COX-1 activity maintains basal production of prostanoids in cultured astrocytes but does not have a major role in the increased production of prostanoids after LPS.

DISCUSSION
These results show that the production of prostanoids induced by LPS in glial cells is essentially mediated by COX-2. The MyD88-dependent pathway and the transcription factor NFB were involved in Cox-2 gene expression. In addition, p38 and JNK MAPK pathways influenced Cox-2 expression, thus revealing a complex regulation of the expression of this gene in response to TLR-4 activation in astroglia. COX-2 induction was accompanied by strong production of PGE 2 and, to a lesser extent, other prostanoids. Several lines of evidence suggest that the COX isoforms are coupled to the activity of the various prostaglandin isomerases favoring the production of certain prostanoids in a cell type-dependent manner (32). The strong production of PGE 2 in astrocytes after LPS is in concordance with the enhanced expression of the microsomal isoform of prostaglandin E synthase-1 (mPGES-1), in agreement with previous findings (33). In addition, here we report that the expression of the mPges-1 gene after LPS was up-regulated through the MyD88 pathway, like Cox-2. Therefore, LPS induces the coordinated expression of COX-2 and mPGES-1, which appeared to be functionally coupled and accounts for the high production and release of PGE 2 in astrocytes, in a manner similar to the responses described in macrophages (34).
In a previous study, induction of COX-2 and mPGES-1 was mainly found in microglia in the substantia nigra 48 h after intracerebral injection of LPS (35). In contrast, we also observed the induction of COX-2 in astrocytes 8 h after LPS injection into the striatum. Besides any regional differences in the reaction to LPS, it is likely that the time course of the glial reaction to this challenge accounts for the observed differences. Increased expression of mPGES-1 has been reported under pathological conditions, e.g. in the brain of Alzheimer disease patients (36), and the enzyme is up-regulated in astrocytes stimulated with ␤-amyloid (37). Also, after intracerebral hemorrhage, strong induction of COX-2 and mPGES-1 was found in astrocytes (38). Therefore, the findings reported here in cultured cells might be relevant to certain neuropathological conditions.
In contrast to the increased expression of COX-2 and mPGES-1, the expression of COX-1 was down-regulated in astrocytes after TLR-4 activation. Reduced expression of COX-1, together with increased COX-2 expression, was previously found in the lungs and hearts of LPS-treated rats (39). In addition, we found that LPS down-regulated the expression of the Ts gene in astrocytes, suggesting some link in the control of the expression of Cox-1 and Ts. Despite down-regulation of Pgis and Ts mRNA, LPS enhanced the production of PGI 2 and TxA 2 but to a lower extent than it increased PGE 2 . This apparently contradictory effect (i.e. reduction of mRNA but increase in enzymatic products) could be due to the time delay needed for an effective reduction of protein content following decreases of constitutive mRNA expression. It is feasible that down-regulation of these genes limits the production of PGI 2 and TxA 2 in astrocytes, while favoring the production of PGE 2 because of up-regulation of mPGES-1. Under basal nonstimulated conditions, Cox-2-deficient astrocytes produced less PGI 2 and TxA 2 FIGURE 8. LPS-induced prostanoid production is dependent on COX-2 although COX-1 exerts selective regulatory effects. Purified cultures of mouse astrocytes were treated with LPS (10 ng/ml), and the concentration of prostanoids in the culture medium was studied by ELISA. Cells were collected at 4 and 8 h (A-C), or 8 h (D-I) after LPS. A-C, production of prostanoids is strongly attenuated in MyD88-deficient cells. D-F, LPS does not induce PGE 2 (G), PGF1␣ (H), or TxB 2 (I) in Cox-2 KO cells (Ϫ/Ϫ). G-I, production of PGE 2 induced by LPS is enhanced in Cox-1-deficient astrocytes (Ϫ/Ϫ) (G). In contrast, the production of PGF1␣ (H) and TxB 2 (I), as indirect assessments of PG and TxA 2 , respectively, is smaller in Cox-1 KO cells (Ϫ/Ϫ) compared with the wild type (ϩ/ϩ), both in the presence or absence of LPS. J-L, silencing COX-1 expression with siRNA slightly enhances the production of PGE 2 (J) but does not modify the production of PGI 2 (K) or TxA 2 (L). n ϭ 3 in at least two independent experiments. One symbol, p Ͻ 0.05; two symbols, p Ͻ 0.01; three symbols, p Ͻ 0.001. * indicates comparison versus control. & indicates comparison versus LPS in WT or after treatment with nonsilencing (ns) RNA. # indicates comparison versus control KO cells.
than the wild type cells, and COX-2-deficient astrocytes showed unaltered concentrations of these prostanoids, suggesting that COX-1 was involved in the basal production of PGI 2 and TxA 2 . However, after LPS, the production of PGI 2 and TxA 2 increased in Cox-1-deficient cells but not in Cox-2-deficient cells, demonstrating that COX-2 was mainly responsible for their up-regulation after this challenge. Likewise, the production of PGE 2 after LPS was fully dependent on COX-2. However, Cox-1-deficient cells produced more PGE 2 after LPS than the corresponding WT cells, suggesting that COX-1 exerts some negative control on COX-2-dependent PGE 2 production after LPS.
Taken altogether, these results demonstrate the key role of COX-2 in prostanoid production after LPS in astrocytes and show that the production of PGE 2 also depends on down-regulation of Cox-1 gene expression. Finally, these findings show that astrocytes respond to proinflammatory triggers with a strong generation of vasoactive PGE 2 that might exert effects on the adjacent brain microvasculature and contribute to modulate CBF responses to neuroinflammation.