Anti-inflammatory effects of prostaglandin E2 in the central nervous system in response to brain injury and circulating lipopolysaccharide


Address correspondence and reprints requests to Serge Rivest, Laboratory of Molecular Endocrinology, CHUL Research Center and, Department of Anatomy and Physiology, Laval University, 2705, boul. Laurier, Québec, G1V 4G2, Canada. E-mail:


Prostaglandin E2, a product of the cyclooxygenation of arachidonic acid released from membrane phospholipids, plays major roles in regulating brain injury and inflammation. Although prostaglandin E2 has frequently been considered as a possible inducer of brain damage and degeneration, it may exert beneficial effects in the CNS. Indeed, in spite of its classic role as a pro-inflammatory molecule, several recent in vitro observations indicate that prostaglandin E2 can inhibit microglial activation. This study investigated the effect of central prostaglandin E2 injection on circulating lipopolysaccharide-induced gene expression of different pro-inflammatory molecules in both vascular and parenchymal elements of the brain. Localized, but strong, expression of tumor necrosis factor-α and interleukin-1β mRNA was found at the edge of the intracerebroventricular tract, which was largely prevented by the central prostaglandin E2 injection. Systemic lipopolysaccharide injection caused a profound transcriptional activation of cyclooxygenase-2 and the inhibitory factor κBα (IκBα, index of NF-κB activity) in the cerebral endothelium and tumor necrosis factor-α in microglial cells across the brain parenchyma. Although exogenous prostaglandin E2 increased lipopolysaccharide-induced NF-κB activity and cyclooxygenase-2 transcription in vascular-associated elements, it significantly reduced microglial activation and tumor necrosis factor-α expression in the brain parenchyma. These results indicate that prostaglandin E2 may play an important role in modulating the immune response occurring at the injured site and the pro-inflammatory signaling events taking place in both vascular- and microglial-associated elements of the CNS.

Abbreviations used

blood–brain barrier




diethyl pyrocarbonate


distrene plasticizer xylene


glucocorticoid-responsive elements


hypothalamic pituitary adrenal


IκB kinase






nuclear factor κB


nitric oxide


peroxisome proliferator-activated receptor γ


refraction density in arbitary units




tumor necrosis factor


von Willebrand Factor

Efficient immune surveillance and tissue repair within the CNS are achieved through a complex and highly regulated network of interactions between resident brain cells and the systemic immune system. Microglial cells, the resident macrophages of the CNS, are important effectors in brain inflammatory, immune and degenerative processes. Depending on the type and intensity of the stimulus and the concurrence of other factors, microglial activation takes place in a graded fashion and can include morphological changes (from ramified to ameboid morphology), increased phagocytosis and the biosynthesis of cytotoxic and/or immunoregulatory molecules. The signaling events involved in these effects have yet to be fully clarified, although accumulating evidence leads to the nuclear factor κB (NF-κB) as being essential in the pro-inflammatory signal transduction pathways occurring in the systemic as well as cerebral myeloid and endothelial cells (Laflamme et al. 1999; Laflamme and Rivest 2000).

One potential candidate for NF-κB is the gene encoding cyclooxygenase-2 (COX-2), the limiting enzyme for the formation of prostaglandins. Of interest is the fact that systemic lipopolysaccharide (LPS) and other experimental models of systemic inflammation cause a profound and transient increase in the transcription of the gene encoding COX-2 and inhibitory factor κBα (IκBα) in the endothelium of the brain capillaries (Lacroix and Rivest 1998; Laflamme and Rivest 1999; Laflamme et al. 1999). It is therefore possible that circulating inflammatory molecules stimulate prostaglandin production via transcriptional activation of COX-2 through NF-κB signaling pathways. One of the main consequences of such a signaling event is the production of different prostaglandins by the cells lining the blood–brain barrier (BBB), which may diffuse across the brain parenchyma and act as inflammatory mediators. Prostaglandin E2 has the ability to activate different populations of hypothalamic neurons that are directly responsible for triggering essential physiological responses during the acute-phase reaction (Zhang and Rivest 1999, 2000). Induction of fever and increase in plasma release of glucocorticoids are representative examples of such physiological outcomes in which prostaglandin E2 plays a key role (for reviews, see Elmquist et al. 1997; Rivest et al. 2000). This prostaglandin was also found to inhibit LPS-induced microglial cell activation and cytokine gene expression and production in vitro (Caggiano and Kraig 1998, 1999; Minghetti and Levi 1998). These data, together with the influence of prostaglandin E2 in stimulating production of anti-inflammatory cytokines (Aloisi et al. 1999), support the concept that it may have a dual role in the CNS, being pro- and anti-inflammatory depending on the target cells and the insults.

A clear understanding of the role of prostaglandins in vivo is made difficult by the fact that activation of the cyclooxygenase cascade is often accompanied by the generation of a broad range of other active molecules, including cytokines, nitric oxide (NO), the prostanoid precursor arachidonic acid itself and its metabolites. Individual prostaglandins may also exert opposite or synergetic effects on common targets and thus, the final contribution of prostaglandins to tissue damage or repair is likely to depend on the balance between the different prostaglandins released by non-neuronal or neuronal cells during specific circumstances. In this study, we investigated the effect of central prostaglandin E2 injection on circulating LPS-induced gene expression of different pro-inflammatory molecules in both vascular and parenchymal elements of the brain. Here, we show that prostaglandin E2 increased LPS-induced NF-κB activity and COX-2 transcription in vascular-associated elements and reduced TNF-α expression in brain microglia. However, prostaglandin E2 largely prevented cytokine expression along the injured site caused by the placement of the guide cannula close to the right lateral ventricle. We therefore suggest that prostaglandin E2 may have a neuroprotective role in modulating the immune response occurring at the injured site and the pro-inflammatory signaling events within endothelial cells of the cerebral microvasculature as well as the parenchymal microglia.

Materials and methods


Fifty-nine adult male Sprague-Dawley rats (175–225 g) were acclimated to standard laboratory conditions (14 h light/10 h dark cycle, lights on at 06.00 and off at 20.00) and free access to rat chow and water. Each rat was used once only for experimentation and all protocols were approved by Laval University's Animal Welfare Committee. Three to four rats were used for each treatment and post-injection time.

Experimental protocols

Animals were anesthetized and the right ventricle was reached stereotaxically (David Kopf Instruments, Tujunga, CA, USA) using the Paxinos and Watson Atlas (Paxinos and Watson 1986). With the incisor bar placed 3.3 mm below the interaural line (horizontal zero), the co-ordinates from bregma for the guide cannula were −0.6 mm anteroposterior, −1.4 mm lateral and −3.1 mm dorsoventral. A 22-gauge stainless-steel guide cannula was implanted close to the right lateral ventricle and secured with screws and cranioplastic cement [cranioplastic powder (Plastic One, Roanoke, VA, USA) and Densply repair material (Densply International, York, PA, USA)]. Before, and two consecutive days after, surgery rats received an anti-inflammatory agent (ketoprofen, 10 mg/kg, i.p.). They were housed individually for a 14-day recuperation period and handled every day.

On the day of the experiment (≈ 09.00), an internal cannula (28 gauge, 1 mm projection beyond the tip of the guide cannula) was connected to the guide cannula. Thereafter, either 2 µg prostaglandin E2 (lot 39H3916, Sigma, St Louis, MO, USA) diluted in 5 µL pyrogen-free, sterile, distilled water or vehicle solution was injected into the right lateral ventricle over 2 min by means of a microinjection pump (model A-99; Razel Scientific Instruments, Stamford, CT, USA). Five minutes later, 500 µg of the bacterial endotoxin LPS/kg of body weight (from Escherichia coli; serotype 055:B5; catalog L-2880, lot 58H4076; Sigma) diluted in 300 µL sterile, pyrogen-free saline or vehicle solution was administrated into the peritoneal cavity. The rats were conscious and freely moving at all times throughout the procedure. Three hours after the systemic LPS injection, animals were anesthetized deeply via an i.p. injection (500 µL) of a mixture of ketamine hydrochloride (91 mg/kg) and xylazine (9.1 mg/kg) and then rapidly perfused transcardially with 0.9% saline, followed by 4% paraformaldehyde in 0.1 m borax buffer (pH 9.5 at 4°C). Another group of animals received an i.c.v. injection of 2 µg prostaglandin E2 followed by an i.p. LPS administration (1.66 mg/kg) and were killed 6 h afterward. This high dose of LPS and time 6 h post injection were selected to induce cytokine gene expression in microglial cells across the brain parenchyma (Nadeau and Rivest 1999), whereas the time 3 h post-treatment with a lower dose of endotoxin was employed to trigger NF-κB activity and COX-2 transcription in the endothelium of the brain capillaries (Lacroix and Rivest 1998; Laflamme and Rivest 1999).

Single in situ hybridization histochemistry

After the transcardiac perfusions, brains were rapidly removed from the skull, post-fixed for 1–3 days and then placed in a solution containing 10% sucrose diluted in 4% paraformaldehyde–borax buffer overnight at 4°C. Frozen brains were mounted on a microtome (model SM2000R; Leica Instruments GmbH, Nussloch, Germany) and cut into 30-µm coronal sections from the olfactory bulb to the end of the medulla. The slices were collected in a cold cryoprotectant solution (0.05 m sodium phosphate buffer pH 7.3 containing 30% ethylene glycol and 20% glycerol) and stored at −20°C. Hybridization histochemical localization of each transcript was carried out on every sixth section of the whole rostrocaudal extent of each brain using 35S-labeled cRNA probes. All solutions were treated with diethyl pyrocarbonate (Depc) and sterilized to prevent RNA degradation. Tissue sections were mounted onto gelatin- and poly(l-lysine)-coated slides, vacuum dried, fixed in 4% paraformaldehyde for 20 min and digested using proteinase K (10 µg/mL in 100 mm Tris/HCl, pH 8.0 and 50 mm EDTA, pH 8.0) at 37°C for 25 min. Brain sections were then rinsed in sterile Depc-treated water followed by a solution of 100 mm triethanolamine (TEA, pH 8.0), acetylated in 0.25% acetic anhydride in 100 mm TEA and dehydrated through graded concentrations of alcohol (50, 70, 95 and 100%). After vacuum drying, 90 µL of the hybridization mixture (106 c.p.m.) was spotted onto each slide, sealed under a coverslip and incubated at 60°C overnight (15–20 h) in a slide warmer. Coverslips were then removed and the slides were rinsed in 4× NaCl/Cit at room temperature. Sections were digested by Rnase A (20 µg/mL, 37°C, 30 min), rinsed in descending concentrations of NaCl/Cit (2×, 1×, 0.5×), washed in 0.1× NaCl/Cit for 30 min at 60°C (1× NaCl/Cit: 0.15 m NaCl, 15 mm trisodium citrate buffer, pH 7.0) and dehydrated through graded concentrations of alcohol. After being dried under vacuum, the sections were exposed at 4°C to X-ray films for 15–60 h (depending on the mRNA), defatted in xylene and dipped in NTB-2 nuclear emulsion (Kodak (Rochester, NY, USA); diluted 1 : 1 with distilled water). Slides were exposed for 7–15 days, developed in D19 developer (Kodak) for 3.5 min at 14–15°C, washed for 15 s in water and fixed in rapid fixer (Kodak) for 5 min. Thereafter, tissues were rinsed in running distilled water for 1–2 h, counterstained with thionin (0.25%), dehydrated through graded concentrations of alcohol, cleared in xylene and coverslipped with distrene plasticizer xylene (DPX) mounting medium.

cRNA probe synthesis and preparation

The bluescript SK plasmid containing the 1.114 kb full-length coding sequence of the mouse IκBα cDNA (kindly provided by Dr Alain Israel, Institut Pasteur, Paris, France; Laflamme and Rivest 1999) was linealized with BamHI and HindIII for the antisense and sense riboprobes, respectively. The pGEM4 plasmid containing the COX-2 cDNA fragment (kindly provided by Dr K. Peri, Ste-Justine Hospital Research Center, Montreal, Canada) was linealized with HindIII and EcoRI for the antisense and sense riboprobes, respectively. The length of COX-2 cDNA fragment was 176 bp consisting of nucleotides 124–300 of the complete published cDNA sequence (Lacroix and Rivest 1998). The plasmid bluescript SK containing a DNA fragment of 716 bp comprising the complete coding sequence of the rat TNF-α (Nadeau and Rivest 1999) was linealized with EcoRI and BamHI for the antisense and sense probes, respectively. The PCRII plasmid containing the 1.360 kb full-length coding sequence of the mouse IL-1β (kindly provided by Dr P. W. Gray, Genentech Inc., CA, USA) was linealized with KpnI and XhoI for the antisense and sense riboprobes, respectively (Herx et al. 2000). Radioactive cRNA copies were synthesized by incubation of 250 ng linealized plasmid in 6 mm MgCl2, 40 mm Tris (pH 7.9), 2 mm spermidine, 10 mm NaCl, 10 mm dithiothreitol, 0.2 mm ATP/GTP/CTP, 200 µCi [α-35S]UTP (Dupont NEN, #NEG039H), 40 U RNAsin (Promega, Madison, WI, USA) and 20 U of either SP6 (COX-2 antisense and IL-1β sense probes), T7 (IκBα and IL-1β antisense, COX-2 and TNF-α sense probes) or T3 (TNF-α antisense and IκBα sense probes) RNA polymerase for 60 min at 37°C. Unincorporated nucleotides were removed using the ammonium-acetate method; 100 µL of Dnase solution (1 µL Dnase, 5 µL of 5 mg/mL tRNA, 94 µL of 10 mm Tris/10 mm MgCl2) was added and 10 min later, extraction was accomplished using a phenol–chloroform solution. The cRNA was precipitated with 80 µL of 5 m ammonium acetate and 500 µL 100% ethanol for 20 min on dry ice. After centrifugation, the pellet was washed with 500 µL 70% ethanol, dried and resuspended in 100 µL of 10 mm Tris/1 mm EDTA. A concentration of 107 c.p.m. probe was mixed into 1 mL of hybridization solution [500 µL formamide; 60 µL 5 m NaCl; 10 µL 1 m Tris pH 8.0, 2 µL 0.5 m EDTA pH 8.0, 50 µL 20× Denhart's solution, 200 µL 50% dextran sulfate, 50 µL 10 mg/mL tRNA, 10 µL 1 m dithiothreitol (118 µL Depc water minus volume of probe used)]. This solution was mixed and heated for 5 min at 65°C prior to being spotted on the slides.

Combination of immunocytochemistry with in situ hybridization

Immunocytochemistry was combined with the in situ hybridization histochemistry protocol to determine the types of cells that express IκBα, COX-2 and TNF-α transcripts. Among the antibodies selected for this study, anti-von Willebrand Factor (vWF) was used to stain the endothelial cells of the microvasculature, whereas anti-ionized calcium binding adapter molecule 1 (iba1) labeled cells of myeloid lineage (macrophages and microglia). Every sixth brain section was processed using the avidin–biotin bridge method with peroxidase as a substrate. Briefly, slices were washed in sterile Depc-treated 50 mm potassium phosphate-buffered saline and incubated 2 h at room temperature with either vWF or iba1 antibody diluted in sterile potassium phosphate-buffered saline + 0.4% Triton X-100 + 1% BSA (fraction V, Sigma) + 0.25% heparin sodium salt USP (ICN Biomedicals Inc., Aurora, OH, USA). Sheep anti-vWF (Cederlane Laboratory Ltd, Can., # Cat CL20176A-R, lot AB22-74) and rabbit anti-iba1 (generously provided by Dr Y. Imai, National Institute of Neuroscience, Kodaira, Tokyo, Japan; Imai et al. 1996) was diluted at 1 : 1000 and 1 : 8000, respectively. After incubation with the primary antibodies, brain slices were rinsed in sterile potassium phosphate-buffered saline and incubated with a mixture of potassium phosphate-buffered saline + 0.4% Triton-X + 1% BSA + 0.25% heparin + biotinylated secondary antibodies (rabbit anti-sheep IgG for vWF or goat anti-rabbit for iba1, 1 : 1500 dilution; Vector Laboratories, CA, USA) for 60 min. Sections were then rinsed with potassium phosphate-buffered saline and incubated at room temperature for 60 min with an avidin–biotin–peroxidase complex (Vectastain ABC elite kit, Vector Laboratories). After several rinses in sterile potassium phosphate-buffered saline, brain slices were reacted in a mixture containing sterile potassium phosphate-buffered saline, the chromagen 3,3′-diaminobenzidine tetrahydrochloride (0.05%) and 0.003% hydrogen peroxide.

Thereafter, tissues were rinsed in sterile potassium phosphate-buffered saline, immediately mounted onto gelatine and poly(l-lysine)-coated slides, desiccated under vacuum for 30 min, fixed in 4% paraformaldehyde for 20 min and digested by proteinase K (10 µg/mL in 100 mm Tris/HCl pH 8.0 and 50 mm EDTA pH 8.0) at 37°C for 25 min. Prehybridization, hybridization and posthybridization steps were performed according to the above description with the difference of dehydration (alcohol 50, 70, 95 and 100%), which was shortened to avoid decoloration of immunoreactive cells (brown staining). After being dried for 2 h under the vacuum, sections were exposed at 4°C to X-ray film (Kodak) for 17 h, defatted in xylene and dipped into NTB-2 nuclear emulsion (Kodak; diluted 1 : 1 with distilled water). Slides were exposed for 10 (IκBα mRNA) or 21 (COX-2 and TNF-α mRNA) days, developed in D19 developer (Kodak) for 3.5 min at 15°C and fixed in rapid fixer (Kodak) for 5 min. Tissues were then rinsed in running distilled water for 1–2 h, rapidly dehydrated through graded concentrations of alcohol, cleared in xylene and coverslipped with DPX. The presence of IκBα, COX-2 and TNF-α transcript was detected by the agglomeration of silver grains in perikarya, whereas vWF and iba1 immunoreactivity within the cell cytoplasm and ramifications (microglia) was indicated by a brown homogeneous coloration. Determination of the double-labeled cells was performed visually for each cell exhibiting clear brown cytoplasm and a number of silver grains within the cell body delineating convincing hybridized message.

Data analysis

The relative intensity of mRNA signals throughout the brain of each animal was assessed on X-ray films and graded according to the scale of undetectable (–), low (+), moderate (++), strong (+++) or very strong (++++) signal. Dipped emulsion slides were examined under microscopic evaluation to ascertain the subcellular localization of the transcript. Semi-quantitative analysis of COX-2 and IκBα mRNA hybridization signals was carried out on nuclear emulsion-dipped slides with an Olympus Optical System (BX-50, Bmax) coupled to a Macintosh computer (Power PC 7100/66) and image software (version 1.59, non-FPU; W. Rasband, NIH). The refraction density in arbitrary units (RDAU) of the hybridization signal was measured under darkfield illumination at a magnification ×100. Sections from experimental and control animals were digitized and subjected to densitometric analysis, yielding measurements of RDAU for each positive cell. An average of 5–10 blood vessels containing 50–60 endothelial cells were quantified for each rat. The RDAU of each cell was then corrected for the average background signal, which was determined by sampling parenchymal tissue immediately outside the cell group of interest.

Data are reported as mean values (± SEM) for experimental and control animals. Statistical analysis was performed by a 2 × 2 analysis of variance (anova), followed by a Bonferroni/Dunn test procedure as posthoc comparisons using the statview program (version 4.01, Macintosh).


Effects of prostaglandin E2 in response to brain injury

Intracerebroventricular cannulation caused localized and robust expression of TNF-α and IL-1β mRNA in the tissue surrounding the lesion of vehicle-treated rats (Fig. 1, left column). The pattern and intensity of both cytokine transcripts were essentially the same, i.e. lining the edge of the tract and spreading into the corpus callosum. Animals that received prostaglandin E2 infusion into the indwelling device failed to exhibit a positive signal for either cytokine mRNA suggesting that the prostaglandin has a potent anti-inflammatory properties on CNS injury-induced cytokine expression. The extent of IL-1β and TNF-α mRNA signal was attenuated by the systemic LPS challenge (Fig. 1, LPS/Veh). A more localized expression pattern was indeed found at the edge of the cannula tract in LPS-treated rats, but the message vanished following prostaglandin E2 administration (Fig. 1, right). The fact that circulating LPS caused a profound increase in plasma glucocorticoids, the most powerful endogenous immunosuppressors, may explain the LPS-induced downregulation of the microglial cells that were activated by the lesion.

Figure 1.

Expression of pro-inflammatory cytokines in rats implanted with a guide cannula in the lateral ventricle. These dark-field photomicrographs show in situ hybridization signals for TNF-α mRNA and IL-1β mRNA at the site of injury in animals that received prostaglandin E2 or the vehicle solution into the internal cannula before being challenged intraperitoneally (i.p.) with the endotoxin lipopolysaccharide (LPS, 3 h; 500 µg/kg of body weight). Note that central prostaglandin E2 injection largely prevented cytokine expression at the injured site in both vehicle- and LPS-treated rats.

Effects of prostaglandin E2 in LPS-induced cellular activation

Systemic LPS injection activated COX-2 and IκBα transcription across the cerebral endothelium. Indeed, positive COX-2 and IκBα hybridization signal was detected along the blood vessels penetrating the brain parenchyma 3 h after the i.p. LPS injection (Fig. 2). Central prostaglandin E2 injection increased the effects of the endotoxin on gene expression within vascular-associated elements (Fig. 2, right).

Figure 2.

Expression of the genes encoding cyclooxygenase-2 (COX-2) and inhibitory factor IκBα in vascular-associated elements of the brain in rats receiving central prostaglandin E2 injection before systemic LPS administration. These dark-field photomicrographs of NTB-2 emulsion-dipped slides depict positive in situ hybridization signals for COX-2 and IκBα mRNAs along the microvasculature 3 h after the i.p. LPS insult (500 µg/kg of body weight). Note that exogenous prostaglandin E2 intensified LPS-induced NF-κB activity and COX-2 transcription in vascular-associated elements. Magnification, ×25.

RDAU was measured on an average of five to ten blood vessels containing 50–60 endothelial cells for each rat. As shown in Table 1, IκBα and COX mRNA signals were significantly higher in the LPS/prostaglandin E2 group than the others. Prostaglandin E2 alone also caused a significant increase in IκBα expression within microvascular-associated elements, although it failed to trigger transcription of the gene encoding COX-2 (Fig. 2 and Table 1). Monitoring IκBα expression is an effective tool for investigating NF-κB activity in the CNS, especially when de novo expression of the transcript is associated with a parallel increase of the possible target genes within the same cells. In this case, however, prostaglandin E2-induced IκBα mRNA may not necessarily be reflecting activation of NF-κB pathways. The increase in the inhibitory factor may also act directly as an endogenous feedback for the pro-inflammatory signal transduction events without previous engagement of the NF-κB signaling.

Table 1.  Average refraction density in arbitrary units of COX-2 and IκBα hybridization signal within brain microvasculature in response to intracerebroventricular prostaglandin E2 and intraperitoneal lipopolysaccharide injections (3 h; 500 µg/kg of body weight)
  1. LPS, lipopolysaccharide; PGE2, prostaglandin E2.
    Analysis of hybridization signals was performed on NTB-2 emulsion-dipped slides as described in Materials and methods. Data are means ± SEM of three or four rats per group.
    *p < 0.01, Veh/prostaglandin E2 versus Veh/Veh; **p < 0.01, LPS/Veh versus Veh/Veh; ***p < 0.01, LPS/prostaglandin E2 versus all other groups.

IκBα032.9 ± 1.4*39.1 ± 1.3**76.5 ± 1.9***
COX-20045.1 ± 1.6**77.8 ± 2.5***

Systemic LPS treatment increased TNF-α expression first in the circumventricular organs and thereafter within microglial cells across the brain parenchyma (Nadeau and Rivest 1999, 2000). The circumventricular organ organum vasculosum of the lamina terminalis, subfornical organ, median eminence and area postrema exhibited a strong hybridization signal for the pro-inflammatory cytokine 6 h after the systemic challenge with the endotoxin (Fig. 3, left). Small scattered positive cells were also found at that time throughout the entire parenchymal brain (Fig. 3, lower left). This phenomenon was significantly altered by the central prostaglandin E2 administration; TNF-α hybridization signal was lower in rats that received a double-treatment with LPS and prostaglandin E2 than those injected centrally with the vehicle solution before being challenged with the bacterial endotoxin (Fig. 3, right).

Figure 3.

Effects of prostaglandin E2 on LPS-induced TNF-α expression in the circumventricular organs (CVOs) and the brain parenchyma. These dark-field photomicrographs of dipped coronal sections into nuclear emulsion NTB-2 exhibit positive TNF-α signals in the CVO subfornical organ (SFO), median eminence (ME), area postrema (AP) and within small scattered cells of the brain parenchyma (bottom) 6 h after the i.p. LPS injection (1.66 mg/kg of body weight). Note the lower TNF-α hybridization signal in the brain of the rat that received prostaglandin E2 infusion before being challenged systemically with the endotoxin (right). Magnification, ×25.

Dual-labeling data

To determine the phenotype of COX-2, IκBα and TNF-α expressing cells, immunocytochemistry was combined with in situ hybridization histochemistry together on the same coronal sections. Antisera directed against iba1 and vWF were used to stain microglial and endothelial cells, respectively. Figure 4 shows different examples of such dual-labeling procedures and presents evidence that COX-2 and IκBα transcripts were colocalized within vWF-ir cells, whereas TNF-α mRNA overlapped with iba1-ir cells in the brain parenchyma of LPS-injected animals. Dense iba1 staining was also detected at the edge of the cannula tract, which showed a convincing agglomeration of silver grains delineating TNF-α-expressing cells (Fig. 4, lower right). However, it is not possible to determine whether these cells are infiltrating macrophages or parenchymal microglia as iba1 antibody may also label systemic cells of monocytic lineage. A mixed population of both infiltrating and parenchymal cells is likely to express the mRNA encoding TNF-α at the site of injury, whereas only microglia were positive for the pro-inflammatory cytokine in the brain parenchyma of LPS-treated rats (Fig. 4, lower left).

Figure 4.

Phenotype of COX-2, IκBα and TNF-α-expressing cells in different structures of the rat brain. Endothelial cells were labeled by immunocytochemistry using an antisera directed against the von Willebrand factor (vWF-ir), whereas cells of myeloid origin (macrophages/microglia) were labeled by means of an antisera directed against the ionized calcium binding adapter molecule (iba1-ir). COX-2, IκBα or TNF-α mRNAs were hybridized thereafter on the same sections using a radioactive in situ hybridization technique (silver grains). Note the presence of the mRNA encoding COX-2 and IκBα within endothelial cells (upper and middle panels, black arrowheads) and the expression of TNF-α mRNA within cells positive for iba1 protein (lower panels, black arrowheads). The upper and middle panels were taken from animals that received i.c.v. prostaglandin E2 infusion and perfused 3 h after i.p. LPS injection (500 µg/kg of body weight). The lower left panel depicts microglial cells expressing TNF-α mRNA in the brain parenchyma of a rat perfused 6 h after a single i.p. LPS bolus (1.66 mg/kg of body weight). The lower right depicts iba1-ir/TNF-α cells lining the tip of the i.c.v. cannula implanted chronically in a rat that received only the vehicle solutions 3 h before being killed. Magnification, ×100.


This study shows that central prostaglandin E2 infusion largely prevented expression of the pro-inflammatory cytokine TNF-α and IL-1β mRNA along the injured site. Systemic LPS injection caused profound transcriptional activation of COX-2 and IκBα in the cerebral endothelium and TNF-α in microglial cells across the brain parenchyma. Although exogenous prostaglandin E2 increased LPS-induced NF-κB activity and COX-2 transcription in vascular-associated elements, it significantly reduced microglial activation and TNF-α expression in the brain parenchyma. These results indicate that prostaglandin E2 may have a neuroprotective role in modulating the immune response occurring at the injured site and the pro-inflammatory signaling events within both vascular- and myeloid-associated elements of the CNS.

Endotoxin LPS, a component of gram-negative bacteria, is a powerful immune challenge associated with an increase in the circulatory levels of numerous cytokines and provokes fever, septic shock and acute-phase reaction in several animals species, including humans. However, brain trauma initiates a cascade of biochemical and molecular changes that activates secondary responses and promotes either regeneration or further damage the nervous system. Inflammation is a key component in this process and is thought to be a major contributor to traumatic brain injury pathophysiology (Kreutzberg 1996). Microglia are the immune cells of the brain parenchyma that respond rapidly to injury and antigenic molecules (Herx et al. 2000). Activated microglia have the ability to produce a wide variety of cytokines that are all involved in orchestrating the inflammatory response at the site of injury (Streit 2000). Appropriate and timely controlled production of anti-inflammatory molecules is essential in this process to prevent an exaggerated response that may be detrimental to the neural element. Prostaglandins, the arachidonic acid metabolites of the cyclooxygenase pathway, are potent local mediators and play major roles in regulating immune responses (Marnett and Kalgutkar 1999). Prostaglandin E2 was found to inhibit lesion formation in dextran sodium sulfate-induced colitis and reduce the levels of mucosal inflammatory cytokines (Sasaki et al. 2000). Prostaglandin E2 also has the ability to inhibit several microglial/macrophage functions, including the expression of major histocompatibility complex class II, NO production and synthesis of pro-inflammatory cytokines (Minghetti et al. 1996; Aloisi et al. 1997; Caggiano and Kraig 1999). Consistent with these reports, our data provide clear evidence that prostaglandin E2 is a powerful molecule to downregulate and/or prevent cytokine gene expression in response to brain injury and circulating LPS.

The exact mechanisms mediating the effects of prostaglandin E2 on the immune response are not fully understood, although the interaction between the ligand with its Gs-coupled EP2 receptor was shown to participate in this process. Indeed, cAMP-elevating agents, forskolin and cAMP analogs, such as dibutyryl cAMP are both effective to enhance LPS-induced COX-2 expression (Levi et al. 1998). Prostaglandin E2 may repress microglial activation by increasing intracellular cAMP levels, because the prostaglandin and cAMP-mimicking agent 8-bromo-cAMP has a similar influence on TNF-α and IL-6 production (Petrova et al. 1999a, 1999b). In cultured rat microglia, the inhibitory effects of prostaglandin E2 were found to be mediated via the EP2 receptor subtype via the cAMP signaling pathway (Caggiano and Kraig 1998, 1999). As shown in this study, the NF-κB-signaling pathway may also be involved in the anti-inflammatory properties of prostaglandins, as exogenous prostaglandin E2 intensified LPS-induced NF-κB activity and COX-2 transcription in vascular-associated elements of the CNS. This may further increase prostaglandin biosynthesis from the endothelium of the cerebrocapillaries to act as endogenous anti-inflammatory molecules on parenchymal microglial cells (Fig. 5).

Figure 5.

Hypothetical mechanisms involved in the anti-inflammatory effects of prostaglandin E2 and other prostaglandins in response to circulating lipopolysaccharide (LPS) and brain injury. See Discussion for details.

Prostaglandin J2 and its metabolite 15d-PGJ2 are naturally occurring derivatives of prostaglandin D2 that have been suggested to exert anti-inflammatory effects by suppressing microglial activation and downregulating inducible NO synthase, TNF-α, IL-1β, IL-6 and COX-2 expression (Petrova et al. 1999a, 1999b; Combs et al. 2000). 15d-PGJ2 is a high affinity ligand for the peroxisome-proliferator-activated receptor (PPARγ) and has been demonstrated to inhibit inflammatory gene transcription in a PPARγ-dependent manner (Jiang et al. 1998; Ricote et al. 1998). Several other lines of evidence suggest that cyclopentenone prostaglandins, such as prostaglandin A1 (Rossi et al. 1997) and 15d-PGJ2 (Straus et al. 2000), may directly inhibit NF-κB-dependent gene expression through covalent modification of critical cysteine residues in IκB kinase and the DNA-binding domains of NF-κB subunits. These findings indicate that 15d-PGJ2 may provide negative feedback in the pro-inflammatory signal transduction pathways and prostaglandin biosynthesis. This is consistent with the presence of 15d-PGJ2in vivo during the critical phase of inflammation (Ricote et al. 1998; Gilroy et al. 1999).

Other major regulators of the immune response are the glucocorticoids that are increased by central prostaglandin E2 injection via a complex neuronal circuit involving corticotropin-releasing factor (CRF) neurons of the endocrine hypothalamus (Zhang and Rivest 1999, 2000). Known as potent suppressors of the immune response, glucocorticoids act on the inflammatory loci to reduce the production and/or secretion of cytokines (Besedovsky and DelRey 1996). They modulate gene expression by interacting with their cytosolic receptors that function as a ligand-dependent transcription factor on genes that contain glucocorticoid-responsive elements (GREs) in their promoters. However, numerous pro-inflammatory genes (such as those encoding for cytokines) that are suppressed by glucocorticoids have no such responsible elements in their promoters (McKay and Cidlowski 1999). Glucocorticoids were found to downregulate NF-κB activity in activating IκBα transcription (Auphan et al. 1995; Scheinman et al. 1995), though mechanisms independent on IκB have also been shown to contribute in the inhibition of NF-κB (Heck et al. 1997). It is therefore possible that the increased levels of glucocorticoids mediate some of the anti-inflammatory properties of prostaglandin E2 on cytokine gene expression in the brain. The role of prostaglandin E2 in activating the neurons that control the corticotroph axis and its paracrine inhibitory action on activated microglia clearly support the concept that this prostaglandin is powerful anti-inflammatory molecule in the CNS and may be an essential player in the mechanisms that prevent an exaggerated immune response, which can be detrimental for the neural tissue. This is in contrast with the general belief that prostaglandin E2 is mainly a pro-inflammatory agonist, at least within the systemic immune system.

Although prostaglandin E2 is generally considered to be a pro-inflammatory molecule, it is rather anti-inflammatory in the CNS. The effects of prostaglandin E2 in preventing cytokine gene expression may be accomplished via different routes (Fig. 5): (i) direct activation of EP2 receptors that trigger cAMP pathway; (ii) increasing NF-κB activity and COX-2 transcription in vascular-associated elements, which in turn may lead to further increase biosynthesis of prostaglandins and their metabolites, such as prostaglandin J2. Endogenous prostaglandin E2 may then act via its EP2 receptor to exert its anti-inflammatory action on the cerebral myeloid cells and through its EP4 receptor to activate the neural circuit involved in the activation of hypothalamic pituitary adrenal (HPA) axis and the plasma release of glucocorticoids; (iii) prostaglandin J2 metabolite may be anti-inflammatory in interfering with NF-κB activity through PPARγ or inhibition of IκB kinase (IKK) phosphorylation and DNA binding. These circuits are obviously quite simplistic and may not be attributable to all models of CNS disorders evolving inflammation, especially with recent data which show that inhibition of prostaglandin synthesis increased the expression of pro-inflammatory cytokines and exacerbated neurodegeneration in Trypanosoma brucei brucei-infected mice (Quan et al. 2000). However, sodium salicylate was used in the latter study and this drug is not a specific COX inhibitor, but interferes with NF-κB activity that is known to play a key role in allowing or not inflammation-induced apoptosis and cell death (Baichwal and Baeuerle 1997). More studies would therefore be needed to provide clear evidence that selective inhibitors of COX activity (ketorolac, indomethacin), and not those associated with interference of the NF-κB signaling, exacerbate or attenuate protozoa infection-induced inflammation in the cerebral tissue.


This research was supported by the Canadian Institutes of Health Research [CIHR; was the Medical Research Council of Canada (MRCC)]. Serge Rivest is an MRCC scientist. Ji Zhang is supported by an MRCC Studentship. The authors thank Dr Alain Israel (Institut Pasteur, Paris, France) for the gift of the plasmid containing the mouse IκBα cDNA, Dr K. Peri (Ste-Justine Hospital Research Center, Montreal, Canada) for COX-2 cDNA, Dr p.W. Gray (Genentech Inc., CA, USA) for the mouse IL-1β cDNA and Dr Y Imai (National Institute of Neuroscience, Kodaira, Tokyo, Japan) for the iba1 antisera.