PGE₂-EP₂ signalling in endothelium is activated by haemodynamic stress and induces cerebral aneurysm through an amplifying loop via NF-κB

Abstract

BACKGROUND AND PURPOSE Cerebral aneurysm is a frequent cerebrovascular event and a major cause of fatal subarachnoid haemorrhage, but there is no medical treatment for this condition. Haemodynamic stress and, recently, chronic inflammation have been proposed as major causes of cerebral aneurysm. Nevertheless, links between haemodynamic stress and chronic inflammation remain ill-defined, and to clarify such links, we evaluated the effects of prostaglandin E₂ (PGE₂), a mediator of inflammation, on the formation of cerebral aneurysms.

EXPERIMENTAL APPROACH Expression of COX and prostaglandin E synthase (PGES) and PGE receptors were examined in human and rodent cerebral aneurysm. The incidence, size and inflammation of cerebral aneurysms were evaluated in rats treated with COX-2 inhibitors and mice lacking each prostaglandin receptor. Effects of shear stress and PGE receptor signalling on expression of pro-inflammatory molecules were studied in primary cultures of human endothelial cells (ECs).

KEY RESULTS COX-2, microsomal PGES-1 and prostaglandin E receptor 2 (EP₂) were induced in ECs in the walls of cerebral aneurysms. Shear stress applied to primary ECs induced COX-2 and EP₂. Inhibition or loss of COX-2 or EP₂in vivo attenuated each other's expression, suppressed nuclear factor κB (NF-κB)-mediated chronic inflammation and reduced incidence of cerebral aneurysm. EP₂ stimulation in primary ECs induced NF-κB activation and expression of the chemokine (C-C motif) ligand 2, essential for cerebral aneurysm.

CONCLUSIONS AND IMPLICATIONS These results suggest that shear stress activated PGE₂-EP₂ pathway in ECs and amplified chronic inflammation via NF-κB. We propose EP₂ as a therapeutic target in cerebral aneurysm.

Abbreviations

ACA

anterior cerebral artery

CA

cerebral aneurysm

COX

cyclooxygenase

EC

endothelial cell

EP₂

prostaglandin E receptor 2

IEL

internal elastic lamina

IL

interleukin

iNOS

inducible nitric oxide synthase

MMP

matrix metalloproteinase

NF-κB

nuclear factor κB

OA

olfactory artery

PGES

prostaglandin E synthase

Ptger2

EP₂ receptor gene

SMC

smooth muscle cell

Introduction

Subarachnoid haemorrhage is a serious cardiovascular event. It is fatal in 45% of patients within 30 days of onset, whereas 30% suffer from moderate to severe morbidity (van Gijn et al., 2007). The main cause of subarachnoid haemorrhage is a rupture of a pre-existing cerebral aneurysm, which is seen in 1–5% of the general public (Wiebers et al., 2003). Given this high prevalence and susceptibility to subarachnoid haemorrhage, treatment of cerebral aneurysm before rupture is important. Currently, there is no medical treatment that would directly interfere with cerebral aneurysm formation because the pathogenesis of these aneurysms remains unknown. In addition, surgical procedures for cerebral aneurysm have a risk of complication, even though it is low. Therefore, most patients are only given treatments to control some risk factors such as hypertension, rather than any direct treatment for the aneurysm itself.

To elucidate the molecular mechanisms of cerebral aneurysm formation, we established a rodent model of cerebral aneurysm (Hashimoto et al., 1978; Morimoto et al., 2002) through inducing haemodynamic stress at bifurcation sites of cerebral arteries. Haemodynamic force is considered to be a trigger of cerebral aneurysm formation by computational simulation (Alnaes et al., 2007; Takeuchi and Karino, 2009), which demonstrated a high haemodynamic force at the site of cerebral aneurysms. Using the above model, we have been examining events leading to cerebral aneurysm formation. Recently, we have shown that chronic inflammation, mediated by nuclear factor κB (NF-κB) activation, increased the formation of cerebral aneurysms (Fukuda et al., 2000, Sadamasa et al., 2003; Moriwaki et al., 2006; Aoki et al., 2007a,b; Aoki and Nishimura, 2010). However, factors that link haemodynamic forces to NF-κB activation in the walls of cerebral aneurysms remain unclear, and thus, compounds aimed at medical treatment of these aneurysms are still to be discovered.

The prostanoids are a group of lipid mediators and are produced and released in response to various stimuli. Prostaglandin E₂ (PGE₂) is a prostanoid involved in various inflammatory diseases. PGE₂ is synthesized from arachidonic acid by sequential actions of two enzymes, cyclooxygenase (COX) and PGE synthase (PGES) (Tilley et al., 2001). COX has two isoforms, COX-1 and COX-2, whereas PGES has three isoforms – microsomal PGES1 (mPGES1), mPGES2 and cytosolic PGES (cPGES) (Murakami et al., 2002). Among them, COX-2 and mPGES1 are inducible enzymes in inflammatory events and are responsible for PGE₂ production in various inflammatory diseases. PGE₂ exerts its actions through a family of G protein-coupled receptors, EP₁ to EP₄ receptors (Narumiya et al., 1999; Breyer et al., 2001; Sugimoto and Narumiya, 2007; nomenclature follows Alexander et al., 2009). Given that COX-2 is known to be induced in response to haemodynamic force and that PGE₂ elicits a variety of actions in inflammation, we hypothesized that PGE₂ signalling was involved in the pathogenesis of cerebral aneurysms and linked these haemodynamic forces with NF-κB activation (Topper et al., 1996; Ogasawara et al., 2001).

In the present study, we have examined PGE₂ production, expression of related enzymes and PGE receptor subtypes, and demonstrated that PGE₂ signalling activated NF-κB through EP₂ and triggers chronic inflammation in cerebral aneurysm walls. We suggest COX-2 and EP₂ as candidate drug targets against cerebral aneurysms.

Methods

Animals

All animal care and experimental use complied with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and was approved by the Institutional Animal Care and Use Committee of Kyoto University Graduate School of Medicine. Sprague Dawley rats and C57BL/6NCrSlc mice were purchased from Japan SLC (Shizuoka, Japan). Mice lacking the gene for each PG receptor, that is Ptger1, Ptger2, Ptger3, Ptger4, Ptgir (also known as IP) or Tbxa2r (also known as TP), were generated, backcrossed and bred as previously described (Kobayashi and Narumiya, 2002). Animals were maintained on a light/dark cycle of 14 h/10 h with free access to chow and water. In total, 130 rats and 190 mice were used in the present study.

Rodent cerebral aneurysm models and histological analysis of cerebral aneurysms

To induce cerebral aneurysms, 7-week-old rats or mice were subjected to the ligation of the left carotid artery, and systemic hypertension was induced by salt overloading and the ligation of left renal artery under general anaesthesia following pentobarbital injection (50 mg·kg⁻¹ i.p.), without any neuromuscular blocking agents. This procedure is designed to increase the haemodynamic stress on arterial walls of a bifurcation from the right carotid artery to the bifurcation site of the anterior cerebral artery (ACA) and olfactory artery (OA). Sham-operated animals were treated in the same manner except for ligation of the carotid arteries and the renal artery. Following initiation of the aneurysm model, animals were fed special chow containing 8% sodium chloride and 0.12% β-aminopropionitrile (Tokyo chemical industry, Tokyo, Japan), an inhibitor of lysyl oxidase that catalyzes the cross-linking of collagen and elastin. Induction of cerebral aneurysms at the right ACA–OA bifurcation was assessed in rats and mice at 3 and 5 months after the start of the procedures described above respectively. In the histological analysis, rats and mice were perfused with 4% paraformaldehyde, and serial sections of cerebral aneurysm walls were made. Then, sections were subjected to Elastica van Gieson staining. Induced cerebral aneurysms were histologically defined as follows. Internal elastic lamina (IEL) disruption is a lesion without outward bulging of arterial walls. Advanced aneurysm is an outward bulging of the arterial wall with IEL fragmentation. Aneurysm size was calculated as a mean of the largest transverse diameter and the height of the aneurysmal dome. In some experiments, celecoxib, 4-[5-(4-methylphenyl)-3-(trifluoromethyl)-1H-pyrazol-1-yl] benzenesulfonamide, a selective COX-2 inhibitor and a generous gift from Pfizer (New York, NY, USA), was mixed in the chow and fed during the indicated period (150 mg·kg⁻¹·day⁻¹).

Measurement of PGE₂ in cerebral aneurysms

The right ACA–OA bifurcation was dissected, rapidly frozen in liquid nitrogen and pulverized. PGE₂ was extracted, and its content was measured by ELISA (Biotrak EIA system, GE Healthcare, Buckinghamshire, UK) according to the manufacturer's instructions. Prostanoid concentrations were normalized to the wet weight of the tissue.

RNA extraction and quantitative real-time polymerase chain reaction (RT-PCR)

At 1 or 3 months after the initiation of cerebral aneurysm, rats were killed as described above. To obtain enough RNA to detect EP₂ mRNA expression in RT-PCR analysis, total RNA was extracted from the left half of the anterior portion of the circle of Willis, which is composed of the cerebral aneurysm lesion and surrounding ‘non-cerebral aneurysm’ region, and converted into cDNA using RNeasy Fibrous Tissue Mini Kit and Sensiscript reverse transcriptase (Qiagen, Hilden, Germany). Quantitative RT-PCR was done with QuantiTect SYBR Green PCR Kit (Qiagen) and LightCycler quick system 330 (Roche, Basel, Switzerland). β-actin was used as an internal control. The primer sets used for rat samples were: 5’-gtccgtgaagatgcgctacc-3’ and 5’-ccacatttatggagatagtc-3’ for Cox-1, 5’-cggtggagaggtgtatcctc-3’ and 5’-ggagcacaacagagtgtgtg-3’ for Cox-2, forward 5’-gttcacccatcagtttcc-3’, reverse 5’-caagacacagcaggcatagg-3’ for Ptger1, forward 5’-taatggccaggagaatgagg-3’, reverse 5’-aggaccgcataccttcagc-3’ for Ptger2, forward 5’-cagtccgaacactgtcatgg-3’, reverse 5’-ctttcctgctgtgcattgg-3’ for Ptger3, forward 5’-gaagtaggcgtggttgatgg-3’, reverse 5’-cctactgggcacattgttgg-3’ for Ptger4, forward 5’-aggcatttgtgaggtgaagg-3’, reverse 5’-catgaaggctgaaccagagg-3’ for mPges1, forward 5’-gactgccctacaagcacagc-3’, reverse 5’-gatgcgacactcacacatcc-3’ for mPges2. The primer sets used for human cells were: 5’-cttcaatgagtaccgcaagagg-3’ and 5’-agtccagggtagaactccaacg-3’ for Cox-1, 5’-acaccctctatcactggcatcc-3’ and 5’-aacattcctaccaccagcaacc-3’ for Cox-2, forward 5’-tttgctgcactcttctctctgg-3’, reverse 5’-actgaagggaccagaaagttcc-3’ for mPges1, forward 5’-aagtactggctcatgctcaacg-3’, reverse 5’-gtagtcaaaggacgccagagc-3’ for mPges2, forward 5’-agcttctttggacacttgc-3’, reverse 5’-atagcagccaccttcattcc-3’ for CCL2 (also known as Mcp-1). For quantification, the second derivate maximum method was used for crossing point determination, using LightCycler Software 3.3 (Roche).

Antibodies, immunohistochemistry, Western blot analysis and gelatin zymography

Primary antibodies used in this study are as follows: anti-COX-1 (SC-7950, SantaCruz Biotechnolgy, Santa Cruz, CA, USA), anti-COX-2 (SC-1747, SantaCruz Biotechnolgy), anti-EP₂ (101750, Cayman, Ann Arbor, MI, USA), anti-mPGES1 (160140, Cayman), anti-mPGES2 (160145, Cayman), anti-cPGES (10209, Cayman), anti-phospho-p65 NF-κB subunit (Ser536) (3033, Cell Signaling, Danvers, MA, USA) and anti-human CCL2 (MAB2791, R&D, Minneapolis, MN, USA). Immunohistochemistry, Western blot analysis and gelatin zymography were performed as previously described (Aoki et al., 2007b; 2009b). Briefly, 1 or 3 months after cerebral aneurysm induction, all rats were deeply anaesthetized and perfused transcardially with 4% paraformaldehyde (immunohistochemistry) or phosphate-buffered saline (PBS) (Western blot analysis and gelatine zymography). For immunohistochemistry, the ACA–OA bifurcation was dissected and embedded in optimum cutting temperature compound. Thin sections (5 µm) were cut and mounted on silane-coated slides. After blocking with 5% donkey serum (Jackson ImmunoResearch, Baltimore, MD, USA), the sections were incubated with primary antibodies, followed by the incubation with fluorescence-labelled secondary antibodies (Jackson ImmunoResearch). Fluorescent images were acquired through a conventional epifluorescent microscope (BX51N-34-FL-1, Olympus, Tokyo, Japan). Whole cell lysate was extracted from the anterior portion of the circle of Willis, composed of cerebral aneurysm lesions and ‘non-cerebral aneurysm’ regions, using Complete Lysis M Kit (Roche). Although cerebral aneurysms were formed in the left side, it was necessary to pool tissues from both the left and the right sides to detect EP₂ reproducibly. As a result, the collected tissue included the tissue from the right side, which may have diluted or masked a change in EP₂ expression in cerebral aneurysm walls. A sample (15 µg) of the whole cell lysate was used for each assay. After sodium dodecyl sulphate polyacrylamide gel electrophoresis, separated proteins were transferred to polyvinylidene fluoride membranes (Hybond-P, GE Healthcare) and blocked with enhanced chemiluminescent (ECL) plus blocking agent (GE Healthcare). The membranes were then incubated with primary antibodies followed by incubation with horseradish peroxidase-conjugated anti-IgG antibody (GE Healthcare). The signal was detected by chemiluminescent reagent (ECL Plus Western Blotting Detection System, GE Healthcare). α-Tubulin served as an internal control. In gelatin zymography, whole cell lysate was applied to a Gelatin Zymo Electrophoresis Kit (Primary cell, Sapporo, Japan) according to the manufacturer's instructions. The gelatinolytic activity of each group was measured by densitometry and calculated as a ratio to that of the sham-operated group.

Human cerebral aneurysm samples were obtained with informed consent and approval of the ethical committee at Kyoto University from five patients with unruptured cerebral aneurysms during neurosurgical procedures. Control arterial walls were dissected from cadavers that had been preserved by freezing as quickly as possible after death to limit degenerative change (n= 3).

Primary cultures of endothelial cells from human carotid artery

Primary endothelial cells (ECs) from human carotid artery were purchased from Cell Applications (San Diego, CA, USA). Characteristics of these cells were confirmed to be compatible with those of ECs by morphological examination, Western blot analysis for von Willebrand factor, immunostaining for vascular endothelial-cadherin and uptake of acetyl low-density lipoprotein (LDL) (Supporting Information Figure S1A–D). In some experiments, ECs were incubated with selective agonists to each EP receptor subtype. Selective EP receptor agonists (EP₁: ONO-D1-004, Ki = 0.15 µM, EP₂: ONO-AE1-259, Ki = 0.003 µM, EP₃: ONO-AE-248, Ki = 0.0075 µM, EP₄: ONO-AE1-329, Ki = 0.0097 µM) were provided by Ono Pharmaceutical Co. Ltd (Osaka, Japan) (Sugimoto and Narumiya, 2007).

Loading shear stress

Primary ECs were cultured on gelatin-coated glass slides and were loaded with shear stress at 1.5 Pa (15 dyne·cm²), which was equivalent to a stress experimentally proven to be loaded on cerebral aneurysm walls of rats in vivo (Nakatani et al., 1991) and also frequently used in shear stress experiment (Inoue et al., 2002), with a custom-made apparatus, as previously described (Yamamoto et al., 2006). After 24 h of shear stress loading, cells were subjected to RT-PCR analysis or immunohistochemistry.

EP₂ receptor gene (Ptger2) knockdown by RNAi in ECs

Primary ECs were transfected with Ptger2 siRNA (5’-caguauuacucuuacaagagu-3’ and 5’-ucuuguaagaguaauacugua-3’) or scrambled siRNA at 100 nM, which had been shown in preliminary studies to be effective in depleting Ptger2 mRNA. From 72 h after the transfection, ECs were incubated with the EP₂ agonist (ONO-AE1-259) for 24 h and were then subjected to RT-PCR and Western blot analyses.

Primary culture of smooth muscle cells

Primary culture of smooth muscle cells (SMCs) from rat cerebral arteries was established as previously described (Ishibashi et al., 2010). Briefly, the whole circle of Willis was isolated from rats under deep anaesthesia. Dissected tissue was cut into small pieces and incubated with collagenase and elastase under gentle shaking. Dispersed cells were collected and incubated in a fibronectin-coated dish with medium containing epithelial growth factor, fibroblast growth factor, insulin and fetal bovine serum.

NF-κB decoy oligodeoxynucleotide treatment

NF-κB decoy oligodeoxynucleotide (ODN) and scrambled ODN were synthesized as previously described (Morishita et al., 2004). In rat cerebral aneurysm experiments, 40 µg of decoy ODN dissolved in PBS was injected into the cerebrospinal fluid space from the foramen magnum, as previously described (Aoki et al., 2007b). In primary EC experiments, decoy ODN or scrambled ODN (400 nM) was transfected. At 4 h after the transfection, treatment with the EP₂ agonist (ONO-AE1-259) was initiated.

Statistical analysis

Data are shown in mean ± SEM. Differences between two groups were statistically examined using Mann–Whitney test. Statistical comparisons among more than two groups were conducted using Kruskal–Wallis test followed by post hoc Dunn's test. The cerebral aneurysm incidence across groups was statistically compared with Fisher's exact test.

Results

Induction of COX-2–PGE₂–EP₂ signalling in cerebral aneurysm walls

We first examined the cerebral artery of patients with cerebral aneurysms and found extensive expression of COX-2 in the cerebral aneurysm walls but not in other regions (Figure 1A; Supporting Information Figure S2A). To further confirm the involvement and actions of PGs in cerebral aneurysm development, we used the rodent model, in which rats were subjected to unilateral carotid ligation and systemic hypertension by salt overloading (Aoki et al., 2007b). After 3 months, cerebral aneurysm develops at the contralateral arterial bifurcation site in the circle of Willis, where cerebral aneurysm is also observed in most human patients (Wiebers et al., 2003; van Gijn et al., 2007) (Figure 1B). We found that, for the two COX isoforms, COX-2 but not COX-1 was induced at both mRNA and protein levels in cerebral arteries during cerebral aneurysm induction, a finding similar to that observed in walls of human cerebral aneurysm (Figure 1C and D; Supporting Information Figure S3A–C). In addition, mPGES1, a major inducible form of PGES, but not mPGES2, was induced in cerebral arteries (Figure 1C and D; Supporting Information Figure S3D). Consistent with the induction of this enzyme, the amount of PGE₂ significantly increased in cerebral aneurysm walls compared with that found in the sham-operated artery (Figure 1E). PGE₂ primarily binds to G protein-coupled receptors, EP₁–EP₄, for its action (Sugimoto and Narumiya, 2007). We analysed expression of EP receptor subtypes and found that only EP₂ expression was induced at cerebral aneurysms (Figure 1F; Supporting Information Figure S4), suggesting that PGE₂ produced by the COX-2–mPGES1 pathway acted on EP₂. Immunostaining revealed that expression of COX-2, mPGES1 and EP₂ was predominantly present in the EC layer of cerebral aneurysm walls at 2 weeks, the early stage of cerebral aneurysm formation (Figure 1G). At 3 months, when cerebral aneurysm formation was apparent, COX-2 and mPGES1 expression was extended to other regions such as SMCs in the aneurysm walls, suggesting augmentation of the production of PGE₂ at a later stage (Figure 1G).

Figure 1

PGE₂-EP₂ signalling was induced during cerebral aneurysm formation. (A) COX-2 expression in cerebral aneurysm walls of human patients (left, middle) and the wall of a control cerebral artery (right). Bar = 50 µm. (B) Location and time course of cerebral aneurysm induction in the rat model. The circle of Willis of the rat is shown in the left panel. The boxed region shows the ACA–OA bifurcation where cerebral aneurysm was induced. Microscopic images, stained with Elastica van Gieson, of dissected arterial walls of this region are shown in the right panels. Arrows show the boundaries of disruption of internal elastic lamina. OA: olfactory artery. ACA, MCA: anterior or middle cerebral artery. ICA: internal carotid artery. 0,1 or 3 M: before and at 1 or 3 months after induction of cerebral aneurysm. Bar = 20 µm. (C, D) RT-PCR (C, n= 6) and Western blot analysis (D, n= 5) for expression of isoforms of COX and PGES in rat cerebral artery during the formation of cerebral aneurysm. Data represent mean ± SEM. *P < 0.05, **P < 0.01. (E) PGE₂ content measured by ELISA in the ACA–OA bifurcation of rats after induction of cerebral aneurysm (n= 5). Data represent mean ± SEM. *P < 0.05, **P < 0.01. (F) RT-PCR (n= 6) and Western blot analysis (n= 5) for expression of PGE receptor (EP) subtypes in rat cerebral artery during the formation of cerebral aneurysm. Data represent mean ± SEM. *P < 0.05 compared with the level before induction of cerebral aneurysm (0 M). **P < 0.01. (G) Immunostaining for COX-2, mPGES1 and EP₂ in cerebral aneurysm walls during cerebral aneurysm formation. The cerebral aneurysm walls of rats 2 weeks (2 wks) or 3 months (3 M) after induction of cerebral aneurysm were dissected and stained. EvG: Elastica van Gieson staining. The endothelial cell layer is defined as a cell layer outside the internal elastic lamina. Bar = 20 µm. (H) RT-PCR analysis for Cox-2 and Ptger2 mRNA expression in cultured endothelial cells exposed to shear stress (n= 7). Data represent mean ± SEM. *P < 0.05, **P < 0.01. (I) Immunostaining for EP₂ in the cerebral aneurysm walls of human patients (left) and the corresponding region of control subjects (right). Bar = 50 µm.

These results indicated that hypertensive stress acts first on ECs at the prospective site of a cerebral aneurysm in the cerebral artery and induces COX-2, mPGES1 and EP₂. To confirm this possibility, we used primary culture of ECs from human carotid artery (Supporting Information Figure S1A–D), exposed them to shear stress in vitro and examined the effect of shear stress on the expression of these molecules. Shear stress induced COX-2 mRNA and protein in these cells (Figure 1H; Supporting Information Figure S1E and F). Shear stress further significantly elevated Ptger2 mRNA in these cells (Figure 1H). Consistent with these findings, we noted strong expression of EP₂ in the cerebral aneurysm walls but not in other cerebral arterial regions of human patients (Figure 1I; Supporting Information Figure S2B). These data support the proposal that shear stress induced PGE₂-EP₂ receptor signalling in ECs at the site of cerebral aneurysm formation and suggest that PGE₂-EP₂ signalling could link hypertensive stress to the formation of cerebral aneurysms.

Involvement of COX-2-EP₂ signalling in the pathogenesis of cerebral aneurysm formation

To examine any causal relationship between EP receptors and the development of cerebral aneurysms, we subjected mice deficient in each EP receptor subtype (Kobayashi and Narumiya, 2002; Aoki et al., 2007b) to the procedure for inducing cerebral aneurysms. Consistent with the altered expression of EP₂ in cerebral aneurysm walls (Figure 1F and I), cerebral aneurysm formation was almost absent in Ptger2-deficient mice, but not in the other three knockout strains (Figure 2A). Ptger2 heterozygous mice showed intermediate incidence of cerebral aneurysm formation between wild-type and Ptger2-deficient mice, suggesting the gene dosage effect of Ptger2. Ptger2-deficient mice did not show any apparent abnormality in the structure of the circle of Willis or of the systemic blood pressure (Supporting Information Figure S5). In addition, mice deficient in the receptors for thromboxane (TX) A₂ and PGI₂, prostanoids involved in atherosclerosis (Kobayashi et al., 2004), developed cerebral aneurysms similarly to wild-type mice (Supporting Information Figure S6).

Figure 2

Effects of EP₂ deletion (knockout) and COX-2 inhibition on cerebral aneurysm formation. (A) Induction of cerebral aneurysm in mice lacking each EP receptor subtype. Aneurysmal changes were histologically assessed at 5 months after induction of cerebral aneurysm. Numbers of animals are shown in parentheses. *P < 0.05 compared with the incidence in wild-type mice (WT). (B) Effects of COX-2 inhibition. Rats were treated with or without celecoxib during the whole experimental period (0–3 months) or its former (0–1.5 months) or latter (1.5–3 months) half. Aneurysmal changes (left) and the size of aneurysm (right) were histologically assessed and are shown for each group. The number of animals is shown in parentheses under each bar. M, month. Data represent mean ± SEM. *P < 0.05, **P < 0.01. (C) Positive feedback induction between COX-2 and EP₂. Rats were treated with or without celecoxib for 3 months and examined for expression of EP₂ by immunostaining (left) and Western blot analyses (middle). Ptger2-heterozygous (Ptger2^+/−) and Ptger2-knockout (Ptger2^−/−) mice were subjected to induction of cerebral aneurysm for 5 months and examined for expression of COX-2 by immunostaining (right). Bar = 20 µm.

To confirm the involvement of the PGE₂-EP₂ signalling in the pathogenesis of cerebral aneurysm, we administered celecoxib, a selective COX-2 inhibitor (Everts et al., 2000), to rats subjected to cerebral aneurysm induction. Celecoxib, given orally for 3 months, significantly reduced the incidence of advanced aneurysm (Figure 2B) without affecting the systemic blood pressure (Supporting Information Figure S7A). Note that the celecoxib treatment during the earlier half (0–1.5 months), but not during the latter half (1.5–3.0 months), significantly reduced both the incidence and the size of cerebral aneurysm to levels similar to those found in the whole-period treatment (Figure 2B). Given the expression of COX-2, mPGES1 and EP₂ in ECs at the early stage of cerebral aneurysm induction (Figure 1G), these data suggest that PGE₂-EP₂ signalling acts primarily in ECs at the initial stage and triggers cerebral aneurysm formation. We next wondered whether COX-2 and EP₂ formed a positive feedback loop because each mutually affects the expression of the other under various physiological and pathological conditions (Sonoshita et al., 2001). Indeed, the celecoxib treatment suppressed EP₂ expression in cerebral aneurysm walls (Figure 2C). Conversely, COX-2 expression in cerebral aneurysm walls was suppressed in Ptger2-deficient mice (Figure 2C). Thus, a positive feedback between COX-2 and EP₂ also operates in cerebral aneurysm walls.

Regulation of chronic inflammation in cerebral aneurysm walls via COX-2-EP₂ signalling in vivo

Earlier studies have demonstrated that chronic inflammation characterized by infiltration of macrophages was present in the cerebral aneurysm wall and played a causative role in cerebral aneurysm formation (Chyatte et al., 1999; Fukuda et al., 2000; Moriwaki et al., 2006; Aoki et al., 2007a,b; 2009b). In our present experiments, we noted that macrophage infiltration was suppressed in cerebral aneurysm walls of both Ptger2-deficient mice and rats treated with celecoxib (Figure 3A; Supporting Information Figure S7B). As NF-κB is critical for the transcription of various pro-inflammatory mediators in cerebral aneurysm lesions (Aoki et al., 2007b), we next examined the effects of EP₂ deficiency or COX-2 inhibition on the phosphorylation of NF-κB p65 subunit and expressions of downstream molecules, such as the chemokine CCL2, interleukin (IL)-1β, inducible nitric oxide synthase (iNOS) and matrix metalloproteinase (MMP)-2 (Fukuda et al., 2000; Sadamasa et al. 2003; Moriwaki et al., 2006; Aoki et al., 2007a,b; 2008a; 2009b). EP₂ deficiency significantly suppressed NF-κB activation and down-regulated induction of all of these molecules as shown by both immunostaining and Western blot analyses (Figure 3B and C). The celecoxib treatment also suppressed NF-κB activation and the induction of MMP-2, CCL2 and IL-1β (Supporting Information Figure S7C, D and F). We also visualized MMP-2 activity using gel zymography and confirmed the inhibitory effect of celecoxib (Supporting Information Figure S7E). Consistent with the fact that COX-2 is transcriptionally regulated by NF-κB activation (Schmedtje et al., 1997), NF-κB activation and COX-2 expression were co-localized in cerebral aneurysm walls of the rats (Supporting Information Figure S7G). Furthermore, NF-κB inhibition by decoy ODN, a specific inhibitor of NF-κB transcriptional activity (Morishita et al., 2004), suppressed COX-2 expression in cerebral aneurysm walls, suggesting the presence of a positive feedback loop between COX-2 and NF-κB (Supporting Information Figure S7H) via EP₂ (see below). These results, combined together, suggest that shear stress-induced PGE₂-EP₂ receptor signalling activates NF-κB in ECs at an early stage and contributes to chronic inflammation in arterial walls for cerebral aneurysm formation.

Figure 3

EP₂ signalling mediates chronic inflammation in cerebral aneurysm walls. (A) Effects of EP₂ deficiency on macrophage infiltration in cerebral aneurysm walls. The number of CD68-positive macrophages was determined in immunostained sections. The number of animals is shown in parentheses under each bar. Data represent mean ± SEM. **P < 0.01. Representative images are shown in the right panel. Bar = 20 µm. (B) Immunostaining for pro-inflammatory mediators in walls of cerebral aneurysms in littermates of Ptger2 heterozygous (Ptger2^+/−) and Ptger2-deficient (Ptger2^−/−) mice. Black arrows in Elastica van Gieson (EvG) staining (left) indicate the centre of the cerebral aneurysm or its corresponding position. White arrows indicate the internal elastic lamina of arterial walls. Bar = 20 µm. p-p65: phosphorylated NF-κB p65 subunit. (C) Effect of EP₂ deficiency on NF-κB activation and expression of pro-inflammatory mediators in cerebral arteries. Representative images of Western blot analyses are shown in the upper panels and the quantification by densitometric analysis is shown in the lower panel (n= 5). α-Tubulin was used as an internal control. 0 or 5 M: before and at 5 months after induction of cerebral aneurysm. Data represent mean ± SEM. *P < 0.05.

Regulation of NF-κB mediated CCL2 expression in ECs via PGE₂-EP₂ signalling in vitro

Given that NF-κB is activated first in ECs and then in various cell component in cerebral aneurysm walls (Aoki et al., 2007b) and that the inhibition of PGE₂-EP₂ signalling suppressed NF-κB activation (Figure 3; Supporting Information Figure S7), we hypothesized that EP₂-mediated NF-κB activation induced CCL2 expression in ECs to trigger cerebral aneurysm formation. To corroborate this hypothesis, we again used primary culture of ECs from human carotid artery and examined the effect of the EP₂ agonist (ONO-AE1-259) on the phosphorylation of NF-κB and expression of CCL2. We found that PGE₂ significantly induced CCL2 in ECs (Supporting Information Figure S8A, B). ONO-AE1-259, but not agonists for other EP receptor subtypes, mimicked this increase in CCL2 (Figure 4A and B). We then examined whether NF-κB is involved in EP₂-mediated induction of CCL2. ONO-AE1-259 induced NF-κB phosphorylation in these cells (Figure 4B), which was blocked by RNAi for Ptger2, the gene coding for EP₂ (Figure 4C; Supporting Information Figure S8C–E). Ptger2 depletion by RNAi also blocked the CCL2 increase (Figure 4C). We also found that NF-κB decoy ODN blocked the EP₂-mediated CCL2 induction (Figure 4D). Thus, EP₂ receptor stimulation induced CCL2 via an NF-κB-dependent pathway in primary ECs. We further examined the effect of EP receptor agonists on the expression of pro-inflammatory mediators in primary SMCs from rat cerebral arteries (Ishibashi et al., 2010) because EP₂ receptors are also expressed in SMCs at the late stage of cerebral aneurysm formation and cerebral aneurysm walls of human patients (Figure 1G and I; Supporting Information Figure S2B). ONO-AE1-259, but not agonists of other EP receptor subtypes, induced the expression of Il-1β and iNos (Fukuda et al., 2000; Sadamasa et al., 2003; Moriwaki et al., 2006) in these cells (Supporting Information Figure S9). These data suggest that PGE₂-EP₂ signalling in EC and SMC is involved in the inflammation at distinct stages of cerebral aneurysm formation.

Figure 4

Induction of CCL2 expression by EP₂ stimulation via NF-κB pathway in ECs. (A) Effects of EP agonists on the mRNA level of CCL2 in primary ECs from human carotid artery (EP₁ agonist; ONO-D1-004, EP₂ agonist; ONO-AE1-259, EP₃ agonist; ONO-AE-248, EP₄ agonist; ONO-AE1-329). After 24 h stimulation with the respective receptor agonists (0.5 µM), mRNA levels of CCL2 were determined by RT-PCR (n= 6). Data represent mean ± SEM. *P < 0.05, **P < 0.01. (B) Effects of ONO-AE1-259, an EP₂ agonist, on phosphorylation of p65 subunit of NF-κB (p-p65) and CCL2 protein. The left panel shows representative images of Western blot analyses. Middle and right panels show the quantification by the densitometric analyses of phospho-p65 signals (middle) and CCL2 signals (right), respectively (n= 5). Data represent mean ± s,e,mean. *P < 0.05. **P < 0.01. (C) Effects of Ptger2 depletion by RNAi on NF-κB phosphorylation and CCL2 expression induced by ONO-AE1-259. The left panel shows representative phosphorylated p65 subunit (p-p65) and EP₂ signals in Western blot analyses under the indicated conditions. The middle panel shows quantification by the densitometric analyses of phosphorylated-p65 signals (n= 5). The right panel shows mRNA levels of CCL2 from ECs pretreated with siRNA for Ptger2 or control siRNA without or with subsequent EP₂ stimulation (n= 6). Data represent mean ± SEM. **P < 0.01. (D) Effect of NF-κB inhibition by decoy oligodeoxynucleotide on EP₂-induced CCL2 expression (n= 6). Data represent mean ± SEM. *P < 0.05, **P < 0.01.

Discussion

In this study, we have presented several lines of evidence that PGE₂-EP₂ signalling functions as a link between haemodynamic stress and cerebral aneurysm formation through the activation of NF-κB and evokes chronic inflammation in rodent models of cerebral aneurysm. We have further provided evidence that this mechanism also operates in human patients. Recent analyses on samples from patients with cerebral aneurysm have strongly implicated chronic inflammation in the pathogenesis of this condition (Chyatte et al., 1999; Takagi et al., 2002; Jayaraman et al. 2005) and various findings in rodent cerebral aneurysm models support this notion (Aoki et al., 2007b; 2009b; Aoki and Nishimura, 2010). Meanwhile, based on the findings that mechanical force is high at the site of cerebral aneurysm formation and that increased shear stress by surgical procedures in dogs induces pathological changes resembling cerebral aneurysm, the mechanical force exerted by blood flow (shear stress) has been proposed to be a cause for cerebral aneurysm formation (Hashimoto et al., 1980; Nakatani et al., 1991; Meng et al., 2007; Takeuchi and Karino, 2009). However, how haemodynamic stress leads to chronic inflammation has thus far remained unknown. Here, we have further disclosed that, once shear stress triggers PGE₂-EP₂ signalling, it amplifies the inflammatory processes through a positive feedback loop containing EP₂, NF-κB and COX-2. Therefore, it appears that inflammation in cerebral aneurysm walls is boosted and maintained by a pathway of COX-2–PGE₂-EP₂ NF-κB signalling, resulting in cerebral aneurysm formation (Figure 5).

Figure 5

A proposed model for the formation of cerebral aneurysm induced by shear stress.

Shear stress and PGE₂ signalling have also been suggested to be involved in the pathogenesis of aortic aneurysms (Holmes et al., 1997; Miralles et al., 1999; King et al. 2006). However, previous studies suggest that the underlying pathogenesis of aortic aneurysms and cerebral aneurysms is different. Epidemiological studies have shown that the risk of cerebral aneurysm and aortic aneurysm are gender dependent as the relative risk of cerebral aneurysm is 1.6 in females (Linn et al., 1996) while that of aortic aneurysm is 3.96 in males (Cornuz et al., 2004). Atherosclerosis is now understood to be an inflammatory disease of the arteries. It has been accepted that aortic aneurysm occurs as a consequence of aortic degeneration by atherosclerosis. Hypercholesterolaemia, a major risk factor for atherosclerosis, also increases the incidence of aortic aneurysm, but not that of cerebral aneurysm (Cornuz et al., 2004; Feigin et al., 2005; van Gijn et al., 2007). Consistent with these findings, hypercholesterolaemia induced by apolipoprotein E deficiency in mice, which causes atherosclerosis, does not increase the incidence of cerebral aneurysm (Aoki et al., 2008c). Moreover, we found that PGI₂-IP and TXA₂-TP signalling, which suppress and facilitate atherosclerosis, respectively, failed to influence cerebral aneurysm formation (Supporting Information Figure S6). Furthermore, previous studies have implicated EP₄ and not EP₂ in the pathogenesis of aortic aneurysm (Bayston et al., 2003). Therefore, it seems that different PGE receptor subtypes contribute to the formation of cerebral and aortic aneurysms.

Currently, there is no medical treatment for cerebral aneurysm except for the treatment of risk factors. From our present findings, COX-2 inhibition would be one option for cerebral aneurysm treatment as demonstrated in the rat experiments. However, the clinical use of COX-2 inhibitors for cerebral aneurysm has not been reported to date. We are currently performing a case-control study for evaluating the preventive effect of 3-hydroxy-3-methylglutaryl-coenzyme A reductase inhibitors (statins) on cerebral aneurysm rupture (subarachnoid haemorrhage-statin study). Statins have an anti-NF-κB activity through a pleiotropic effect (Liao and Laufs, 2005) and we have previously reported the preventive effect of statins on cerebral aneurysm formation in rat models (Aoki et al., 2008b; Aoki et al. 2009a). Our clinical study also allows us to examine effects of other drugs on cerebral aneurysm rupture. In this survey, we have found that the administration of non-steroidal anti-inflammatory drugs (NSAIDs) that suppress PG synthesis tended to suppress cerebral aneurysm rupture in four out of 49 cases (8.2%) in the subarachnoid haemorrhage group of NSAID users, whereas 26 out of 149 cases (17.4%) in the unruptured cerebral aneurysm group are NSAID users (chi-square test, P= 0.11). Although the findings from the case-control study suggest a potential of COX inhibitors as therapeutic agents for cerebral aneurysm, selective COX-2 inhibition produces a relative increase of TXA₂ over PGI₂ levels in vivo and increases the risk of cardiovascular events (Mukherjee et al. 2001; Juni et al., 2004; Funk and FitzGerald, 2007). EP₂ antagonism could thus circumvent this cardiovascular risk of COX-2 inhibition and could be a better pharmaceutical target for the medical treatment of cerebral aneurysm. This idea is reinforced by our findings that EP₂ expression was detected in SMC in the walls of advanced cerebral aneurysms. Thus, EP₂ antagonism could be useful not only in the prevention of cerebral aneurysm but also in the attenuation of the progression of cerebral aneurysms.

In summary, results from the present study suggest that shear stress activated the PGE₂-EP₂ pathway in ECs at the early stage of cerebral aneurysm formation and triggered chronic inflammation in arterial walls leading to cerebral aneurysm formation. This seems to occur through an amplification loop via NF-κB. In addition, given that selective COX-2 inhibitors carry the risk of cardiovascular events, our findings provide a possible alternative of EP₂ antagonists as a new class of drug against cerebral aneurysm.

Acknowledgements

This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan and a collaborative grant to Kyoto University from ONO Pharmaceuticals (S.N.). T.A. is a recipient of the Research Fellowship for Young Scientists from Japan Society for the Promotion of Science. We thank Pfizer and Ono Pharmaceuticals for their kind gifts of celecoxib and EP agonists, respectively.

Conflicts of interest

The authors declare that the research by S.N. is supported in part by a collaborative grant to Kyoto University from Ono Pharmaceuticals, which also supports partially the mouse colonies and has provided several compounds used in this study.