Vascular wall–produced prostaglandin E2 exacerbates arterial thrombosis and atherothrombosis through platelet EP3 receptors

We tested the ability of the arterial wall to produce PGE2 from locally delivered AA. We observed that the PGE2 content of a healthy carotid wall (72 ± 9 pg/carotid; n = 15) increased significantly after we topically applied 50 mg/ml AA onto the adventitia (317 ± 34 pg/carotid; n = 13; P < 0.0001; Fig. 1 A). Thus, the arterial wall can convert AA into PGE2. To test the production of PGE2 by arterial tissue in more pathophysiological conditions, we placed a collar loosely maintained around the carotid for 4 wk. This procedure, known to locally induce a chronic inflammatory lesion (17), predominantly enhanced the tissue content of PGE2 by 22-fold, whereas the PGI2 and TXB2 increased by only 7- and 6-fold, respectively (Table I). These data show that inflammation stimulates the arterial wall to produce PGE2.

Endothelium produces potent inhibitors of platelet aggregation, such as PGI2 or nitric oxide (NO), to prevent any inappropriate local thrombosis. To test whether PGE2 produced by the arterial wall can oppose PGI2 and NO to alter the local arterial hemostatic balance, we adapted a previously described model of inflammatory venous thrombosis (14). Topical superfusion of 100 mg/ml AA onto the mouse carotid induced intraarterial thrombosis. The inability of eicosatrienoic acid, which differs from AA by only a double bond, to induce thrombosis suggests that thrombus formation in this model depends on formation of AA metabolites (unpublished data). Indeed, this thrombosis was triggered by TXA2, because the visual semiquantitative thrombotic score (18) dropped from 2.62 ± 0.32 (n = 8) in WT mice to 0.01 ± 0.01 (n = 8; P < 0.0001) in mice lacking thromboxane prostanoid (TP), the receptor for TXA2. Hence, periadventitial delivery of AA leads to its conversion into both TXA2 and PGE2 by the arterial wall. To examine whether the TXA2-induced thrombosis is modulated by PGE2 in this model, we quantified the extent of thrombosis in mice deficient for EP3. The targeted disruption of the EP3 locus did not alter the production of prostaglandins by the carotid wall (Table II). After the mice were injected with fluorescently labeled platelets, the entire clotting process was measured by counting the number of green fluorescent pixels on images acquired through a fluorescence macroscope (Fig. 1 B). The extent of arterial thrombosis measured in the absence of EP3 (29.5 ± 7 × 10⁶ pixels/min; n = 13) was significantly reduced when compared with WT mice (57.4 ± 6.2 × 10⁶ pixels/min; n = 18; P = 0.006; Fig. 1 C). Thus, PGE2 produced by local periadventitial delivery of AA facilitated arterial thrombosis induced by TXA2 (Video S1), suggesting that it successfully opposed PGI2 and NO to shift the hemostatic balance toward a prothrombotic state.

Because its conversion by cyclooxygenases leads to PGE2 biosynthesis, the local delivery of AA to the arterial wall might have enforced PGE2 production and exaggerated its role in thrombosis formation. To test whether endothelial injury induced by inflammation produces enough PGE2 to alter local hemostasis, we topically applied ferric chloride (5%) onto the carotid. After its delivery, ferric chloride crosses the arterial wall and triggers thrombosis by inducing endothelial cell death (Fig. 1 D), subsequent release of ADP, and exposure of underlying collagen (19). In addition, the lesion induces local inflammation and increased the arterial wall production of AA metabolites, as shown in Table II. In this setting, fluorescent thrombosis appeared ∼8 min after ferric chloride application. It reached 12.7 ± 3.3 × 10⁶ pixels/min (n = 11) in WT mice but was significantly reduced to 2.2 ± 10⁶ pixels/min (n = 8; P = 0.011; Fig. 1 E) in mice lacking EP3. These results show that PGE2 produced in response to aggression of the arterial wall is sufficient to facilitate platelet activation induced by ADP, TXA2, and/or collagen in arterial flow.

PGE2 facilitates thrombosis by decreasing the activation threshold of platelets, making them more sensitive to their agonists (14). In vivo, PGE2 is produced by the arterial wall, but perhaps also by activated platelets themselves (20). In the latter case, PGE2 might facilitate the effect of ADP and TXA2 secreted by activated platelets to recruit more platelets. This would suggest that PGE2 plays a role in thrombosis amplification rather than in thrombosis driven by inflammation of the vascular wall. To address this possibility, we examined whether in vitro aggregation of isolated platelets is EP3 dependent. Platelets stimulated with low concentrations of collagen elicited partial aggregation that was not modified by the presence or the absence of EP3 (47.6 ± 2.1% vs. 47.9 ± 4.2% [n = 4] at 1.5 μg/ml collagen [P = 0.95] and 58.8 ± 4.2% vs. 57.2 ± 3% [n = 9] at 2 μg/ml collagen [P = 0.75], respectively; Fig. 2). The absence of potentiation shows that activated platelets do not produce enough PGE2 to amplify aggregation in vitro.

To further ascertain that platelet production of PGE2 does not alter aggregation in vivo, we looked for a model of thrombosis in which mural PGE2 is not produced when thrombosis starts. We tested a model of endothelial injury induced by oxidative stress resulting from local excitation of Rose bengal by a laser beam. Under standard conditions (21), we observed that a few clots were already visible at 4 s and that 80% of the arterial diameter was visually obstructed at 45 s (Fig. 3 A). PGE2 levels in these injured carotids were found in the range of control values and were not significantly different between 4 s (55 ± 6 pg/carotid; n = 6) and 45 s (73 ± 11 pg/carotid; n = 6; P > 0.05; Fig. 3 B). Thus, we used this model deprived of PGE2 at the initiation of thrombosis to examine whether aggregating platelets could produce enough PGE2 to alter amplification of thrombosis. Consistent with our in vitro data, carotids of Ep3^−/− mice infused with EP3-deficient platelets were totally occluded after 30 min of laser exposure, as were WT mice (n = 5 in each group). We concluded that platelets did not produce enough PGE2 in vivo to facilitate amplification. Collectively, our data show that thrombosis is facilitated primarily by arterial wall–produced PGE2.

To further substantiate its role in pathophysiological conditions, we examined whether PGE2 modulates thrombosis on atherosclerotic plaques. To confirm previous data that suggested the ability of plaque to produce PGE2 (8, 12), we quantified it directly in mouse plaques. ApoE^+/+ aorta contained 661 ± 98 pg PGE2 (n = 12), whereas its amount in ApoE^−/− aorta was found to be about fourfold higher, at 2,483 ± 485 pg (n = 14; P < 0.05). After we fed ApoE^−/− mice a high fat diet, which is known to increase the size of atherosclerotic lesions, the difference was even more impressive (5,018 ± 705 pg/aorta; n = 23; P < 0.001; Fig. 4 A). Thus, atherosclerotic plaques produce PGE2.

Because numerous macrophages are recruited in the plaque (22, 23) and produce reactive oxygen species (24), we examined whether PGE2 produced by the plaque was functionally altered. Homogenates of whole atherosclerotic plaques added to platelet suspension in vitro trigger their aggregation, as a consequence of their high content of collagen (25). We observed that low concentrations of homogenized plaques added to EP3-deficient platelets induced aggregation that reached only 14.6 ± 3% of WT platelet aggregations (n = 6; Fig. 4 B). We concluded that mouse atherosclerotic plaques contain a molecule able to activate EP3. To rule out that a molecule other than PGE2 could activate EP3 (26), we treated mice with high doses of aspirin (500 mg/kg for 8 d) to inhibit cyclooxygenases. As expected, the treatment depressed the PGE2 levels in plaques (401.1 ± 163.8 pg/aorta; n = 4) down to the range of ApoE^+/+ aorta values. Aggregation elicited by EP3-deficient platelets in response to homogenized suspensions of these aspirin-treated plaques reached 101 ± 15.1% of the aggregation elicited by WT platelets (Fig. 4 C). Thus, the EP3 receptor was not activated when PGE2 production is inhibited, indicating that the potentiating effect of homogenized plaques was indeed caused by PGE2. Hence, PGE2 produced by atherosclerotic plaques can activate EP3.

We wondered whether PGE2 can exit from the plaque to act efficiently on blood platelets. To address the question, we enforced the plaque production of PGE2 by exposing its endoluminal side to low concentrations of AA (2 mg/ml; Fig. 5 A). We observed thrombi forming spontaneously when blood flow returned after the plaque has been incubated with AA (Fig. 5 B, middle). Importantly, thrombosis was detected only at the contact of the plaque, indicating that the plaque produced a platelet agonist that triggered local aggregation. Conversely, healthy the arterial wall was left unchanged by AA incubation, showing that the chosen dose of AA was too low to induce thrombosis at its contact (Fig. 5 B, top). We examined whether the plaque-produced PGE2 could modulate the local thrombosis elicited by the plaque, using Ep3^−/− mice crossed with ApoE^−/− mice. In these >55-wk-old double-mutant mice (ApoE^−/− × Ep3^−/−), the lack of EP3 did not alter the extent of atherosclerosis, because their plaques covered 49.3 ± 4.1% (n = 10) of the total aortic surface versus 49.8 ± 5.1% (n = 9; P = 0.94) in ApoE^−/− × Ep3^+/+ controls. The extent of thrombosis induced by intraluminal AA delivery reached 5.3 ± 1.5 × 10⁶ pixels/min (n = 10) in the presence of WT platelets but was drastically reduced to 0.9 ± 0.2 × 10⁶ pixels/min (n = 10; P < 0.05) when injected fluorescent platelets lacked EP3 (Fig. 5 C and Video S2). This experiment shows that AA-induced atherothrombosis in these mice was modulated by PGE2, which indicates that the plaque-produced PGE2 was able to act on blood platelets through EP3.

In the last step, we tested the role of PGE2 in pathophysiological conditions. Rupture of the fibrous cap allows blood platelets to be activated by their agonists present inside the lesion, such as collagen (25), lysophosphatidic acid (27), or tissue factor (28) that generates thrombin. We therefore examined whether a mere rupture of the plaque, the most frequent cause of myocardial infarction in human pathology, could induce atherothrombosis. After a tiny curved needle was introduced through a collateral of internal carotid and scratched the plaque (Fig. 6, A–C), we observed small, nonoccluding, but nevertheless detectable thrombosis on the surface of the ruptured plaque. Thrombi associated platelets (seen in fluorescence), red blood cells (seen on cross section; Fig. 6 C), and fibrin (Fig. 6 B, bottom). Although the ruptured areas measured by scanning electronic microscopy were similar in both groups of mice (34,200 ± 7,691 μm² [n = 13] vs. 35,600 ± 7,454 μm² [n = 10], respectively; P = 0.89), we observed an impressive reduction in the development of thrombosis according to the presence (0.16 ± 0.05 × 10⁶ pixels/min; n = 13) or absence (0.007 ± 0.005 × 10⁶ pixels/min; n = 10; P = 0.008) of EP3 on platelets injected in ApoE^−/− mice (Fig. 6 D). Thus, our data support the hypothesis that PGE2 produced in the plaques aggravates atherothrombosis.

It was not possible to predict what the in vivo action of PGE2 would be, if any, from its in vitro effects, which are both inhibition and potentiation of platelet aggregation. Studies identified EP3 as the receptor mediating the potentiating effect and showed that in vivo PGE2 might exert its potentiating effect on venous thrombosis (14, 16). However, these studies did not establish whether PGE2 is produced in pathophysiopathological conditions nor whether PGE2 modifies the hemostatic balance in arterial flow, where thrombosis is tightly controlled by endothelial production of potent inhibitors. Our present experiments show that PGE2 is produced by the arterial wall in response to inflammation and shifts the local hemostatic balance toward a prothrombotic state. These data allow us to conclude that PGE2 has an in vivo prothrombotic role. Moreover, our studies of the atherosclerotic plaques demonstrate that thrombosis induced by a rupture of the fibrous cap is clearly enhanced by the plaque-produced PGE2. This suggests that PGE2 modulates the thrombogenic potential of the plaque.

However, the impact of low PGE2 levels on thrombosis was examined in a recent study using mice deficient for one of the enzymes catalyzing the last step in PGE2 biosynthesis, mPGES-1 (21). To our surprise, arterial thrombosis was not decreased in mPGES-1^−/− mice, suggesting that PGE2 did not potentiate in vivo platelet aggregation, at least when thrombosis was induced by endothelial oxidative stress resulting from local excitation of Rose bengal. Using the same model of endothelial injury, we could not observe any difference in the extent of fluorescent thrombosis detected in mice injected with WT versus EP3-deficient platelets. However, we were also unable to detect substantial levels of PGE2 in the arterial walls of WT mice subjected to this oxidative stress. The Rose bengal–laser model induces thrombosis before inflammatory mechanisms could produce the PGE2 level required for its potentiating effect, and it is likely that inflammation takes place secondarily when PGE2 can no longer influence an already maximal thrombosis. Thus, the inability of this model to produce PGE2 when thrombosis develops explains the lack of difference between WT and mPGES-1^−/− mice. On another hand, this limitation highlights the fact that some models of thrombosis might not have a clear pathophysiological significance. This prompted us to examine the effect of PGE2 in a clearly more pathophysiological condition such as atherothrombosis.

A role for PGE2 in atherothrombosis has been previously suggested, but through an indirect mechanism. From association studies showing colocalization of macrophages, mPGES-1, and metalloproteases, it was deduced that PGE2 might stimulate metalloproteases to thin the fibrous cap. The subsequent increased vulnerability of the plaque might favor atherothrombosis. This concept links PGE2 to atherothrombosis through a suggested action on plaque stability (8). Our study used direct tissue detection and confirmed that atherosclerotic plaques produce PGE2. Above all, our work links PGE2 to atherothrombosis through its direct action on platelets and thrombosis, showing that it increases the platelet sensitivity to their agonists in vivo.

To establish this direct link, we set up a model of atherothrombosis in which we simply scratched the plaque without interrupting the blood flow (Fig. 6 B). This mechanical disruption induces local thrombosis at the contact of the plaque (Fig. 6 C). Although our model looks histologically close to ruptures observed in autopsy studies, creating a defect in the fibrous cap and exposing the thrombogenic core to circulating blood, its main limitation is linked to the mechanism of rupture. The current paradigm links inflammation to the plaque vulnerability through the action of metalloproteases. This mechanism implies that the most inflammatory plaques are the most vulnerable. As a consequence, rupture happens mainly on high-grade inflamed plaques. In contrast, we scratched plaques without knowledge of their inflammatory status, and some of them were very likely low-grade inflamed. This potential limitation might explain the variable response to the scratch in the control group (Fig. 6 D). However, it conversely strengthens our conclusion that PGE2 enhances atherothrombosis, because the more inflamed a plaque, the more PGE2 it might produce.

We established this direct link in mice, not in humans. However, a high homology (in the range of 80–90%) was found between mouse and human gene sequences coding for the PGE2 receptors (15). In addition, mouse and human distribution patterns of EP receptors on platelets are very similar, because RT-PCR showed obvious EP3 and EP4 bands in both species, whereas the EP2 band was really faint (16, 29). This is consistent with in vitro platelet responses to PGE2, which are very similar in mice (14) and in humans (13, 30, 31). Collectively, these data show a high similarity between mouse and human platelet behavior in response to PGE2, suggesting that PGE2 also facilitates arterial thrombosis in humans.

PGE2 appears in our study as a molecule playing a role in vascular wall homeostasis through its action in limited healing thrombosis. Indeed, the fact that PGE2 facilitated thrombosis in the arterial flow in response to inflammation indicates that its in vivo effect is sufficient to oppose PGI2. Therefore, even when a small vascular lesion releases or exposes low amounts of agonists, the concentrations of which are below the platelet activation threshold, the presence of PGE2 produced by local inflammation may sensitize platelets and foster healing thrombosis. Conversely, PGI2 produced upon inflammation by neighboring functional endothelial cells limits the extent of thrombosis (32) to restrict it to the lesion. Hence, PGE2 catalyzes healing thrombosis on small lesions, whereas PGI2 restricts its amplification to keep it focalized (Fig. 7).

On the contrary, PGE2 produced by the plaques must be regarded as a harmful molecule, because it aggravates atherothrombosis. In this study, PGE2 present in subendothelial connective tissue sensitizes platelets that adhered to a mere erosion and facilitates local thrombosis. Providing that the surrounding endothelium on the plaque is dysfunctional (33, 34), the nascent thrombosis is incompletely controlled by the deficient local production of NO and/or PGI2. Thus, PGE2 might trigger a massive and occluding thrombosis on a mere endothelial erosion.

To prevent this dreadful sequence, the effect of PGE2 should be inhibited. PGE2 has numerous functions in various physiological systems, implying that inhibition of its production might induce unexpected side effects. Inhibiting the EP3 receptor instead might be safer, especially considering that the EP3 receptor has multiple isoforms (35). Therefore, a drug targeting the platelet EP3 isoform, which inhibits adenylate cyclase, might be much more specific than a drug inhibiting PGE2 production. Such a drug should improve the current limited efficiency in preventing myocardial infarction (36, 37). Indeed, aspirin at low doses, the treatment of reference, inhibits COX-1 in platelets but does not alter the COX-2 that produces PGE2 in plaques. In addition, specific inhibition of COX-2 that could be beneficial in decreasing the PGE2 level has in fact been detrimental because of concomitant inhibition of PGI2 (38). Specific inhibition of the platelet EP3 receptor will shift the hemostatic balance toward an antithrombotic state by inhibiting the facilitating effect of PGE2 and allowing PGI2 to predominate. Finally, keeping in mind that PGE2 is produced by the arterial wall in response to inflammation, such a treatment would theoretically decrease only thrombosis triggered by inflammation.

In conclusion, we have shown that the potentiating effect of PGE2 has a clear in vivo significance. Through its platelet EP3 receptor, PGE2 is a full actor in intravascular hemostasis in inflammatory conditions, appearing as another link between inflammation and thrombosis. As a major consequence, we demonstrated that PGE2 exacerbates atherothrombosis. This unexpected role for PGE2 opens new possibilities for preventing myocardial infarction or stroke.

Homozygous ApoE^−/− mice were obtained from the Jackson Laboratory. Ep3^−/− mice (provided by B.H. Koller, University of North Carolina at Chapel Hill, Chapel Hill, NC) and Tp^−/− mice (provided by T.M. Coffman, Duke University, Durham, NC) were previously described (39, 40). Ep3^−/− and ApoE^−/− mice have been intercrossed to obtain double-mutant mice. All of these strains were maintained on a C57BL/6 genetic background. Animal procedures were approved by the local committee for animal ethics (Comité Régional d'éthique en Matière d'Expérimentation Animale de Strasbourg, Strasbourg, France). The investigator was unaware of the genotype of mice throughout the experiments.

Prostaglandin levels were determined from mashed tissues, using specific enzyme immunoassays (GE Healthcare). In the first experiment, the left common carotid of mice was dissected free, and a strip of parafilm was slipped below the carotid for topical application of 100% EtOH (vehicle) or 40 μL AA (50 mg/ml in EtOH; Sigma-Aldrich) for 10 min. The carotid was harvested and snap frozen 30 min later. In a second series of experiments, a nonconstrictive polyethylene collar (0.38-mm inner diameter; Biotrol) was placed loosely around the left common carotid, as previously described (17), for 4 wk. In a third group, aortas of at least 35-wk-old ApoE^−/− mice were harvested from the aortic valves to the renal bifurcation and snap frozen. In the final group, ApoE^−/− mice were fed a high fat diet (D12336i; Research Diets, Inc.) for 4 wk.

Carotids were exposed to 100 mg/ml AA for 10 min, and thrombosis was scored 2 h later under magnification using a visual scale from 0 to 4 (18).

Whole blood collected on 0.38% sodium citrate was centrifuged at 100 g to allow separation of platelets. Washed platelets were incubated with 300 ng/ml calcein AM (Invitrogen) in Tyrode's buffer for 15 min in the dark. Washed labeled platelets were injected in the jugular vein of receiver mice. The treated carotid was placed under a fluorescent macroscope (MacroFluo; Leica) for video recording of thrombosis at 480 nm. Images extracted from the 40-min recorded video were computer processed for counting the number of green pixels. Thrombosis extent was expressed as the total number of green pixels divided by the duration of the experiment in minutes (pixels/min).

The carotid was exposed to 100mg/ml AA or eicosatrienoic acid for 25 min, and thrombosis was recorded. Thrombosis induced by 5% ferric chloride (1 μL; Sigma-Aldrich) was recorded immediately after the topical application. To demonstrate ferric chloride–induced denudation, three mice received 200 μl of 2% Evans blue and 30 μl heparin (100 U/ml). The photochemical injury followed the procedure previously described (21), except that thrombosis was quantified by fluomacroscopy, as described in the previous section. Rose bengal (Fischer Scientific) was diluted in PBS and injected in the jugular vein in a volume of 0.12 ml at a concentration of 50 mg/kg. The 1.5-mW, 540-nm laser (model 05-LGR-193; Melles Griot) was applied from a distance of 5 cm for 4 s, 45 s, or 30 min. The carotid was then harvested and snap frozen for PGE2 detection.

Aggregation tests were performed in an optical aggregometer (model 570-VS; Chrono-log Corporation), using 250 μl of platelet-rich plasma adjusted to 300,000 platelets/μL. Collagen was obtained from Chrono-log Corporation. Homogenates were obtained by mashing plaques in saline at a concentration of 80 mg/ml. Aspirin-treated ApoE^−/− mice received 500 mg/kg aspegic (Sanofi-Synthelabo) i.p. daily for 8 d. Each homogenized plaque was used for both PGE2 detection and aggregation tests.

Mice >55 wk old received fluorescent platelets, and one carotid was exposed. An injection chamber was made by isolating an arterial segment bearing at least one plaque from blood flow with temporary ligatures (Fig. 5 A). Once washed, the chamber was filled with DMSO for 10 min before the blood flow returned. A 10-min video ensured that DMSO did not induce thrombosis. After ligatures were placed again, the vascular chamber was filled with a 2-μL solution of 2 mg/ml AA (in DMSO) for 10 min. Once blood reflowed, thrombosis was recorded through a double bandpass optical filter (525 and 605 nm) for 30 min.

After its tip was curved, a clean 30-gauge needle was introduced in 60-wk-old ApoE^−/− mice through a small collateral of internal carotid (Fig. 6 A). Under magnification, the tip of the needle was positioned next to the distal edge of the plaque, applied to the plaque, and pulled backward under visual control. The collateral artery was ligatured, and the mouse was placed under the fluomacroscope for 20 min for video recording. The carotid was washed, harvested, fixed in 2.5% glutaraldehyde overnight, fixed in cacodylate buffer for a few minutes, postfixed for 60 min in 4% osmium tetroxide, dehydrated in increasing ethanol concentrations, critical point dried with hexamethyldisilazane, mounted, sputter coated with palladium, and examined with a scanning electron microscope (XL20; Philips). Rupture areas were measured using Scion Image software. In two additional cases, the carotid was snap frozen, and 7-μm cross sections were stained by hematoxylin and eosin.

Anesthetized mice were infused with 10% neutral buffered formalin. Aorta were dissected, placed overnight in formalin at 4°C, opened longitudinally, and stained with 0.5% Sudan black. Measurements of the plaque area were performed using Scion Image software.

Statistical tests were bilateral and nonpaired student t tests. When variances were not comparable, we used the bilateral nonpaired Welch-corrected t test. To compare more than two groups with non-Gaussian distribution, we used the Kruskal-Wallis test, followed by Dunn's multiple comparison test. Means are expressed ±SEM. All tests were performed using GraphPad Prism (version 3.0 for Windows; GraphPad Software). P < 0.05 denotes statistical significance.

Video S1 shows the effect of EP3 on the thrombotic response induced by periadventitial delivery of AA onto the mouse carotid artery. Video S2 shows atherothrombosis developing at the surface of a mouse carotid plaque in response to intraluminal delivery of AA, when EP3 is expressed or not.

We want to thank B.H. Koller for the EP3 mice and T.M. Coffman for the TP mice.

This work was supported in part by grants (to J.-E. Fabre) from the Fondation de France and the Fondation pour la Recherche Médicale (FRM) and in part by four Ph.D. grants (to S. Gross) from the Nouvelle Société Française d'Athérosclérose associated with Fournier laboratories, the Groupe d'Etudes sur l'Hémostase et la Thrombose, FRM, and the Association de Recherche contre le Cancer.

The authors have no conflicting financial interests.