COX-2-Derived Prostacyclin Confers Atheroprotection on Female Mice

See allHide authors and affiliations

Science  10 Dec 2004:
Vol. 306, Issue 5703, pp. 1954-1957
DOI: 10.1126/science.1103333


Female gender affords relative protection from cardiovascular disease until the menopause. We report that estrogen acts on estrogen receptor subtype alpha to up-regulate the production of atheroprotective prostacyclin, PGI2, by activation of cyclooxygenase 2 (COX-2). This mechanism restrained both oxidant stress and platelet activation that contribute to atherogenesis in female mice. Deletion of the PGI2 receptor removed the atheroprotective effect of estrogen in ovariectomized female mice. This suggests that chronic treatment of patients with selective inhibitors of COX-2 could undermine protection from cardiovascular disease in premenopausal females.

Age-dependent increases in cardiovascular disease are less pronounced in women than in men, but this difference narrows after the menopause (1). Estrogen (E2) retards atherogenesis in models (2) and improves endothelial function in hyperlipidemic women (3). However, the mechanisms of atheroprotection are largely unknown.

The prostaglandin PGI2 exhibits properties of relevance to atheroprotection, inhibiting platelet activation, vascular smooth muscle contraction and proliferation (4), leukocyte–endothelial cell interactions (5), and cholesteryl ester hydrolase (6). PGI2 analogs retard atherogenesis (7), and atherosclerotic lesions exhibit a decreased capacity to produce PGI2 ex vivo (8). Cyclooxygenase 2 (COX-2), expressed in vascular endothelial cells (9), catalyzes prostaglandin endoperoxide formation from arachidonic acid. This is subsequently transformed to PGI2 by PGI2 synthase (PGIS), a process that occurs in endothelial cells (10). Selective COX-2 inhibitors depress PGI2 metabolite excretion (11). Inhibition of the COX-2–PGIS pathway may have particular relevance to atheroprotection in females because E2 increases COX-2 expression in vascular tissues and augments PGI2 production in vitro (12).

Mice lacking the low density lipoprotein (LDL) receptor (LDLR KO) were examined to address the role of PGI2 in gender-dependent atheroprotection; males develop atherosclerosis more rapidly than females (13). No differences in plasma total cholesterol or high density lipoprotein (HDL) cholesterol were observed among groups (table S1). Body weight increased as the animals aged, but there were no differences with genotype. En face aortae analyses showed that the extent of atherosclerosis was greater at 3 and 6 months in males than in females (Fig. 1, A and B). Female mice lacking both the PGI2 receptor (IP) and the LDLR (IP/LDLR DKO) developed greater aortic lesions than LDLR KO females after 3 months (5.6 ± 1.1% versus 2.2 ± 0.2%) and 6 months (23.2 ± 1.4% versus 15.0 ± 1.2%). This effect was gene dose dependent (Fig. 1C) and was not observed in male mice (Fig. 1D). There was a trend of greater lesion area in the IP/LDLR KO males than in LDLR KO males (5.95 ± 1.1%, n =12, versus 3.83 ± 0.5%, n =12) at 3 months, but this did not attain significance and was not evident at 6 months.

Fig. 1.

Increased atherosclerosis and thromboxane (Tx) biosynthesis in IP/LDLR DKO female mice. (A and B) Representative en face aortae approximately equal to the mean % lesion area for each group are shown. (C and D) Percentage aortic lesion areas (stained red) in female and male LDLR KO and IP/LDLR DKO mice showed no significant difference between male IP/LDLR DKO and LDLR KO mice fed a high-fat diet for similar durations. IP deletion accelerated atherosclerosis at both time points in female mice only. IP/LDLR DKO females have a greater % lesion area (*P < 0.001 versus LDLR KO at 3 months, #P < 0.05 at 6 months). Data are given as the mean ± SEM, n = 10 to 13 mice per group. (E) Excretion of urinary 2,3-dinor TxB2 (Tx-M) in LDLR KO and IP/LDLR DKO mice at baseline and after being fed a high-fat diet for 3 and 6 months. Coincidental IP deletion increases Tx-M excretion in female mice (F = 14.83, P < 0.001). Pairwise comparisons reveal significant differences at baseline (*P < 0.05) and at 3 months (**P < 0.01). (F) Male IP/LDLR DKO mice do not have elevated Tx-M biosynthesis at baseline, but upon being fed a high-fat diet, show a significant increase (all males: F = 6.14, P < 0.01; all females: F = 8.17, P < 0.01). This is accounted for by a difference after 3 months (**P < 0.01). Data are given as the mean ± SEM, n = 10 to 13 mice per group.

Platelet activation facilitates early atherogenesis and is reflected by the presence of the thromboxane metabolite 2,3-dinor TxB2 (Tx-M) in urine. IP deletion increases urinary Tx-M in response to vascular injury (4) and increased urinary Tx-M in IP/LDLR DKO mice, compared to LDLR KO mice (Fig. 1, E and F). This effect was evident in both genders at 3 months. Although urinary Tx-M was higher in male than in female LDLR KO mice (32.8 ± 3.2 versus 20.5 ± 3.0 ng/mg creatinine), additional deletion of the IP elevated Tx-M excretion in females to exceed those in male IP/LDLR DKO mice at baseline. Thus, PGI2 decreases the platelet activation that accompanies early atherogenesis in female mice.

Oxidative stress increases lipid peroxidation, an effect reflected by the increased excretion of urinary isoprostanes (iPs) (14). One of the most abundant iPs in urine, 8,12-iso-iPF-VI, increased in LDLR KO mice during atherogenesis (Fig. 2). Although IP deletion had no effect in male IP/LDLR DKO mice (F = 1.77, P = 0.2) (Fig. 2A), its absence increased lipid peroxidation in female IP/LDLR DKO mice (F = 15.8, P < 0.0005). This was particularly evident at baseline (8.7 ± 2.6 versus 2.8 ± 0.3 ng/mg creatinine, P < 0.05) and at 3 months (8.8 ± 1.1 versus 2.8 ± 0.4 ng/mg creatinine, P < 0.001) (Fig. 2B), indicating that PGI2 serves an antioxidant function in female LDLR KO mice prior to, and at initiation of, atherogenesis.

Fig. 2.

Increased oxidant stress in female IP/LDLR DKO mice and an antioxidant effect of the IP in vitro. Urinary 8, 12-iso-iPF-VI levels in (A) male and (B) female LDLR KO and IP/LDLR DKO mice. Urine collections were performed at 24-hour intervals before (baseline) and during (3 and 6 months) atherogenesis. Analysis of variance reveals a significant effect of IP deletion on lipid peroxidation in female (F = 15.82, P < 0.001) but not male LDLR KO mice (F = 2.70, P = 0.1082). Elevation in iP generation in female IP/LDLR DKO mice was evident at baseline (*P < 0.05) and at 3 months (**P < 0.01). Data are given as the mean ± SEM, n = 5 to 8 mice per group. (C) Increased oxidative stress in mouse aortic smooth muscle cells (MASMCs) lacking the IP. IP KO MASMCs show an exaggerated oxidative response to H2O2 exposure (30 min, 100 μM) [**P < 0.001 compared to wild-type (WT) MASMCs]. (D) As shown with the fluorescent probe, dichlorofluorescein, IP KO MASMCs were more fluorescent than WT MASMCs before and after stimulation with 100 μm H2O2. (E) Cicaprost (100 nm, 8 hours) treatment increases HO1 protein expression in WT MASMCS but not in IP KO MASMCs (*P < 0.05). (F) HO1 protein expression is attenuated in IP/LDLR DKO atherosclerotic aortas (*P < 0.05).

Given the importance of oxidant stress in atherogenesis (14), we examined the effect of PGI2 in cultured mouse aortic smooth muscle cells (MASMCs) exposed to hydrogen peroxide (H2O2) (15). This treatment increased PGI2 synthesis (16) and lipid peroxidation, and the latter was augmented by IP deletion (Fig. 2C). COX-2 expression was not affected (16). H2O2 treatment increased reactive oxygen species in MASMCs, as shown by the fluorescent probe dichlorofluorescein (17). Absence of IP expression increased this effect (Fig. 2D), suggesting that IP modulates oxidant stress under basal conditions. Cicaprost (100 nM), an IP agonist, limited the response of MASMCs to H2O2 with subsequent reduction of lipid peroxidation (fig. S1A) but did not affect unstimulated cells. Cicaprost also increased expression of the antioxidant heme oxygenase 1 (HO1) in an IP-dependent manner (Fig. 2E), whereas IP deletion decreased aortic HO1 expression in female IP/LDLR DKO mice (Fig. 2F).

MASMCs subjected to long-term E2 exposure (10-8 M, 18 hours) were examined to determine whether E2 alters the antioxidant effect of PGI2 (15). E2 stimulated COX-2 expression and PGI2 formation as reflected by an increase in the PGI2 hydrolysis product, 6-keto PGF (from 2.4 ± 0.33 to 3.8 ± 0.38 ng/mg protein; *P < 0.05). Pretreatment of cells with E2 (10-8 M, 18 hours) attenuated the response to H2O2, as reflected by diminished release of the iP (Fig. 3B), whereas direct coincubation of H2O2 with E2 had no effect (16). E2 administration (2 or 8 μg/day for 7 days) to ovariectomized LDLR KO mice increased PGI2 biosynthesis, as reflected by increased excretion of the urinary PGI2 metabolite 2,3-dinor 6-keto PGF (PGI-M) (Fig. 3C) (** P < 0.01 and *P < 0.05, respectively, versus their own respective baseline controls) and depressed urinary iP (Fig. 3D) (*P < 0.05). Similarly, E2 (3 μg/day for 7 days) administration to ovariectomized wild-type mice (C57Bl6) increased urinary PGI-M and depressed excretion of the iP (16). When MASMCs were treated with agonists selective for E2 receptor α (ERα) (propyl pyrazole triol, PPT) (18) and ERβ (WAY-200070) (19), only the ERα-selective agonist increased 6-keto PGF (Fig. 3E) (**P < 0.05). Furthermore, E2 (0.5 μg/day for 7 days) increased urinary PGI-M in female ERβ KO mice but not in female ERα KO mice (Fig. 3F). Thus, estrogen increases PGI2 biosynthesis by activating ERα. LDLR KO and IP/LDLR DKO mice were bilaterally ovariectomized to assess the atheroprotective effect of exogenous E2 in the absence of PGI2. One week after surgery, mice began feeding on a high-fat diet in the absence or presence of exogenous E2 (8 μg/day) (15). After 3 months, aortic lesion area was significantly increased in the IP/LDLR DKO compared to the LDLR KO females (Fig. 4). E2 significantly reduced lesion burden (by ∼80%) in LDLR KO female mice (Fig. 4A). However, coincidental deletion of the IP reduced this effect significantly (P < 0.0001) to 32%.

Fig. 3.

E2 increases PGI2 biosynthesis in vitro and in vivo. (A) E2 increased COX-2–mediated PGI2 production from WT MASMCS exposed to E2 (10–8 M, 18 hours). PGI2 production was measured in the conditioned medium as 6-keto PGF. Values are expressed as fold over basal increase in 6-keto PGF1 α (pg/mg protein) (*P < 0.05). (B) Pretreatment with E2 mediates an antioxidant effect on H2O2-induced oxidant injury. The increase in iP generation from WT MASMCs exposed to 100 μM H2O2 (30 min) was attenuated when cells were preincubated with 10–8 M E2 for 18 hours (*P < 0.05). Treatment with 10–8 M E2 alone does not significantly alter iP generation from MASMCs. (C) Subcutaneous E2 administration increases urinary PGI2 biosynthesis in ovariectomized LDLR KO mice (2 μg/day: **P < 0.05; 8 μg/day: *P < 0.05), coincident with (D) diminished urinary iP (2 μg/day: *P < 0.05; 8 μg/day: *P < 0.05). (E) An ERα-selective agonist (1.0 μM) elevates PGI2 biosynthesis in MASMCs (**P < 0.01, fold over basal). (F) E2 fails to increase PGI2 biosynthesis in ovariectomized ERα KO mice. E2 increases PGI2 biosynthesis in ERβ KO mice (*P < 0.05). Data are given as the mean ± SEM, n = 6 to 10 mice per group.

Fig. 4.

The atheroprotective effect of estrogen is significantly reduced in the absence of the IP. (A) The atheroprotective effect of a subcutaneous regimen of E2 (8 μg/day) in ovariectomized LDLR KO mice (***P < 0.001) is constrained in IP/LDLR DKO females. As expected, the % lesion area is greater in IP/LDLR DKO females than in LDLR KO females (*P < 0.05). (B) Representative en face aortas approximately equal to the mean % lesion area (stained red) are shown. Data are given as the mean ± SEM, n = 11 to 19 mice per group.

Rofecoxib, a selective inhibitor of COX-2, was withdrawn from clinical use when a placebo-controlled trial revealed an increase in myocardial infarction and stroke (20). These patients were initially at low cardiovascular risk, and the hazard was detected after 18 months of treatment. Selective inhibitors of COX-2 depress PGI2, but not platelet COX-1–derived TxA2, which affords a mechanism whereby they might elevate blood pressure, accelerate atherogenesis, and augment the thrombotic response to plaque rupture (5, 21, 22). Depression of PGI2 may accelerate atherogenesis by multiple mechanisms, including augmenting platelet and neutrophil interactions with the vasculature (4, 5). PGI2 also exerts an antioxidant effect, which may retard atherogenesis (14). IP deletion augments lipid peroxidation, whereas its overexpression in human embryonic kidney cells has the opposite effect (fig. S1B). An antioxidant role for PGI2, which may reflect induction of HO1, is consistent with exacerbation of reperfusion injury by IP deletion (23).

Although extrapolation of results in mice to humans is performed with caution, the experience with rofecoxib has focused attention on the role of atherogenesis in transformation of cardiovascular risk during chronic treatment with selective inhibitors of COX-2 (20). We reveal a substantial contribution of ERα-mediated, COX-2–derived PGI2 to atheroprotection in female LDLR KOs. This finding raises concern about the use of COX-2 inhibitors in juvenile arthritis, a disease that predominantly affects females. It may also have implications for the design and interpretation of trials of hormone-replacement therapy.

Supporting Online Material

Materials and Methods

Fig. S1

Table S1

References and Notes

View Abstract

Navigate This Article