Raloxifene Acutely Relaxes Rabbit Coronary Arteries In Vitro by an Estrogen Receptor–Dependent and Nitric Oxide–Dependent Mechanism
Background—Selective estrogen receptor modulators (SERMs) have been defined as compounds that display tissue specificity with regard to estrogenic effects. They appear to share the beneficial effects of estrogen on bone and lipids but are not associated with an increased risk of breast or uterine carcinoma. Estrogen relaxes coronary arteries and has long-term protective effects on the vascular system. The effect of SERMs on the coronary vasculature is unknown. We therefore investigated the effects of the SERM raloxifene on isolated rabbit coronary arteries.
Methods and Results—–Rings of coronary artery from adult male and nonpregnant female New Zealand White rabbits were suspended in organ baths containing Krebs solution; isometric tension was then measured. Raloxifene induced coronary arterial relaxation in male and female coronary arteries by an endothelium-dependent and estrogen receptor–dependent mechanism involving nitric oxide. Raloxifene also had a direct calcium antagonistic effect on the coronary myocyte. Estrogen, however, induced only endothelium-independent coronary arterial relaxation. The endothelium-dependent component of relaxation induced by raloxifene 10−6 mol/L resulted in almost 100% (79±7% versus 43±3%, P<0.001) more relaxation than that induced by estrogen 10−6 mol/L.
Conclusions—These data demonstrate that raloxifene has vascular relaxing properties. The surprising finding is that the receptor-dependent effects via the endothelium are observed in coronary arteries from both male and female animals.
The risk of coronary heart disease and osteoporosis in postmenopausal women is reduced if they are current users of hormone replacement therapy.1 2 Estrogen has direct relaxing properties on blood vessels, including coronary arteries from animals and humans in vivo.3 4 5 Some of the mechanisms discovered may contribute to a protection against atheroma development.6
In rabbit coronary artery rings in our laboratory, acute estrogen-induced relaxation has been shown to be endothelium-independent.3 7 Rings without endothelium relaxed to a degree similar to rings with endothelium, and inhibition of nitric oxide synthase (NOS) did not alter the relaxation response.3 Calcium antagonism at the level of the coronary myocyte was shown to be a mechanism of relaxation by estrogen.8 This effect of estrogen was not dependent on the classic estrogen receptor.4 Studies have also shown an enhancement of endothelium-dependent responses in a variety of vessels.9 10 11 12 13
Chronic exposure to estrogen can result in an increase in the release of NO from the vascular endothelium.11 14 15 16 This effect is estrogen receptor–dependent,16 supported by the identification of estrogen receptor in human coronary endothelial cells.17 Estrogen therefore has both receptor-dependent and non–receptor-dependent effects on blood vessels that contribute to its vasorelaxing properties and possibly its long-term cardioprotective effects.
Despite the epidemiological evidence that postmenopausal hormone therapy is cardioprotective,1 18 it is estimated that <10% of women who might benefit from a reduction of cardiovascular disease are actually taking it.19 The major reasons for this are fear of estrogen-induced breast and uterine cancer, as well as resumption of menses, mastodynia, and weight gain.20 Raloxifene hydrochloride (LY139481), a selective estrogen receptor modulator, prevents bone loss without producing uterine proliferation.21 22 23 24 25 Raloxifene is known to share the lipid-lowering effects of estrogen,22 26 27 and it inhibits aortic accumulation of cholesterol in ovariectomized cholesterol-fed rabbits.28 There was, however, a lack of effect of raloxifene on coronary artery atherosclerosis in a cynamolgus monkey model.28a The effects of raloxifene on coronary artery vasoreactivity have not been investigated. The mixed estrogen agonist/antagonist profile that it exhibits in different tissues may also result in its action on different sites on the vessel wall, which may contribute to vasodilation. The purpose of this study was to investigate the possible relaxing effects of raloxifene on the coronary artery in vitro and the role of endothelial modulation and calcium antagonism in isolated rabbit coronary arterial preparations.
Animals and Tissues
Adult male or nonpregnant female New Zealand White rabbits (2.5 to 3.0 kg) were killed by an overdose of pentobarbitone (60 mg/kg) and heparin (150 U/kg). The heart was removed, and epicardial coronary arteries were dissected free of connective tissue. Arterial rings were prepared, and in alternate rings, the endothelium was removed by gentle rubbing with a wooden probe. Each ring (3 to 4 mm long) was suspended horizontally between 2 parallel stainless steel hooks for the measurement of isometric tension in individual organ baths containing 10 mL modified Krebs solution at 37°C, bubbled with 95% O2 and 5% CO2. The composition of the modified Krebs solution was as follows (mmol/L): NaCl 118.3, KCl 4.7, CaCl2 2.5, MgSO4 1.2, K2PO4 1.2, and glucose 11.1. Coronary arterial rings were allowed to stabilize for 90 minutes under a resting tension of 1 g before being contracted. Preparations were exposed to maximally effective concentrations of a contractile agonist (K+, 30×10−3 mol/L) to ensure stabilization of the vascular smooth muscle. The agonist was then removed and the ring reequilibrated. The presence or absence of endothelium was always verified by observation the relaxation response to acetylcholine.
Effect of Raloxifene and 17β-Estradiol on Precontracted Rabbit Coronary Arteries
Coronary arterial rings with or without endothelium from male and female rabbits were contracted with potassium (K+, 30×10−3 mol/L) or prostaglandin F2α (PGF2α, 3×10−6 mol/L). Raloxifene or the equivalent volume of DMSO solvent was added 7 minutes after the addition of constrictor agents or when the contraction was stable. Raloxifene was added in increasing concentrations at half-log increments, and dose-response curves were performed. Segments not exposed to raloxifene but exposed to the DMSO solvent acted as time-matched controls. In different rabbit coronary arterial rings, 17β-estradiol was added at log increments from 10−7 to 10−5 mol/L.
Effect of Nω-Nitro-l-Arginine Methyl Ester on Relaxation Induced by Raloxifene
Nω-Nitro-l-arginine methyl ester (L-NAME) is an inhibitor of synthesis of endothelium-derived relaxing factor from l-arginine in vascular endothelial cells.29 Rings with endothelium were incubated with L-NAME 10−4 mol/L for 20 minutes before precontraction with PGF2α or K+. A concentration-response curve to increasing concentrations of raloxifene was repeated.
Role of Classic Intracellular Estrogen Receptor in Raloxifene-Induced Relaxation
To examine the possible role of the classic estrogen receptor in mediating raloxifene-induced relaxation, rings with and without endothelium were incubated in ICI 182,780, a specific estrogen receptor antagonist, at 10−5 mol/L for 20 minutes before precontraction in PGF2α. Measurements of the responses to increasing concentrations of raloxifene were then repeated.
Role of Potassium Channels in Raloxifene-Induced Relaxation
To examine the possible role of potassium conductance on raloxifene-induced coronary relaxation, barium chloride, a nonspecific inhibitor of potassium channels,30 at 3×10−3 mol/L was added to coronary arterial rings with and without endothelium 20 minutes before being contracted with PGF2α. The response to increasing concentrations of raloxifene was then measured.
Effect of Raloxifene on Calcium Channels
Rabbit coronary arterial rings without endothelium were incubated in calcium-free solution containing 0.5 mmol/L EGTA for 10 minutes. Calcium concentration–dependent contraction curves were then performed in high-K+ (80×10−3 mol/L) depolarization medium. Rings were readjusted in modified Krebs solution for 20 minutes before being incubated with raloxifene for 30 minutes. The calcium concentration–dependent contraction curves were then repeated. Both the control and raloxifene calcium-dependent contraction curves were also repeated after incubation in barium chloride for 20 minutes
The following drugs were used: raloxifene hydrochloride (LY139481; dissolved in DMSO, gift from Eli Lilly Research Laboratories, Indianapolis, Ind), L-NAME (Sigma), PGF2α (Sigma), pentobarbitone (Sigma), ICI 182,780 (ICI). All drugs were analytical grade.
All results are expressed as mean±SEM. Relaxation is expressed as percentage relaxation of contraction induced by PGF2α 3×10−6 mol/L or K+ 30×10−3 mol/L. The results were analyzed by ANOVA. Each group was compared with the time-matched DMSO solvent control, with the Student-Newman-Keuls test for multiple comparisons. A value of P≤0.05 was considered statistically significant.
Effect of Raloxifene on Precontracted Coronary Arteries and the Effect of Endothelium Removal
Raloxifene induced significant dose-dependent relaxation of coronary arterial rings with endothelium compared with time-matched DMSO solvent controls precontracted by PGF2α (Figure 1⇓) or K+. Relaxation occurred in rings without endothelium; however, relaxation in rings without endothelium was significantly less than in rings with endothelium (Figure 2⇓, top). There was no significant difference in relaxation between rings precontracted by PGF2α and K+ (n=8 to 13; P>0.05) (n indicates number of animals).
17β-Estradiol induced significant dose-dependent relaxation of coronary arterial rings with and without endothelium. There was no difference in the relaxation in rings with or without endothelium (Figure 2⇑, bottom).
K+ 30×10−3 mol/L and PGF2α 3×10−6 mol/L induced comparable contractile responses in rings with endothelium (0.93±0.1 and 0.98±0.1 g, respectively; P>0.05) and without endothelium (0.78±0.1 and 0.86±0.1 g, respectively; P>0.05). The difference between contraction in rings with or without endothelium was not significant for rings contracted by either K+ or PGF2α (P>0.05).
The time course of the relaxation response was longer for raloxifene than for 17β-estradiol. Relaxation curves were performed over ≈90 minutes, whereas similar relaxation occurred over 10 to 15 minutes with estrogen (Figure 3⇓).
There were no differences between relaxation in arteries from male or female rabbits in rings with endothelium contracted in PGF2α (P>0.05) (Figure 4⇓). Rings from male and female rabbits showed similar significant reduction of relaxation in rings denuded of endothelium.
Effect of L-NAME on Raloxifene-Induced Relaxation
Incubation with L-NAME 30×10−6 mol/L for 20 minutes inhibited raloxifene-induced relaxation in rabbit coronary arterial rings with endothelium precontracted with PGF2α (Figure 5⇓, top). There was no significant difference between rings with endothelium incubated in L-NAME and endothelium-denuded rings (P>0.05) (Figure 5⇓, bottom).
Effect of ICI 182,780 on Raloxifene-Induced Relaxation
Incubation with the specific estrogen receptor antagonist ICI 182,780 at 10−5 mol/L for 20 minutes inhibited raloxifene-induced relaxation in rings with endothelium but not in rings without endothelium (Figure 6⇓).
Effect of Barium Chloride on Raloxifene-Induced Relaxation
Incubation with the nonspecific potassium channel inhibitor barium chloride 3×10−6 mol/L for 20 minutes inhibited raloxifene-induced relaxation in rings with an intact endothelium precontracted with PGF2α but not in rings without endothelium (Figure 7⇓).
Effect of Raloxifene on Calcium Concentration–Dependent Contractile Responses
In rings without endothelium, incubation in raloxifene (−6.5 and −6 log mol/L) shifted the calcium concentration–dependent contraction curves in high-K+ (80 mmol/L) depolarization medium to the right compared with control. Maximal contraction was also reduced (Figure 8⇓). This shift was not reversed by incubation in the nonspecific potassium channel inhibitor barium chloride 3×10−6 mol/L for 20 minutes (n=4; P>0.05).
We have demonstrated that raloxifene induces relaxation in precontracted male and female rabbit epicardial coronary arteries, with and without endothelium. This relaxation was significantly greater in rings with endothelium than in rings without endothelium, which is a distinct difference from relaxation induced by estrogen in this preparation (Figure 2⇑). At a concentration of 10−6 mol/L, raloxifene caused ≈100% more relaxation in rings with endothelium than in rings without endothelium (Figure 2⇑). Inhibition of raloxifene-induced relaxation in endothelium-intact rings was achieved by incubation with L-NAME, ICI 182,780, and barium chloride. Raloxifene shifted the calcium concentration–dependent contraction curve to the right in endothelium-denuded rings. The endothelium-dependent contribution to relaxation was greatest at lower concentrations of raloxifene (−6.5 and −6 log mol/L). However, as the concentration of raloxifene was increased, relaxation occurred by a combined effect on the endothelium and smooth muscle. Endothelium-denuded arterial rings relaxed to ≈100% at greater concentrations by a direct effect of raloxifene on the vascular smooth muscle myocyte.
Studies have also shown an enhancement of estrogen-induced endothelium-dependent responses in a variety of vessels.9 10 11 12 13 It is important to stress, however, that in all these studies, the animals were pretreated for a number of days before the demonstration of an enhancement of endothelium-dependent relaxation of vessels in vitro in organ bath experiments. This is in distinct contrast to these data on raloxifene, which resulted in acute responses after exposure to raloxifene in the organ bath. There was no requirement for pretreatment of the animals to produce an endothelium-dependent vascular relaxing effect. NO is synthesized in the endothelium from l-arginine by NO synthase31 and can diffuse rapidly to smooth muscle, causing relaxation via stimulation of soluble guanylate cyclase and a subsequent increase in cGMP.32 By blocking NO synthesis with L-NAME, we were able to completely abolish the difference in relaxation to raloxifene between rings with and without endothelium, implying that the endothelial contribution to raloxifene-induced relaxation is dependent on NO.
The specific estrogen receptor antagonist ICI 182,780 also inhibited raloxifene-induced relaxation in rings with endothelium, again to a level identical to relaxation in rings without endothelium. ICI 182,780 had no effect on raloxifene-induced relaxation in coronary artery rings denuded of endothelium, suggesting that the classic estrogen receptor mediates endothelium-dependent but not endothelium-independent relaxation. Raloxifene appears to interact with the estrogen receptor and may increase the expression of NOS, thereby increasing endothelium-derived NO production. We used an equivalent maximal concentration of ICI 182,780 (10−5 mol/L), which in single-cell studies has specific anti–estrogen receptor antagonism.33 Support for a short-term (5-minute) activation of eNOS, mediated by estrogen receptor-α functioning in a nongenomic manner via mitogen-activated protein, has recently been published.33a The mechanism of action of raloxifene in our experiments could easily involve this mechanism. We were unable to distinguish between an effect on the classic estrogen receptor-α and the newly discovered estrogen receptor-β.34
Potential-sensitive calcium channels are activated by depolarization of the plasma membrane when the extracellular potassium concentration is increased. Raloxifene relaxed coronary arterial rings contracted by high extracellular potassium both with and without endothelium. This suggests that it may have an inhibitory effect on potential-sensitive calcium channel activation in the smooth muscle cell. Raloxifene also induced relaxation of coronary arteries precontracted with PGF2α (an agonist of receptor-operated calcium channels), indicating that raloxifene has relaxing effects on contraction induced by activation of both receptor-operated and potential-operated calcium channels. The shift of the calcium concentration–dependent contraction curve to the right by raloxifene suggests that raloxifene may be a calcium antagonist in these preparations and may be a mechanism involved in the direct smooth muscle relaxation induced by raloxifene. Calcium antagonism may have been a result of decreased probability of voltage-gated calcium channels being open secondary to a change in other ion conductances and consequent hyperpolarization. To investigate this, rings were incubated in the nonspecific potassium channel blocker barium chloride. The shift in calcium concentration–dependent contraction curve to the right was not reversed by incubation in barium chloride, suggesting that modulation of potassium channel conductance does not play a role in the direct calcium antagonistic effect of raloxifene.
Another possible mechanism of relaxation by raloxifene may involve modulation of ion channel conductance in the smooth muscle in response to NO release from the endothelium. This could cause hyperpolarization of the cell, decreasing calcium influx through voltage-gated calcium channels and resulting in relaxation. In this study, we demonstrated that raloxifene-induced, endothelium-dependent relaxation is sensitive to inhibition by the nonspecific potassium channel inhibitor barium chloride, but endothelium-independent relaxation is not. This occurred only at a raloxifene concentration of 10−6 mol/L, a dose that induces maximal NO-dependent relaxation. Investigations with estrogen have shown that its ability to open potassium channels in vascular smooth muscle cells may be partly dependent on the endothelium and increased production of NO.16 NO activates guanylate cyclase, elevates intracellular cGMP, and stimulates G kinase in the coronary myocyte.35 36 G kinase is able to activate large-conductance Ca2+-activated K+ (BKCa) channels.35 37 The inhibition of NO-induced coronary artery dilation by the BKCa blocker iberiotoxin16 16 38 further supports the idea that endothelium-dependent estrogen-induced relaxation may depend on NO activation of BKCa channels. A similar mechanism may explain the sensitivity of endothelium-dependent relaxation by raloxifene to the nonspecific potassium channel blocker barium chloride. An increase in conductance of the BKCa channels by NO is a means by which raloxifene could indirectly decrease calcium influx. Hyperpolarization would then result in a decreased probability of voltage-operated calcium channels being open and relaxation.
Raloxifene took longer to induce relaxation than comparable studies using estrogen.3 39 Relaxation curves were performed over ≈90 minutes versus 10 to 15 minutes for similar studies with estrogen (Figure 3⇑). The relaxant effect of raloxifene was not reversed over periods of up to 5 hours, compared with estrogen, for which initial precontraction levels were achieved after ≈30 minutes. These observations are consistent with the finding of estrogen-receptor involvement in the endothelium-dependent relaxation by raloxifene. This is an effect that is not observed in vitro for estrogen in this model3 and is not involved in the calcium antagonistic effect. Further studies are needed to investigate this phenomenon, because it may represent an advantage in long-term treatment. Steroid-induced events, which are mediated via classic intracellular receptors, occur over periods of minutes to hours. Thus, in our investigation, initial relaxation may be due to early direct actions of raloxifene, followed by the prolonged, estrogen receptor–mediated relaxation. It is also well recognized that tissues such as breast tumor can actively accumulate similar moieties, such as tamoxifen, to concentrations much greater than those found in the plasma.40 41 No such data exist in vascular tissue; however, this could partially explain the extended time course of action compared with estrogen in this coronary arterial model, which may involve stimulation of the receptor and transcription of mRNA.
In vitro and in vivo concentrations do differ with regard to potency. In the present study, we have demonstrated a relaxant effect of raloxifene on coronary arterial ring preparations at concentrations ≈50 times greater than the plasma concentrations achieved by administration of raloxifene to patients. However, this model does not include effects of other circulating hormones, blood volume, or coronary circulation. In the case of 17β-estradiol, the lowest concentrations shown to be effective in relaxing rabbit coronary arterial rings in vitro are less than for raloxifene 10−6 mol/L.3 42 Plasma levels of estrogen similar to those with administered raloxifene (≈10−9 mol/L) are achieved during hormone replacement therapy, and long-term beneficial effects in the cardiovascular system have been reported at these concentrations.43 44 The actions demonstrated are short-term effects of raloxifene; long-term effects in vivo may differ.
Coronary heart disease is one of the most important causes of morbidity and mortality in developed countries, in terms of both its prevalence and its human and economic cost. We have shown that raloxifene is able to act both via the endothelium and directly on the vascular smooth muscle to induce relaxation of epicardial coronary arterial rings from both male and female animals. The former mechanism involves interaction with the classic estrogen receptor at the level of the endothelium and stimulation of NO. Raloxifene also has the ability to directly antagonize calcium influx in the smooth muscle cell, achieving smooth muscle cell relaxation independent of the endothelium. These in vitro findings with raloxifene in epicardial coronary arterial preparations would support further investigation of this compound in vivo. The mechanisms discovered in the present study raise the possibility that raloxifene may have the potential to be a cardioprotective agent, with the benefit of no increased risk of cancer and other side effects. Further work in humans may identify a therapeutic role for raloxifene in reducing cardiovascular disease in postmenopausal women and the significant mortality and morbidity associated with it.
- Received December 15, 1998.
- Revision received April 20, 1999.
- Accepted April 28, 1999.
- Copyright © 1999 by American Heart Association
Adams MR, Kaplan JR, Manuck SB, Koritnik DR, Parks JS, Wolfe MS, Clarkson TB. Inhibition of coronary artery atherosclerosis by 17-beta estradiol in ovariectomized monkeys: lack of an effect of added progesterone. Arteriosclerosis. 1990;10:1051–1057.
Han SZ, Karaki H, Ouchi Y, Akishita M, Orimo H. 17β-Estradiol inhibits Ca2+ influx and Ca2+ release induced by thromboxane A2 in porcine coronary artery. Circulation. 1995;91:2619–2626.
Vedernikov YP, Liao QP, Jain V, Saade GR, Chwalisz K, Garfield RE. Effect of chronic treatment with 17beta-estradiol and progesterone on endothelium-dependent and endothelium-independent relaxation in isolated aortic rings from ovariectomized rats. Am J Obstet Gynecol. 1997;176:603–608.
Bell DR, Rensberger HJ, Koritnik DR, Koshy A. Estrogen pretreatment directly potentiates endothelium-dependent vasorelaxation of porcine coronary arteries. Am J Physiol. 1995;268:H377–H383.
Collins P, Shay J, Jiang C, Moss J. Nitric oxide accounts for dose-dependent estrogen-mediated coronary relaxation following acute estrogen withdrawal. Circulation. 1994;90:1964–1968.
Gisclard V, Miller VM, Vanhoutte PM. Effect of 17β-estradiol on endothelium-dependent responses in the rabbit. J Pharmacol Exp Ther. 1988;244:19–22.
Hayashi T, Fukuto JM, Ignarro LJ, Chaudhuri G. Basal release of nitric oxide from aortic rings is greater in female rabbits than in male rabbits: implications for atherosclerosis. Proc Natl Acad Sci U S A. 1992;89:11259–11263.
Collins P, Rosano GMC, Sarrel PM, Ulrich L, Adamopoulos S, Beale CM, McNeill J, Poole-Wilson PA. Estradiol-17β attenuates acetylcholine-induced coronary arterial constriction in women but not men with coronary heart disease. Circulation. 1995;92:24–30.
Wellman GC, Bonev AD, Nelson MT, Brayden JE. Gender differences in coronary artery diameter involve estrogen, nitric oxide, and Ca2+-dependent K+ channels. Circ Res. 1996;79:1024–1030.
Kim-Schulze S, McGowan KA, Hubchak SC, Cid MC, Martin MB, Kleinman HK, Greene GL, Schnaper W. Expression of an estrogen receptor by human coronary artery and umbilical vein endothelial cells. Circulation. 1996;94:1402–1407.
Judd HL, Meldrum DR, Deftos LJ, Henderson BE. Estrogen replacement therapy: indications and complications. Ann Intern Med. 1983;98:195–205.
Black LJ, Sato M, Rowley ER, Magee DE, Bekele A, Williams DC, Cullinan GJ, Bendele R, Kauffman RF, Bensch WR, Frolik CA, Termine JD, Bryant HU. Raloxifene (LY139481 HCI) prevents bone loss and reduces serum cholesterol without causing uterine hypertrophy in ovariectomized rats. J Clin Invest. 1994;93:63–69.
Kauffman RF, Bensch WR, Roudebush RE, Cole HW, Bean JS, Phillips DL, Monroe A, Cullinan GJ, Glasebrook AL, Bryant HU. Hypocholesterolemic activity of raloxifene (LY139481): pharmacological characterization as a selective estrogen receptor modulator. J Pharmacol Exp Ther. 1997;280:146–153.
Bjarnason NH, Haarbo J, Byrjalsen I, Kauffman RF, Christiansen C. Raloxifene inhibits aortic accumulation of cholesterol in ovariectomized cholesterol-fed rabbits. Circulation. 1997;96:1964–1969.
Pohl U, Busse R. EDRF increases cyclic GMP in platelets during passage through the coronary vascular bed. Circ Res. 1989;65:1798–1803.
Wakeling AE, Dukes M, Bowler J. A potent specific pure antiestrogen with clinical potential. Cancer Res. 1991;51:3867–3873.
Kuiper GG, Enmark E, Pelto-Huikko M, Nilsson S, Gustafsson JA. Cloning of a novel receptor expressed in rat prostate and ovary. Proc Natl Acad Sci U S A. 1996;93:5925–5930.
Murad F. Regulation of cytosolic guanylyl cyclase by nitric oxide: the NO-cyclic GMP signal transduction system. Adv Pharmacol. 1994;26:19–33.
Murad F. The nitric oxide-cyclic GMP signal transduction system for intracellular and intercellular communication. Recent Prog Horm Res. 1994;49:239–248.
Jiang C, Sarrel PM, Poole-Wilson PA, Collins P. Acute effect of 17β-estradiol on rabbit coronary artery contractile responses to endothelin-1. Am J Physiol. 1992;263:H271–H275.
Johnston SD, Haynes BP, Sacks NM, McKinna JA, Griggs LJ, Jarman M, Baum M, Smith IE, Dowsett M. Effect of oestrogen receptor status and time on the intra-tumoural accumulation of tamoxifen and N-desmethyltamoxifen following short-term therapy in human primary breast cancer. Breast Cancer Res Treat. 1993;28:241–250.
Gerhard M, Walsh BW, Tawakol A, Haley EA, Creager SJ, Seely EW, Ganz P, Creager MA. Estradiol therapy combined with progesterone and endothelium-dependent vasodilation in postmenopausal women. Circulation. 1998;98:1158–1163.