Association Between Estrogen Receptor α Gene Variation and Cardiovascular Disease FREE

Atherothrombotic cardiovascular diseases (CVDs) such as myocardial infarction (MI) and stroke are multifactorial disorders with substantial heritable components. Genetic constitution may contribute to underlying risk factors such as hypertension, diabetes mellitus, and hypercholesterolemia and also may act through other, unidentified routes to alter susceptibility to CVD events.^1- 2 While the relation between estrogen and CVD events remains unclear, there is a substantial body of epidemiological evidence that points toward endogenous estrogen as having a protective role, although in several recently reported clinical trials, combination hormone therapy was associated with an increased incidence of MI and stroke.^3- 5

Estrogens and estrogen receptors have important physiological roles in men as well as in women. There are 2 known estrogen receptors: estrogen receptor α (ESR1) and estrogen receptor β (ESR2). Both receptors are expressed in a wide range of tissues, including macrophages, vascular smooth muscle, and vascular endothelial cells.⁶ Estrogen receptors regulate gene expression by both estrogen-dependent and estrogen-independent mechanisms that result in activation of transcription. There have been several genetic association studies, each limited to a few hundred individuals, of ESR1 gene variants in relation to coronary artery disease,^7- 11 coronary artery wall atherosclerosis,¹² and variation in high-density lipoprotein cholesterol or E-selectin levels in response to estrogen therapy.^13- 14 Three of the 4 positive association findings examined the IVS1-401 T/C variant (also known as c.454-397T>C and the PvuII restriction site) in intron 1 of ESR1.^12- 14 The C allele but not the T allele forms part of a potential binding site for the myb family of transcription factors,¹⁴ which are among the many genes whose transcription is activated by estrogen.¹⁵ None of these reports, either singly or in combination, provided the confirmation that is required for the result of an association study to be deemed conclusive; however, they provided a strong rationale to carry out a more definitive study of the association of ESR1 c.454-397T>C with atherothrombotic CVD events. We therefore analyzed the association of ESR1 c.454-397T>C with CVD events in a sample of 1739 unrelated men and women from the Framingham Heart Study.

The Framingham Heart Study began in 1948 with enrollment of 5209 men and women aged 28 to 62 years from Framingham, Mass. Repeat examinations were performed every 2 years. Details of study design and selection criteria are described elsewhere.^16- 17 In 1971, 5124 offspring of original cohort members and spouses of offspring were recruited. After their initial evaluations, these individuals have undergone repeat examinations approximately every 4 years. Additional details of the Framingham offspring cohort selection criteria are described elsewhere.¹⁸

Participants included in the current study (n = 1739, including 875 men and 864 women) are from a subset of unrelated individuals from the Framingham offspring cohort who provided blood samples for DNA extraction at the sixth examination cycle (n = 2933). Eligible participants had to be unrelated (ie, 1 person randomly selected from each extended family or biologically unrelated to any other Framingham Heart Study participant), and selection was designed to include equal numbers of men and women. These criteria were established to provide a panel of DNA samples that would be suitable for multiple studies. Genotyping was carried out in 1811 participant DNA samples; 1739 of these had ESR1 c.454-397T>C genotypes. Data for these 1739 individuals are used throughout the current study. According to data from the sixth examination cycle (Table 1), the 1739 selected participants were similar to the other Framingham offspring cohort members with respect to prevalence of CVD events and risk factors, except for a larger percentage of men (50% vs 44%; P<.001), slightly older age (mean age, 60 years vs 58 years; P<.001), higher prevalence of hypertension (44% vs 39%; P = .001), higher likelihood of diabetes mellitus (13% vs 11%; P = .045), and higher likelihood of taking cholesterol-lowering medications (14% vs 12%; P = .049).

At each visit to the Framingham Heart Study clinic, offspring cohort participants underwent extensive evaluations. Examining physicians measured seated blood pressure twice with a mercury-column sphygmomanometer. The 2 blood pressure measurements were averaged to derive the systolic and diastolic measurements for that examination. Hypertension was defined as systolic blood pressure of 140 mm Hg or more, diastolic blood pressure of 90 mm Hg or more, or current use of medication for treatment of hypertension. Diabetes was defined as a fasting blood glucose level that exceeded 125 mg/dL (6.9 mmol/L) or current use of medication for treatment of diabetes. Total and high-density lipoprotein cholesterol were measured at a Centers for Disease Control and Prevention standardized laboratory. Participants were categorized as smokers if they currently smoked cigarettes or if they had quit within 1 year prior to the clinic visit. Body mass index was calculated as weight in kilograms divided by the square of height in meters.

At each clinic visit, CVD history was obtained and hospitalization records were collected routinely for participants with suspected interim CVD events. For individuals who did not attend a clinic examination, a health history update was obtained by telephone and records from interim hospitalizations were obtained and reviewed. Prior to genotyping participant DNA samples, a panel of 3 physicians established the diagnosis of CVD end points after reviewing all available medical records. A diagnosis of recognized acute MI required simultaneous presence of at least 2 of the following 3 criteria: symptoms consistent with MI, diagnostic electrocardiographic changes of MI, and diagnostic elevation of biomarkers. Clinical end points were defined according to standardized Framingham Heart Study criteria. A diagnosis of unrecognized MI was made when an electrocardiogram revealed new pathological Q waves in comparison with the last available tracing, and MI was not known to have occurred in the interim. Angina was defined as occurrence of typical ischemic chest discomfort of brief duration that was precipitated by exertion and relieved by rest or nitroglycerine. Coronary insufficiency was diagnosed when prolonged ischemic chest discomfort prompted a medical evaluation that found electrocardiographic evidence of transient ischemic abnormalities and there were no elevations in biomarkers indicative of infarction. Intermittent claudication was diagnosed when a participant reported recurrent lower-extremity discomfort characteristic of ischemia occurring with exertion and relieved with rest. Atherothrombotic stroke was diagnosed when an unequivocal neurologic deficit lasting more than 24 hours occurred and an embolic cause was not found. In more than 90% of strokes, magnetic resonance or computed tomographic imaging was available to aid in the diagnosis.

All participants gave written informed consent (including that for DNA analysis) at each clinic visit, and the examination protocol was approved by the Boston Medical Center Institutional Review Board, Boston, Mass.

Genomic DNA was extracted from peripheral blood leukocytes from offspring cohort participants attending their sixth examination cycle (1995-1998) using standard methods.

We used the current recommendations of the Human Genome Variation Society for the description of sequence variation and described the polymorphism in relation to a specific human ESR1 cDNA sequence (accession number NM_000125), in which position 454 of the protein coding sequence was the first nucleotide of the start of the next closest exon to the marker we term ESR1 c.454-397T>C, and the variation was 397 nucleotides (according to the current chromosome 6 reference sequence, NT_023451, version 12) upstream in the intron. According to these guidelines, the polymorphism that has in previous studies been named according to its detection method (the PvuII restriction site in intron 1) or by older, equivocal recommendations (IVS1-401 T/C or IVS1-397T/C) may be described as c.454-397T>C. This single-nucleotide polymorphism (SNP) is also described at http://www.ncbi.nlm.nih.gov/SNP under identification number rs2234693. Three other polymorphisms,^{7- 14,19- 22} 1 intronic and 2 in exons, were studied to determine whether any association identified was specific to ESR1 c.454-397T>C. These SNPs were named in a similar fashion to ESR1 c.454-397T>C.

The ESR1 c.454-397T>C and ESR1 c.454-351A>G SNPs were detected by polymerase chain reaction amplification and PvuII or XbaI restriction fragment length analysis.^23- 24 Genotyping was performed without knowledge of participant characteristics and allele calling was carried out independently by 2 investigators. The T allele of ESR1 c.454-397T>C codes for presence of the PvuII site and has been termed the p allele in some previous reports; likewise, the C allele, which codes for absence of the site, has been termed the P allele. The A allele of ESR1 c.454-351A>G codes for presence of the XbaI site and has been termed the x allele in some previous reports; likewise, the G allele, which codes for absence of the site, has been termed the X allele.

The ESR1 c.30T>C (rs2077647) genotype was detected by a TaqMan assay (Applied Biosystems, Foster City, Calif), with the following amplification conditions: 95°C for 10 minutes; 40×: 95°C for 15 seconds, 62°C for 1 minute, 72°C for 1 minute (5′ to 3′ probe sequences: VIC-ACC AAA GCA TCT GGG ATG GCC C-TAMRA, 6FAM-CCA AAG CAT CCG GGA TGG CC-TAMRA; 5′ to 3′ primer sequences: GAC CAT GAC CCT CCA CAC, CCT GGA TCT GAT GCA GTA GG).

For each ESR1 c.975C>G (rs1801132) allele, a 17-mer oligonucleotide centered on the polymorphism was synthesized (5′ to 3′ oligo sequences: GAG CCC CCC ATA CTC TA, GAG CCC CCG ATA CTC TA), end-labeled with [γ³³P]adenosine triphosphate using polynucleotide kinase (New England Biolabs, Beverly, Mass), and hybridized at 52°C in a tetramethyl ammonium chloride–based buffer to Hybond N⁺ membranes (Amersham Pharmacia Biotech, Buckinghamshire, England) supporting polymerase chain reaction products. Polymerase chain reaction conditions were as follows: 95°C for 5 minutes; 35×: 95°C for 30 seconds, 54°C for 30 seconds, 72°C for 1 minute; 72°C for 10 minutes (5′ to 3′ primer sequences: TTC ACC TGT GTT TTC AGG GA, GCT GCG CTT CGC ATT CTT AC).

The observed genotype frequencies were compared using a χ² test to determine if they were in Hardy-Weinberg equilibrium.

Although enrollment of participants in the offspring cohort of the Framingham Heart Study began in 1971, we were able to study only individuals who survived until the sixth examination, at which blood samples were taken for DNA extraction (approximately 27 years after study enrollment). Thus, we used a case-control study design presenting results for prevalence of CVD.

Association analyses examined the relation of the ESR1 c.454-397T>C genotype and prevalence of atherothrombotic CVD events. Three CVD end points were studied: (1) total atherosclerotic CVD events, defined as MI (both recognized and unrecognized), angina pectoris, coronary insufficiency, intermittent claudication, coronary heart disease death, or atherothrombotic stroke); (2) major atherosclerotic CVD (recognized acute MI, atherothrombotic stroke, coronary insufficiency, or coronary heart disease death); and (3) recognized acute MI.

Among the 1739 men and women, initial qualifying events for total atherosclerotic CVD included 52 MIs (42 of which were recognized), 75 cases of angina pectoris, 8 cases of coronary insufficiency, 29 cases with intermittent claudication, 0 CHD deaths, and 14 atherothrombotic strokes. Initial major atherosclerotic CVD included 57 MIs, 10 cases of coronary insufficiency, 0 CHD deaths, and 16 atherothrombotic strokes. There were 59 cases of recognized MI (including those qualifying as initial events for the other 2 outcome categories as well as subsequent events) in analyses restricted to that end point.

Multivariate logistic regression was carried out with adjustment for age, sex, and CVD risk factors measured at the first examination (body mass index, hypertension, diabetes mellitus, total cholesterol, high-density lipoprotein cholesterol, and cigarette smoking). Correction was made for overestimation of odds ratios in common events.²⁵ The nominal threshold for significance was set at P<.05 for comparison of baseline traits and P<.01 for association analyses. Statistical analyses were performed using SAS software, version 8.2 (SAS Institute Inc, Cary, NC).

The characteristics of the 1739 unrelated study participants at the baseline examination of the Framingham Heart Study offspring cohort are shown in Table 2. The distribution of ESR1 c.454-397T>C genotypes was in Hardy-Weinberg equilibrium (Table 3) (P>.05) except in the group of individuals with recognized MI (P = .03).

The results from analyses treating each genotype as a distinct group (genotype model) suggested a recessive effect of the C allele (Table 4 and Figure 1); when analyses were computed for the recessive model, the level of statistical significance increased for both major atherosclerotic CVD events and recognized MI (Table 4). Individuals homozygous for the C allele comprised 20% of the study sample and, compared with those carrying either 1 or 2 copies of the T allele, they had substantially higher odds of major atherosclerotic CVD and recognized MI (Table 4). For individuals with the CC genotype, a fully adjusted recessive C model yielded an odds ratio of 2.0 (95% confidence interval [CI], 1.3-3.2; P = .004) for major atherosclerotic CVD and 3.0 (95% CI, 1.7-5.2; P<.001) for recognized MI. Survival analysis was performed for comparison and the results were very similar; for example, after full adjustment and in a recessive model, association of MI with the ESR1 c.454-397T>C polymorphism gave a hazard ratio of 3.0 (95% CI, 1.7-5.1), with P<.001 in men and women combined, almost identical to the odds ratio reported in Table 4 and discussed herein. The deviation from Hardy-Weinberg equilibrium within individuals with recognized MI (Table 3) is consistent with a genuine association; 37% of individuals with recognized MI and only 20% of the general study population had a CC genotype. When we limited the regression analyses to men, the results remained significant (Table 5); small numbers of events among women precluded studying them separately. In contrast, no differences in total atherosclerotic CVD by genotype were observed. For comparison, we subsequently extended our investigation to 3 other highly informative and commonly studied SNPs in ESR1^{7- 14,19- 22}: c.454–351A>G in intron 1 (also called IVS1-354 A/G or the XbaI restriction site), c.30T>C in exon 1 (T30C, Ser10Ser, a silent change), and c.975C>G (C1335G, +975C/G, Pro325Pro) in exon 4. The pairwise linkage disequilibrium coefficients between ESR1 c.454-397T>C and the 3 other polymorphisms, ESR1 c.454-351A>G, ESR1 c.30T>C, and ESR1 c.975C>G, are 0.992, 0.823, and 0.362, respectively. The genotype frequencies of the 4 markers were similar to those reported previously (Table 3).^13,26- 27 With the exception of ESR1 c.975 C>G among the total atherosclerotic CVD group, the distribution of genotypes was in Hardy-Weinberg equilibrium for all 3 SNPs and results for association with all 3 end points were negative after adjustment for covariates (Table 6).

The CC genotype of ESR1 c.454-397 was present in one fifth of our population-based sample and it was associated with a 3-fold odds of MI. These findings underscore a potentially important role of ESR1 in influencing the development of atherosclerosis and/or in accelerating the transition from subclinical atherosclerosis to plaque rupture and acute thrombotic CVD events such as MI and stroke. The frequency of the CC genotype was 20% in the entire study sample, but it was 24% in individuals with the broadest definition of atherosclerotic CVD, 31% in individuals with major atherosclerotic CVD, and 37% in individuals with recognized MI. For the latter 2 end points, there was statistically significant evidence of association.

Although several prior investigations of the ESR1 c.454-397T>C variant and CVD risk factors have been reported, only 1, a postmortem analysis, found a significant association of this variant with coronary artery disease.¹² In the Helsinki Sudden Death Study of 300 white Finnish male autopsy cases aged 33 to 69 years at death, the ESR1 c.454-397T>C variant was significantly associated with coronary artery disease in the subset of men aged 53 years or older. In this group of 142 men, 69 had coronary thromboses and there was a significant 10.2-fold (95% CI, 1.1-103.5; P = .04) higher frequency of coronary thrombosis among the CC individuals compared with the TT individuals. The heterozygotes had an intermediate risk that was not significantly different from that of the TT homozygotes. In addition, the CC cases had, on average, a 5-fold greater area of complicated lesions than the TT cases (age-adjusted P = .001).¹² Because of the population-based design of the Framingham Heart Study and the limitations inherent in postmortem studies, our results greatly extend those of the Finnish investigation. There may be some interpopulation heterogeneity of environmental or genetic factors that resulted in our study finding no difference in CVD susceptibility in heterozygotes compared with the TT homozygotes. A study with a larger number of CVD events is required to examine questions of gene-gene and gene-environment interactions.

It is important to consider how likely it is that our findings can be replicated in future studies. A recent meta-analysis of more than 300 association studies of common variants investigated this question and provided results that led to the conclusion that an initial significant finding, when followed by a study with P<.001, is strongly predictive of future replication,²⁸ providing support that the results reported here are likely to be replicated. The design and results of our study include many of the features that are considered desirable components of an ideal association study.²⁹ These characteristics include a large sample size, small P values, an association that makes biological sense, and alleles that may affect the gene product in a way that is of potential physiological relevance. The consistency and completeness of follow-up in the Framingham Heart Study are additional strengths of this work. Despite the highly significant association derived from a biologically based a priori hypothesis, the possibility of population stratification and related limitations of all association studies cannot be excluded. The risk of population stratification resulting in spurious allelic associations appears to have been overestimated,³⁰ and in our study, the absence of association in the additional 3 ESR1 polymorphisms that we analyzed is consistent with an absence of serious stratification. Further studies are needed to determine if the substantial genetic risk that we have identified for MI can be generalized to other, genetically distinct populations.

Our study of unrelated men and women from the Framingham Heart Study included few events in women. Further investigations are required to determine if the association of ESR1 c.454-397T>C with MI applies to women and if there is a significant interaction with hormone therapy.

There are many reports of the ESR1 c.454-397T>C variant, mostly in bone mineral density or cancer studies. We observed a C allele frequency of 45% and other reports of white US individuals have provided very similar data (45%-49%),^13,26- 27 while studies of other races/ethnicities have provided a wider range of results (31%^31- 33 to 55%; see http://www.ncbi.nlm.nih.gov/SNP/).

Although the mechanism by which ESR1 variants may influence risk of MI requires further investigation, there are several biologically plausible explanations.⁶ Given the Findings reported here, a number of hormone-sensitive alterations in hemostatic and fibrinolytic variables^6,34- 36 may be worth exploring. Estrogen receptors are now understood to have an important role in normal vascular physiology in both men and women and in other species.^6,37- 40 The ESR1 gene has been shown to mediate 3 direct effects of estrogen on the vessel wall: acceleration of re-endothelialization,³⁷ alteration of endothelial nitric oxide production,³⁹ and inhibition of the vascular injury response.³⁸ These animal studies also demonstrated the importance of estrogen receptors in cardiovascular physiology in both males and females. This is consistent with human studies that provide further support for the importance of estrogen receptors in both sexes.^41- 42

In summary, our results reveal an association between a common estrogen receptor genotype, ESR1 c.454-397CC, and increased odds of MI. This association persists after adjustment for traditional CVD risk factors and provides support for the importance of estrogen receptors in CVD susceptibility, especially in men. Estrogen receptor variation also has the potential to explain recent data regarding the effects of combination hormone therapy on CVD risk in women.^3,43