Author + information
- Received August 11, 2009
- Revision received September 23, 2009
- Accepted October 29, 2009
- Published online August 3, 2010.
- Sonny Dandona, MD,
- Alexandre F.R. Stewart, PhD⁎ (, )
- Li Chen, MSc,
- Kathryn Williams, MS,
- Derek So, MD,
- Ed O'Brien, MD,
- Christopher Glover, MD,
- Michel LeMay, MD,
- Olivia Assogba, BA,
- Lan Vo, MSc,
- Yan Qing Wang, BS,
- Marino Labinaz, MD,
- George A. Wells, PhD,
- Ruth McPherson, MD, PhD and
- Robert Roberts, MD
- ↵⁎Reprint requests and correspondence:
Dr. Alexandre F. R. Stewart, University of Ottawa Heart Institute, 40 Ruskin Street, H3100, Ottawa, Ontario K1Y 4W7, Canada
Objectives The purpose of this study was to test the hypothesis that 9p21 gene dosage determines the severity of coronary artery disease (CAD).
Background The 9p21 locus is the first common genetic variant to associate with risk of CAD and/or myocardial infarction in multiple studies.
Methods A cross-sectional study examined nondiabetic patients with CAD defined by coronary angiography to have at least 1 epicardial stenosis >50%. In all, 950 patients with early onset CAD (age 56.1 ± 9.6 years) and an independent sample of 764 patients with late onset CAD (age 70.0 ± 8.0 years) were enrolled from the cardiac catheterization laboratories at the University of Ottawa Heart Institute from April 15, 2006, to August 15, 2008, and genotyped for the single nucleotide polymorphism rs1333049 9p21 risk variant. Angiographers were blinded to genotype. The association between 9p21 risk genotype and the proportion of patients with 3-vessel disease, 1-vessel disease, left main trunk disease, and coronary artery bypass graft surgery was tested, as was its association with the modified Gensini and Duke coronary scoring indexes.
Results Among younger CAD cases, 3-vessel disease demonstrated a strong, direct association with 9p21 gene dosage (p = 4.26 × 10−4). Conversely, 1-vessel disease demonstrated a strong inverse association with increasing gene dosage (p = 2.41 × 10−5). In the replication sample, gene dosage also predicted 3-vessel disease (p = 6.51 × 10−6). Left main trunk disease and coronary artery bypass graft surgery demonstrated a direct strong association with gene dosage (p = 3.66 × 10−4) and (p = 2.42 × 10−2), respectively. Gene dosage demonstrated a strong, direct association with both the modified Gensini (p < 0.0001) and modified Duke (p = 3 × 10−4) coronary scores. Risk variant 9p21 did not associate with myocardial infarction once stratified for disease severity.
Conclusions Gene dosage of the common risk variant 9p21 predicts the severity of coronary atheromatous burden.
Coronary artery disease (CAD) is the leading cause of mortality in the western world (1,2). It will soon be the pre-eminent killer in the developing world (3). Coronary artery disease manifests in many forms from chronic stable angina to that of sudden death in asymptomatic persons. It is a disease that has both genetic and environmental determinants. Until recent advances in high throughput genome-wide association studies, attempts to define the genetics of CAD were undertaken almost exclusively by the candidate gene approach. Genes identified by this approach have not been replicated in subsequent populations.
In 2007, the first genetic variant, 9p21, was discovered using genome-wide association studies that predisposes to CAD in Caucasians and was almost simultaneously reported by 3 other groups (4–7). Our group identified the 9p21 variant in samples derived from patients with angiographically defined CAD (6). Icelandic, German, and United Kingdom samples were largely from myocardial infarction (MI)-based cohorts (4,5,7). More recently, the risk locus was shown to confer cross-race risk by demonstrating an association with CAD in Korean and Japanese patients (8–10). Extending the concept of vascular phenotype, Helgadottir et al. (11) demonstrated an association with both abdominal aortic and intracranial aneurysms.
While the 9p21 risk for CAD and its independence of conventional risk factors (i.e., diabetes mellitus, hypercholesterolemia, hypertension, smoking, and so forth) have been confirmed worldwide by many investigators (6,12), of major interest is how 9p21 exerts its effect on the coronary vasculature. The 9p21 risk locus spans 58,000 base pairs and contains a noncoding ribonucleic acid antisense sequence whose function has not been elucidated. The 9p21 variant is extremely common with approximately 50% of the population possessing 1 copy and a further 20% to 25% possessing 2 copies of the risk variant. One copy of 9p21 increases risk for CAD by approximately 20% whereas 2 copies augment risk by 40% (6). If 9p21 were to mediate its effect primarily by promoting deposition of coronary atheroma, one would expect homozygotes for the risk allele to have more severe CAD than heterozygotes. To test this hypothesis, we genotyped 9p21 in patients (n = 950) with angiographically documented CAD stratified by disease burden and replicated the findings in an independent sample (n = 764) of older patients also with angiographically documented CAD.
Study design and eligibility
The study was conducted at the University of Ottawa Heart Institute. Patients were enrolled from the coronary catheterization laboratory between the dates of April 18, 2006, and August 15, 2008. Males <55 years of age and females <65 years of age with at least 1 epicardial stenosis >50% were eligible for the early onset sample of angiographic cases (Table 1).A second sample of late onset cases (Table 2) was obtained from all patients who were older than the age cutoff for the early onset sample. The Ottawa Heart Genomics Study (13) was designed to identify genetic variants that predispose to CAD. At its inception, it was decided to exclude diabetic patients to enrich the population for genetic variants that uniquely pre-dispose to CAD.
This study was approved by the Research Ethics Board of the University of Ottawa Heart Institute, and written informed consent was obtained from all study participants. An arterial blood sample was taken at the time of the catheterization for deoxyribonucleic acid extraction and subsequent genotyping.
Genomic deoxyribonucleic acid was purified from peripheral blood lymphocytes from patients with CAD and genotyped using the Affymetrix (Santa Clara, California) 500K and 6.0 arrays. All samples analyzed by the 500K arrays had call rates >93% by the dynamic module algorithm and 98% by the Bayesian Robust Linear Model with Mahalanobis distance classifier algorithm. Samples analyzed by the 6.0 arrays yielded an average call rate by quality control of 95.7% and by the Birdseed algorithm of 99.3%. Population stratification was analyzed by the principal components method (14), and outliers who may be of non-European origin were removed from the analysis. After removing outliers, genotypes were analyzed in 950 early onset and 764 late onset patients for whom complete angiographic studies were available for review. Conformity of the genotypes to Hardy-Weinberg equilibrium was confirmed by the exact test. Patients were designated AA (risk/risk), AB (risk/nonrisk), and BB (nonrisk/nonrisk) according to the number of inherited variant alleles (rs1333049).
The coronary angiograms were reviewed by 2 independent angiographers who were both blinded to the results of the genotype. For a vessel to be scored, stenosis >50% had to be noted in an epicardial coronary vessel of interest or in one of its major branches. Ramus intermedius lesions were attributed to the circumflex system for the purposes of this study. Left main disease was defined as stenosis >50% in a vessel that was not bypassed, namely, protected left main disease was not included for the purposes of analysis, given accelerated atherosclerosis in segments proximal to a bypass graft anastomosis (15). Patients were classified as coronary artery bypass graft surgery (CABG) if it had been performed before enrollment (i.e., those patients whose angiograms at study entry led to CABG were not included for the purposes of analysis). In the event of discordance of the number of vessels scored between the 2 reviewers, angiograms were scored by a third independent reviewer. The angiogram performed on the day of entry was classified as either inaugural (first diagnostic catheterization) or repeat. This was done so that the inaugural subgroup could be interrogated separately; this would provide information with respect to atheromatous-burden acquired in the asymptomatic phase of disease.
Modified Gensini and modified Duke coronary scores were derived in subjects who were not previously bypassed. Furthermore, subjects who had previously undergone percutaneous intervention for a lesion for which the degree of stenosis (pre-intervention) was not known were not analyzed. If in-stent restenosis occurred producing a lesion of greater significance than the lesion that necessitated initial intervention, then these subjects were not analyzed. Modified Gensini scores were derived by a method previously described (16). In short, the number of lesions, their severity, and their respective locations were evaluated. The coronary vasculature was divided into 27 segments, and each segment was weighted from 0.5 to 5.0, reflecting the importance of location. The severity was scored as follows: <25% = 2; 26% to 50% = 4; 51% to 75% = 8; 76% to 90% = 16; 91% to 99% = 32: total occlusion = 64. The product of the severity score and location score constitutes the score for any given segment, and their sum is the score for a given subject. Modified Duke scores were derived by a method previously described (17). The number of lesions and their location (left main trunk and proximal left anterior descending artery [LAD]) were weighted as follows: stenosis >50% 1 vessel = 23; stenosis >50% 2 vessels = 37; >50% 3 vessels = 56; stenosis >50% 1 vessel and proximal LAD >50% = 48; stenosis >50% 2 vessels and proximal LAD >50% = 56; stenosis >50% 3 vessels and >50% proximal LAD = 74; >50% lesion in left main trunk = 80; >80% lesion in left main trunk = 100.
A power calculation was conducted to determine whether the early onset CAD (340 cases with 3-vessel disease, and 610 cases with 2- and 1-vessel disease) and late onset CAD (293 cases with 3-vessel disease, and 471 cases with 2- and 1-vessel disease) sample sizes were sufficient to detect important linear trends across genotypes. With a level of significance of 0.05, minor allele frequency ranging from 0.3 to 0.4, the power ranges from 85% to 93% to detect the genotypic odds ratios (ORs) of 1.4 based on an additive model.
For baseline patient characteristics, continuous variables were presented as means and standard deviations, and categorical variables were presented as frequencies with percentages. The Wilcoxon rank-sum test was used to compare continuous variables, and Fisher's exact test was used for categorical variables. The associations between the proportion of patients with 3-vessel disease, 1-vessel disease, left main trunk involvement, and CABG and gene dosage of the rs1333049 risk variant (i.e., the number of risk copies of rs1333049 inherited) were compared using additive tests adjusted for sex, age, hypertension, smoking, and body mass index (BMI) in patients with early and late onset CAD. The associations between modified Duke and modified Gensini scores and gene dosage of the rs1333049 risk variant were also compared using additive tests after adjustment for sex, age, hypertension, smoking and BMI in patients with early onset disease. ORs per copy of risk allele for the rs1333049 with 95% confidence intervals (CIs) were calculated for all end points. To increase the power, a combined analysis of 2 cohorts was performed to assess the overall effect of the association between gene dosage of the rs1333049 risk variant and the proportion of patients with 3- and 1-vessel disease using additive tests after adjustment for sex, age, hypertension, smoking, BMI, and baseline lipids. Statistical analyses were performed using SAS software (version 9.1.3, SAS Institute, Cary, North Carolina), and statistical significance was defined as p < 0.05.
Effect of 9p21 genotype on disease severity
There was a strong direct association between the proportion of early onset patients with 3-vessel disease and increasing gene dosage of the rs1333049 risk variant (p = 4.26 × 10−4) (Fig. 1A).The corresponding OR per copy of the risk allele for rs1333049 was 1.45 (95% CI: 1.18 to 1.79). Conversely, there was a strong inverse association between the proportion of patients with 1-vessel disease and increasing gene dosage of the risk variant (p = 2.41 × 10−5) (Fig. 1B). The corresponding OR per copy of the risk allele for rs1333049 was 0.64 (95% CI: 0.52 to 0.79). Thus, patients homozygous for the 9p21 risk allele are more likely to have severe CAD.
Replication of 9p21 genotype association with CAD severity in elderly cases
Given the strong predictive association between gene dosage and severity of CAD in the early onset sample, we asked whether this association would hold in older patients with angiographically documented CAD. A strong direct association between the proportion of older patients with 3-vessel disease and increasing gene dosage of the risk variant was seen (p = 6.51 × 10−6) (Fig. 1C). The corresponding OR per copy of the risk allele for rs1333049 was 1.70 (95% CI: 1.35 to 2.14). Similarly, there was an inverse association between the proportion of older patients with 1-vessel disease and increasing gene dosage of the risk variant (p = 4.97 × 10−2) (Fig. 1D). The corresponding OR per copy of the risk allele for rs1333049 was 0.79 (95% CI: 0.62 to 1.00). Thus, even in older patients, an association between 9p21 gene dosage and CAD severity was observed.
Effect of 9p21 genotype on CAD severity at inaugural catheterization
For patients in whom the catheterization was inaugural, there was a direct association between the proportion of patients with 3-vessel disease and increasing gene dosage of the risk variant (p = 1.14 × 10−2) (Fig. 2A).The corresponding OR per copy of the risk allele for rs1333049 is 1.43 (95% CI: 1.08 to 1.89). Conversely, there was a strong inverse association between proportion of patients with 1-vessel disease and increasing gene dosage of the risk variant (p = 3.21 × 10−4) (Fig. 2B). The corresponding OR per copy of the risk allele for rs1333049 is 0.64 (95% CI: 0.51 to 0.82).
Effect of 9p21 genotype on left main trunk disease and CABG
There was a strong direct association between the proportion of early onset patients with left main trunk disease and increasing gene dosage of the risk variant (p = 3.66 × 10−4) (Fig. 2C). The corresponding OR per copy of the risk allele for rs1333049 is 2.38 (95% CI: 1.48 to 3.85). There was also a significant association between the proportion of early onset patients with CABG and increasing gene dosage of the risk variant (2.42 × 10−2) (Fig. 2D). The corresponding OR per copy of the risk allele for rs1333049 is 1.37 (95% CI: 1.04 to 1.79).
Effect of 9p21 genotype on modified Gensini and modified Duke coronary scoring
There was a strong, direct association of increasing gene dosage with both modified Gensini coronary scores (p < 0.0001) and modified Duke coronary scores (p = 3.0 × 10−4) in the early onset cohort (Table 3).
To increase the power, a combined analysis of early and late onset CAD cases was performed to assess the overall effect of the association between rs1333049 gene dose and disease severity using additive tests after adjustment for sex, age, hypertension, smoking, BMI, and baseline lipids. The combined analysis also demonstrated a significant association between 9p21 genotype and disease severity (Figs. 1E and 1F, Table 4).Another SNP at 9p21 in linkage disequilibrium with rs1333049, rs9632884 (r2= 0.832) showed a similar association of gene dosage with proportion of 3-vessel disease (p = 2.2 × 10−3, OR: 1.38, 95% CI: 1.12 to 1.70) and 1-vessel disease (p = 8.9 × 10−5, OR: 0.67, 95% CI: 0.55 to 0.82; data not shown). When the 143 angiographic cases with <50% stenosis in any coronary were included and called 0-vessel disease, the minor allele of rs1333049 showed a highly significant association with CAD severity (p = 0.008) (Fig. 3).
Genotype 9p21 does not associate with MI
Given that 9p21 associates with CAD severity, we asked whether it also associates with myocardial infarction (MI). The prevalence of 3-vessel disease is higher among patients who had sustained an MI than among patients who had not. Conversely, the prevalence of 1-vessel disease is lower among MI patients (Table 5).Therefore, to examine whether there is an association between 9p21 and MI, we stratified for number of diseased vessels. No association between 9p21 and MI was observed (Table 6).
For the first time, we show that gene dosage of the 9p21 risk allele predicts CAD burden. Patients (n = 950) with early onset CAD homozygous for the 9p21 risk allele had a twofold higher risk for the development of 3-vessel disease at inaugural catheterization when compared to the nonrisk genotype BB. Markedly increased frequencies of both significant left main trunk disease and CABG further confirmed homozygosity to be associated with more severe disease. Furthermore, we have demonstrated a robust direct relationship between 9p21 gene dosage and modified Gensini and Duke prognostic scores. An independent sample (n = 764) of older patients with angiographically documented CAD was subsequently genotyped, and the finding of 9p21 gene dose as a predictor of disease severity was replicated. Thus, there exists a dose response between the number of inherited copies of the risk allele and CAD severity. These findings indicate that 9p21 mediates its risk through deposition of coronary atheroma rather than plaque rupture or thrombosis. Screening for 9p21 gene dosage provides a strategy to separate individuals into high- and low-risk profiles for CAD.
In the Wellcome Trust Case Control Consortium study, 9p21 was more strongly associated with revascularization than MI (7). This is consistent with the observation of Broadbent et al. (18) that, in their CAD population, patients without a history of MI showed stronger association with 9p21 than did those with a history of MI. A Chinese study confirmed the 9p21-CAD association (19). Interrogation of these data by us revealed an association between the rs2383207 gene dosage (in linkage disequilibrium with rs1333049, r2= 0.915) and 3-vessel disease (p = 0.031). Fischer et al. (20) have indicated heritability of proximal coronary disease and provided evidence of familial aggregation of left main disease, suggesting a genetic predisposition for this anatomic variant of CAD (21). Our data are consistent with this observation and implicate 9p21 in the risk for left main disease. These studies are consistent with our observation that gene dosage of the risk variant predicts the degree of coronary atheromatous burden, suggesting that 9p21 promotes formation of atheroma.Furthermore, we have demonstrated that once stratified for the number of vessels with at least 1 50% lesion, there is no significant difference in minor allele frequency between the group who had sustained an MI versus the group who had not (Table 6). That is, the effect of 9p21 on the risk for MI is a consequence of greater atherosclerotic burden rather than a propensity for plaque rupture and/or thrombosis. The implications of this point are worth emphasizing. Despite 3 years since the initial discovery, the mode by which 9p21 exerts its risk is unknown. Its elucidation will most certainly reveal a novel mechanism that will provide a novel therapeutic target. Our results suggest that efforts to determine the mechanisms that underlie the action of 9p21 should focus on processes that promote atherogenesis rather than those related to plaque instability and thrombosis. Recently presented data in an Italian population that had experienced an index MI demonstrated that, while gene dosage was predictive of subsequent revascularization, it had no bearing on subsequent MI (22), supporting the notion that 9p21 acts on plaque progression rather than plaque rupture and/or thrombosis.
Recent studies confirming an association of 9p21 and CAD risk in Koreans, Japanese, and Utah Caucasians did not show a gene-dose relationship with disease severity (8–10,23). Several possibilities might account for their findings. First, the study populations were heavily skewed toward single-vessel disease with the remaining sample size for 2- and 3-vessel disease inadequate for determination of a gene dose response. Second, the Utah study used rs2383206 to genotype the 9p21 locus. Although rs2383206 is in linkage disequilibrium with rs1333049 (r2= 0.875 in Utah Caucasians), it might represent a less robust marker than rs1333049 for the 9p21 risk locus. Because rs2383206 is not on our array, we examined the association between the gene dosage of rs9632884, another SNP in linkage disequilibrium with rs1333049 (r2= 0.832 in Utah Caucasians), and proportion of patients with 1-vessel and 3-vessel disease, respectively. In this case, the association, while still significant, was less robust. Third, 42% of the patients in 1 Japanese study (10) and 23.4% of the Utah patients were diabetic, whereas diabetic patients were excluded from the Ottawa Heart Genomic Study (13). Recent evidence suggests that 9p21 exacerbates the risk conferred by poor glycemic control in patients with type 2 diabetes (24). Thus, diabetes might have obscured the association of 9p21 gene dosage with CAD severity. These negative reports provided, in part, the impetus for us to confirm our findings in an independent sample.
The 9p21 gene dose response for CAD severity was evident even among patients at their inaugural catheterization. This finding raises the possibility that patients with a homozygous risk genotype acquire a greater atheromatous burden in the asymptomatic phase of the disease. Thus, screening with 9p21 may capture anatomically important disease before symptoms. It should be noted that testing for a SNP located at 9p21, rs10757274, which is in linkage disequilibrium with rs1333049, added significantly to the ability of the Framingham risk score to discriminate subjects at risk for cardiac events (25).
Both the modified Gensini and the modified Duke prognostic scores demonstrate a robust, direct relationship with 9p21 gene dosage. It should be noted that this relationship exists within a CAD population. That is, whereas several studies have documented that 9p21 confers risk for CAD versus controls, we describe the ability of 9p21 gene dosage to predict anatomically important disease that carries an adverse prognosis.
The 9p21 risk allele is an extremely common variant that is present at birth and not subject to the variation characteristic of the standard, modifiable risk factors (e.g., diurnal variation in blood pressure, the dietary effects on cholesterol measurement and the effects of physiologic stress on the level of low-density lipoproteins and blood pressure). Given its ability to predict risk within a CAD population (i.e., the direct correlation between 9p21 gene dosage and both prognostic indices), genotyping 9p21 may be useful not only in determining risk for development of disease but also for risk stratification among patients with documented CAD.
The authors thank Patricia Gerard, Heather Doelle, RN, Gwen Ewart, RN, Julie Rutberg, MS, and Rosemary LaRose, RN, for sample collection and data management.
Drs. Stewart, McPherson, Wells, and Roberts are supported by grants from the Canadian Institutes of Health Research. Infrastructure support is provided by the Canada Foundation for Innovation and the Ontario Research Fund.
- Abbreviations and Acronyms
- body mass index
- coronary artery bypass graft surgery
- coronary artery disease
- confidence interval
- left anterior descending artery
- myocardial infarction
- odds ratio
- Received August 11, 2009.
- Revision received September 23, 2009.
- Accepted October 29, 2009.
- American College of Cardiology Foundation
- Thom T.,
- Haase N.,
- Rosamond W.,
- et al.
- Helgadottir A.,
- Thorleifsson G.,
- Manolescu A.,
- et al.
- McPherson R.,
- Pertsemlidis A.,
- Kavaslar N.,
- et al.
- Shen G.Q.,
- Li L.,
- Rao S.,
- et al.
- Schunkert H.,
- Gotz A.,
- Braund P.,
- et al.
- Stewart A.F.,
- Dandona S.,
- Chen L.,
- et al.
- Montorsi P.,
- Ravagnani P.M.,
- Galli S.,
- et al.
- Broadbent H.M.,
- Peden J.F.,
- Lorkowski S.,
- et al.
- Fischer M.,
- Broeckel U.,
- Holmer S.,
- et al.
- Fischer M.,
- Mayer B.,
- Baessler A.,
- et al.
- Lina D.,
- Notarangelo M.F.,
- Merlini P.A.,
- et al.
- Talmud P.J.,
- Cooper J.A.,
- Palmen J.,
- et al.