Novel genetic variants linked to coronary artery disease by genome-wide association are not associated with carotid artery intima-media thickness or intermediate risk phenotypes.

BACKGROUND: It is uncertain whether the novel single nucleotide polymorphisms (SNPs) that have recently been associated with coronary artery disease (CAD) in genome-wide studies also influence carotid atheroma and stroke risk. The mechanisms of their association with CAD are unknown; relationships to other cardiovascular phenotypes may give mechanistic clues. Carotid artery intima-media thickness (CIMT) is a subclinical marker of atherosclerosis associated with stroke. We investigated association of reported CAD risk variants with CIMT, and with other intermediate phenotypes that may implicate causative pathways.
METHODS: We studied 1425 members of 248 British Caucasian families ascertained through a hypertensive proband. We genotyped CAD risk SNPs on chromosomes 9 (rs1333049, rs7044859, rs496892, rs7865618), 6 (rs6922269) and 2 (rs2943634) using TaqMan. Merlin software was used for family-based association testing.
RESULTS: No significant association was found between genotype at any SNP and CIMT in 846 individuals with acceptable measurements. Nor were SNPs significantly associated with blood pressure, obesity, cholesterol, CRP, interleukin-6, TNF-alpha, or leptin.
CONCLUSIONS: These novel CAD variants are not associated with CIMT and do not appear to mediate the risk of atherothrombosis through known risk factors.

Background

A number of common single nucleotide polymorphisms (SNPs) have been robustly associated with coronary artery disease (CAD) in recent genome-wide studies [1-3], but, in contrast to previous associations derived from candidate gene analyses, the mechanisms underlying the associations with CAD are unknown. A relationship of these variants to other intermediate cardiovascular phenotypes may give important mechanistic clues. Carotid artery intima-media thickness (CIMT) is a subclinical marker of atherosclerosis associated with stroke [4]. Blood pressure, plasma cholesterol and obesity are established cardiovascular risk factors, while plasma levels of interleukin-6 (IL-6), tumour necrosis factor-α (TNF-α), C-reactive protein (CRP), and leptin have also been implicated in atherogenesis [5].

We determined whether these novel CAD risk variants are associated with carotid atheroma, assessed by CIMT, and other intermediate risk markers which might implicate causative pathways, in a large panel of extended families that has been shown to have adequate power to detect small genetic influences on quantitative traits, including CIMT [6].

Methods

Participants

1425 members of 248 British Caucasian families were ascertained through hypertensive probands and phenotyped for a quantitative genetic study of cardiovascular risk factors from 1993 to 2001. The local ethics committee approved the study; the detailed ascertainment strategy has been described previously [6]. A clinical history, body mass index (BMI), waist–hip ratio (WHR), 24-hour ambulatory blood pressure recordings, and blood samples were obtained. Bilateral common carotid ultrasonography and CIMT measurement was performed in a subset of 953 participants, as previously described [6]. Levels of plasma IL-6, TNF-α, high-sensitivity CRP, leptin and total non-fasting cholesterol were measured using commercially available assays.

Genotyping

The chromosome 9p21.3 region associated with CAD in previous genome-wide studies consists of two haplotype blocks which can be tagged in Caucasian populations with three SNPs from block 1 (rs7044859, rs496892 and rs7865618) and one SNP from block 2 (rs1333049) [3,7]. We therefore genotyped these four SNPs in that region. We also genotyped the lead SNPs most strongly associated with CAD in previous studies on chromosome 6q25.1 (rs6922269) and chromosome 2q36.3 (rs2943634) [3]. All SNPs were genotyped using TaqMan. Mendelian inheritance and correspondence of genotype frequencies to Hardy–Weinberg proportions was checked using PEDSTATS [8].

Statistical analysis

Phenotypes were examined for normality and log-transformed if required. Adjustments were made for significant covariates determined by linear regression, considering age, sex, cardiovascular medications, smoking status (current/former/never), alcohol consumption (units per week) and exercise habit (frequency per week). Association between genotypes and adjusted phenotypes was assessed using Merlin v1.1.2 [9]. This evaluates the evidence for association under an additive genetic model, which was selected in view of the data from previous publications showing clear trends in CAD risk between different genotypes at these SNPs (i.e. heterozygotes having CAD risk intermediate between high-risk and low-risk homozygotes) [3]. In order to estimate the upper bound of the genetic effect that is plausibly associated with each SNP for each phenotype, linear regression models (with robust “sandwich” variance estimation to compensate for family clustering) were fitted to model additive genetic effects using Stata™9.2. Narrow-sense heritability estimates (and 95% confidence intervals thereof) were calculated by standard population genetic formulae.

Results

1425 participants were included: mean age 49.5 (±15.5) years, 747 (52.4%) female. General characteristics of the study participants are shown in Table 1.

Table 1

General characteristics of the study population

Characteristic	N	Median (interquartile range)
Previous ishemic heart disease	61/1425 (4.3%)	–
Previous stroke	33/1425 (2.3%)	–
Previous peripheral vascular disease	13/1425 (0.9%)	–
Diabetes	35/1425 (2.5%)	–
Current or former smoker	507/1425 (35.6%)	–
Hypertension	512/1425 (35.9%)	–
Daytime systolic blood pressure (mmHg)	958	131 (121.1–144.1)
Daytime diastolic blood pressure (mmHg)	958	78.6 (72.0–88.0)
Plasma total cholesterol (mmol/l)	1289	5.6 (4.8–6.4)
BMI (kg/m2)	1402	25.4 (23.1–28.2)
WHR	1357	0.85 (0.78–0.91)

854 participants had measurable CIMT, as shown in Table 2. Median maximal CIMT values of 0.858 mm in men and 0.803 mm in women were within the normal population range (0.36–1.07 mm).

Table 2

Risk phenotypes in the study population and estimated maximum genetic effect on total phenotypic variance for the typed SNPs

Variable	N	Median (interquartile range)	Maximum genetic effect for typed SNPsa
BMI (kg/m2)b	1402	25.4 (23.1–28.2)	0.4–1.7%
WHR	1357	0.85 (0.78–0.91)	0.6–1.4%
Daytime systolic blood pressure (mmHg)b	958	131 (121.1–144.1)	0.5–0.7%
Mean CIMT (mm)b	854	0.76 (0.65–0.91)	0.7–1.5%
Max CIMT (mm)b	856	0.83 (0.71–1.00)	0.7–1.4%
Plasma total cholesterol (mmol/l)b	1289	5.6 (4.8–6.4)	0.5–1.4%
Plasma IL-6 (pg/ml)b	1186	0.78 (0.48–1.38)	0.4–0.7%
Plasma TNF-α (pg/ml)b	1186	0.74 (0.35–1.66)	0.6–1.3%
Plasma CRP (mg/l)b	1314	1.40 (0.55–3.2)	0.4–1.7%
Plasma Leptin (ng/μl)b	1319	8.6 (4.6–15.3)	0.5–1.0%

Numbers represent the range of maximum plausible genetic contribution to total phenotypic variance of each trait for typed SNPs.

Variables log-transformed before analysis to approximately normalise the distributions.

Genotyping was complete for more than 95% of the 1393 participants with DNA available at all loci. Allele frequencies and linkage disequilibrium for typed SNPs are shown in Fig. 1; these were similar to the HapMap CEU population and did not deviate significantly from Hardy–Weinberg proportions.

Fig. 1

Linkage disequilibrium and minor allele frequencies of typed SNPs. Linkage disequilibrium plot for chromosome 9p21 SNPs showing in each diamond the D’ (on top) and r2 values (on bottom) for typed SNP combinations in our population. mAF = Minor allele frequency.

There was no significant association between genotype and covariate-adjusted log-transformed maximal CIMT at the loci studied. Neither was there evidence of a significant association between genotypes and plasma cholesterol, CRP, IL-6, TNF-α, or leptin concentrations; nor clinical measurements of blood pressure, BMI or WHR (using a global p-value of p = 0.05, adjusted for multiple testing). The upper bound of the contribution to the total population variance in maximal CIMT was between 0.7% and 1.4% for the typed SNPs, and of a similar order for the other traits tested (Table 2, right-hand column).

A stepwise backward-elimination multivariate regression procedure failed to find any multilocus models showing association (removal p-value > 0.2) to the CIMT phenotype. This is a powerful approach for examining tightly linked SNPs in association models that avoids the computational complexity and degrees of freedom that accompany a formal haplotype analysis [10]. It is therefore unlikely that chromosome 9p21 haplotypes are associated with CIMT, recognising that the power to detect subtle associations is very low for minor haplotypes with frequencies < 5%.

Discussion

The lack of association with CIMT measurements suggests that these novel SNPs, while convincingly associated with CAD risk, do not influence development of carotid atherosclerosis. Our findings extend the results of a recent study that found no significant association between a SNP associated with CAD risk on chromosome 9 (rs10757278, for which rs1333049 we typed is a perfect proxy) and ischemic stroke, once potentially-confounding cases of known CAD were excluded from the analysis [11].

The lack of association with intermediate biochemical and clinical phenotypes associated with atherosclerosis (including inflammatory markers) extends the findings of two previous studies which found no association between CAD risk variants on chromosome 9 (in strong linkage disequilibrium with the SNPs we typed, r2 > 0.8) and risk factors including blood pressure, BMI, cholesterol and glucose [2]; nor with biochemical parameters including lipoprotein(a), fibrinogen, albumin, bilirubin and homocysteine [12].

The maximum plausible genetic effect for the typed SNPs on CIMT and the other phenotypes, which we calculated based on the observed mean effect and 95% confidence interval, was low in all cases. Therefore, although this study cannot exclude any effect of the typed SNPs on these phenotypes, it demonstrates that such effects could only be of small magnitude, and are unlikely to be clinically significant. Selection through hypertensive probands could potentially introduce bias if there were an interaction between hypertension and the variants studied, but only 36% of the participants were hypertensive and the variants did not associate with hypertension. On the other hand, ascertainment for hypertension, a trait strongly related to cardiovascular risk, resulted in selection of individuals with mean CIMT 33% higher than the general population, which should increase the power for the detection of variants with small effects.

CIMT has been commonly used as a surrogate marker for atherosclerosis [13,14]. It is associated with cardiovascular disease risk factors and is an independent predictor of atherosclerosis and clinical outcomes including stroke [4,15]. CIMT progression may be reduced or reversed with interventions that reduce the risk of future cardiovascular events, supporting the concept that CIMT measurement can be used as a marker for atherosclerosis [15]. Although a close association between CIMT and carotid plaques has been demonstrated [16], however, increased CIMT is not synonymous with carotid atheroma [15]. A potential limitation of this study is that the presence of carotid plaque, which is more specific for atherosclerosis, was not directly assessed.

We examined SNP associations with a number of intermediate phenotypes to identify causative pathways through which genotypes at these SNPs influence risk. No associations were found, suggesting that the effects are mediated through unknown pathways. Further investigations will be required to determine the mechanisms by which these SNPs are associated with CAD.

Conclusion

These novel CAD variants are not associated with variation in CIMT and do not appear to mediate CAD risk through known risk factors.

Conflict of interest

The authors declare no conflicts of interest relating to this research.