NOS3 polymorphisms, cigarette smoking, and cardiovascular disease risk

Abstract

Objective

Endothelial nitric oxide synthase (NOS3) activity and cigarette smoking significantly influence endothelial function. We sought to determine whether cigarette smoking modified the association between NOS3 polymorphisms and risk of coronary heart disease or stroke.

Methods

All 1085 incident coronary heart disease cases, all 300 incident ischemic stroke cases, and 1065 reference individuals from the Atherosclerosis Risk in Communities study were genotyped for the T-786C and E298D polymorphisms in NOS3. Using a case–cohort design, associations between genotype/haplotype and disease risk were evaluated by multivariable proportional hazards regression. Multiplicative scale interaction testing evaluated the influence of cigarette smoking history at baseline on these associations.

Results

In Caucasians, association between E298D genotype and risk of coronary heart disease was significantly modified by current smoking status (interaction P = 0.013), with the highest risk observed in smokers carrying the variant D298 allele relative to nonsmokers carrying two E298 alleles (adjusted hazard rate ratio 2.07, 95% confidence interval 1.39−3.07). In African-Americans, association between T-786C genotype and risk of ischemic stroke was significantly modified by pack-year smoking history (interaction P = 0.037), with the highest risk observed in ≥ 20 pack-year smokers carrying the variant C-786 allele relative to < 20 pack-year smokers carrying two T-786 alleles (adjusted hazard rate ratio 4.03, 95% confidence interval 1.54−10.6).

Conclusions

An interaction between the E298D and T-786C polymorphisms in NOS3, cigarette smoking, and risk of incident coronary heart disease and ischemic stroke events appears to exist, suggesting a potential complex interplay between genetic and environmental factors and cardiovascular disease risk.

Introduction

Ischemic cardiovascular disease is a major public health problem. It is estimated that 1.2 million Americans will experience an acute coronary heart disease (CHD) event and 0.7 million will experience an acute stroke event this year [1]. Endothelial dysfunction is integrally involved in the pathogenesis of ischemic cardiovascular disease, and has been associated with an increased risk of myocardial infarction and stroke [2,3].

Nitric oxide is synthesized by endothelial nitric oxide synthase (eNOS, NOS3) in the vasculature, where it possesses potent vasodilatory [4] and anti-inflammatory [5] effects. Impairment in eNOS-mediated nitric oxide biosynthesis contributes to endothelial dysfunction and promotes the development of atherosclerosis [6,7]. Cigarette smoking also contributes to endothelial dysfunction and atherosclerosis [8], eliciting many of its deleterious endothelial effects via attenuation of nitric oxide biosynthesis and activity [9,10].

Multiple genetic polymorphisms in NOS3 have been discovered, including the T-786C and E298D variants that have been associated with reduced nitric oxide biosynthesis [11,12]. Numerous clinical studies evaluating potential associations between NOS3 polymorphisms and cardiovascular disease risk have reported inconsistent results [13]; however, few have specifically evaluated the contribution of cigarette smoking, even though smoking may modify the influence of these polymorphisms on endothelial function [14−17]. The primary aim of our investigation was to determine whether exposure to cigarette smoke significantly modified the association between presence of the T-786C or E298D polymorphisms in NOS3 and risk of incident CHD or ischemic stroke events in individuals enrolled in the biethnic Atherosclerosis Risk in Communities (ARIC) study.

Methods

Study population

Participants were selected from the ARIC study, a longitudinal, population-based cohort study of 15 792 men and women aged 45−64 years from four US communities (Forsyth County, North Carolina; Jackson, Mississippi; Minneapolis, Minnesota; and Washington County, Maryland) enrolled between 1987 and 1989 [18]. Since enrollment, participants have been followed prospectively via annual phone interviews, clinic examinations approximately every 3 years through 1998, and ongoing abstraction of hospital and death certificate records. The study protocol was approved by the Institutional Review Board of each center, and consent was obtained from each participant.

Ascertainment of incident coronary heart disease and stroke cases

All incident cases that occurred between baseline and December 31, 1998 were evaluated (median follow-up 9.1 years), excluding participants with a history of CHD or stroke at baseline. Incident CHD (n = 1085) was defined as (1) definite or probable myocardial infarction (n = 520), (2) electrocardiographic evidence of silent myocardial infarction (n = 112), (3) definite CHD death (n = 110), or (4) coronary revascularization procedure (n = 343). Incident stroke was defined as a definite or probable ischemic stroke (n = 300).

The ascertainment of cases and criteria for classification have been previously described [19,20]. All potential events were systematically reviewed and adjudicated by the ARIC Morbidity and Mortality Classification Committee [18−20]. Hospitalized myocardial infarction was classified as definite or probable on the basis of chest pain symptoms, cardiac enzyme levels, and electrocardiographic changes. Definite CHD death was classified on the basis of chest pain symptoms, underlying cause of death, hospitalization records, and medical history. Coronary revascularization procedures included coronary artery bypass grafting and percutaneous coronary intervention. Using the National Survey of Stroke criteria [21], a stroke event was classified as definite or probable if there was (1) rapid onset of neurological symptoms that lasted > 24 h or led to death within 24 h, (2) no evidence of pathology that could have mimicked stroke, and (3) one major (e.g. aphasia or hemiparesis) or two minor (e.g. diplopia or dysarthria) neurological deficits. Qualifying cases were further classified as ischemic (thrombotic brain infarction, cardioembolic stroke) or hemorrhagic (subarachnoid hemorrhage, intracerebral hemorrhage) on the basis of neuroimaging studies and autopsy, when available [20].

Baseline measurements

Detailed demographic, clinical, and biochemical data were obtained from each participant at baseline. Race was self-reported. Prevalent CHD and stroke were defined as history of a physician-diagnosed event. Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, or current antihypertensive medication use. Diabetes was defined as fasting blood glucose ≥ 7.0 mmol/l, nonfasting blood glucose ≥ 11.1 mmol/l, physician diagnosis, or pharmacologic treatment. Detailed information on cigarette smoking was obtained through an interview-administered questionnaire, and was defined by current smoking status (yes/no) and pack-year smoking history ( ≥ 20 or < 20 pack-years).

Cohort random sample

A random sample of all ARIC participants without the history of CHD or stroke at baseline was assembled to serve as the reference group for the case–cohort comparisons [n = 1065; 85 (57 Caucasians, 28 African-Americans) and 29 (12 Caucasians, 17 African-Americans) of which are also incident CHD and stroke cases, respectively]. Sampling of the cohort was stratified on age ( ≥ 55 or < 55 years), sex, and race (Caucasian or African-American). Sampling proportions varied across each stratum.

Genotyping

Genomic DNA from all incident CHD and ischemic stroke cases and the cohort random sample was genotyped for the T > C polymorphism at nucleotide position −786 in the proximal promoter (T-786C) and the G > T polymorphism at nucleotide position 5246 in exon 7 (E298D) of NOS3, using multiplex matrix-assisted laser desorption/ionization time-of-flight mass spectrometry methods, as described [22]. Briefly, targeted sequences were amplified via polymerase chain reaction using specific forward (T-786C: 5′-ACGTTGGATGACTGTAGTTTCCCTAGTCCC-3′; E298D: 5′-ACGTTGGATGACCTCAAGGACCAGCTCGG-3′) and reverse (T-786C: 5′-ACGTTGGATGAGGTCAGCAGAGAGACTAGG-3′; E298D: 5′-ACGTTGGATGAAACGGTCG CTTCGACGTGC-3′) primers for each polymorphism, followed by extension reactions utilizing specific oligonucleotide primers that annealed immediately upstream of each polymorphic site and extended the amplification product by a single base pair in the forward direction (T-786C: 5′-CATCAAGCTCTTCCCTGGC-3′; E298D: 5′-GCTGCAGGCCCCAGATGA-3′). The mass of the extension reaction products was determined and translated into genotype using the MassARRAY RT software (Sequenom Inc., San Diego, California, USA). Blind replicates were included for quality control. Missing genotypes were present in < 5% of individuals.

Data analysis

Inverse sampling fractions from each stratum were used as weights in variance estimation of adjusted covariate means and proportions by linear and logistic regression, respectively. Cohort random sample allele frequencies were evaluated for deviation from Hardy–Weinberg equilibrium, and pairwise linkage disequilibrium statistics were calculated (Haploview 3.2; Broad Institute of Harvard and MIT Cambridge, Massachusetts, USA) [23]. Hazard rate ratios (HRRs) and 95% confidence intervals (CIs) for the development of incident CHD or ischemic stroke in relation to genotype were calculated by weighted proportional hazards regression, using Barlow’s method to account for the stratified random sampling and case–cohort design [24]. Model 1 included age, sex, and study center as covariates. Model 2 also included significant clinical covariates, which were current smoking status, diabetes, hypertension, high-density lipoprotein cholesterol, total cholesterol, and body mass index for the CHD end point, and current smoking status, diabetes, and hypertension for the stroke end point. Assuming an autosomal dominant mode of inheritance, individuals with one or two variant alleles were combined for comparison with wild-type individuals. All analyses were completed separately in Caucasians and African-Americans. Assuming a case–control design and type I error α = 0.05, we had approximately 99 and 90% power for the CHD end point and 90 and 80% power for the stroke end point to detect an odds ratio of 2.0 in Caucasians and African-Americans, respectively.

Haplotypes and their frequencies were estimated using the phase reconstruction method (PHASE 2.1), which assigned the most probable haplotype pair to each individual [25], and compared across case status by χ². Associations between haplotype (0 or 1) and risk of CHD or stroke were evaluated using Barlow’s method [24], by simultaneously modeling each haplotype relative to the most common haplotype using the model 2 covariate adjustment strategy.

Gene–environment interaction testing was completed on a multiplicative scale between genotype and current smoking status (yes/no) or pack-year smoking history (≥ 20 or < 20 pack-years) at baseline using a Wald χ² test for significance of the estimated β-coefficient for the interaction term [26]. Haplotype–smoking interaction testing was also completed in Caucasians for the reconstructed haplotype containing both the variant C-786 and D298 alleles; however, this was not completed in African-Americans because of sample size limitations. As interaction hypothesis testing on a multiplicative scale is underpowered, the critical value for statistical significance was set to α = 0.15, two-sided [27]. To minimize the impact of the multiple statistical tests conducted in our analysis, we estimated the false discovery rate (FDR) q-value for each interaction test separately for the CHD and stroke end points (QVALUE). The FDR q-value is defined as the proportion of statistical tests deemed significant that are actually false-positives [28]. Stratified weighted proportional hazards regression was also completed according to baseline current smoking status or pack-year smoking history using the model 2 covariate adjustment strategy (minus current smoking status) to further explore potential interactions.

Results

Study population

Significant baseline differences in various risk factors were observed between CHD and stroke cases and noncases included in the cohort random sample (Table 1). Incident CHD and stroke cases were significantly older and more likely to be men, smokers, diabetic, hypertensive and have abnormal fasting lipid panels compared with noncases. Stroke cases were also more likely to be African-Americans. In the cohort random sample, 71.5% of current smokers at baseline also had a ≥ 20 pack-year history.

Table 1.

Baseline characteristics by incident case status

Characteristicsa	CRS	CHD casesb	Stroke casesc
N	1065	1085	300
Sex (%male)	41.4	67.4	53.0
Race (%African-American)	27.2	23.4	45.0
Age (years)	53.8 ± 0.10	55.8 ± 0.17	56.7 ± 0.32
Current smoker (%yes)	24.8	36.1	36.8
Pack-years ≥ 20 (%yes)	29.0	48.6	41.4
Diabetes (%)	11.1	25.1	34.6
Hypertension (%)	29.8	49.4	66.0
SBP (mmHg)	120.4 ± 0.58	128.3 ± 0.63	134.8 ± 1.30
DBP (mmHg)	73.3 ± 0.37	76.2 ± 0.38	79.4 ± 0.81
BMI (kg/m2)	27.7 ± 0.20	28.3 ± 0.15	29.0 ± 0.31
HDL cholesterol (mmol/l)	1.38 ± 0.02	1.12 ± 0.01	1.25 ± 0.02
LDL cholesterol (mmol/l)	3.46 ± 0.03	3.92 ± 0.03	3.75 ± 0.07
Total cholesterol (mmol/l)	5.48 ± 0.03	5.86 ± 0.03	5.76 ± 0.08
Triglycerides (mmol/l)	1.41 ± 0.03	1.87 ± 0.04	1.73 ± 0.07

CRS, cohort random sample; Pack-years, (cigarettes smoked per day/20) × (years smoked); SBP, systolic blood pressure; DBP, diastolic blood pressure; BMI, body mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; CHD, coronary heart disease.

^aCharacteristics weighted according to the sampling fraction. Data presented as mean ± standard error of the mean.

^bP < 0.05 for comparison of CHD cases to 980 non-cases in the CRS for all characteristics.

^cP < 0.05 for comparison of stroke cases to 1036 non-cases in the CRS for all characteristics.

NOS3 genotype

In the cohort random sample, the T-786C (0.371 versus 0.151) and E298D (0.330 versus 0.112) minor allele frequencies were significantly different in Caucasians and African-Americans, respectively (P < 0.001). Minimal evidence of pairwise linkage disequilibrium was observed between these polymorphisms in both Caucasians (D′ = 0.47, r² = 0.19) and African-Americans (D′ = 0.41, r² = 0.12). The allelic distribution of each polymorphism was in Hardy–Weinberg equilibrium (P > 0.05).

E298D polymorphism

The variant D298 allele was not significantly associated with CHD or ischemic stroke incidence in either Caucasians or African-Americans (Tables 2 and 3), nor were these relationships modified by sex. In Caucasians, the relationship between E298D genotype and CHD risk was, however, significantly modified by baseline current smoking status (interaction P = 0.013, q = 0.078) and tended to be modified by pack-year smoking history (interaction P = 0.126, q = 0.247), with the highest risk of CHD observed in smokers carrying the variant D298 allele (Table 4). In the stratified analysis, current smokers with the variant D298 allele were at significantly greater risk of CHD than smokers with two E298 alleles (HRR 1.84, 95% CI 1.10−3.09, P = 0.021); however, no association between the variant D298 allele and CHD risk was observed in nonsmokers (HRR 0.81, 95% CI 0.59−1.10, P = 0.168). Current smoking status (interaction P = 0.762) and pack-year history (interaction P = 0.641) did not modify the relationship between E298D genotype and CHD risk in African-Americans (Table 4). Moreover, neither cigarette smoking exposure estimate modified the relationship between E298D genotype and risk of ischemic stroke.

Table 2.

T-786C and E298D genotype frequency by incident coronary heart disease (CHD) or stroke case status

	Caucasian			African-American
Genotypea	Noncases	Incident cases	P-value	Noncases	Incident cases	P-value
CHD
E298D
G/G	279 (47.2%)	365 (46.7%)		242 (77.3%)	182 (80.2%)
G/T + T/T	336 (52.8%)	416 (53.3%)	P=0.862	70 (22.7%)	45 (19.8%)	P=0.437
T-786C
T/T	248 (42.3%)	276 (37.4%)		212 (68.4%)	155 (74.9%)
T/C + C/C	362 (57.7%)	463 (62.7%)	P=0.079	91 (31.6%)	52 (25.1%)	P=0.126
Stroke
E298D
G/G	296 (47.0%)	70 (46.7%)		255 (78.5%)	86 (74.1%)
G/T + T/T	364 (53.0%)	80 (53.3%)	P=0.939	68 (21.5%)	30 (25.9%)	P=0.362
T-786C
T/T	268 (42.5%)	56 (39.2%)		223 (69.7%)	66 (60.6%)
T/C + C/C	386 (57.5%)	87 (60.8%)	P=0.466	90 (30.3%)	43 (39.5%)	P=0.099

^aGenotype data presented as absolute (percent) genotype frequency, and weighted according to sampling fraction.

Table 3.

Hazard rate ratio between T-786C and E298D polymorphisms and risk of incident CHD or stroke

	Caucasian			African-American
Polymorphism	HRR	95% CI	P-value	HRR	95% CI	P-value
CHD
E298D	G/T + T/T versus G/G			G/T + T/T versus G/G
Model 1a	0.94	0.75−1.17	P=0.586	0.87	0.57−1.34	P=0.526
Model 2b	1.03	0.80−1.33	P=0.819	0.66	0.40−1.09	P=0.105
T-786C	T/C + C/C versus T/T			T/C + C/C versus T/T
Model 1a	1.13	0.90−1.42	P=0.210	0.80	0.52−1.20	P=0.279
Model 2b	1.24	0.96−1.62	P=0.106	0.83	0.53−1.31	P=0.421
Stroke
E298D	G/T + T/T versus G/G			G/T + T/T versus G/G
Model 1a	0.94	0.65−1.36	P=0.742	1.50	0.89−2.53	P=0.133
Model 2c	0.97	0.65−1.44	P=0.884	1.35	0.76−2.41	P=0.311
T-786C	T/C + C/C versus T/T			T/C + C/C versus T/T
Model 1a	1.06	0.72−1.55	P=0.783	1.70	1.05−2.75	P=0.032
Model 2c	1.09	0.73−1.64	P=0.665	1.63	0.93−2.88	P=0.090

CHD, coronary heart disease; HRR, hazard rate ratio; CI, confidence interval.

^aAdjusted for age, sex, and study center.

^bAdjusted for age, sex, study center, current smoker, diabetes, hypertension, HDL cholesterol, total cholesterol, and body mass index.

^cAdjusted for age, sex, study center, current smoker, diabetes, and hypertension.

Table 4.

E298D genotype by smoking interaction and risk of incident CHD and stroke^a,b

	Caucasian			African-American
Smoking exposure	E298D genotype		Interactionc	E298D genotype		Interactionc
CHDd	G/G n = 365/279	G/T + T/T n = 416/335		G/G n = 182/241	G/T + T/T n = 44/70
Current smoker
No	1 (referent)	0.83 (0.61−1.13)	2.10 (1.17−3.76)	1 (referent)	0.63 (0.34−1.16)	1.17 (0.43−3.18)
Yes	1.19 (0.77−1.83)	2.07 (1.39−3.07)	P=0.013	1.90 (1.17−3.10)	1.40 (0.62−3.18)	P=0.762
Pack-years
< 20 years	1 (referent)	0.84 (0.61−1.17)	1.50 (0.89−2.53)	1 (referent)	0.61 (0.35−1.06)	1.30 (0.43−3.90)
≥ 20 years	1.23 (0.83−1.83)	1.56 (1.08−2.24)	P=0.126	1.19 (0.73−1.93)	0.94 (0.35−2.50)	P=0.641
Strokee	G/G n = 70/296	G/T + T/T n = 80/363		G/G n = 85/254	G/T + T/T n = 30/68
Current smoker
No	1 (referent)	0.81 (0.51−1.30)	1.70 (0.71−4.06)	1 (referent)	1.26 (0.57−2.77)	1.24 (0.37−4.19)
Yes	1.58 (0.83−3.00)	2.18 (1.19−3.99)	P=0.234	2.00 (1.08−3.69)	3.12 (1.22−7.99)	P=0.726
Pack-years
< 20 years	1 (referent)	0.80 (0.46−1.38)	1.48 (0.66−3.31)	1 (referent)	1.36 (0.67−2.74)	1.00 (0.25−4.01)
≥ 20 years	1.49 (0.81−2.75)	1.77 (0.98−3.21)	P=0.337	1.39 (0.70−2.75)	1.88 (0.53−6.62)	P=0.999

CHD, coronary heart disease; HRR, hazard rate ratio; CI, confidence interval.

^aNumber of cases/noncases in each genotype cell.

^bData presented as HRR (95% CI).

^cMultiplicative E298D by smoking interaction term HRR (95% CI), and P-value.

^dAdjusted for age, sex, study center, diabetes, hypertension, HDL cholesterol, total cholesterol, and body mass index.

^eAdjusted for age, sex, study center, diabetes, and hypertension.

T-786C polymorphism

In Caucasians, the variant C-786 allele tended to be more common in CHD cases compared with noncases (62.7 versus 57.7%, P = 0.079) (Table 2), and was associated with a trend toward greater CHD risk relative to T-786 homozygotes (HRR 1.24, 95% CI 0.96−1.62, P = 0.106) (Table 3); although, this did not attain statistical significance. Pack-year smoking history appeared to modify this association, such that in the stratified analysis ≥ 20 pack-year smokers carrying the variant C-786 allele were at significantly greater risk of CHD relative to ≥ 20 pack-year smokers with two T-786 alleles (HRR 1.52, 95% CI 1.01−2.27, P = 0.042). No association was observed between the variant C-786 allele and CHD risk in < 20 pack-year smokers (HRR 1.06, 95% CI 0.75− 1.51, P = 0.731). In African-Americans, no association between the variant C-786 allele and risk of CHD was observed (P = 0.421) (Table 3); however, this relationship tended to be modified by current smoking status (interaction P = 0.142, q = 0.444) (Table 5).

Table 5.

T-786C genotype by smoking interaction and risk of incident CHD and stroke^a,^b

	Caucasian			African-American
Smoking exposure	T-786C genotype		Interactionc	T-786C genotype		Interactionc
CHDd	T/T n = 276/247	T/C + C/C n = 463/362		T/T n = 154/212	T/C + C/C n = 52/90
Current smoker
No	1 (referent)	1.33 (0.97−1.81)	0.83 (0.46−1.49)	1 (referent)	0.60 (0.34−1.07)	2.04 (0.79−5.31)
Yes	1.92 (1.21−3.05)	2.11 (1.41−3.17)	P=0.527	1.68 (1.00−2.83)	2.08 (0.95−4.54)	P=0.142
Pack-years
< 20 years	1 (referent)	1.08 (0.77−1.50)	1.44 (0.85−2.44)	1 (referent)	0.66 (0.39−1.11)	1.88 (0.68−5.15)
≥ 20 years	1.22 (0.79−1.88)	1.89 (1.32−2.70)	P=0.179	1.07 (0.63−1.82)	1.32 (0.58−3.00)	P=0.222
Strokee	T/T n = 56/267	T/C + C/C n = 87/386		T/T n = 65/223	T/C + C/C n = 43/89
Current smoker
No	1 (referent)	1.09 (0.66−1.79)	0.94 (0.39−2.28)	1 (referent)	1.34 (0.66−2.72)	1.49 (0.45−4.96)
Yes	2.17 (1.06−4.43)	2.22 (1.20−4.11)	P=0.892	1.86 (0.94−3.68)	3.72 (1.39−9.99)	P=0.515
Pack-years
< 20 years	1 (referent)	0.94 (0.54−1.65)	1.33 (0.59−3.02)	1 (referent)	1.06 (0.55−2.06)	3.73 (1.08−12.9)
≥ 20 years	1.56 (0.80−3.04)	1.97 (1.09−3.57)	P=0.488	1.02 (0.49−2.13)	4.03 (1.54−10.6)	P=0.037

CHD, coronary heart disease; HRR, hazard rate ratio; CI, confidence interval.

^aNumber of cases/noncases in each genotype cell.

^bData presented as HRR (95% CI).

^cMultiplicative T-786C by smoking interaction term HRR (95% CI), and P-value.

^dAdjusted for age, sex, study center, diabetes, hypertension, HDL cholesterol, total cholesterol, and body mass index.

^eAdjusted for age, sex, study center, diabetes, and hypertension.

In African-Americans, the variant C-786 allele tended to be more common in incident stroke cases compared with noncases (39.5 versus 30.3%, P = 0.099) (Table 2), and was associated with a trend toward greater risk of stroke relative to T-786 homozygotes (HRR 1.63, 95% CI 0.93− 2.88, P = 0.090) (Table 3); although this did not attain statistical significance. This relationship was not modified by sex. Cigarette smoking exposure, however, significantly modified this association (interaction P = 0.037, q = 0.148), such that the greatest risk of ischemic stroke was observed in ≥ 20 pack-year smokers carrying the variant C-786 allele (Table 5). In ≥ 20 pack-year smokers, stratified analysis demonstrated that the C-786 allele was more common in stroke cases compared with noncases (41.9 versus 23.4%, P = 0.083), and was associated with significantly higher risk of stroke relative to ≥ 20 pack-year smokers with two T-786 alleles (HRR 4.23, 95% CI 1.34−13.4, P = 0.014). No relationship between the variant C-786 allele and risk of stroke was observed in < 20 pack-year smokers (HRR 1.08, 95% CI 0.56−2.10, P = 0.813). A similar association was also observed in current smokers carrying the variant C-786 allele (HRR 2.65, 95% CI 0.99−7.1, P = 0.052), but not nonsmokers (HRR 1.16, 95% CI 0.55−2.46, P = 0.694). In Caucasians, the variant C-786 allele was not significantly associated with the risk of ischemic stroke (P = 0.665) (Table 3), nor was this relationship modified by current smoking status (interaction P = 0.892) or pack-year history (interaction P = 0.488) (Table 5).

NOS3 haplotypes

Four potential haplotypes were reconstructed, and in Caucasians only haplotype CG (uniquely tagged by the variant C-786 allele) was significantly more frequent in CHD cases compared with noncases (16.1 versus 13.1%, P = 0.034) (Table 6) and was associated with a trend toward greater CHD risk relative to the most common haplotype TG (HRR 1.34, 95% CI 0.97−1.85, P = 0.072); although this did not attain statistical significance. No significant differences in haplotype copy frequency were observed across CHD case status in African-Americans, or across stroke case status in Caucasians and African-Americans (Table 6).

Table 6.

NOS3 haplotype frequency by incident coronary heart disease (CHD) or stroke case status

	CHD			Stroke
Haplotypea	Noncaseb	Incident caseb	P-valuec	Noncaseb	Incident caseb	P-valuec
Caucasiand	n = 606	n = 735		n = 650	n = 142
TG	54.4%	53.0%	P=0.494	54.2%	53.9%	P=0.935
CT	23.0%	23.0%	P=0.987	23.0%	22.5%	P=0.870
CG	13.1%	16.1%	P=0.034	13.2%	14.8%	P=0.496
TT	9.6%	8.0%	P=0.160	9.7%	8.8%	P=0.648
African-Americand	n = 303	n = 207		n = 313	n = 109
TG	78.8%	79.2%	P=0.876	79.3%	72.9%	P=0.069
CT	6.5%	3.9%	P=0.076	6.0%	7.3%	P=0.519
CG	9.9%	10.1%	P=0.891	9.8%	13.8%	P=0.145
TT	4.9%	6.8%	P=0.230	4.8%	6.0%	P=0.536

^aHaplotype frequencies weighted according to cohort random sample strata. Tagging polymorphisms are in bold.

^bEstimated copy frequency of each haplotype.

^cχ² P-value for the distribution of each haplotype by CHD or stroke case status.

^dHaplotype includes the T-786C and E298D polymorphisms, respectively.

Haplotype CT (uniquely tagged by both the variant C-786 and D298 alleles) was not significantly associated with risk of CHD relative to haplotype TG (HRR 1.16, 95% CI 0.88−1.53, P = 0.296). Although no statistically significant interaction between haplotype CT and either current smoking status (interaction P = 0.179) or pack-year history (interaction P = 0.206) was observed, smokers carrying at least one copy of haplotype CT appeared to demonstrate higher risk of CHD than nonsmokers without this haplotype (Table 7). In the stratified analysis, ≥ 20 pack-year smokers carrying haplotype CT were at significantly higher risk of developing CHD relative to ≥ 20 pack-year smokers with two haplotype TG copies (HRR 1.55, 95% CI 1.02−2.36, P = 0.041). No association between haplotype CT and CHD risk was, however, observed in < 20 pack-year smokers (HRR 0.91, 95% CI 0.62−1.33, P = 0.469). A similar but nonsignificant association was also observed in current smokers carrying haplotype CT (HRR 1.73, 95% CI 0.99−3.02, P = 0.054), but not nonsmokers (HRR 1.01, 95% CI 0.72−1.42, P = 0.937).

Table 7.

NOS3 haplotype by smoking interaction and risk of incident CHD and stroke in Caucasians^a,^b

Smoking exposure	Haplotype		Interactionc
CHDd	Non-CT n = 430/350	CT n = 305/256
Current smoker
No	1 (referent)	1.00 (0.74−1.35)	1.50 (0.83−2.70)
Yes	1.46 (1.00−2.13)	2.18 (1.40−3.38)	P=0.179
Pack-years
< 20 years	1 (referent)	0.95 (0.68−1.33)	1.39 (0.84−2.30)
≥ 20 years	1.32 (0.93−1.87)	1.74 (1.21−2.51)	P=0.206
Strokee	Non-CT n = 84/374	CT n = 58/276
Current smoker
No	1 (referent)	0.87 (0.53−1.45)	1.59 (0.65−3.86)
Yes	1.67 (0.92−3.04)	2.32 (1.18−4.51)	P=0.309
Pack-years
< 20 years	1 (referent)	0.92 (0.52−1.63)	1.24 (0.54−2.80)
≥ 20 years	1.70 (0.98−2.94)	1.93 (0.99−3.74)	P=0.610

CHD, coronary heart disease; HRR, hazard rate ratio; CI, confidence interval.

^aNumber of cases/noncases in each haplotype cell.

^bData presented as HRR (95% CI).

^cMultiplicative haplotype by smoking interaction term HRR (95% CI), and P-value.

^dAdjusted for age, sex, study center, diabetes, hypertension, HDL cholesterol, total cholesterol, and body mass index.

^eAdjusted for age, sex, study center, diabetes, and hypertension.

Haplotype CT was not significantly associated with risk of ischemic stroke relative to haplotype TG (HRR 1.03, 95% CI 0.67−1.58, P = 0.642). Moreover, no significant interaction between haplotype CT and either current cigarette smoking status (interaction P = 0.309) or pack-year history (interaction P = 0.610) was observed (Table 7). In the stratified analysis, statistically significant associations were not observed between haplotype CT and risk of ischemic stroke in either current smokers (HRR 2.09, 95% CI 0.76−5.77, P = 0.156) or nonsmokers (HRR 0.85, 95% CI 0.52−1.41, P = 0.536) relative to haplotype TG.

Discussion

Our analysis demonstrated no statistically significant associations between the variant C-786 or D298 alleles and risk of either incident CHD or ischemic stroke in Caucasians and African-Americans enrolled in the ARIC study; although a trend towards higher risk was observed with the variant C-786 allele for CHD in Caucasians and stroke in African-Americans. Importantly, our analysis identified a potential interaction between genetic variation in NOS3, cigarette smoking history, and risk of incident CHD and ischemic stroke events in these individuals. Specifically, the variant C-786 and D298 alleles appeared to be associated with significantly higher risk of developing CHD or ischemic stroke in current cigarette smokers and/or those with a ≥ 20 pack-year history at baseline, whereas these variant alleles were not significantly associated with disease risk in the absence of the smoking exposure. This interaction was particularly evident with E298D genotype and CHD risk in Caucasians, and T-786C genotype and ischemic stroke risk in African-Americans.

The role of endothelial dysfunction in the pathogenesis of cardiovascular disease has become increasingly appreciated. Endothelial dysfunction is typically manifested by impairment in endothelial-dependent vasodilation, and is associated with increased risk of myocardial infarction and stroke [2,3]. The vasodilatory, anti-inflammatory, and anti-proliferative properties of nitric oxide are vital mediators of this process, such that impairment in nitric oxide biosynthesis and/or activity elicits a substantial worsening of endothelial function and promotes the development of atherosclerosis [4−7]. Cigarette smoking also contributes significantly to endothelial dysfunction and atherosclerosis [8], in part through its reduction of eNOS-mediated nitric oxide biosynthesis via enhanced oxidative stress and eNOS uncoupling [9,10].

Numerous investigations have evaluated associations between NOS3 polymorphisms and cardiovascular disease. Most have focused on the T-786C and E298D polymorphisms owing to associations between the variant alleles and reduced nitric oxide biosynthesis [11,12]; however, results have been inconsistent [13]. Importantly, few investigations have evaluated incident clinical events. Moreover, few have considered the potential influence of cigarette smoking despite their potential synergistic contribution to worsening endothelial function. For instance, the variant C-786 allele has been significantly associated with risk of coronary spasm [15] and decreased cerebral blood flow [16] in smokers, but not nonsmokers. The variant D298 allele has also been associated with significantly lower endothelial-dependent vasodilation in smokers, with no significant effect in nonsmokers [17]. Interestingly, the D298 allele did not influence endothelial-dependent vasodilation in healthy volunteers despite being associated with reduced systemic nitric oxide levels [29]. These data suggest that presence of established underlying endothelial dysfunction, as observed in cigarette smokers, may be necessary for these polymorphisms to significantly attenuate endothelial function and predispose patients to increased cardiovascular risk. Our findings demonstrating a potential interaction between the T-786C and E298D polymorphisms, cigarette smoking, and risk of CHD and stroke events are consistent with this hypothesis.

We observed that associations between the variant D298 and C-786 alleles and CHD incidence may be significantly modified by exposure to cigarette smoke, such that smokers carrying these variant alleles were at significantly higher risk of developing CHD compared with nonsmokers without these variants. Smokers carrying the reconstructed haplotype containing both variant alleles also appeared to be at greater risk of CHD. Moreover, pack-year history significantly modified the association between the C-786 allele and ischemic stroke risk in African-Americans. Collectively, we believe that these data suggest potential existence of a complex interplay between NOS3 polymorphisms, cigarette smoking exposure, endothelial function, and cardiovascular disease risk.

Although our study evaluated rigorously ascertained incident events, we are unable to elucidate mechanisms underlying the observed statistical interactions. We also cannot rule out that the T-786C and E298D polymorphisms are simply markers in linkage disequilibrium with the true causative loci. We confirmed that these polymorphisms are, however, not in significant linkage disequilibrium with each other [13]. A previous epidemiological study reported a significant interaction between the NOS3 27-base pair repeat polymorphism in intron 4, smoking history, and risk of angiographically documented CHD [30]. Although we did not evaluate this polymorphism in our population, previous studies have demonstrated that it is not in significant linkage disequilibrium with either the E298D or T-786C polymorphisms [13]. Our analysis also identified a potential haplotype–smoking interaction. As haplotypes are inferred, some uncertainty exists in the haplotype assigned to each individual. Only two polymorphisms were, however, included in haplotype reconstruction, posterior haplotype probabilities exceeded 0.83 in all individuals (approximately 75% had a probability of 1.0), and similar haplotype distributions were observed using the expectation-maximization algorithm [23]. Unfortunately, validation of reported smoking status via biomarkers of tobacco exposure, such as urinary cotinine, was not completed in the ARIC study; however, misclassification of smoking status was expected to be minor and more likely to involve a smoker who failed to report cigarette consumption in small amounts. Moreover, previous ARIC analyses evaluating smoking and risk of cardiovascular disease outcomes have been consistent with other populations [31]. Thus, it is unlikely that misclassification bias, if present, would have a detectable impact on our estimates.

We acknowledge it is difficult to gauge the statistical significance of these findings considering the number of comparisons completed in our analysis, and recognize the undesirable consequences of reporting false-positive findings. Moreover, some of our reported HRRs were imprecise because of small sample sizes and limited power, particularly in our African-American and stroke subsets. Even though statistical significance was, however, not demonstrated with every comparison, we believe our evaluation of two functionally relevant polymorphisms (T-786C and E298D), two measures of cigarette smoking exposure (current smoking status and pack-year history), and two incident cardiovascular disease end points (CHD and ischemic stroke) in Caucasian and African-American individuals enrolled in the rigorously characterized ARIC study cohort provided a conservative mechanism to characterize this potential gene–environment interaction at the population level. Overall, we believe that the internal consistency of our results argue in favor of their validity, albeit at the cost of an increased number of statistical comparisons. In order to minimize the impact of the multiple tests completed in our analysis, we also assessed the FDR of our interaction test results. The estimated q-values for the interactions observed with E298D genotype and CHD risk in Caucasians (q = 0.078) and T-786C genotype and ischemic stroke risk in African-Americans (q = 0.148) enhance our confidence in the conclusions drawn from these findings. Moreover, previous studies have demonstrated that the T-786C and E298D polymorphisms are particularly associated with endothelial dysfunction in cigarette smokers [15−17], suggesting that existence of a causal interaction between genetic variation in NOS3 and smoking in the pathogenesis of CHD and stroke is biologically plausible. Our observation of significant gene–smoking interactions in the absence of significant genotype–disease associations independent of the interaction (i.e. main effects), however, suggests that these findings should be interpreted with caution. Future studies in different populations will undoubtedly be required to validate our findings and improve our understanding of the complex relationship between NOS3, cigarette smoking, and ischemic cardiovascular disease.