|Home | About | Journals | Submit | Contact Us | Français|
Elevated low-density lipoprotein cholesterol (LDL-C) levels are a major cardiovascular disease risk factor. Genetic factors are an important determinant of LDL-C levels.
To identify single nucleotide polymorphisms associated with LDL-C and subclinical coronary atherosclerosis, we performed a genome-wide association study of LDL-C in 841 asymptomatic Amish individuals aged 20 to 80 years, with replication in a second sample of 663 Amish individuals. We also performed scanning for coronary artery calcification (CAC) in 1018 of these individuals.
From the initial genome-wide association study, a cluster of single nucleotide polymorphisms in the region of the apolipoprotein B-100 gene (APOB) was strongly associated with LDL-C levels (P < 10−68). Additional genotyping revealed the presence of R3500Q, the mutation responsible for familial defective apolipoprotein B-100, which was also strongly associated with LDL-C in the replication sample (P < 10−36). The R3500Q carrier frequency, previously reported to be 0.1% to 0.4% in white European individuals, was 12% in the combined sample of 1504 Amish participants, consistent with a founder effect. The mutation was also strongly associated with CAC in both samples (P < 10−6 in both) and accounted for 26% and 7% of the variation in LDL-C levels and CAC, respectively. Compared with noncarriers, R3500Q carriers on average had LDL-C levels 58 mg/dL higher, a 4.41-fold higher odds (95% confidence interval, 2.69–7.21) of having detectable CAC, and a 9.28-fold higher odds (2.93–29.35) of having extensive CAC (CAC score ≥400).
The R3500Q mutation in APOB is a major determinant of LDL-C levels and CAC in the Amish.
Elevated low-density lipoprotein cholesterol (LDL-C) levels are associated with an increased risk of cardiovascular disease (CVD). Twin and family studies1–3 suggest that 40% to 80% of the population variation in levels of LDL-C is attributable to genetic factors. The variability in LDL-C levels in the general population is likely polygenic and affected by environmental factors. Rare forms of monogenic hypercholesterolemia have been identified, including familial hypercholesterolemia due to mutations in the LDL receptor gene and familial defective apolipoprotein B-100 due to mutations in the apolipoprotein B (OMIM 107730) (APOB).4,5
To identify genes associated with variations in LDL-C levels and their relation with sub clinical coronary atherosclerosis, we performed a genome-wide association study (GWAS) on LDL-C levels in a socially and culturally homogeneous Old Order Amish population residing in Lancaster County, Pennsylvania. Almost all the participants descended from fewer than 300 founders who emigrated from Switzerland during the 1700s. Their diet is relatively homogeneous, and their use of cholesterol-lowering medications is low, minimizing the effect of environmental factors that might obscure or modify genetic determinants of LDL-C levels.
This study began with a GWAS of LDL-C levels in an initial cohort of 841 individuals (stage 1). Stage 1 individuals were participants in the Heredity and Phenotype Intervention (HAPI) Heart Study,6 which was initiated in 2003 to identify genes that interact with short-term environmental exposures to modify risk factors for CVD. Participants were members of the Old Order Amish community aged 20 years or older and considered to be relatively healthy based on the exclusion criteria of severe hypertension (blood pressure > 180/105 mm Hg), malignancy, and kidney, liver, or untreated thyroid disease. Further description of the HAPI Heart Study design is available on request from the authors.
The HAPI Heart Study participants underwent measurement of CVD risk factors and questioning about their history of CVD. Physical examinations were conducted at the Amish Research Clinic in Strasburg, Pennsylvania. Blood samples were collected after an overnight fast. Those taking lipid-lowering medications at enrollment (7 participants) discontinued use 7 days before examination. Serum lipid and high-density lipoprotein cholesterol levels were assayed by Quest Diagnostics (Horsham, Pennsylvania).All the participants had triglyceride levels lower than 400 mg/dL (to convert to millimoles per liter, multiply by 0.0113), and LDL-C levels were calculated according to the formula of Friedewald et al.7 Lipoprotein sub fractions were measured using Vertical Auto Profile technology (VAP; Atherotech, Birmingham, Alabama). Hypertension was defined as a systolic blood pressure of 140 mm Hg or higher, a diastolic blood pressure of 90 mm Hg or higher, or use of prescription blood pressure–lowering medications. Diabetes mellitus was defined as a fasting glucose level of 126 mg/dL or higher (to convert to millimoles per liter, multiply by 0.0555) or current use of prescription medications for diabetes. Smoking habits were recorded; current smoking status included the use of cigarettes, pipes, or cigars. The measurements for coronary artery calcification (CAC) by electron beam computed tomography were available in 355 HAPI Heart Study participants.
In stage 2, we genotyped an independent sample of 663 Old Order Amish individuals aged 30 years or older in whom we had measured serum lipid levels and CAC. These individuals had previously been recruited into the Amish Family Calcification Study between 2002 and 2006 to identify the genetic determinants of CAC.8 Participants were recruited without regard to their CVD status. The study protocol was approved by the institutional review board at the University of Maryland and participating institutions. Informed consent was obtained from each of the study participants.
Genotyping was performed in all stage 1 study participants using the Affymetrix GeneChip Human Mapping 500K Array Set (Affymetrix, Santa Clara, California), including 500 568 single-nucleotide polymorphisms (SNPs). The Affymetrix GeneChip Genotyping Analysis Software and BRLMM genotype-calling algorithm were used to generate SNP data files. The mean genotype call rate in the stage 1 sample was 98.3%. The SNPs (n = 98 806) with minor allele frequencies less than 2% in the overall sample were removed from further analyses. Finally, 369 241 SNPs that passed quality control and Hardy-Weinberg equilibrium checks (at P < .001) were retained for analysis.
A region on chromosome arm 2p showed strong evidence of association, implicating the nearby gene, APOB, as an attractive positional candidate gene. To determine whether the observed association could be explained by the presence of the APOB R3500Q (OMIM 107730.0009) (rs5742904) mutation, which is responsible for familial defective APOB-100, we genotyped this variant in stage 1 and 2 participants using Custom TaqMan SNP Genotyping Assays (Applied Biosystems, Foster City, California).
Coronary artery calcification was measured by electron beam computed tomography using an Imatron C-150 scanner (Imatron Inc, San Francisco, California) (scanning protocol available on request from the authors). Coronary artery calcification was quantified using the Agatston method, which incorporates density and area.9 The sum of the Agatston scores in the 4 epicardial arteries was considered the CAC score.
Association analyses of LDL-C levels and other quantitative traits were performed under a variance component model that assesses the effect of genotype, as an additive effect, on the quantitative trait while simultaneously estimating the effects of age, age2, sex, and a polygenic component to account for phenotypic correlation due to relatedness. The polygenic component was modeled using the relationship matrix derived from the complete pedigree structure because all the participants are related. Association analysis using the complete pedigree structure was performed using mixed-model analysis for pedigree software developed by one of us (J.R.O.).We considered a genome-wide significance threshold to be P < 10−7.
Pairwise linkage disequilibrium correlation statistics (r2) were computed using Haploview software (http://www.broadinstitute.org/haploview/haploview). Power calculations using the Genetic Power Calculator program10 indicated we would have 80% power to detect SNPs in the stage 1 sample (n = 841) accounting for approximately 4% of phenotypic variation at α = 10−7. For example, a SNP accounting for 4% of the trait variance in LDL-C levels that had an allele frequency of 0.20 would be associated with a 13- to 14-mg/dL (to convert to millimoles per liter, multiply by 0.0259) increase in LDL-C per copy of the mutant allele.
Because the distribution of CAC scores was positively skewed, the scores were natural log-transformed after adding 1.11 Age- and sex-adjusted CAC scores were calculated using linear regression, and standardized residuals were obtained. The residuals, which were approximately normally distributed, were used for analysis of CAC quantity. Odds ratios describing the association between the R3500Q mutation and the presence of CAC and having extensive CAC (CAC score ≥400) were estimated using generalized estimating equations incorporated into SAS version 9.1 (SAS Institute, Inc, Cary, North Carolina), with sibship membership included as a random effect. All reported P values are 2-sided and are not adjusted for multiple comparisons.
The clinical characteristics of stage 1 and 2 participants are given in Table 1. The mean ages of the stage 1 and 2 samples were 43.7 and 59.6 years, respectively. The mean (SE) heritability of LDL-C, adjusted for age, age2, and sex, was 0.60 (0.07) (P = 1.2 × 10−20) in the stage 1 samples and 0.51 (0.08) (P = 2.8 × 10−14) in the stage 2 samples.
In the GWAS, the strongest SNP associations with LDL-C levels were observed for a cluster of 68 SNPs spanning a 12-megabases region on chromosome arm 2p, with P values ranging from P < 10−9 to a peak association of P < 10−68 for the SNP rs4971516 (300 kilobases [kb] downstream of the APOB gene) (additive model, minor allele frequency = 0.056). (Detailed GWAS results of LDL-C levels are available on request from the authors.)
One gene in the associated region, APOB, located 300 kb downstream of rs4971516, was considered the most compelling positional candidate gene for several reasons. First, the phenotypic distribution of LDL-C and the strength of the observed associations indicated the presence of a highly penetrant allele. Second, rare mutations in APOB have been found in patients with familial defectiveAPOB-100.5 Third, examination of haplotypes among presumed carriers suggested the presence of an extended region of haplotype sharing that included APOB compatible with a single mutation transmitted through the population from a common ancestor. Thus, we hypothesized that a mutation in the APOB gene may have entered the Amish population and increased in frequency through genetic drift. We then genotyped R3500Q (dbSNP designation: rs5742904) in the stage 1 sample.
R3500Q was in near-complete linkage disequilibrium with rs4971516 (D’ = 0.99 and r2 = 0.96) and was strongly associated with total cholesterol (P = 1.60 × 10−40) and LDL-C (P = 1.84 × 10−52) levels and with the presence (P < .001) and quantity (P = 6.67 × 10−7) of CAC comparing R3500Q heterozygote carriers with noncarriers (Table 2). The frequency of the R3500Q allele was 0.064; the 12% of the population that carried the mutant allele had LDL-C levels approximately 61 mg/dL higher than those of noncarriers. We performed an additional GWAS that included APOB R3500Q in the model as a covariate but did not identify any further SNPs significantly associated with LDL-C concentrations (data not shown).
In stage 2 participants, the frequency of the R3500Q allele was 0.06. R3500Q was strongly associated with LDL-C levels (P < 10−30) and accounted for an approximately 55-mg/dL increase in LDL-C levels. The mutation was associated with a higher prevalence (P < .001) and quantity (P = 9.2 × 10−7) of CAC in this independent replication sample. R3500Q genotypes were not associated with LDL-C subclass particle patterns (data not shown) or appreciably with high-density lipoprotein cholesterol levels (Table 2).
Because of the consistency of effect (61- and 55-mg/dL increases in LDL-C levels per allele in stages 1 and 2, respectively), stage 1 and 2 samples were combined to permit estimation of age- and sex-specific effects of the R3500Q mutation on LDL-C levels. R3500Q carriers had higher LDL-C levels in each age group and in both sexes (Figure 1). In the combined sample, the R3500Q mutation accounted for 26% of the variation in age- and sex-adjusted LDL-C levels.
The R3500Q mutation also accounted for 7% of the variation in adjusted CAC quantity in the combined sample. The prevalence of any detectable CAC and of extensive CAC (defined as a CAC score ≥400)11 by age in the combined sample is shown in Figure 2 according to carrier status. Before age 40 years, the prevalence of CAC was low in carriers and noncarriers. After age 50 years, R3500Q carriers had greater prevalence of both any CAC and extensive CAC in every group. The R3500Q mutation accounted for 4.5% of CAC presence and 12.8% of extensive CAC presence in this population. After adjusting for age, sex, lipid lowering medication use, and sibship, R3500Q carriers had a 4.41-fold higher odds (95% confidence interval [CI], 2.69–7.21) of having detectable CAC compared with non carriers. Moreover, carriers had an increased risk of having a CAC score of at least 400 compared with noncarriers (odds ratio [OR], 9.28; 95% CI, 2.93–29.35) (Table 3).
To determine whether the increase in CAC in APOB R3500Q carriers could be explained by increased LDL-C levels measured at a single time point, we assessed the effects of R3500Q on the degree of CAC after additional adjustment for LDL-C levels in the combined sample. After this adjustment, the R3500Q mutation explained 4.1% of the residual variance in the degree of CAC. Even with LDL-C in the model, R3500Q remained a strong independent predictor of CAC presence (OR, 3.18; 95% CI, 1.81–5.59) and of extensive CAC (5.43; 1.54–19.18) (Table 3).
Results of the association analyses were confirmed using the family-based association test (FBAT).12 These analyses revealed excess R3500Q transmission to be associated with higher LDL-C levels (P = 4.0 × 10−11) and a greater degree of CAC (P = 10−5).
The surprising finding from this study was the identification of the R3500Q mutation in APOB as a major determinant of LDL-C levels and CAC in the Old Order Amish. This mutation accounted for approximately 13% of extensive CAC in the population, a substantial effect given that CAC is strongly associated with the risk of future cardiovascular events. The present sample included 181 carriers of the R3500Q allele (ie, approximately 1 of every 8–9 participants), a much higher frequency for this mutation than in any other single population previously reported.13 Because the Old Order Amish in Lancaster County are descendents of approximately 300 founders who emigrated to Pennsylvania from Switzerland in the latter part of the 1700s,14 we speculate that the R3500Q mutation was introduced into the population by a single founding ancestor and has been maintained in the population at relatively high frequency through genetic drift.
The R3500Q mutation, which is responsible for familial defective apolipoprotein B-100,5 is believed to prevent the proper folding of apolipoprotein B by altering the interactions between 2 amino acids, thereby reducing the ability of the LDL-C particle to bind to the LDL receptor.15 The precise effects of this mutation on CVD risk have been hard to quantify because previous studies have either included few participants sampled from the general population or have focused on samples enriched with prevalent ischemic heart disease. In the largest published population sample, the APOB R3500Q mutation was associated with total cholesterol levels 100 mg/dL higher and LDL-C levels 82 mg/dL higher based on only 7 carriers identified through a population screen of 9255 Danes16 compared with the present estimated effect sizes of 55 mg/dL for total cholesterol and 58 mg/dL for LDL-C based on a sample of 181 carriers. Previous studies16–18 based on identification of individuals with hypercholesterolemia or ischemic heart disease have reported even larger effect sizes. R3500Q has been associated with ischemic heart disease, although a clear effect of this mutation on early coronary artery atherosclerosis has not been established. These findings reinforce the idea that the APOB R3500Q mutation may increase ischemic heart disease risk by increasing LDL-C levels and CAC. This view is supported by evidence that the APOB R3500Q mutation is a direct cause of elevated LDL-C levels, the known associations of APOB R3500Q and LDL-C levels with ischemic heart disease, and previous studies,8,19–21 including from this population,8 showing that LDL-C concentrations are associated with the degree and presence of CAC.
The association between the R3500Q mutation and CAC could not be accounted for entirely by a 1-time measure of LDL-C levels because the R3500Q mutation remained strongly associated with CAC even after adjusting for or stratifying by LDL-C. A possible explanation is that R3500Q carriers have been exposed to a lifelong increase in LDL-C levels, which cannot be captured by a single cross-sectional LDL-C measurement. Alternatively, the mutation may be associated with features related to LDL-C metabolism that affect CAC development independently of circulating LDL-C levels. However, through this analysis of LDL sub particle distributions, we found no evidence to suggest that the mutation was associated with more atherogenic LDL.
Controlling for R3500Q in the GWAS abolished almost all other associations with SNPs in the chromosome 2 region (data not shown). However, many SNPs in this region are in high-linkage disequilibrium, and carriers of the R3500Q mutation share a common extended haplotype. We, thus, cannot exclude the possibility that there might be other variants in linkage disequilibrium with R3500Q that are functional and contribute to elevated LDL-C levels and increased CAC levels. However, given what is known about the R3500Q mutation,15,22 there does not seem to be a strong rationale for positing the presence of additional functional variants in the region that co-segregate with the R3500Q mutation. Even if such variants did exist, it would be difficult to detect and attribute functional consequences to them because of the limited recombination along the at-risk haplotype.
The R3500Q mutation is relatively uncommon in non-Amish populations. Population-based surveys have reported carrier frequencies of 1 per 500 to 1 per 1250,23 and a carrier frequency of 1 per 209 has been reported in a Swiss population.13 Of 1840 individuals genotyped in the Baltimore-Washington Stroke Prevention in Young Women Study,24 we identified 2 heterozygotes, for a carrier rate of 0.11% (range, 0.02%–0.44%), in line with prior estimates of 0.08% to 0.41% reported from population-based studies.23 Importantly, even this low rate corresponds to a large number of carriers at the population level. For example, a carrier rate of 0.1% in the United States (2008 estimated population size of 305 million25) would translate into an estimated 305 000 R3500Q carriers. The markedly increased level of LDL-C and degree of CAC due to this single mutation have potentially important implications for personalized medicine in the full US population and in the Amish population, in whom this mutation is enriched. Because R3500Q carriers are more likely than noncarriers to have an increased burden of sub clinical coronary atherosclerosis even at the same LDL-C levels, such individuals may benefit from earlier and more aggressive treatment with lipid-lowering medications. Is there value in screening for R3500Q carriers at the population level? If so, how would this be most effectively accomplished, for example, by screening for the mutation or by screening for LDL-C levels? Prospective clinical trials are necessary to address these questions and to determine the most efficacious treatment strategies for carriers of this mutation, including at what age to start treatment.
In summary, we demonstrated that carriers of the R3500Q mutation in APOB are frequent in the Old Order Amish. In this population, the R3500Q mutation is a major determinant of LDL-C levels and CAC.
Funding/Support: This work was supported by research grants R01 HL69313, R01 088119, R01 AR046838, U01 HL72515, and U01 HL084756 from the National Institutes of Health and Scientist Development grant 0830146N (Dr Shen) and grant-in-aid 0855400E (Dr Mitchell) from the American Heart Association; grant M01 RR 16500 from the University of Maryland General Clinical Research Center; grant P30 DK072488 from the Mid-Atlantic Nutrition and Obesity Research Center; a grant from the General Clinical Research Centers Program, National Center for Research Resources, National Institutes of Health; and the Baltimore Veterans Administration Geriatric Research and Education Clinical Center. Dr Post was supported by the Paul Beeson Physician Faculty Scholars in Aging Program.
Additional Contributions: We acknowledge our Amish liaisons and field workers and the cooperation and support of the Amish community, without which these studies would have been impossible.
Author Contributions: Study concept and design: Post, Rumberger, Shuldiner, and Mitchell. Acquisition of data: Shen, Damcott, Horenstein, Post, Shelton, Shuldiner, and Mitchell. Analysis and interpretation of data: Shen, Damcott, Rampersaud, Pollin, McArdle, Peyser, Bielak, Post, Chang, Ryan, Miller, Rumberger, Sheedy, O’Connell, Shuldiner, and Mitchell. Drafting of the manuscript: Shen, Damcott, Miller, Sheedy, O’Connell, Shuldiner, and Mitchell. Critical revision of the manuscript for important intellectual content: Shen, Damcott, Rampersaud, Pollin, Horenstein, McArdle, Peyser, Bielak, Post, Chang, Ryan, Miller, Rumberger, Shelton, O’Connell, Shuldiner, and Mitchell. Statistical analysis: Shen, Pollin, McArdle, Ryan, Sheedy, O’Connell, and Mitchell. Obtained funding: Shen, Post, Shuldiner, and Mitchell. Administrative, technical, and material support: Damcott, Rumberger, Shelton, and Mitchell. Study supervision: Peyser and Mitchell.
Financial Disclosure: None reported.