Identification of Genes Contributing to Cardiovascular Disease in Overweight and Obese Individuals from West Virginia

Abstract

Excess weight is a known risk factor for coronary artery disease (CAD) and a large percentage of overweight and obese individuals ultimately develop CAD. The objective of this study was to identify human genes associated with CAD in a subgroup of overweight and obese individuals using population-based association methods. Logistic regression analyses were used to test the association between single nucleotide polymorphisms (SNPs) in 34 candidate genes and the CAD phenotype with age, gender, and BMI as covariates. Two SNPs in the Apolipoprotein B (Apo B) gene [rs1042031 and rs1800479], one in the Cholesterol Ester Transfer Protein (CETP) gene [rs5880], and one in the Low Density Lipoprotein Receptor (LDLR) gene [rs2569538] met the 0.01 significance level for association with CAD. Based on these findings, we conclude that variants within the CETP and Apo B genes conferred susceptibility to CAD in overweight individuals and that a variant with the LDLR gene conferred susceptibility in an obese group.

Introduction

Coronary artery disease (CAD) is the leading cause of death and premature disability in the United States.^{1, 2,3} In its most recent update on heart disease, the American Heart Association reported that 16.8 million US adults are afflicted with CAD and that the estimated CAD patient care costs for 2009 are $165 billion.³ The primary cause of CAD is atherosclerosis, a progressive multifactorial disease characterized by the accumulation of lipids and fibrous elements in the large arteries.⁴ Evidence from animal studies, epidemiological studies and clinical correlations strongly suggest a role for dietary cholesterol as an environmental promoter of atherosclerosis.^2,3 A genetic etiology for atherosclerosis is indicated by familial aggregation of the trait and heritability estimates of contributing factors (e.g. LDL- and HDL-cholesterol levels) that range from 20% to 80%.^2,5

Several significant causal genes have been identified by studying Mendelian disorders that either cause or confer susceptibility to CAD.⁵ Genome-wide linkage scans have revealed CAD susceptibility loci on chromosomes 1p34-36, 2q, 3q13, 13q12-13, 14q32, 16p13, and 17.^{(6, 7 and reviewed in 8)} Several genome-wide association studies for CAD have been performed based on single nucleotide polymorphisms (SNPs).⁶ Although some identified regions were replicated (3q12-13, 9p21), most of the known findings of regions associated with CAD could not be replicated. A recent replication study on 85 previously identified variants in genetic risk factors for acute coronary syndrome found that only one of these variants (in the β-fibrinogen promoter) was validated as a risk factor.⁹ Additional candidate gene association studies in different subpopulations are needed in order to confirm the associations with these SNPs and identify new SNPs associated with CAD.

Overweight and obesity have become global epidemics in both children and adults¹⁰ and are known to contribute to CAD, heart failure, and sudden cardiac death.^4,5,11 Since a significant number of overweight individuals go on to develop CAD, we hypothesized that one or more genes could cause or confer susceptibility to the development of CAD in this group. We sought to identify genes associated with CAD in overweight and obese individuals by screening for SNPs in a set of cardiovascular disease candidate genes using case control association methods. We identified SNPs within three genes [Apo B (apolipoprotein B), LDLR (low density lipoprotein receptor) and CETP (cholesterol ester transfer protein)] that were associated with CAD phenotype. The Apo B and CETP SNPs have been identified in previous CAD association studies,^12,13 while the LDLR SNP is a novel association.

Methods

Human Subjects and CAD Phenotype

Human subjects were enrolled according to Institutional Review Board protocols approved by the following West Virginia institutions: Marshall University Department of Cardiovascular Services, Charleston Area Medical Center, Lincoln Primary Care Center, Tri-County Clinic, Tug River Clinic, and Valley Health Systems.

The CAD phenotype was predicated on known indicators of the disease: angina, myocardial infarction, cardiac artery stenosis, and acute coronary syndrome. Individuals were classified as affected with CAD if one or more of the following conditions were present: (1) at least one vessel with at least 50% stenosis as determined by cardiac catheterization, (2) acute myocardial infarction diagnosed either by elevation of creatine phosphokinase myocardial band or by diagnostic electrocardiogram, (3) coronary artery bypass graft, (4) percutaneous coronary intervention, (5) documented cerebral vascular disease, or (6) documented peripheral arterial disease. Individuals without any of these symptoms or interventions who did not report any history of CAD and who were at least 40 years of age were classified as controls.

Participants with Body Mass Index (BMI) ≥ 25 were placed in the overweight category while those with BMI ≥ 30 were considered obese. Of the 261 Caucasian participants, we identified 243 who were overweight. In this group of individuals, 104 had evidence of CAD and 92 were controls. Forty seven individuals did not meet our criteria for cases or controls and were excluded from analysis. Of these 47 exclusions, 33 clinically normal individuals were younger than age forty and 14 individuals had an indeterminate disease phenotype. Obese (BMI ≥ 30) individuals accounted for 141 of the 196 overweight individuals.

Candidate Genes and SNP Selection

Thirty four candidate genes for CAD were selected from three different sources of cardiovascular disease susceptibility genes: Online Mendelian Inheritance in Man (www.ncbi.nlm.nih.gov/omim), CardioGenomics (cardiogenomics. med.harvard.edu), and Comparative Genomic Resources for Cardiovascular Research (Berkeley Programs for Genomic Applications; pga.jgi-psf.org/) (Table 1).

Table 1.

Candidate genes for CAD with corresponding number of genotyped SNPs

Candidate Gene	Number of SNPs
ABCA1 (ATP-binding cassette)	6
AGT (angiotensinogen preprotein)	8
ALOX5A (arachidonate 5-lipoxygenase-activating protein) protein)	1
Apo B (Apolipoprotein B)	8
Apo C2 (Apolipoprotein C-II)	1
Apo E (Apolipoprotein E)	1
BHMT (betaine-homocysteine methyltransferase)	3
CETP (cholesteryl ester transfer protein)	14
F5 (coagulation factor V)	5
HDLBP ( high density lipoprotein binding protein)	3
HNF4A (hepatocyte nuclear factor 4, alpha)	2
LDLR (low density lipoprotein receptor)	7
LIPA (lipase A, lysosomal acid, cholesterol esterase)	11
LIPG (lipase, endothelial)	2
LRP1 (low density lipoprotein-related protein 1)	3
LRP1B (low density lipoprotein-related protein 1B)	7
LRP5 (low density lipoprotein receptor-related protein 5)	8
LPR8 (low density lipoprotein receptor-related protein 8)	7
MPO (myeloperoxidase)	1
MSR1 (macrophage scavenger receptor 1)	2
MTHFR (5,10-methylenetetrahydrofolate reductase)	5
MTR (5-methyltetrahydrofolate-homocysteine)	1
MTRR (5-methyltetrahydrofolate-homocysteine)	1
NOS3 (nitric oxide synthase 3)	3
PLTP (phospholipid transfer protein)	2
PPARA (peroxisome proliferative activated receptor)	6
SHMT1 (serine hydroxymethyltransferase 1)	2
SCARB1 (scavenger receptor class B, member 1)	5
SERPINB2 (serpin peptidase inhibitor, clade B)	1
TCN2 (transcobalamin II)	4
TNFSF4 (tumor necrosis factor superfamily, member 4)	2
USF1 (upstream stimulatory factor 1)	1
USF2 (upstream stimulatory factor 2)	1
VWF (von Willebrand factor)	2
Total: 34	136

Selection of candidate gene SNPs was performed in a two phase approach. In the first phase, we selected a minimal set of tagging SNPs within each candidate gene using the HapMap Genome Browser and Haploview software.^14,15 Only SNPs with a minor allele frequency (MAF) ≥ 5% were selected for analysis. When MAFs were not available from the SNP databases, we calculated the frequency based on our study population. SNPs for which a TaqMan Allelic Discrimination assays were available were pre-selected prior to application of the aggressive tagging protocol. Haplotype blocks were identified by Haploview analysis of HapMap data. In the second phase of the SNP selection, additional SNPs were genotyped for genes in which statistically significant associations were identified in the first phase. These additional SNPs were genotyped in order to maximize the coverage of the gene and better define the area of interest. One hundred thirty six SNPs from thirty-four selected candidate genes were tested for an association with CAD as described below.

Genotyping Methods

Genomic DNA was purified from patient whole blood samples using Qiagen Blood and Cell Culture DNA Maxi Kits. SNPs were genotyped using either pyrosequencing or TaqMan Allelic Discrimination methods. In the pyrosequencing method, specific loci were amplified from 25 nanograms of genomic DNA by the polymerase chain reaction (PCR). Pyrosequencing reactions were carried out as described by Nyren et al.¹⁶ using a Biotage PSQ96MA Pyrosequencing System. Allelic Discrimination analyses were performed on either an ABI Model 7000 or Model 7500 Sequence Detection Systems as described by Livak et al.¹⁷ ABI Sequence Detection System Absolute and Allelic Discrimination software was used to determine the genotype.

Statistical Methods

A series of conditional logistic regression analyses with age, gender, and BMI as covariates were performed to identify SNPs associated with the CAD phenotype. Statistical analyses were performed using statistical analysis software (SAS) release 9.2. p-values less than or equal to 0.01 were considered to be statistically significant. A chi-square test was used to determine if SNP alleles were in Hardy-Weinberg Equilibrium (HWE). Three of the 136 genotyped SNPs (rs289714, rs4823613, and rs891512) were found to be not in HWE at significance level of 0.05, and were excluded from further statistical analyses. The remaining 133 SNPs passed the HWE significance test at 0.05.

Results

Demographic data of the individuals used in this study are given by age, gender, and BMI in Table 2. Out of 196 overweight individuals, 94 (48%) were males and 102 (52%) were females. The average age of the patients entering the study was 60.1 years with a standard deviation of 10.9 years. Both case and control groups were composed of non-Hispanic, Caucasian populations from West Virginia. Since there were significant differences in age, gender, and BMI between cases and controls, logistic regression analyses were performed with age, gender, and BMI as covariates.

Table 2.

Mean age (in years), BMI, and gender differences between CAD control subjects. Data are given as means ± standard deviation (SD). stands for not significant.

	CAD Subjects	Control Subjects	p-value
Total number of subjects (n)	104	92
Mean age (in years)	63.5 ± 10.7	56.3 ± 9.8	< 0.001
Mean BMI	33.4 ± 6.2	35.6 ± 7.5	< 0.05
Number of males/females	70/34	24/68	< 0.001
Mean age for males	63.4 ± 9.8	57.1 ± 9.7	< 0.01
Mean age for females	63.8 ± 12.7	55.9 ± 9.9	< 0.01
Mean BMI for males	32.3 ± 5.4	32.5 ± 4.6	NS
Mean BMI for females	35.7 ± 7.2	36.7 ± 8.1	NS

Using the allelic model, we identified two SNPs (rs1042031 and rs1800479) within the Apo B gene that were associated with CAD at significance level below 0.01 (Table 3). SNP rs1042031 is located within Apo B exon 29 and substitution of the reference allele G by the minor allele A results in a substitution of the wild type glutamine (GLU) with lysine (LYS) at amino acid 4181. For overweight individuals, each occurrence of A at rs1042031 conferred an increased likelihood of 2.44 fold for developing CAD compared to individuals with a G substitution. Apo B SNP rs1800479 is an intronic SNP about 2kb up stream of rs1042031 SNP. For overweight individuals, each occurrence of the minor allele, G, at rs1800479 conferred an increased likelihood of 2.33 fold for developing CAD compared to individuals with a C substitution. Six other SNPs analyzed within the Apo B gene region did not show significant association with CAD (Table 4).

Table 3.

Association with Atherosclerosis (Allelic Model). The at-risk allele is given in bold. CI stands for confidence interval, OR stands for odds ratio.

BMI ≥ 25
SNP	Gene	Allele Frequencies				p-value	OR [95% CI]
		Cases (n = 96)		Controls (n = 91)
rs1042031	Apo B	A	G	A	G	0.0054	A 2.44 [1.30, 4.57]
		0.30	0.70	0.13	0.87
		Cases (n = 100)		Controls (n = 90)
		C	G	C	G
rs1800479	Apo B	0.71	0.29	0.87	0.13	0.0079	G 2.33 [1.25, 4.34]
BMI ≥ 30
		Cases (n = 65)		Controls (n = 65)
rs2569538	LDLR	A	G	A	G	0.0049	A 4.29 [1.56, 11.8]
		0.18	0.82	0.06	0.94

Table 4.

Apo B (Chromosome 2), LDLR (Chromosome 19), and CETP (Chromosome 16) SNPs. Location of each SNP was identified through the HapMap project Database. Minor Allele Frequencies (MAFs) were either found through the HapMap project or the dbSNP web site. SNPs in bold were genotyped in the second phase of SNPs selection.

Apo B SNP	Position (bp)	MAF	p-value	SNP type	Location	Substitution
rs1367117	21,175,542	0.267	NS	nonsynonymous	exon 4	T98I
rs520354	21,171,254	0.475	NS		intron 6
rs679899	21,162,556	0.490	NS	nonsynonymous	exon 14	A618V
rs11676704	21,156,000	0.158	NS		intron 18
rs693	21,143,837	0.492	NS	synonymous	exon 26	T2515T
rs1800479	21,139,025	0.174	0.008		intron 27
rs1042031	21,137,395	0.208	0.005	nonsynonymous	exon 29	E4181K
rs1042034	21,136,923	0.192	NS	nonsynonymous	exon 29	S4338N
LDLR SNP	Position (bp)	MAF	p-value	SNP type	Location	Substitution
rs2228671	11,071,902	0.100	NS	synonymous	exon 2	C27C
rs1799898	11,088,544	0.150	NS	synonymous	exon 12	L575L
rs5927	11,094,931	0.275	NS	synonymous	exon 15	R744R
rs2738456	11,097,794	0.250	NS		intron 15
rs2738459	11,099,463	0.431	NS		intron 15
rs2569538	11,099,538	0.108	0.0049		intron 15
rs1433099	11,103,648	0.300	NS		3’UTR
CETP SNP	Position (bp)	MAF	p-value	SNP type	Location	Substitution
rs1800775	55,552,727	0.425	NS		promoter
rs9926440	55,560,154	0.250	NS		intron 2
rs1532625	55,562,792	0.380	NS		intron 7
rs12708974	55,563,041	0.133	NS		intron 7
rs289717	55,566,879	0.333	NS		intron 10
rs289718	55,567,423	0.370	NS		intron 10
rs289719	55,567,432	0.279	NS		intron 10
rs4784744	55,568,676	0.333	NS		intron 10
rs5880	55,572,582	0.050	0.008	nonsynonymous	exon 12	A373P
rs5882	55,573,583	0.305	NS	nonsynonymous	exon 14	V405I
rs9923854	55,574,493	0.125	NS		intron 14
rs289741	55,574,965	0.288	NS		intron 15
rs1801706	55,575,153	0.200	NS		3’UTR

SNP rs2569538 in the LDLR gene was significantly associated with CAD in a subgroup of obese individuals under both the allelic (p-value = 0.005, Table 3) and the genotypic model (p-value = 0.007, Table 5). Obese individuals who are A/G heterozygotes at rs2569538 were four times more likely to develop CAD than G/G homozygotes. This SNP is located inside intron 15 within the second haplotype block of LDLR gene. Six other LDLR SNPs were genotyped but did not show significant relationship with CAD (Table 4).

Table 5.

Association with Atherosclerosis (Genotypic Model). The at-risk genotype is given in bold. CI stands for confidence interval, OR stands for odds ratio.

BMI ≥ 25
SNP	Gene	Genotype Frequencies						p-value	OR [95% CI]
		Cases (n = 101)			Controls (n = 89)
rs5880	CETP	GG	CG	CC	GG	CG	CC	0.0080	GG:CG 5.56 [1.57, 19.6]
		.94	.05	.01	.84	.16	.00
BMI ≥ 30
rs2569538	LDLR	Cases (n = 65)			Controls (n = 65)			0.0072	AG:GG 4.14 [1.47, 11.7]
		AA	AG	GG	AA	AG	GG
		.03	.31	.66	.00	.12	.88

Using the genotypic model in the overweight group, we identified a SNP within the CETP gene on chromosome 16 (rs5880) that was associated with CAD at a significance level of 0.008 (Table 5). SNP rs5880 represents a nonsynonymous change which results in substitution of proline for the normally occurring alanine at position 373 in the processed protein (position 390 in the precursor protein). Overweight individuals who were G/G homozygotes at rs5880 were 5.56 times more likely to develop CAD than C/G heterozygotes. SNP rs5880 was also found to be associated (p-value = 0.006) with CAD in a sample of overweight individuals in which participants with type II diabetes were excluded (data not shown). None of twelve additional CETP SNPs were associated with the CAD phenotype in overweight and obese groups (Table 4).

Discussion

Two Apo B SNPs, rs1042031 and rs1800479, met our criterion for association with CAD. Based on their physical map location both SNPs are located in the first haplotype block of Apo B.^14,15,18 Two other SNPs (rs1042034 and rs693) within the first haplotype block of Apo B were not associated with CAD.

Apo B is the principal protein component of several low density lipoproteins (LDLs) and is the main apolipoprotein of chylomicrons. Defects in the Apo B gene are the cause of familial ligand-defective apolipoprotein b-100 (fdb), an autosomal dominant disorder that leads to hypercholesterolemia and increased susceptibility to coronary artery disease. Causal mutations map to exon 26 of Apo B and lead to substitution of glutamine for arginine 3500 or substitution of cysteine for arginine 3532.¹⁹ In contrast to the fdb genotype-phenotype correlation, a meta-analysis of 30 case control studies failed to find an association between CAD and the exon 26 XbaI restriction fragment length polymorphism (RFLP).¹² This study did report a significant association with an EcoR1 RFLP located in exon 29 which is identical to SNP rs1042031. Based on the location of the two associated Apo B SNPs found in this study and the meta-analysis, we hypothesize that a susceptibility locus for CAD is located near Apo B exon 29.

Binding of LDLs to the LDLR is mediated by interactions between the LDLR and the Apo B-100 lipoprotein. Over 1000 mutant alleles within the LDLR gene are known to cause familial hypercholesterolemia, and a class of these alleles is known to disrupt the binding of LDLR to Apo B.²⁰ We identified a single SNP (rs2569538) within intron 15 of the LDLR gene that was associated with CAD in a group of obese individuals under both the allelic and genotypic models (Tables 3 and 5). Although allele A was found to be a risk allele for susceptibility to CAD under the allelic (additive) model, the A/A genotype was not significantly different from the G/G genotype under the genotypic model. It is possible that small sample size may not have provided sufficient power for the analysis in the case of the genotypic model. Additional studies are needed to determine if SNP rs2569538 directly affects LDLR function or is linked to a variant that reduces LDLR function.

CETP catalyzes the transfer of cholesterol esters and triglycerides from high density lipoprotein (HDL) to ApoB containing particles like LDL and very low density lipoprotein. The role of CETP in CAD has been vigorously debated by a number of reviewers and its ultimate action may depend on genetic, metabolic and environmental factors.²¹ Several studies point to a model in which lower CETP levels are anti-atherogenic because HDL-C is more abundant and results in improved cholesterol efflux via ABCG1.²² For example, in a large meta-analysis, three CETP genotypes that showed moderate reductions in CETP activity were associated with reduced coronary risk.²³ However, the connection between increased CETP and CAD risk is weakened by other studies. In a prospective community-based sample,²⁴ lower plasma CETP activity was associated with greater CAD risk.

SNP rs5880 results in the substitution of proline for the normally-occurring alanine at position 373 (A373P) in the processed CETP protein.²⁵ In a linker scanning mutagenesis study, insertional substitutions of amino acid positions 373–379 severely impaired cholesterol ester transfer activity without dramatic impairment of secretion.²⁶ This study suggests that the A373P substitution would result in a CETP variant with reduced activity that could raise HDL-C levels and lower CAD risk. However, three studies^13,25,27 showed that the rs5880 minor allele (A373P) and/or a linked polymorphism form a CETP variant that reduces HDL-C levels and presumably result in higher CAD risk. We show that the normal allele (G) of rs5880 is associated with the disease phenotype. This association has not been reported in previous studies but is consistent with the pro-atherogenic effects of a fully functional CETP.

Although obesity is recognized as a risk factor for cardiovascular disease,²⁸ the underlying physiological or metabolic basis for the increased risk is unknown. Several authors have speculated that this may be due to abnormal lipid profiles, hypertension, type II diabetes or increased cardiovascular workload in obese individuals.^11,28 Although we did not measure serum lipid levels in our patient population, relative weight is positively correlated with both LDL cholesterol and triglyceride levels.^29,30 The observation that the deleterious effects of elevated CETP levels are manifested only in patients with elevated triglycerides^21,30 may explain the association of the CETP variant with CAD in our overweight group.

Conclusions

We have shown that variants within two CAD candidate genes (CETP and Apo B) play a role in CAD in overweight individuals. Apo B SNP rs1042031 or a linked variant confers susceptibility to CAD in both the overweight population in our study and several general populations.¹² CETP SNP rs5880 appears to confer susceptibility to CAD in overweight individuals perhaps because elevated triglycerides in this group are more rapidly converted to atherogenic LDL cholesterol. The LDLR SNP was only associated in our obese population and has not been identified in previous studies. These results suggest that genetic mutations in Apo B, CETP, and LDLR genes contribute to development of CAD in overweight individuals. Further investigations of these genes needed in order to identify the specific mutations and understand their diagnostic values. Although the diagnostic value of individual genetic markers identified in association studies of complex traits is generally very low, the identification of rare variants and use of combinations of genetic markers could potentially improve predictive value.

Author Information

Yulia Dementieva, PhD

Department of Mathematics, Emmanuel College, Boston, MA

Todd L. Green, PhD

Department of Pharmacology, Physiology and Toxicology, John C. Edwards School of Medicine, Marshall University, Huntington

Donald A. Primerano, PhD, Liping Wei, MS, James Denvir, PhD

Department of Biochemistry and Microbiology, John C. Edwards School of Medicine, Marshall University, Huntington

Paulette Wehner, MD

Department of Cardiovascular Services, Joan C. Edwards School of Medicine, Marshall University, Huntington

Sarah Dodson PhD, Mark R. Flood, PhD, Bonnie A. Pollock, Melinda Huff, Contessa Hill

Department of Biology, Chemistry and Geoscience, Fairmont State University, Fairmont

Robert Kreisberg, PhD, Amanda Francis, Katie Morrison

Department of Biology, West Liberty University, West Liberty

Holly Blackwood, RN

Charleston Area Medical Center Health Education and Research Institute, Charleston

Mary Davis, PhD

Department of Physiology and Pharmacology, School of Medicine, Robert C. Bird Health Sciences Center, West Virginia University, Morgantown

Huey Miin Lee, PhD

Department of Statistics, West Virginia University, Morgantown

Stafford Warren, MD

Anne Arundel Medical Center, Annapolis, MD

Glossary

Linkage analysis

Genetic linkage describes the tendency of gene loci to be inherited together. Linkage analysis relies on this tendency to determine the location of genes on chromosomes

Association studies

determine whether a specific allele is present at an increased frequency in individuals with a given disease compared to the frequency in individuals without the disease

Single Nucleotide Polymorphisms

a variation in a DNA sequence consisting of a single base