Local Ancestry and Clinical Cardiovascular Events Among African Americans From the Atherosclerosis Risk in Communities Study.

BACKGROUND: Local ancestry in relation to clinical cardiovascular events (CVEs) among African Americans can provide insight into their genetic susceptibility to the disease. METHODS AND
RESULTS: We examined local European ancestry (LEA) association with CVEs among 3000 African Americans from the Atherosclerosis Risk in Communities Study (ARIC). We estimated LEA using Local Ancestry Inference in adMixed Populations using Linkage Disequilibrium (LAMP-LD) and examined its association with myocardial infarction, stroke, coronary heart disease and its composite and cardiovascular disease composite using logistic regression. Genome-wide significance was achieved by 121 LEA regions in relation to myocardial infarction and 2 in relation to the cardiovascular disease composite. The LEA region downstream of 4q32.1 was significantly associated with 2 times higher odds of myocardial infarction (P=1.45×10-6). The LEA region upstream of 6q11.1 was associated with 0.37 times lower odds of fatal coronary heart disease (P=7.34×10-4), whereas the LEA region downstream of 21q21.1 was associated with 1.55 times higher odds of composite coronary heart disease (P=3.45×10-4). Association of LEA with stroke was observed in the region upstream of 6p22.3 with a 1.57 times higher odds of stroke (P=9.69×10-4). Likewise, the LEA region on 4q32.3 was associated with a 1.53 times higher odds of composite cardiovascular disease (P=3.04×10-4). We also found 20 of the LEA regions at previously significant cardiovascular disease single-nucleotide polymorphisms to be associated with CVE in our study.
CONCLUSIONS: Future studies are needed to replicate and/or determine the causal variants driving our associations and explore clinical applications for those consistently associated with CVEs.

Introduction

Coronary heart disease (CHD) and stroke are the largest contributors to cardiovascular disease (CVD) deaths (46.2% and 16.1%, respectively),1 the leading cause of death in the United States for over three quarters of a century.2 Prevalence of CVD is highest among non‐Hispanic African Americans as compared to other race/ethnicities, and a high prevalence of CVD risk factors, such as hypertension and diabetes mellitus, contribute to higher disease rates in this racial/ethnic group.1 Given the high heritability of both CHD and stroke,3, 4 genetic factors have been explored to explain the variability in disease not accounted for by traditional CVD risk factors. Subsequently, several genome‐wide association studies (GWAS) were conducted that identified single‐nucleotide polymorphisms (SNPs) associated with myocardial infarction (MI) and CHD,5, 6 with a handful of studies also examining stroke and its subtypes.7, 8 Most of these studies were almost exclusively conducted in populations of European descent, with the focus shifting to the study of minority populations only recently.

The first large‐scale GWAS among African Americans from the Candidate gene Association Resource (CARe) Project explored CHD as well as CHD risk factors and combined association and ancestry results to obtain narrow and localized gene regions and detect African American specific susceptibility loci.9 GWAS or meta‐analyses focused on African Americans have reported few novel genetic associations, but have been successful in replicating single SNP associations identified from genetic studies of European populations.9, 10, 11 Some of the consistently replicated results include variants in or near the zinc finger homeobox 3 (ZHFX3), cyclin‐dependent kinases 2A and 2B (CDK2A/CDK2B), and histone deacetylase 9 (HDAC9) genes associated with CHD, stroke, or their combined phenotypes. Genetic association studies among African Americans were initially hampered by smaller sample sizes and limited coverage of SNPs on genotyping platforms that either lacked variations among African Americans or resulted in weak/undetectable associations.12 However, recent technical advances in genotyping and statistical models have alleviated these issues and identified other methods, such as admixture mapping, that help delineate the disparities in genetic susceptibility and risk of disease among admixed populations.13

Admixture mapping is used to detect genetic loci that are believed to be inherited from the ancestral population with a higher prevalence of the disease. This approach has so far been used to detect ancestry associations in relation to atherosclerosis using subclinical measures, such as coronary artery calcification,14 or CVD risk factors, such as hypertension,15 diabetes mellitus,16 lipids,17 or body mass index (BMI),18 among admixed populations. In particular, studies have not yet evaluated the association of local ancestry in relation to clinical cardiovascular events (CVEs), such as MI, stroke, CHD, and CVD. The objective of this study was to examine the association of local European ancestry (LEA) with clinical CVD events—MI, stroke, CHD, and CVD among African Americans from the Atherosclerosis Risk in Communities Study (ARIC) using admixture mapping.

Methods

Clinical and genotype data for the ARIC study were obtained from the National Center for Biotechnology Information's (NCBI) database for Genotypes and Phenotypes (dbGaP) through the accession number phs000090.v1.p1. The conduct of this study was approved by the University of Alabama at Birmingham (Birmingham, AL) Institutional Review Board.

Study Population

ARIC is a prospective cohort study designed to examine the natural history and progression of atherosclerosis to clinical cardiovascular disease as well as disparities attributed to race, sex, geography, and time.19 The study began in 1987 and had both cohort and community surveillance components. Participants, aged 45 to 64 years and representative of their respective communities, were recruited from 4 field centers (Forsyth County, NC; Minneapolis, MN; Washington County, MD; and Jackson, MS). Follow‐up included examinations every 3 years, annual phone interviews to document overall health and event status, and surveillance for hospitalized and nonhospitalized events and deaths. A greater proportion of African Americans were recruited from Jackson, Mississippi, and comprised 27% (n=4314) of the total number of cohort participants at baseline. In the current study, a total of 3000 African Americans without prevalent CVD at baseline visit, and with complete information on sociodemographic, lifestyle, clinical, and genomic data, were included.

Data Collection and Clinical Examination

The baseline examination in ARIC included a home interview as well as a clinical exam.19 The interview collected information on the participant's medical/family history, including hospitalizations in the past year as well as other lifestyle and socioeconomic factors. The clinic visit included procedures such as anthropometry and physical exam; sitting blood pressure; hematology and lipid profile; medical history and record review; echocardiography; and ultrasound. Classification of CVD‐related risk factors included: hypertension defined as systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or antihypertensive medication use; diabetes mellitus defined based on medication use, fasting plasma glucose of ≥126 mg/dL or casual blood glucose level of ≥200 mg/dL; and low‐density lipoprotein (LDL) and high‐density lipoprotein (HDL) cholesterol (mg/dL) measured and calculated from fasting blood samples.

Clinical Events

Data on incident events were obtained from the baseline visit through December 31, 2004 for inclusion in the current analysis. Information on events was collected through annual phone interviews, 3‐year examinations, and active surveillance.19 Events were initially determined using discharge diagnoses codes for CVD‐related hospitalizations. The hospital records were abstracted and reviewed if the initial check included CVD discharge codes, with the final diagnosis being assigned by a computer and accompanied by a physician review in uncertain or complicated cases. Events in the cohort component also included nonhospitalized diagnoses obtained through the participant's clinic records or questionnaires. CVD‐related deaths were also initially determined through discharge codes for in‐hospital deaths whereas information on out‐of‐hospital death was obtained from next of kin, physicians, coroners, or medical examiners. Death certificates and medical records were then reviewed to confirm cause of death and arrive at a final diagnosis, and further adjudication conducted in case of disagreements. All cardiovascular clinical events have been defined in the parent ARIC study.20, 21

Briefly, MI was classified as definite, probable, suspect, or no MI and identified based on electrocardiogram (ECG), cardiac biomarker levels, and symptoms of cardiac pain. Fatal CHD was defined as death resulting from atherosclerotic cardiac disease, death within 72 hours of the onset of cardiac pain, or history of CHD in cases where cause of death was unknown.22

Stroke was classified similar to MI and was based on sudden onset of focal neurological deficit that lasted >24 hours or until death. The diagnostic criteria included the presence of 1 or more neurological symptoms of hemiparesis or unilateral numbness of ≥2 body parts, homonymous hemianopia, aphasia, diplopia, dysarthria, dysphagia, dysphonia, or vertigo or gait disturbance; diagnostic scans indicating areas of infarct, embolus or bleed in the brain; or autopsy findings suggestive of infarct, embolus, or bleed in the brain. Focal neurological deficits attributed to nonvascular causes, such as brain trauma, tumor, infection, and complications attributed to metabolic or hematological disorders, were not considered as stroke.23

The CHD composite was defined as the presence of either MI, fatal CHD, silent MI (defined as ECG evidence of MI occurring in between visits and lacking clinical documentation), or revascularization procedures such as angioplasty, stenting, or bypass surgery. Furthermore, the CVD composite included a definitive diagnosis of MI, stroke, or death following MI or stroke.

Genotyping and Quality Control

Participants from the ARIC study who consented to genetic analysis and provided DNA samples were genotyped using the Affymetrix 6.0 SNP array (hg18 build) at the Broad Institute Center for Genotyping and Analysis, and the genotype calls were determined using the Birdseed algorithm.24 The 3000 African Americans included in the current study were genotyped for 909 622 SNPs. Quality control (QC) resulted in the removal of 65 637 SNPs because of genotype call rates <95% and 2165 SNPs because of plate associations (P<1×10−10), leaving 841 820 SNPs for further analysis. Further QC resulted in the removal of 390 SNPs because of missing >10% SNP data individually, 20 975 because of deviation from Hardy–Weinberg equilibrium (HWE) <0.001 (using exact test in PLINK), and 99 103 SNPs because of minor allele frequency (MAF) <5%. Furthermore, SNPs were also removed because they were not available in HapMap (87 397 SNPs), 22 495 had ambiguous A/T or G/C SNPs with MAF >35% and could not be resolved for strand annotation, and 1727 were found to be duplicates; this resulted in 579 847 autosomal SNPs remaining for inclusion in the admixture estimation and association analyses (Table S1).

Estimation of LEA

The Local Ancestry in adMixed Populations using Linkage Disequilibrium (LAMP‐LD) program was used to estimate LEA.25 The ancestry at each SNP for each individual was determined by using haplotype sets from ancestral populations. We used the HapMap phase II and III data from 234 Utah residents with North‐West European ancestry (CEU) and 230 Yorubans in Ibadan, Nigeria (YRI) with African ancestry to serve as our reference populations (HAPMAP phase 3, release 2). The reference phased haplotype data were used to determine parameters for the Hidden Markov Model (HMM) with 15 state spaces. The local ancestry was estimated using the HMM parameters within a 300‐SNP‐long, window‐based framework. Each chromosome was analyzed separately and the local ancestry was coded as the number of European ancestry alleles at each SNP (ie, 0, 1, or 2 European alleles). Global ancestry was calculated from local ancestry by averaging it across all chromosomes for each individual.

Association With Clinical CVE

Cases were defined as individuals who had a clinical event based on the criteria specified above. Controls included the rest of the participants who were still free of these events at the time when follow‐up ended. Ancestry association testing in relation to the clinical events was performed using additive logistic regression in PLINK (v1.9) with default settings to compare the cases (<10%) and controls.26 The association of LEA was conducted by coding ancestry at an SNP as homozygous European ancestry (coded as 11), heterozygous European/African ancestry (coded as 12), or homozygous African ancestry (coded as 22), which served as a reference. Using the Cochran‐Armitage trend test, the analysis tests whether the probability of disease increases with increase in number of “European ancestry” present. All analyses were adjusted for global ancestry and other known CVD risk factors, including age, sex, current smoking status, hypertension, diabetes mellitus, and LDL and HDL cholesterol. To correct for multiple testing, we used the method proposed by Shriner et al to calculate the total number of independent tests.27 The method uses an autoregressive model, which accounts for the high correlation between SNPs given that ancestry over extended portions of a chromosomal region tends to be the same. The total number of independent tests estimated based on this approach were 146, resulting in a threshold significance level of 3.41×10−4 after applying Bonferroni correction to yield an experiment‐wise type I error rate of 5%.

A SNP most significantly associated with the clinical events along with the block of neighboring SNPs with similar European ancestry (“0”, “1”, or “2” where “0” means both chromosomal regions had African ancestry, “1” means the chromosomal region was admixed, European ancestry and the other African ancestry, and “2” means both chromosomal regions had European ancestry) were denoted as a “LEA region.” Additionally, blocks of same ancestry LEA regions were further defined based on whether the SNPs were in the genic or intergenic regions. A genic region included the 5′ and 3′ untranslated regions, the exons and the introns; the intergenic regions were those that fell between 2 genic regions as described above. Gene nomenclatures were obtained from the NCBI Reference Sequence database.

Association of LEA Regions With Previously Reported CVD‐SNPs

Based on previous GWAS, we identified 341 SNPs that were found to be associated with MI, CHD, stroke, or CVD. We searched the National Human Genome Research Institute's (NHGRI) GWAS catalog for the above 4 mentioned CVEs and found a total of 222 SNPs.28 We also used the coronary artery disease (CAD) gene database (CADgene) to identify SNPs not found in our search through NHGRI's GWAS catalog. CADgene is a web resource that provides a comprehensive and curated list of genes that have been previously associated with CAD or related traits.29 This led to the identification of 76 SNPs. The remaining SNPs (n=43) were identified through recent GWAS/meta‐analyses on CHD and stroke performed in African Americans,11, 30 which, together with the 298 SNPs, were examined for significant LEA associations with clinical events. Because these individual results are replications of previous studies, we set the statistical significance at P<0.05 for each test in the current study.

Results

The baseline characteristics and follow‐up events of the 3000 African Americans included in the current analysis are shown in Table 1. The average age was 53.2 (±5.8) years, and the average estimated global European ancestry was 17.4% (±11.0%) among all participants. Females comprised 63.2% of the participants and were similar to males in age, proportion of estimated LEA, and plasma LDL cholesterol levels. Similarly, the number of participants with hypertension and diabetes mellitus did not differ by sex. Females and males were significantly different with respect to their current smoking status and plasma HDL cholesterol levels. Clinical CVD events were significantly higher in males as compared to females.

Table 1

Clinical Characteristics and CVEs Among 3000 African American Participants From the ARIC Study

Characteristics	All (n=3000)	Males (n=1103)	Females (n=1897)
European ancestry, %	17.4 (11.0)	17.5 (11.4)	17.3 (11.0)
Age (y), mean (SD)	53.2 (5.8)	53.5 (6.0)	53.1 (5.7)
Current smoker, n (%)	877 (29.2)	418 (37.9)†	459 (24.2)
Hypertension, n (%)	1642 (54.7)	583 (52.9)	1059 (55.8)
Diabetes mellitus, n (%)	551 (18.4)	184 (16.7)	367 (19.3)
HDL cholesterol (mg/dL), mean (SD)	52.5 (20.8)	50.6 (16.6)†	58.0 (17.3)
LDL cholesterol (mg/dL), mean (SD)	130.3 (52.8)	137.2 (40.9)	139.1 (44.1)
MI, n (%)	219 (7.3)	96 (8.7)*	123 (6.5)
Stroke, n (%)	251 (8.4)	109 (9.9)*	142 (7.5)
Fatal CHD, n (%)	115 (3.8)	60 (5.4)†	55 (2.9)
CHD composite,a n (%)	351 (11.7)	171 (15.5)†	180 (9.5)
CVD composite,b n (%)	420 (14.0)	187 (16.9)†	233 (12.3)

ARIC indicates Atherosclerosis Risk in Communities; CHD, coronary heart disease; CVD, cardiovascular disease; CVEs, cardiovascular events; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; MI, myocardial infarction.

*P<0.05 and † P<0.001 variables statistically signficantly different between males and females.

CHD composite: includes MI, silent MI, revascularization post‐MI, or fatal CHD.

CVD composite: includes MI, stroke, or deaths following either MI or stroke.

The genome‐wide associations of the LEA regions with each of the clinical events are summarized in the Manhattan plots in Figure A through E). The LEA at the top 5 regions associated with each clinical CVE are listed in Table 2 (LEA regions significant at P<0.05 are available in Tables S2 thorugh S6). Genome‐wide significance was achieved by 121 LEA regions with an effective size (as described above) of 14 in relation to MI and 2 of the LEA regions in relation to the CVD. The most significant LEA region was associated with ≈2 times higher odds of MI (P=1.45×10−6) and included the region spanning 158.388 to 159.262 Mb on chromosome 4 downstream to the glutamate ionotropic receptor alpha‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazole propionate (AMPA) type subunit 2 (GRIA2 [4q32.1]) gene. On the other hand, the LEA region from 62.030 to 62.331 Mb on chromosome 6, upstream of the humanin‐like 9 (HN9 [6q11.1]) gene, was associated with 0.37 times lower odds of fatal CHD (P=7.34×10−4), whereas the LEA region spanning from 16.915 to 17.727 Mb downstream of the chromosome 21 open reading frame 37 (C21orf37 [21q21.1]) gene was associated with 1.55 times higher odds of the composite CHD event (P=3.45×10−4). Association of LEA with stroke was observed in the region spanning 18.681 to 19.939 Mb on chromosome 3 upstream of the inhibitor of DNA‐binding 4, helix loop helix (HLH) protein (ID4 [6p22.3]) gene, with a 1.57 times higher odds of stroke (P=9.69×10−4). Likewise, the LEA region from 166.242 to 166.250 Mb on chromosome 4 in the transmembrane protein 192 (TMEM192 [4q32.2]) gene was associated with a 1.53 times higher odds of the composite CVD event (P=3.04×10−4).

Figure 1

Manhattan plot of the association between local European ancestry (LEA) and CVE using the additive model for (A) myocardial infarction, (B) fatal coronary heart disease, (C) coronary heart disease composite, (D) stroke, and (E) cardiovascular disease composite. The gray line corresponds to P=3.41×10−4 based on the effective number of the independent test, and the results above this threshold are considered stasitically significant associations (negative log10 P values are plotted against each SNP's respective position on each chromosome). CHD indicates coronary heart disease; CVD, cardiovascular disease; CVE, cardiovascular events; MI, myocardial infarction; SNP, single‐nucleotide polymorphism.

Table 2

Association Results of Top 5 LEA Regions Associated With Cardiovascular Clinical Events Among African Americans From the ARIC Study

Chromosome	Region (Mb)	First SNP	Last SNP	Gene	OR	95% CIs	P Value
MI
4	158.388 to 159.262	rs11100102	rs10517686	GRIA2	1.97	1.48 to 2.58	1.45×10−6 a
4	160.502 to 164.258	rs6536413	rs10019236	FSTL5	1.93	1.44 to 2.50	2.66×10−6 a
4	160.277 to 160.498	rs17225089	rs6816166	RAPGEF2	1.91	1.44 to 2.48	3.88×10−6 a
4	158.226 to 158.339	rs4615228	rs17035876	GLRB	1.92	1.45 to 2.53	3.90×10−6 a
4	159.895 to 160.176	rs6815665	rs17037938	FNIP2	1.90	1.45 to 2.50	4.05×10−6 a
Fatal CHD
6	62.030 to 62.331	rs1689238	rs4710691	HN9	0.37	0.23 to 0.70	7.34×10−4
6	62.451 to 63.030	rs9350974	rs16882802	KHDRBS2	0.37	0.21 to 0.66	7.37×10−4
6	64.090 to 64.330	rs9448873	rs6936079	LGSN	0.39	0.22 to 0.69	1.16×10−3
6	64.490 to 66.474	rs1884118	rs9294647	EYS	0.39	0.22 to 0.69	1.18×10−3
6	64.350 to 64.350	rs16894460	rs16894460	PTP4A1	0.39	0.22 to 0.69	1.21×10−3
CHD compositeb
21	16.915 to 17.727	rs2823911	rs1689155	C21orf37	1.55	1.17 to 1.90	3.45×10−4
2	228.443 to 228.489	rs10498223	rs13406984	DAW1	1.55	1.12 to 1.83	3.67×10−4
2	228.572 to 228.745	rs11680053	rs13402484	SKIP	1.55	1.22 to 1.97	3.67×10−4
2	227.413 to 227.567	rs4673171	rs10210997	RHBDD1	1.53	1.18 to 1.92	6.07×10−4
2	227.318 to 227.371	rs13431179	rs3731597	IRS1	1.52	1.19 to 1.91	6.66×10−4
Stroke
6	18.681 to 19.939	rs760827	rs12528525	ID4	1.57	1.19 to 2.04	9.69×10−4
6	18.270 to 18.329	rs9477642	rs214506	KDM1B	1.56	1.20 to 2.04	1.01×10−3
6	17.869 to 18.082	rs9350058	rs2328176	KIF13A	1.56	1.19 to 2.02	1.14×10−3
6	17.419 to 17.500	rs1745070	rs2328099	RBM24	1.55	1.18 to 2.01	1.16×10−3
6	18.239 to 18.253	rs9367980	rs11964408	TPMT	1.55	1.19 to 2.02	1.20×10−3
CVD compositec
4	166.242 to 166.250	rs12642069	rs17045969	TMEM192	1.53	1.21 to 1.91	3.04×10−4 a
4	168.398 to 169.248	rs17600543	rs6845007	TES‐3	1.50	1.17 to 1.83	3.32×10−4 a
4	166.173 to 166.176	rs12511595	rs13111010	TRIM60	1.51	1.21 to 1.91	3.49×10−4
4	169.395 to 169.471	rs6552559	rs6846811	DDX60	1.49	1.19 to 1.86	5.19×10−4
4	165.903 to 165.941	rs17045455	rs9999138	LINC01207	1.49	1.18 to 1.87	5.76×10−4

ARIC indicates Atherosclerosis Risk in Communities; CHD, coronary heart disease; CVD, cardiovascular disease; LEA, local European ancestry MI, myocardial infarction; OR, odds ratio.

Statistically significant at genome‐wide association level of 3.41×10−4.

CHD composite: includes MI, silent MI, revascularization post‐MI, or fatal CHD.

CVD composite: includes MI, stroke, or deaths following either MI or stroke.

LEA Associations With Previously Significant CVD‐SNPs

The LEA associations at 20 of the 341 SNPs previously reported to be significant in relation to 1 or more CVD events or CVD‐related traits are presented in Table 3. Among the 20 SNPs, 9 were associated with MI, 6 with stroke, 3 with fatal CHD, 4 with the composite CHD event, and 6 with the composite CVD event. s10777317 downstream of the decorin (DCN) gene on chromosome 12 was the only SNP associated with MI, stroke, and the CVD composite, with ≈0.71 times lower odds for all 3 events. The strongest association of MI and composite CHD was observed with rs10498211 located near the insulin receptor substrate 1 (IRS1) gene on chromosome 2, which was associated with 1.51 times higher odds for both events, whereas fatal CHD was associated with rs16893526 located in the region intergenic to family with sequence similarity 46 member A (FAM46A) and inhibitor of Bruton tyrosine kinase (IBTK) on chromosome 6. Rs7136259 in the ATPase plasma membrane Ca2+ transporting 1 (ATP2B1) gene and rs10777317 near the DCN gene on chromosome 12 were both associated with 0.71 times lower odds of stroke, and the CVD composite was associated with rs879324 in the ZFHX3 gene on chromosome 16, resulting in 1.37 times higher odds of CVD.

Table 3

Association of Clinical CVE With LEA at Previously Reported Significant CVD Single‐Nucleotide Polymorphisms (SNPs)

Chromosome	Position	SNP	Gene	MI		Fatal CHD		CHD Compositea		Stroke		CVD Compositeb
				OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value	OR (95% CI)	P Value
1	109623689	rs599839	PSRC1					1.33 (1.06–1.69)	0.02
1	180348583	rs10911021	ZNF648									0.77 (0.6–0.99)	0.04
2	226776324	rs2943634	IRS1 c	1.50 (1.12–2.00)	0.006			1.47 (1.16–1.87)	0.002
2	227330075	rs10498211	IRS1 c	1.51 (1.13–2.01)	0.005			1.51 (1.19–1.91)	0.0008			1.27 (1.01–1.60)	0.04
4	148620640	rs6841581	EDNRA	1.41 (1.05–1.88)	0.02							1.30 (1.04–1.63)	0.02
6	32449331	rs9268402	C6orf10, HCG23					0.75 (0.58–0.97)	0.03
6	43866851	rs6905288	VEGFA			0.55 (0.33–0.91)	0.02
6	82572034	rs16893526	FAM46A, IBTK			0.48 (0.29–0.81)	0.006
7	12898574	rs732577	ARL4A, ETV1			1.60 (1.08–2.35)	0.02
10	44095830	rs1746048d	CXCL12 c	0.71 (0.52–0.97)	0.03
12	88605319	rs7136259	ATP2B1							0.67 (0.49–0.93)	0.02
12	90504505	rs10777317	DCN, C12orf79	0.71 (0.51–0.99)	0.04					0.68 (0.49–0.93)	0.02	0.73 (0.57–0.94)	0.01
12	119919970	rs2259816	HNF1A							0.73 (0.53–0.99)	0.04
16	71488586	rs16971384	ZFHX3							1.32 (1.01–1.74)	0.04	1.36 (1.09–1.69)	0.006
16	71626179	rs879324	ZFHX3							1.34 (1.03–1.74)	0.03	1.37 (1.10–1.70)	0.005
16	81769899	rs8055236	CDH13							0.74 (0.54–1.00)	0.04

CHD indicates coronary heart disease; CVD, cardiovascular disease; LEA, local European ancestry; OR, odds ratio; SNP, single‐nucleotide polymorphism.

CHD composite: includes MI, fatal CHD, silent MI, or revascularization procedures.

CVD composite: includes MI, stroke, or deaths following MI or stroke.

Refer to genes reported in the literature with a different annotation at present.

Four more SNPs in linkage disequilibrium with the reported SNP had the same magnitude of association—rs622472, rs687175, rs535949, and rs174604.

Discussion

Admixture mapping in relation to clinical cardiovascular events at the genome level has not been evaluated to date. Our study examined and found several LEA regions significantly associated with clinical events, such as MI, stroke, CHD, and CVD, among admixed African Americans in a large, population‐based cohort. Genome‐wide significant LEA regions were detected for MI and the CVD composite. The local ancestry association with MI included a total of 121 LEA regions reaching genome‐wide significance and included several SNPs in strong LD with one another (r 2>0.8). The LEA at these regions was associated with ≈2 times higher odds of MI. The top LEA regions were associated with higher odds for all clinical events except in the case of fatal CHD, where they were associated with lower odds of the disease. Furthermore, we found LEA at 20 SNPs to be significantly associated with the clinical events in our study that were reported in previous GWAS.

A review of our top LEA regions found several SNPs that have been associated with either clinical CVD or CVD‐related traits. These, with the relevant references, are listed in Table S7 along with SNP associations reported in the literature for some of the previously reported genes. Other than the clinical CV events, majority of the SNP associations are related to lipid markers, followed by body weight measures/BMI, blood pressure, tunica media, inflammatory markers, such as monocyte chemoattractant protein‐1, interleukin‐10, basophils, fibrinogen, and diagnostic measures. Multiple associations were noted for the sphingosine kinase type 1 (SPHK1)‐interacting protein (SKIP), inhibitor of DNA‐binding 4 (ID4), and DCN genes, demonstrating that these particular genes are most likely involved in overlapping or complex pathways that lead to manifest disease.

The top LEA regions for the cardiac events, including MI, fatal CHD, and the CHD composite, were located in chromosomes 4, 6, and 2, respectively. These were associated with higher odds of the event, except in the case of fatal CHD. The top LEA region for MI was located downstream of the GRIA2 gene that encodes for the GluR2 protein, a subunit of the glutamate receptor that is sensitive to AMPA and acts as a ligand‐gated cation channel for calcium (Ca2+) and zinc (Zn2+).31 Ischemia‐induced death of neuronal cells has been associated with increased Ca2+ permeability of the channels regulated by the GluR2 subunit.32 Downregulation of the gene and RNA editing has been postulated to change GluR2 ion permeability and subsequently its downstream effects.33 A study of benign uterine smooth muscle cell tumors also localized the receptor subunits to endothelial cells with a possible role in increased vascularization of the tumor. These findings suggest that the gene may either play a role in ischemic‐reperfusion injury as noted in case of CHD or ischemic injuries as observed during MI or stroke.

The protective association of the LEA region near HN9 with fatal CHD is supported by several in vitro, in vivo, and clinical studies demonstrating the protective effect of Humanin.34 The Humanin‐like 9 peptide is an isoform of Humanin and is highly expressed in heart and kidney muscles.35 Humanin was initially shown to inhibit apoptosis in neuronal cells linked to Alzheimer's disease.36 Later studies revealed similar effects in relation to MI and reperfusion injury, stroke, and diabetes mellitus.34, 37, 38 This peptide has been shown to preserve endothelial function, inhibit inflammatory responses, as well as moderate stress from increases in oxidized LDL cholesterol levels. The effects were also observed early during the atherosclerotic process and has led to the peptide being considered as a potential therapeutic agent.39

The top 2 genes and their regions on chromosome 21, and 2 associated with CHD composite, have not yet been characterized and thus their potential role in the pathogenesis of the disease cannot be elucidated. The SKIP protein encoded by the SKIP gene regulates SPHK1, which is involved in the phosphorylation of sphingosine to sphingosine‐1‐phosphate (S1P),40 and is an A‐kinase anchoring protein present in endothelial and vascular smooth muscle cells (VSMCs), and in high concentrations in heart tissue. SKIP is involved in cell proliferation and cell survival, monocyte‐endothelial cell adhesion, and calcium homeostasis through SPHK1 and S1P mechanisms, which could potentially lead to heart disease.40, 41

The odds of stroke were higher in association with the LEA region upstream of the ID4 gene. The ID4 protein binds to the HLH transcription factors and inhibits their binding to DNA.42 The HLH proteins are involved in cell growth and differentiation. Messenger RNA (mRNA) expression levels show that ID4 is expressed, to a certain extent, in the cardiac and VSMCs.41, 42, 43 Cyclic adenosine monophosphate (cAMP) mediates downregulation of ID4 through its downstream effects and inhibits the apoptotic process.43 ID4 methylation reduces mRNA expressions of transforming growth factor β‐1 and vascular endothelial growth factor (VEGF), resulting in reduced angiogenesis. SKIP, on the other hand, facilitates cell proliferation and angiogenesis through the cAMP pathway and interaction of S1P with VEGF. The common signaling pathway and downstream effects of ID4 and SKIP suggests that these genes may be part of the same or extended gene network and possibly interact with each other, regulating mechanisms involved in stroke or other CVD events.

The TMEM192 gene associated with higher odds of the CVD composite encodes for a lysosomal membrane protein that is expressed, to some extent, in heart, brain and skeletal muscles44 and was shown to inhibit autophagy and subsequent apoptosis in a cell‐culture study, but the actual function of the gene has not yet been characterized.45 The TES‐3 region is also associated with the CVD composite and encodes for the testican‐3 protein that has been shown to suppress matrix metalloproteinase‐2 (MMP2) activity in cultured cells.46, 47 MMPs are Ca2+‐ and Zn2+‐dependent proteases that are involved in several physiological and pathological processes, most important of which is the breakdown of the extracellular matrix.48 The role of MMPs in atherosclerosis and CVD is well established and shows that they are involved in cellular and inflammatory responses, including migration and adhesion of immune cells, cytokine/chemokine regulation, angiogenesis as well as cell proliferation, differentiation, and apoptosis.48 A recent review provides evidence of the role of MMPs not only in CVD, but also in CVD‐related risk factors.49 TES‐3 may therefore be 1 of the inhibitors involved in regulation of MMPs and thus modulate the pathophysiological processes involved in CVD.

The association of genetic risk factors with CVD has been extensively studied using candidate gene, GWAS, and meta‐analyses approaches. A search of the NHGRI's GWAS catalog alone lists a total of 32 GWAS for clinical CVE, 27 of which were related to MI or CHD, and the remaining 5 related with stroke.28 The SNPs in certain genes, for example, IRS1, endothelin receptor type A (EDNRA), DCN, and ZFHX3, are frequently associated with CVD or related traits, including our study. Insulin signaling and function was shown to reduce atherosclerosis through the endothelin receptors, whereas downregulation of DCN and ZFHX3 were associated with anti‐inflammatory/antiapoptotic and perturbed Ca2+ homeostasis, respectively.50, 51, 52, 53, 54, 55 Functional studies are available for a few but future research is needed to ascertain the separate or combined effects of these genes that have been consistently associated with CVEs.

The strength of our study lies in the comprehensive evaluation of local ancestry association with individual as well as composite cardiac and CVEs. We used genome‐wide SNP data to estimate local ancestry, which provided higher resolution and allowed us to determine LEA associations at regions that cannot be possibly covered by ancestry informative markers. However, we followed stringent quality control measures (eg, HWE) and excluded SNPs that were not available in HapMap to allow ancestry estimation, which means we may have missed associations with SNPs that were excluded from the current analysis. Even though we may have excluded informative SNPs (eg, low heterozygosity) by using a stringent HWE threshold of P<0.001, we ensured that we had ruled out false positives attributed to genotyping errors. It is possible that this may have also affected our ability to replicate some of the admixture and association results reported in earlier studies.

This study is the first to report the association of LEA regions in relation to clinical CVEs using a genome‐wide admixture mapping approach among African Americans. We also present genomic regions associated with CVEs in the admixed population in our study that corroborate with findings from previous association studies in European ancestry populations, suggesting potential ancestry specific associations that might not have been evident from traditional association studies. However, these are initial findings and as such should be considered hypothesis generating. Nonetheless, these findings provide us with an opportunity to explore some new gene regions and pathways, but also necessitate confirmation of some of the previously reported and known genetic associations. Therefore, future studies are needed to replicate the novel associations, detect the causal variants driving these associations, and explore clinical applications for those gene regions consistently associated with CVEs.

Sources of Funding

Support is provided by the grant R56HL125061 (Shrestha) from NIH National Heart, Lung, and Blood Institute. The ARIC Study is carried out as a collaborative study supported by NIH National Heart, Lung, and Blood Institute contracts (HHSN268201100005C, HHSN268201100006C, HHSN268201100007C, HHSN268201100008C, HHSN268201100009C, HHSN268201100010C, HHSN268201100011C, and HHSN268201100012C). Funding for GENEVA was provided by the NIH National Human Genome Research Institute grant U01HG004402.

Disclosures

None.

Table S1. Quality Control of Genome‐wide Data Among African Americans From the Atherosclerosis Risk in Communities Study

Table S2. Local European Ancestry (LEA) Regions Associated With Myocardial Infarction Among African Americans From the Atherosclerosis Risk in Communities Study (ARIC)

Table S3. Local European Ancestry (LEA) Regions Associated With Fatal CHD Among African Americans From the Atherosclerosis Risk in Communities Study (ARIC)

Table S4. Local European Ancestry (LEA) Regions Associated With CHD Composite Among African Americans From the Atherosclerosis Risk in Communities Study (ARIC)

Table S5. Local European Ancestry (LEA) Regions Associated With Stroke Among African Americans From the Atherosclerosis Risk in Communities Study (ARIC)

Table S6. Local European Ancestry (LEA) Regions Associated With CVD Composite Among African Americans From the Atherosclerosis Risk in Communities Study (ARIC)

Table S7. List of Other Diseases/Traits and Associated Single‐Nucleotide Polymorphisms (SNPs) Among Top Clinical Cardiovascular Genes in the Current Study

Click here for additional data file.