New susceptibility locus for coronary artery disease on chromosome 3q22.3

Abstract

We present a three-stage analysis of genome-wide SNP data in 1,222 German individuals with myocardial infarction and 1,298 controls, in silico replication in three additional genome-wide datasets of coronary artery disease (CAD) and subsequent replication in ~25,000 subjects. We identified one new CAD risk locus on 3q22.3 in MRAS (P = 7.44 × 10⁻¹³; OR = 1.15, 95% CI = 1.11–1.19), and suggestive association with a locus on 12q24.31 near HNF1A-C12orf43 (P = 4.81 × 10⁻⁷; OR = 1.08, 95% CI = 1.05–1.11).

Recent genome-wide association studies (GWAS) of coronary artery disease (CAD) have focused on a few chromosomal regions with strong signals^1-4. We hypothesized that the application of stringent statistical thresholds may have dismissed SNPs with modest effects or low allele frequencies (Supplementary Fig. 1 and Supplementary Table 1 online). For this study, we started r by identifying SNPs meeting a less-stringent cutoff for association (P ≤ 1 × 10⁻³) in a new GWAS for myocardial infarction.

In stage 1, we genotyped 869,224 autosomal SNPs from the Affymetrix Genome-Wide Human SNP Array 6.0 in 1,222 myocardial infarction cases (German MI Family Study II (GerMIFS II); Supplementary Methods online) with premature disease onset and positive family history and 1,298 population-based Germans of European descent. After quality control (Supplementary Fig. 2 online), 567,119 SNPs remained. The inflation factor estimated for all SNPs that passed quality control is 1.04 (s.e.m. = 9.2 × 10⁻⁵). Of these, 694 SNPs showed association with myocardial infarction at P ≤ 1 × 10⁻³ in a two-sided trend test. These SNPs were evaluated by in silico analysis in three additional GWAS datasets (stage 2), comprising a total of 5,768 cases and 7,657 controls (WTCCC CAD study⁴, GerMIFS I¹ and Myocardial Infarction Genetics Consortium⁵ together with the Italian Atherosclerosis, Thrombosis, and Vascular Biology Working Group⁶; the latter two form a single data source and are grouped together (MIGen/IATVB); Supplementary Methods). SNPs that were selected for large-scale replication all met the following two criteria: (i) the same allele was associated in the same direction as in the exploratory GWAS in at least two of the three in silico GWAS datasets using a one-sided trend test (P ≤ 0.05, Table 1), and (ii) the SNP had nearby correlated SNPs (within 25 kb) that also showed a signal (one-sided trend test P ≤ 0.05). A total of 21 SNPs met these criteria. They clustered into five chromosomal regions: two regions known to be associated with CAD (9p21.3 and 1q41)^1-4,7 and three previously unreported regions (3q22.3, 9p24.2 and 12q24.31; Supplementary Table 2 online).

Table 1. Association results for the three new candidate CAD risk loci identified in stage 1, represented by the corresponding lead SNPs.

			rs9818870				rs7048915					rs2259816
	Number of		MAF (%)				MAF (%)				MAF (%)
	Cases	Controls	Cases	Controls	P value	OR (CI)	Cases	Controls	P value	OR (CI)	Cases	Controls	P value	OR (CI)
Stage 1a
GerMIFS  II	1,222	1,298	19.5	15.0	2.38 × 10−5	1.37 (1.18,1.59)	22.8	27.1	0.0004	0.79 (0.69,0.90)	40.4	35.3	0.0002	1.25 (1.11,1.40)
Stage 2b
WTCCC	1,926	2,938	17.0	15.1	0.0048	1.16 (1.05,1.27)	21.6	23.4	0.0201	0.90 (0.82,0.98)	36.3	34.3	0.0215	1.09 (1.02,1.17)
MIGen/  IATVB	2,967	3,075	14.8	13.7	0.0390	1.10 (1.01,1.19)	27.5	29.0	0.0302	0.93 (0.87,0.99)	40.2	38.0	0.0066	1.10 (1.03,1.17)
GerMIFS I	875	1,644	18.6	16.5	0.0427	1.15 (1.01,1.31)	23.4	25.9	0.0185	0.87 (0.78,0.97)	38.1	34.9	0.0118	1.15 (1.04,1.27)
Pooled	5,768	7,657	16.1	14.8	0.0002	1.13 (1.07,1.19)	25.0	26.3	0.0004	0.91 (0.87,0.95)	38.6	35.9	5.89 × 10−5	1.10 (1.06,1.15)
Stage 3c
Population  -based MIs	2,737	4,469	17.4	15.0	0.0004	1.20 (1.10,1.31)	26.0	27.2	0.1520	0.94 (0.85,1.04)	37.5	35.4	0.0106	1.10 (1.03,1.17)
Hospital-  based Mis	772	734	18.9	13.5	2.85 × 10−5	1.52 (1.28,1.81)	26.4	26.3	0.4879	1.00 (0.87,1.15)	33.7	35.2	0.2002	0.94 (0.83,1.06)
Angio  graphic  CAD	8,908	7,208	17.7	16.3	0.0008	1.10 (1.05,1.15)	25.3	25.2	0.1915	1.02 (0.98,1.07)	36.5	36.0	0.0410	1.04 (1.00,1.08)
Pooled	12,417	12,411	17.7	15.9	2.69 × 10−7	1.14 (1.08,1.20)	25.5	25.5	0.9999	1.01 (0.96,1.06)	36.6	35.8	0.0277	1.05 (1.00,1.09)
All stagesd	19,407	21,366	17.3	15.4	7.44 × 10−13	1.15 (1.11,1.19)	25.1	25.9	0.0073	0.95 (0.92,0.99)	37.4	35.8	4.81 × 10−7	1.08 (1.05,1.11)

^aStage 1: P value, odds ratio (OR), 95% confidence interval (95% CI) from two-sided trend test (CAT).

^bStage 2: single study analysis: P value, OR, 90% CI from one-sided CAT; pooled analysis: P value, OR, 90% CI from one-sided Cochran-Mantel-Haenszel test (CMH).

^cStage 3: population-based myocardial infarction studies: P value, OR, 90% CI from one-sided CMH (for rs7048915 only one study: one-sided CAT); hospital-based myocardial infarction studies: P value, OR, 90% CI from one-sided CAT; angiographic CAD studies: P value, OR, 90% CI from fixed-effect logistic regression models (FELRM) with adjustments for study and multiple testing.

^dAll stages: P value, OR, 95% CI from (FELRM) with adjustments for study.

In stage 3, we evaluated the lead SNPs of the three previously unreported loci in 12,417 cases and 12,411 controls. Two loci, represented by rs9818870 in the MRAS gene and rs2259816 in the HNF1A-C12orf43 region, showed replication in this stage at P = 2.69 × 10⁻⁷ and P = 0.0277, respectively, when adjusted for study and multiple testing. Results from stages 1–3 are displayed in Table 1 and Supplementary Figure 3 online.

Our rationale for these three stages was to be liberal in terms of a broad inclusion of SNPs at stage I and conservative with respect to a stringent replication strategy in stages 2 and 3. To determine the genuineness of effects in an efficient manner and to eliminate inconsistent findings, we carried out the second stage using in silico replication. The final stage served as a test for significance while adjusting for multiple testing. Power analyses showed that this three-staged strategy detects relevant effects with high probability (Supplementary Methods).

SNP rs9818870 in MRAS on 3q22.3 showed an aggregate P = 7.44 × 10⁻¹³ (OR = 1.15; 95% CI = 1.11–1.19), whereas SNP rs2259816 located in the HNF1A-C12orf43 region on 12q24.31 showed an aggregate P = 4.81 × 10⁻⁷ (OR = 1.08; 95% CI = 1.05–1.11) in 19,407 cases and 21,366 controls, respectively. Both SNPs met the formal criterion for genome-wide significance of 5 × 10⁻⁷ suggested by the WTCCC⁴ that has been applied to several GWAS. However, SNP rs2259816 in HNF1A-C12orf43 failed to reach the more stringent threshold of 7.2 × 10⁻⁸ suggested for an infinitely dense SNP map⁸.

To further explore the association between SNPs in MRAS and HNF1A-C12orf43, we tested whether the lead SNPs were associated with traditional risk factors for CAD in 3,276 controls from MONICA/KORA Augsburg survey S4; however, no significant association was found (Supplementary Table 3 online). In addition, we carried out exploratory subgroup analyses to test for interaction (Supplementary Fig. 4 online).

SNP rs9818870 is located in the 3′ UTR of MRAS in close proximity to a cluster of miRNA binding sites. This SNP is among a cluster of four associated SNPs (rs1199338, rs2347252, rs3732837, rs9818870) that are in a block of strong LD covering the whole gene (Fig. 1a), which consists of five exons and is ~33 kb in size (MIM608435). The M-ras protein belongs to the ras superfamily of GTP-binding proteins and is widely expressed in all tissues, with a very high expression in the cardiovascular system, especially in the heart (SymAtlas). Previous work has shown that M-ras is involved in TNF-α-stimulated LFA-1 activation in splenocytes by using mice deficient in this process⁹. These findings suggest a role for M-ras in adhesion signaling, which is important in the atherosclerotic process¹⁰.

Figure 1. Association results from Stage 1.

(a) 3q22.3 locus. (b) 12q24.31 locus. Genomic positions refer to the UCSC Genome Browser Human March 2006. Presented are SNPs that passed quality control in stage 1. The GWAS P value for the lead SNPs (rs9818870 (3q22.3), rs2259816 (12q24.31)) is denoted by a red diamond. Blue diamonds indicate P values for the lead SNPs in the replication sample. Proxies are indicated with diamonds of smaller size, with colors determined from their pairwise r² from HapMap CEU (red: high LD with lead SNP (r² > 0.8); orange: moderate LD with lead SNP (0.5 < r² < 0.8); yellow: weak LD with lead SNP (0.2 < r² < 0.5); white: no LD with the lead SNP (r² < 0.2) or no information available). The locations of the UCSC genes in the region and recombination rates and hot spots as defined previously¹⁴ are shown.

SNP rs2259816 (representing the locus on 12q24.31) is located in intron 7 of HNF1A. This and two further associated SNPs cluster (rs1169313 and rs2258287) are in a LD block that covers the coding region of HNF1A and C12orf43 (Fig. 1b). HNF1A (also known as TCF1; MIM142410) encodes a transcription factor that binds to promoters of a variety of genes that are expressed exclusively in the liver¹¹. Variants in HNF1A may cause maturity-onset diabetes of the young (MODY) (MIM600496) and affect plasma concentrations of C-reactive protein (CRP), a powerful risk marker for cardiovascular disease¹². Moreover, and in parallel to the present work, a risk allele at the HNF1A locus (rs2258287) has been mapped to higher plasma levels of low-density lipoprotein cholesterol¹³. SNPs rs1169313 and rs2258287 are located within the C12orf43 gene (alias FLJ12448, hypothetical protein LOC64897), which is located downstream of HNF1A in a tail-to-tail manner, with only 500 bp in between (Fig. 1b). RT-PCR studies showed that MRAS and C12orf43 are ubiquitously expressed, including in mouse and human aorta and heart, tissues that are potentially involved in atherosclerosis (Supplementary Table 4 and Supplementary Fig. 5 online).

In conclusion, analysis of a new GWAS followed by in silico replication in three GWAS datasets and a large-scale replication study identified one new susceptibility locus for CAD on 3q22.3 with compelling statistical evidence and a second locus on 12q24.31 with suggestive evidence. Further functional work is needed to define the mechanisms by which these loci translate into a higher risk of CAD, and whether this information can be used to improve prevention, prediction or treatment of this common condition.