Introduction

In the past 7 years, genome-wide association studies (GWAS) have been successful in mapping the chromosomal location of numerous common coronary artery disease (CAD)-associated alleles. Although important, these explain only approximately 10 % of the predicted heritable risk for CAD. The age-specific incidence of CAD in participants in the Framingham Offspring Study was increased approximately two-fold in subjects with a family history of premature disease after adjustment for conventional CAD risk factors [1]. The Swedish Twin registry followed close to 21,000 subjects for over 35 years and estimated the heritability of fatal CAD events to be 0.57 for men and 0.38 for women, with heritable effects being most manifest in younger individuals [2]. Similarly several GWA studies have demonstrated not unexpectedly that the genetic influence is greatest for early onset CAD events [3•].

This recent progress in the genetics of CAD and other complex disease has been driven by technological advances including high-throughput DNA microarray technology, the availability of 1,000 Genomes data sets to facilitate imputation of less common variants, and a number of bioinformatic approaches including pathway analysis for GWAS. In the commercial arrays used for GWAS, single nucleotide polymorphisms (SNPs; generally 0.5–1 M) are used to tag common variation (SNPs with a frequency of ≥5 %) across the human genome. Importantly, these are ‘tag’ SNPs that point to a causative locus but are rarely in themselves functional variants. This approach makes use of linkage disequilibrium (LD), that is, the nonrandom coinheritance of genetic variants across the human genome.

Segments of DNA known as haplotype blocks are shared within ancestral groups. Due to the recent migration of humans into Europe, individuals of European ancestry have more correlated SNPs and longer haplotype blocks as compared to populations of African ancestry [46]. As reported by Hinds et al., the average haplotype block is ~20.7 kb in European whites, ~8.8 kb in African Americans and ~25.2 kb in Han Chinese. Thus, fewer ‘tag’ SNPs are required for genotyping a population of European or East Asian versus African ancestry. If causal variants are shared across different populations, by taking advantage of differences in LD and allele frequencies, trans-ethnic mapping could facilitate identification of causal variants underlying disease susceptibility. In particular, the significantly shorter haplotype blocks in African American populations can be helpful in fine mapping association signals identified in European populations and also in identifying new signals that may be ethnic specific [7].

Although most GWAS for CAD have been carried out in European white populations, smaller but important studies have been performed in East Asian, South Asian and African American populations (Table 1).

Table 1 CAD Loci reported in European and Non-European populations

GWAS in European White Populations

The first robust association with CAD identified by the GWAS approach, a 53-kb linkage disequilibrium block containing multiple highly correlated single nucleotide polymorphisms (SNPs) at the 9p21.3 locus, was identified by three independent groups in 2007 [810]. The early discovery of this risk locus was facilitated by its large effect size and high-risk allele frequency (approximately 0.48). Approximately 25 % of Europeans carry two copies of the risk allele and have a 50 % increased risk of CAD in general and an even greater increased risk of premature CAD. This was demonstrated in the large Coronary Artery Disease Genome-wide Replication and Meta-Analysis (CARDIoGRAM) of several GWASs, in which the allele-specific odds ratio (OR) for CAD in subjects with CAD onset before the age of 50 years was 1.41 [95 % confidence interval (CI) 1.34–1.48], significantly greater than that for older individuals (OR 1.24; 95 % CI 1.20–1.28) [3•]. The 9p21 locus is also associated with the overall severity of atherosclerosis [11, 12], with a substantially higher risk allele frequency in subjects with multi-vessel disease. Notably, the risk conferred by this locus is independent of known risk factors including diabetes, plasma lipids, blood pressure, adiposity, inflammatory markers, sex and age. Other vascular phenotypes associated with the 9p21 risk alleles include carotid atherosclerosis [13], stroke [14] [15] and peripheral arterial disease [16], as well as abdominal aortic aneurysm [17] and intracranial aneurysms [18], the latter highlighting possible effects on vascular remodeling pathways as well as platelet reactivity [19]. There is also a surprising and confirmed association with periodonitis [20, 21].

The causative alleles at 9p21.3 have not been identified. Fine mapping efforts, including targeted resequencing at high coverage and 1,000 Genome imputation, have failed to identify stronger associations than the original GWAS signals [22, 23].

Although the risk region is devoid of protein coding genes, it overlaps a large nonprotein coding RNA, termed “CDKN2BAS (ANRIL),” and lies adjacent to a cluster of cell cycle-regulating tumor-suppressor genes, including the cyclin-dependent kinase inhibitors, CDKN2A and CDKN2B, and SNPs associated with CAD have been found to associate with the expression of each of these genes [2325].

Since the 9p21 discovery in 2007, large meta-analyses of additional GWASs, the majority of which have been conducted in European populations, have identified over 35 additional loci of smaller effect size but with genome-wide (P < 5 × 10−8) significance. This success has been built on large collaborative efforts, including the Myocardial Infarction Genomics Consortium [26], the CARDIoGRAM consortium [3•], the Coronary Artery Disease (C4D) Genetics Consortium [27], CARDIoGRAMplusC4D [28•] and others.

Novel CAD risk loci identified include genes playing known roles in lipoprotein metabolism, hypertension and other CAD-associated phenotypes, but importantly include several novel loci of unknown function. Highlighting the discovery potential of the GWAS approach, the majority of risk loci harbor genes not previously known to be involved in atherosclerosis or plaque rupture. Several risk regions such as ABO and SH2B3 exhibit pleiotropic effects, associating with multiple CAD and non–CAD-related phenotypes. Overall, as might be expected for common variants affecting a complex trait such as CAD, the effect sizes are small. With the exception of the 9p21.3 locus, the LPA gene [29] encoding lipoprotein(a), and a region of unknown function at 6p24, the allele-specific ORs for CAD of replicated loci are less than 1.15.

Studies in East Asian Populations

In general, most of the disease-related GWAS loci discovered in Europeans have been extensively replicated in populations of European and East Asian ancestry. Marigorta and Navarro [30•] recently demonstrated a strong and significant correlation of odds ratios of specific SNPs for CAD and 27 other diseases across European and East Asian samples, indicating that in general causal variants are shared between the two populations. The SNPs that failed to replicate in East Asian populations mapped to genomic regions with differing linkage disequilibrium patterns.

Shortly after the 9p21.3 finding in Europeans was reported, three groups replicated the association with CAD in Chinese Han [31], Japanese [32, 33] and Korean [34] populations. Fewer large GWAS studies for CAD have been carried out in East Asian as compared to European populations. In 2011, Wang et al. [35] reported a three-stage GWAS for CAD in the Chinese Han population. The numbers were relatively small, consisting of a total of 3,470 cases and 4,583 controls. A novel association between rs6903956 near C6orf105 on chromosome 6p24.1 was found. Of note, the MAF of this SNP is higher in the HapMap European (CEU) population (28 vs. 6.7 % in Chinese), and no association with CAD has been reported in European GWAS.

In 2012, Lu et al. [36•] performed a meta-analysis of two GWAS of CAD in Han Chinese, consisting of 1,515 cases and 5,019 controls with replication in 15,460 cases and 11,472 controls. Four loci originally identified in European populations were confirmed, including 9p21.3, PHACTR1, TCF1 and C12orf51, and four new loci identified, including WDR35, GUCY1A3 (later confirmed in CARDIoGRAMplusC4D [28•]) and C6orf10-BTNL2 and ATP2B1. In the same year, Takeuchi et al. [37] reported on a multistage GWA study performed in the Japanese. The discovery phase consisted of a GWAS in 806 cases and 1,337 controls with wet lab replication of 12 SNPs in 3,052 cases and 6,335 controls. Three loci achieved significance: 12q24 near BRAP and ALDH2, HLA, DRB-DQB on 6p21 and 9p21.3.

More recently (2013), a two-stage CAD GWAS was reported in Korean and Japanese populations [38]. The discovery sample included 2,123 cases and 3,591 controls recruited in Korea with wet lab replication in 3,052 cases and 4,976 controls in Japan. CAD association was replicated for three GWAS-identified loci identified in European populations including the SORT1 locus at 1p13.3 (rs599839), the 9p21.3 CDKN2BAS locus (rs4977574) and PDGFD at 11q22.3 (rs974819).

Guo et al. [39] recently investigated the association of the 9p21.3 locus with CAD in 12 case–control studies of East Asians and undertook a meta-analysis for effect size, heterogeneity, publication bias and strength of evidence. SNPs (rs1333049, rs2383206 and/or rs10757278) were genotyped in 12 case–control studies involving a total of 9,813 patients and 10,710 controls. The mean summary odds ratios for these three 9p21.3 SNPs was 1.29, similar to that observed in European populations. In accord with this finding, a large meta-analysis by Dong et al. [40] reported a similar allele-specific odds ratio for the lead 9p21 CAD risk SNP in East Asian and European populations.

Overall, the GWAS and candidate loci replication studies in East Asian populations have successfully replicated many of the loci identified in much larger European studies. Novel loci, e.g., ATP2B1, are of interest and await confirmation in larger studies.

Studies in South Asian Populations

There have been few studies reported in South Asians despite the apparently higher risk for CAD than would be anticipated based on conventional risk factors. In 2010, Saleheen et al. [41] examined the relationship of 9p21.3 risk variants to acute MI in a case-control study in Pakistan consisting of 1,851 cases and 1,903 controls. For the Pakistani population, the odds ratio of the lead SNP, rs1333049, was 1.13 (CI 1.05–1.22) as compared to 1.31 (1.26–1.37) in a meta-analysis of Europeans.

In 2011, the IBC 50 K CAD Consortium [42] conducted a large candidate gene study of CAD susceptibility, including analysis of 49,094 genetic variants in 2,100 genes using a customized gene array in 15,596 CAD cases and 34,992 controls (11,202 cases and 30,733 controls of European descent; 4,394 cases and 4,259 controls of South Asian origin). Associations of several previously known CAD susceptibility loci including 9p21.3, LPA, COL4A1/COL4A2, ZC3HC1 and CYP17A1 were confirmed. Of note, associations in South Asians did not differ appreciably from those in Europeans, except for 9p21.3 exhibiting an allele-specific odds ratio of 1.14 versus 1.27 (P for heterogeneity  =  0.003).

The first published GWAS for CAD including a substantial number of South Asians was by the C4D Consortium and included a discovery data set of 15,420 CAD cases, of which 6,996 were South Asians [27]. Their replication sample included 21,408 CAD cases with only 3,359 of South Asian descent. A total of 11 previously reported common variants for CAD were confirmed in this study with directionally consistent effects in European and South Asian populations for all 11 loci. However, the odds ratio for several of these including 9p21.3, SORT1 and WDR12 was somewhat lower in the South Asian studies. Five new loci achieved genome-wide significance, including LIPA on 10q23, PDGFD on 11q22, a locus containing multiple genes on 7q22, KIAA1462 on 10p11 and ADAMTS7. More recently, the CARDIoGRAMplusC4D Consortium carried out an association analysis in 63,746 CAD cases and 130,681 controls identifying 15 novel loci reaching genome-wide significance [28•]. This study included additional South Asian cohorts, but subgroup analysis by ethnicity was not reported.

In summary, despite smaller data sets, there is general concordance in the directional effects of CAD risk alleles identified in European populations on CAD risk in South Asians. It is of interest that in contrast to the similar effect size reported in European and East Asian populations, the effect size of many risk alleles appears attenuated in South Asians, possibly due to interaction with unknown genetic or environmental risk modifiers. Other larger GWA studies in this population are underway.

African American Populations

In the recent study by Marigorta and Navarro [30•], SNPs associated with 28 disease phenotypes in Europeans exhibited low replication in individuals of African ancestry despite an average statistical power of 59.2 %. This might suggest limited sharing of causal variants between Europeans and Africans as compared to East Asians. However, given the lower level of LD in African populations, potentially shared casual variants may not be tagged by the index SNP identified in European studies [30•].

In general, CAD-associated loci identified in European populations have failed to replicate or shown attenuated effects in black populations. In the PAGE multiethnic study [43], the association of 13 published CAD SNPs with incident CAD events over a 9–16-year follow-up period was examined in four large US prospective cohorts, including 26,617 white individuals (6,626 events), 8,018 black individuals (914 events), 1,903 Hispanic individuals (113 events), 3,669 American Indian individuals (595 events) and 885 Asian/Pacific Islander individuals (66 events). In white subjects, 9 of the 13 loci were statistically associated with incident CAD events including 9p21, 16q23.1, 6p24.1, 2q36.3, MTHFD1L, APOE, ZNF627, CXCL12 and LPL. Notably, despite an adequate sample size, these SNPs were not associated with CAD in black participants, and 9p21.3 reached nominal significance in the American Indian but not the other small populations.

Kral et al. [44] sought to further characterize the role of genetic variants in 9p21.3 in African American individuals. Healthy siblings of African American patients with documented CAD < 60 years of age (548 sibling pairs) were genotyped and followed for incident CAD for up to 17 years. Of 86 SNPs across the 9p21.3 region, a single SNP within the 3′UTR of the CDKN2B gene met stringent criteria for statistical significance, including permutation-based evaluations. This variant, rs3217989, with a MAF of 0.242, was associated with protection against CAD (OR 0.19, 95 % CI 0.07–0.50, P = 0.0008) and in this study replicated in a combined analysis of two additional case/control studies of prevalent CAD/MI in African Americans (n = 990, P = 0.024, OR 0.779, 95 % CI 0.626–0.968). This was the first report of a CAD association signal in a population of African ancestry within the 9p21 locus. Although common in the African American population, rs3217989 has an MAF of <0.01 in European white and Asian populations, and this finding remains to be confirmed in independent African American cohorts.

The largest GWAS for CAD-related complex traits in African Americans was reported by Lettre et al. [45•] in 2011. The NHLBI CARe study consisted of 8,090 subjects from five population-based cohorts: Atherosclerosis Risk in Communities (ARIC; N = 3,269), Coronary Artery Risk Development in young Adults (CARDIA; N = 1,209), Cleveland Family Study (CFS; N = 704), Jackson Heart Study (JHS; N = 2,200) and Multi-Ethnic Study of Atherosclerosis (MESA; N = 1,737), genotyped on the Affymetrix 6.0 array. However, the number of CAD cases was relatively small. In this study the authors took advantage of the shorter LD blocks in African Americans in an attempt to fine-map some of the associations previously reported in Europeans. To do this, they evaluated SNPs that were correlated with the index SNP in HapMap CEU (r 2 > 0.5) but largely uncorrelated with it in the HapMap African population (YRI, r 2 < 0.1). For many traits, the same signals were responsible for the associations in Europeans and African Americans with important exceptions where the predominant association signals were at SNPs strongly correlated with the index SNPs in HapMap CEU but not with the index SNPs in HapMap YRI including at the 9p21.3 locus for CAD.

The only 9p21.3 SNP previously reported in European populations to reach significance in CARe was rs6475606 at 9p21 position 22081850 in intron 12 of CDKN2BAS (ANRIL) (replication P value 6.4E−4; adds ratio 2.0). Of interest, rs6475606 lies within a smaller HapMap YRI LD block (43 kb) as compared to the major 128-kb LD block defined in the HapMap CEU population. The frequency of the effect allele of rs6475606 was 0.109 in this African American population versus 0.008 in HapMap YRI and 0.52 in HapMap CEU data sets. It was hoped that this finding might define a smaller genomic interval within the 9p21 locus to search for causative alleles. Several linked SNPs overlap DNAse protected sites, harbor enhancer histone marks and are predicted to alter transcription factor binding motifs.

In a more recent study of the genetics of coronary artery calcification (CAC) [46], 166 SNPs in the 9p21.3 region significant for CAD and CAC for EUR population were queried. Of these, 24 SNPs displayed nominal evidence for association (P ≤ 0.05). Ten of these 24 SNPs localized within the 43-kb HapMap YRI region and 14 in the 128-kb HapMap CEU LD block. However, neither the strongest 9p21.3 EUR SNPs for CAC (rs1333049) nor CAD (rs4977574) in Europeans reached significance in the African American population. The peak CAC association was within the smaller 43-kb LD block at rs16905644 (effect allele frequency 0.11, Bonferroni corrected P = 0.0068). No signal for CAC was apparent for either rs6475606 [44] or rs3217989 [45•], reported to associate with CAD in African Americans by Kral et al. [44] and Lettre et al. [45•], respectively.

Conclusion

In conclusion, despite the high prevalence of CAD among African Americans and the potential advantage of interrogating their shorter LD blocks for fine mapping of previously identified CAD loci, progress has been minor. The two signals reported for CAD at the 9p21.3 locus have not yet achieved replication in independent African American data sets, and other signals for CAD per se have not been identified, although replication has been achieved for signals associated with discrete CAD risk factors [45•]. It is possible that the genetic risk for CAD relates more strongly to genetic contribution to discrete risk factors more common in African Americans including hypertension and obesity. Multiple genetic variants of small effect are believed to account for much of the missing heritability of CAD in Europeans, and these may be even greater in number and thus more difficult to detect in populations of African descent.

As Carlson et al. [47••] reported from the Population Architecture using Genomics and Epidemiology (PAGE) study, a consortium of multi-ancestry, population-based studies, 25 % of tag SNPs identified in European GWAS had significantly different effect sizes in non-European populations, and this was particularly evident in African American cohorts. They demonstrated that that differential LD between tag SNPs and functional variants within populations contributed significantly to diluting the effect sizes in this population [47••]. Larger studies in both African American and South Asian populations are underway, and other approaches such as pathway analysis for GWAS may provide new information relevant to the biology of CAD in these populations.