Emerging Genomic Applications in Coronary Artery Disease

Abstract

Over the last four years, an unprecedented number of studies illuminating the genomic underpinnings of common ‘polygenic’ diseases including coronary artery disease have been published. Notably, these studies have established numerous DNA variants within or near chromosome 9p21.3, LPA, CXADR, and APOE genes, to name a few, as key coronary artery disease and sudden cardiac death susceptibility markers. Most importantly, many of these DNA variants confer over a 2-fold increase in risk for coronary artery disease, myocardial infarction, and ventricular fibrillation. Additionally, loss-of-function variants in the hepatic cytochrome 2C19 system have now been found to be the predominant genetic mediator of clopidogrel antiplatelet response, with variant carriers having a greater than 3-fold increase in risk for stent thrombosis. In the near future, many additional rare polymorphisms, structural variants, and tissue-specific epigenetic features of the human genome including DNA methylation, histone modifications, and chromatin state will emerge as significant contributors to disease pathogenesis and drug response. In these findings will have the potential to radically change the practice of cardiovascular medicine. However, only the individual clinician can ultimately enable the translation of these important discoveries to systematic implementation in clinical practice.

The publication of the initial draft of the human genome sequence in 2000 involved a multinational decade-long effort at a price tag of over 3 billion dollars.¹ Now, commercial platforms can sequence whole genomes from several individuals in a day at a tiny fraction of the costs. Instead of $10,000 for sequencing 1 million DNA bases, the cost is 10 cents.^1,2 These technological advances combined with a detailed catalogue of common human variation provided by the HapMap Project has spawned the publication of over 550 genome-wide association studies (GWAS) that have strongly linked nearly 800 hundred gene variants to over 150 common polygenic diseases and complex traits.^3–5 These studies have also identified numerous highly predictive pharmacogenetic markers of drug response and toxicity.^6–9 Together, these findings have the potential to radically change the practice of medicine.⁴ This review will detail the current discoveries in the field of translational genomics that have potential ramifications in the prevention and treatment of patients with ischemic heart disease (IHD).

Analyzing the Genome

Before commencing with a detailed discussion on the specific applications of genome-based medicine, it is important to first highlight several key concepts regarding the heritability of traits. Monogenic, rare, “simple” Mendelian traits segregate in an autosomal dominant, recessive, X-linked, or mitochondrial basis. These relatively infrequent disorders are typically caused by rare, deterministic genetic mutations. Historical examples of such traits in cardiovascular medicine include hypertrophic cardiomyopathy, long QT syndromes, and familial hypercholesterolemia.¹⁰ In contrast, “complex” traits are common and arise from elaborate gene-gene and gene-environmental interactions and confer risk for disease in a probabilistic manner.¹¹ By far, the vast majority of diseases in clinical practice today are complex traits and include disorders such as diabetes, myocardial infarction (MI), atherosclerotic coronary artery disease (CAD), and various cancers.

Incremental decoding of the genomic basis of complex traits will require many years of collaborative research efforts by clinicians and scientists and will never be complete.^{4, 12} This is primarily due to the sheer complexity involved in analyzing the 6 billion base pairs, over 15 million single nucleotide polymorphisms (SNPs), and hundreds of thousands of structural variants such as base pair insertions, deletions, inversions, and gene copy number variants present in the human diploid genome. To date, structural variants have not been well characterized because we are only in the early phases of whole genome sequencing. Their understanding is additionally compromised because all sequencing that is done today relies on using the human genome reference template—or resequencing, rather than de novo assembly. Beyond that there are tissue-specific important epigenetic variation such as patterns of methylation, histone modifications, and chromatin state – all of which are likely contributors to disease susceptibility.¹² Other differences, such as protein splicing and folding, differences in the individual’s metabolome, also need to be taken into account. Despite these substantial challenges, over the last four years, there has been an unprecedented stream of important genomic discoveries illuminating novel pathways involved in disease biology.¹³

These discoveries stem from several recent key advances. First, large-scale efforts have led to the identification of approximately 10 million single nucleotide polymorphisms (SNPs) that carry at least a 5% minor allele frequency (MAF) and are commonly represented in populations under study.¹⁴ In addition, there are another 10 million or more SNPs that are considered rare or low frequency because they fall below the 5% MAF threshold. In aggregate, these SNPs represent only 0.5% of the human genome, but are the most abundant form of human genomic variation.¹⁵

Notably, these SNPs are not inherited independently, but as ‘bins’ or ‘blocks’ that are in linkage disequilibrium (LD). Further, the genotype of one SNP may be sufficient to infer the genotype of all other SNPs within a given LD block (haplotype), thereby “tagging” an entire region of interest (Figure 1).¹⁵ Thus, by assaying for just 1 million of these tag SNPs, a GWAS is essentially assessing hundreds of thousands of independent haplotype blocks for disease and drug response associations. This hypothesis-free approach in scanning the human genome has yielded hundreds of reproducible disease susceptibility markers in independent cohorts involving tens of thousands of cases and controls.¹³

Figure 1. SNP’s, Tag SNP’s, and microsatellites as genomic markers.

(A) Autologous chromosome with evenly spaced microsatellites (B) Segment of DNA between microsatellite markers. Single nucleotide polymorphisms are noted (A,B,C…) within the DNA segment. Tag SNP’s (C,H,K) travel with other noted SNP’s as blocks (haplotypes) and can serve as a surrogate for these haplotypes and more importantly disease causing genes in close proximity. (C) DNA segment with alternative alleles and genomic markers of the same genes designated in part B of the figure. Note that the microsatellite markers are not as close in proximity to the genes as the noted SNP’s.

*Figure reproduced from - (Figure 2) Damani and Topol. Future use of genomics in coronary artery disease. J Am Coll of Cardiol. 2007;50:1933-40

Second, the striking reduction of costs associated with DNA sequencing have enabled targeted resequencing of genomic regions thought to be involved in disease and health.¹⁶ This has resulted in the ability to identify rare genetic variants with a minor allele frequency (MAF) of less than 5%, which complement the common susceptibility SNPs (MAF > 5%) established through GWAS.¹⁶ Now, these rare SNPs present in common, incriminated haplotype blocks by GWAS, are being assayed for along with common SNPs in more comprehensive GWAS studies of complex traits with great early success.¹⁷ An exemplary case is the recent discovery of two apolipoprotein (a) (LPA) polymorphisms – one rare and one common variant – that when present together confer at least a 250% increase in risk for CAD.¹⁷ This finding was facilitated by data generated from GWAS and resequencing of LPA in thousands of individuals with and without CAD.

Along these lines, Musunuru et al were recently able to completely sequence more than 16,000 genes in two family members affected by familial combined hypolipidemia – a Mendelian disorder marked by low cholesterol levels and lifelong protection against CAD.¹⁸ Interestingly, individuals affected by the disorder were compound heterozygotes for two rare nonsense mutations in the angiopoietin-like 3 protein (ANGPTL3). Further, in a separate cohort, a gene dose effect from the variants was observed with single allele carriers having substantially lower LDL cholesterol levels than noncarriers. Notably, ANGPTL3 is primarily expressed and secreted in the liver with inactivation of the gene leading to lower plasma cholesterol in mice. Thus, ANGPTL3 may serve as a valuable new therapeutic target for LDL reduction and prevention of CAD in the future.

Third, novel approaches in genetic association studies using “Mendelian randomization” principles have assisted in firmly establishing the link between key intermediate phenotypes (inflammation, plasma lipoproteins, glucose levels) and many complex traits.^{17,19, 20} This approach leverages the fact that genes undergo random assortment when forming gametes and are transferred in unbiased fashion from parent to offspring during the time of conception.^{20, 21} Thus, a Mendelian randomization study that assesses the impact of a gene product on a biologic outcome is in essence naturally providing the same principle of a randomized trial but in some ways may be considered superior. Potential confounders of these studies include additional potential causative genes that may be nearby and in LD with gene variants being studied, as well as population admixture from ancestral populations that carry different risks for disease and different genotypes. However, both of these confounders can be easily accounted for by proper design of such studies.

An ideal application of Mendelian randomization principles is well illustrated in a recent publication by Zacho and colleagues on C-reactive protein (CRP) – a well-known marker of inflammation strongly tied to CAD.²² Notably, prior to this study, the question of whether CRP was simply a marker for CAD or actually contributed to disease causation remained unanswered. In order to answer this question, the investigators assessed two independent populations involving over 50,000 subjects and found that four defined CRP gene polymorphisms with a known impact on plasma CRP levels did not predict IHD. Conversely, the relationship between elevated CRP and IHD remained strong, thereby indicating that lifelong elevation of CRP levels as indicated by genetic polymorphisms does not confer risk for IHD and that CRP is simply a marker of and not likely a contributor to CAD risk.

These seminal findings recently took on greater meaning with the publication of the high profile JUPITER (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) trial.²³ In JUPITER, individuals with elevated CRP levels treated with rosuvastatin had a 44% reduction in IHD events that was independent of lipid status. Unfortunately, a control group with low CRP and low LDL was not included in the study. Nevertheless, when reconciling JUPITER with genetic data on CRP, it is clear that the benefit from rosuvastatin is most likely secondary to inadvertent decrease in inflammation and not from a direct reduction in plasma CRP levels. Moreover, the genetic data on CRP further validates the previously touted pleiotropic effects of statins with respect to inflammation reduction and potentially supports the use of statins to prevent CAD in high-risk patients with elevated CRP levels and normal plasma lipoproteins.

Lipoprotein A

The lipoprotein (a) [Lp (a)] molecule is composed of a low-density lipoprotein that is covalently linked to the plasminogen-like glycoprotein, apolipoprotein (a) (Figure 2).²⁴ Notably, Lp (a) has been regarded as a putative cardiovascular risk factor for over four decades.²⁴ However, the extent of its contribution to IHD has been controversial and unsettled.²⁵ Now, three independently conducted genomic studies have underscored the importance and clearly defined the role of Lp (a) in coronary artery disease and myocardial infarction (MI).^{17, 19, 26}

Figure 2. Lipoprotein (a) Molecule.

Lipoprotein(a) consists of an LDL particle and a glycoprotein molecule, Apo(a), attached to the ApoB-100 moiety of the LDL particle through a disulfide bond. Apo(a) size is determined by the number of kringle repeats.

*Figure reproduced from - Permission needed (Figure 1) Danesh J and Erqou S (2009) Lipoprotein(a) and coronary disease—moving closer to causality Nat Rev Cardiol doi:10.1038/nrcardio.2009.138

The initial confirmatory study was an extensive three-stage GWAS.²⁶ In the first stage, the investigators assayed over 500,000 SNPs in 2000 cases and controls. Subsequently, the most significant SNPs (p < 10⁻⁵) were then validated in a second cohort consisting of 875 CAD cases and 1644 controls. Interestingly, a single haplotype on chromosome 6q26-q27, which encompasses the LPA gene, remained highly significant and was subsequently validated in a third stage. The final population-adjusted analysis of all three stages demonstrated that two LPA haplotypes were highly significant for CAD (p = 1.0 × 10⁻¹³, p = 1.0 × 10⁻¹⁵) with odds ratios (OR) of 1.2 and 1.8 respectively.

In a parallel study that utilized Mendelian randomization principles, Kamstrup et al. found that common kringle IV type II (KIV-2) copy number polymorphisms (CNPs) in LPA, which are known to impact both apolipoprotein (a) particle size and Lp (a) plasma levels, correlated with a risk for MI (OR = 1.2) independent of plasma Lp (a) levels.^{19, 27} Further, in a subgroup analysis of a 9000 patient prospective cohort with over 16 years of followup, they demonstrated that at-risk KIV-2 CNP carriers had an impressive 150% increase in risk for coronary events.

In the third and final related study, investigators used a novel cardiac gene chip and assayed for 48,742 common and rare SNPs from several candidate genes in over 7,000 CAD cases and controls.¹⁷ Remarkably, the LPA locus on 6q26-27 that was originally identified in the GWAS noted above most strongly correlated with CAD.²⁶ Carriers of a single common variant rs10455872 (MAF = 7%) or an independent rare variant rs3798220 (MAF = 2%) had an OR of 1.7 and 1.92 respectively, or a striking 2.5 when carrying both at-risk variants (Figure 3). Moreover, these variants were shown to tag individual LPA alleles with fewer KIV-2 repeats, thereby confirming data linking smaller apolipoprotein (a) isoforms to a heightened risk for CAD.^{27, 28}

Figure 3. Association of the LPA Genotype Score with the Lp(a) Lipoprotein Level and the risk of Coronary disease in the PROCARDIS cohort.

The odds ratios (squares, with the size inversely proportional to the sampling variation) are for the association of the LPA genotype score (no variant alleles, one variant allele, or two variant alleles) with the risk of coronary disease, as measured with the use of “floating absolute risks” which summarize the sampling variation for the three genotype scores without the selection of an arbitrary baseline genotype score. The vertical lines indicate 95% confidence intervals.

*Figure reproduced from - Permission needed. Figure 3. Clark et al. NEJM 361;26 December 24, 2009

Most recently, the European Atherosclerosis Society (EAS) has recommended routine screening for Lp (a) levels and treatment with niacin for individuals with plasma Lp (a) levels greater than 50mg/dl.²⁰ However, the findings from the genomic studies covered here suggest that LPA screening for susceptibility variants may be a better initial screening tool for many reasons. Current widely used Lp (a) assays do not fully account for important qualitative features including apo (a) particle size, the number of KIV-2 repeats, and are highly variable with respect to sensitivity and specificity. Further, the consistent and reproducible increase in risk for CAD in at-risk LPA variant carriers far surpasses the highly variable and at times conflicting results seen in the substantial proportion of epidemiologic studies conducted to date on plasma Lp (a) levels.^{25, 27, 29–31} In addition, the genetic effect of LPA variants on CAD appear to be independent of age and gender.

9p21

To date, one of the most interesting findings on the genetic basis of CAD has been the discovery of a novel MI risk locus on chromosome 9p21.³² In 2007, 4 GWAS involving over 50,000 cases and controls and ten independent populations reported a robust link between several SNPs on chromosome 9p21 and MI.^{16, 33–35} Importantly, about 25% of the populations studied are homozygous carriers of the at-risk variants and harbor close to a 2-fold increase in risk for MI. Another recent study has illustrated that a significant gene dosage impact exists with the number of 9p21 risk alleles directly correlating with CAD severity.³⁶ Further, these SNPs have also been shown to predict more progressive CAD, abdominal aortic aneurysm (AAA), and remarkably, intracranial aneurysm.^{37, 38} This surprising new finding now raises the possibility that functional effects from 9p21 may stem more from mechanisms related vascular remodeling rather that atherosclerosis.

In another study of over 30,000 AAA cases and healthy controls, the 9p21 locus was once again linked to AAA (p<1.7×10⁻⁷).³⁹ However, a new susceptibility allele (rs7025486) in an intron of the gene DAB2IP – a cell cycle regulator previously linked to many cancers – was also significantly associated with AAA (OR=1.24, p=1.8×10⁻⁹). Surprisingly, this same allele also conferred susceptibility to early onset myocardial infarction, peripheral arterial disease, and pulmonary embolism (ORs=1.18, 1.14, 1.2), thereby further underscoring a common genetic thread between previously unlinked vascular phenotypes.

Currently, the exact mechanism by which the 9p21 locus exerts its impact on these various vascular phenotypes remains unclear. Notably, the at-risk variants lie in a ‘desert’ region of the genome with the closest annotated genes, CDKN2A and CDKN2B, being greater than 100,000 base pairs (bps) away.³² These tumor suppressor genes are cyclin-dependent kinase inhibitors with established regulatory roles in cell-cycle progression and documented anti-proliferative effects.^{32, 40} Recently, Visel et al. have provided preliminary evidence linking 9p21 SNPs to altered CDKN2A and CDKN2B gene expression.⁴¹ By deleting a 70-kilobase (kb) region from mouse chromosome 4, which is orthologous to the 58 Kb region at 9p21.3 in humans, they found a dramatic reduction in expression of Cdkn2a and Cdkn2b in the knockout mice. Further, they observed a doubling in smooth muscle cell proliferation, enhanced weight gain, a substantially larger number of tumors, and an overall increased death rate in the mutant mice when compared to wild-type (wt) mice. Interestingly, there was no evidence for increased atherosclerotic plaque burden in the aortas of the knockout mice.

Although highly informative, these findings raise additional questions regarding the 9p21 risk locus.³² For example, why was there no observed increase of atherosclerotic burden in the mutant mice, when this has clearly been the case in humans? Overall, it appears that the murine model may not be fully adequate for precisely defining the mechanistic impact of 9p21 variants on multiple arterial phenotypes. This theory is supported by the only moderate homology (less than 50%) present between the human and mouse 9p21 locus. Additionally, the effects from fully deleting a 70kb region, as Visel et al. did in their mouse model, versus carrying a few 9p21 SNPs as humans do, could yield entirely different phenotypes. Thus, the mechanistic basis behind 9p21 associated vascular disease remains unclear. Nevertheless, this genomic region continues to be intensely studied with important findings on the link between 9p21 variants and human CAD certain to surface in the years ahead.

On a similar note, potential genomic markers of sudden cardiac death are also emerging. Recently, Bezzina et al. assessed over 500,000 SNPs in 515 individuals with ventricular fibrillation and MI and 457 controls with MI alone.⁴² Strikingly, the most significant associated SNP, rs2824292 (p=2.2×10⁻¹⁰) was present in over 50% of cases and conferred an impressive 180% increase in risk for ventricular fibrillation. Perhaps most compelling is that the susceptibility allele is located near the gene CXADR, which encodes the coxsackievirus and adenovirus receptor protein. This transmembrane tight junction protein has been previously implicated in both virus-mediated cardiomyopathies and in sudden cardiac death in candidate gene-based studies. Additional mechanistic studies will need to be performed in order to precisely define the basis of this variant’s impact on ventricular tachyarrythmias. However, the genomic underpinnings of ventricular fibrillation illuminated by this study may lead to better predictive algorithms and preventative measures in MI patients at high risk for sudden cardiac death.

Apolipoprotein E

Apolipoprotein E (APOE) polymorphisms have been the most widely studied genetic risk factors in humans due to their well-established links to Alzheimer’s dementia, dyslipidemia, and CAD.^{43, 44} Carriers of the E4 allele, which represent approximately 20% of the population of European ancestry, have higher circulating cholesterol levels than their common E2 and E3 counterparts.^{43, 45} Most importantly, these carriers are at increased risk for CAD.^{43, 45}

First discovered in 1977, APOE4 polymorphisms have been firmly linked to dyslipidemia and CAD in over 50 studies.⁴³ Several GWAS have now confirmed this association.⁴⁶ Further, a decade ago, investigators found that ApoE4 carriers had reduced survival rates after MI, which was abolished by treatment with simvastatin.⁴⁷ Notably, the derived benefit from statin therapy did not relate to greater lipid lowering, which, once again, provides further evidence for a pleiotroic effect of statin use.

Based on these data, it may be possible to use the APOE4 genotype for CAD risk prediction and institution of statin therapy in patients with moderate risk factors for IHD. Such a strategy would be suitable for prospective study, or if the data were available from large scale, placebo-controlled statin trials with multi-year follow up, validation might be supported. However, the psychosocial concerns of assessing APOE status for CAD and the need for disclosure of risk for Alzheimer’s has, at least in part, limited its use. Importantly, a recent study demonstrated that disclosing APOE genotype data to adult children of an affected Alzheimer’s disease parent did not result in significant anxiety or psychological distress.⁴⁸ Thus indicating that current concerns related to APOE4 testing and disclosure may be overblown.

In contrast to the APOE4 pharmacogenetic story, recent claims that a KIF6 gene variant predicts statin response have been dubious.⁴⁹ Surprisingly, over 150,000 KIF6 “Statincheck” tests have been ordered based solely on 3 retrospective candidate gene studies that demonstrated a modest increase in CAD risk (OR 1.1-1.5)^50–52 and another study that found statin therapy abrogated this enhanced risk.⁵³ However, several troublesome aspects of KIF6 data exist. First, unlike APOE, KIF6 is not expressed in the vasculature and has no known biologic relevance in dyslipidemia or CAD.⁵⁴ Second, none of the over 10 GWAS on lipids or CAD have linked KIF6 to either phenotype.^{26, 33–35, 46, 55–59} Third, a recent well-conducted meta-analysis in over 17,000 individuals found no link between CAD and KIF6.⁶⁰ Thus, while many of the examples herein exemplify the promises of genomic medicine, the KIF6 story should serve as a valuable reminder of the pitfalls present when prematurely adopting a genetic test in clinical practice.

Antiplatelet Pharmacogenomics

Adjunctive aspirin and clopidogrel use in the management of patients with acute coronary syndromes (ACS) and those receiving coronary stents has substantially reduced the risk for MI, stent thrombosis, and death.^61–63 However, in 2006, a variable antiplatelet effect with clopidogrel was observed in hepatic cytochrome (CYP) 2C19 loss-of-function variant carriers.⁶⁴ Now, several large studies involving thousands of patients have confirmed that genetic resistance to clopidogrel is prevalent even in patients with acute coronary syndromes.^65–68 Importantly, the at-risk variants result in reduced clopidogrel active metabolite formation, diminished antiplatelet effect, and a greater than 3-fold increase in risk for stent thrombosis, MI, and death (Figure 4).^{65–67, 69, 70} Additionally, recent studies have also identified a common CYP2C19 gain-of-function variant that confers a 2-fold increase in risk for bleeding (Figure 5).⁷¹ Remarkably, these gain and loss-of-function variants are highly common with a third of Europeans and close to half of those with African and Asian ancestry harboring the at-risk alleles.

Figure 4. Event-free Survival Over 1 year Follow-up in Sinai Hospital of Baltimore Patients Treated with Clopidogrel Following Percutaneous Coronary Intervention, Stratified by CYP2C19*2 Genotype.

Postdischarge ischemic events included myocardial infarction, ischemic stroke, stent thrombosis, unplanned revascularization, and cardiovascular death. All analysis adjusted for age, sex, and race. Patients were further stratified into those who were taking clopidogrel when the event occurred or at 1 year of follow-up and those who were not. All analysis adjusted for age, sex, and race. For all patients, hazard ratio (HR)=2.42 (95% confidence interval [CI] 1.18-4.99;P=0.02); for patients taking clopidogrel at the time of event, HR=3.4 (95% CI 1.36-8.46; P=0.004); for patients not taking clopidogrel, HR=1.39 (95% CI 0.39-4.88; P=0.60)

*Figure reproduced from *Permission Needed (Figure 4) Shuldiner et al. JAMA. 2009;302(8):849-858

Figure 5. CYP2C19*17 Genotypes and Incidence of TIMI Bleedings.

*Figure reproduced from *Permission needed (Figure 2) Sibbing et al. Circulation.2010;121:512-518

Similar to the APOE and LPA variants, the CYP2C19 variants were originally identified through hypothesis driven candidate gene studies. Now, Shuldiner et al. through GWAS have incontrovertibly confirmed that the predominant mediator of genetic resistance to clopidogrel is the hepatic CYP2C19 locus.⁶⁹ In order to determine this, the investigators first measured platelet aggregation at baseline and within one hour following the last dose of clopidogrel on day 7. Subsequently, over 400,000 SNPs were simultaneously assessed in a GWAS of platelet reactivity. Not surprisingly, the region most significantly associated with clopidogrel response clustered around the CYP2C19 locus. Moreover, CYP2C19 variant carriers had a striking 345% increase in risk for stent thrombosis, along with heightened risk MI, and death, which is consistent with previous candidate gene study results.

The clopidogrel story represents the prototypical scenario for individualizing medicine based on key pharmacogenomic information. It is one of the most highly prescribed drugs in the world and is used routinely to prevent stent thrombosis, MI, and death in the over 1 million people receiving coronary stents in the U.S. annually.⁷² Most importantly, alternatives to standard clopidogrel dosing are readily available for individuals resistant to the drug. These alternatives include the addition of cilostazol, or the use of alternative P2Y12 receptor blocking anti-platelet agents such as prasugrel or ticagrelor.^73–76 Further, genotyping for at-risk CYP2C19 variants can be performed in patients undergoing coronary stenting and enables identification of those individuals at greatest risk for thrombotic complications. These high-risk individuals can then be closely monitored during the post-stenting period with platelet function testing to ensure adequate antiplatelet response. This is especially important given that stent thrombosis, although only occurring in less than 2% of patients, carries a mortality rate of over 40% and occurs most frequently within hours of stenting.^77–79 Consistent with this line of reasoning, a recently released FDA boxed warning has recommended such an individualized approach to antiplatelet therapy in patients receiving coronary stents.⁵⁵

Another recent key discovery in antiplatelet pharmacogenomics relates to the identification of a rare SNP in LPA that predicts with high accuracy those individuals most likely to benefit from aspirin therapy.^{80, 81} Chasman et al. reported that the nonsynonymous SNP, rs3798220 (MAF = 3.5%), which encodes for an isoleucine to methionine substitution (Ile4399Met), resulted in an 8-fold increase in Lp (a) plasma levels and a corresponding 2-fold increase in risk for MI and stroke.⁸² Interestingly, this heightened risk was completely abrogated in the Ile4399Met carriers by aspirin therapy. This report is especially timely given recent meta-analysis data showing an exceedingly small benefit and substantial risk for bleeding with the nondiscretionary use of aspirin in a primary prevention setting.⁸² Moreover, Clarke et al. (see above) recently found that the same rare SNP (rs3798220) conferred a similar relative risk for CAD in a separate study involving over 7,000 CAD cases and controls.¹⁷ Unfortunately, this study did not test the aspirin hypothesis.

Along these lines, in a recent publication of the first clinically annotated whole-genome sequence, over 60 pharmacogenetic variants with an immediate impact on drug efficacy and toxicity were detected in an otherwise healthy 40 year-old male geneticist who had sequenced his own genome.⁸³ Notably, his variants included the at-risk LPA variant (rs3798220), which resulted in a recommendation by the individual’s physician to initiate aspirin therapy for primary prevention of CAD.

Caution in Interpeting Negative SNP Profiling Studies

Earlier this year, Paynter et al. reported on the prognostic capability of 101 SNPs linked to CAD through previous GWAS.⁸⁴ An integrated “genetic risk score” for CAD based on the carrier status of these 101 SNPs did predict a heightened risk for CAD, but failed to incrementally add to models based on traditional risk factors. However, there are several important limitations of this study that require addressing. First, all 101 variants received equal weighting in their model, despite clear differential and more potent disease causing effects from variants such as those in the 9p21 locus, which promotes a washout of any informative potential of the SNPs assessed. Second, many of the SNPs were from GWAS that used variable definitions for CAD cases and healthy controls. Third, some of the most important known CAD SNPs, such as in LPA, were not included. Fourth, a family history of premature CAD continued to be predictive for the development of CAD even after adjustment for the traditional risk factors, indicating important genetic underpinnings of this disease that have yet to be fully defined. Finally, a recent study has indicated that the simultaneous assessment of all disease associated SNPs proportionally increases the prediction of disease heritability.⁸⁵ Thus indicating that current genetic risk scores must include all relevant rare and common SNPs (including those yet to be discovered) with adequate weighting in order to accurately portray the individual’s true genetic risk for disease. To date, no such model has been developed.

Future Directions

Over the last four years, the breakneck pace of discoveries into the genomic underpinnings of complex traits has largely been enabled by genome-wide assessment of common SNPs. Unfortunately, substantial portions of the heritability of many complex traits including CAD remain missing. This missing heritability or “dark matter” of the genome will likely be unraveled in incremental fashion in the years ahead through the identification and validation of rare susceptibility SNPs and disease causing structural variants in properly designed, large-scale, whole-genome sequencing studies.¹² Moreover, comprehensive delineation of tissue-specific epigenetic characteristics such as DNA methylation, histone modification, and chromatin state will provide additional insight into biologic mechanisms of disease. However, access to arterial and myocardial tissue for epigenetic studies will be limited, so less-invasive approaches to access these tissues will be needed. Recent advances in rare cell biology that sequester circulating endothelial cells from the peripheral blood are exemplary of such non-invasive approaches.⁸⁶ Additionally, the concurrent implementation of proteomic and metabolomic technologies in genomic studies will provide vital information on the biologic alterations present in various disease states, while also serving to uncover novel therapeutic targets.

In summary, the emerging applications in CAD genomics and pharmacogenomics covered in this review represent a preview of the transformative discoveries that will continue to surface in the months and years ahead. The challenge will be in determining when and how to implement such data into routine clinical practice. The clopidogrel and LPA scenarios represent two areas where the evidence threshold for individualizing therapy based on genotype data has been clearly surpassed. However, it will ultimately be the individual clinician that will decide when and how to use this data. Hence, systematic education and training of physicians on the benefits and drawbacks to genomic medicine must be performed in order for the full potential of the future era of individualized medicine to be realized.

Table 1.

Recent Advances in Ischemic Heart Disease Genomics

Gene or Locus	Condition	Experimental Methods	Effect Size (OR) (single allele)	Effect Size (OR) (multiple alleles)	Ref.
9p21.3 (CDKN2A, CDKN2B)	MI AAA Intracranial Aneurysm PAD	GWAS	1.2 – 1.4 1.31 1.29 1.14	1.6 – 2.0 1.74 1.72 —	32‐34, 37,39
LPA	CAD Enhanced Aspirin Response	GWAS, Candidate Gene, Resequencing	1.7 - 1.9 2.2ˆ	2.5 – 4.0 —	17,80
APOE	CAD, Dyslipidemia	GWAS, Candidate Gene, Resequencing	1.1 – 1.4	1.2 – 1.6	43,44
CYP2C19	Stent Thrombosis (*2 ‐ *5 alleles) Bleeding (*17 allele)	GWAS, Candidate Gene, Resequencing	3.5 1.8	4.6 3.2	69,70
21q21 (CXADR)	Ventricular Fibrillation	GWAS	1.5 – 1.8	—	42
DAB2IP	Early Onset MI AAA PE PAD	GWAS	1.18 1.21 1.20 1.14	— — — —	39

^ˆOR represents the increased risk for CAD in rs3798220 carriers. The enhanced risk was completely abrogated by aspirin therapy