Personalized medicine in coronary artery disease: insights from genomic research.

Prior clinical studies have demonstrated that a family history of coronary artery disease (CAD) is associated with future cardiovascular events. Although there are several Mendelian disorders that are associated with CAD, most common forms of CAD are believed to be multifactorial and the result of many genes with small individual effects. The identification of these genes and their variation would be very helpful for the prediction, prevention, and management of CAD; linkage analysis or candidate gene case-control studies have been largely unsuccessful. On the contrary, recent advances in genomic techniques have generated a large amount of deoxyribonucleic acid (DNA)-based information. The link between CAD and inflammation and biological pathways has been highlighted. In particular, several genome-wide association studies have replicated a novel gene marker on chromosome 9p21. The information gained from genomic studies, in combination with clinical data, is expected to refine personalized approaches to assess risk and guide management for CAD. Genetic risk scores derived from several functional single nucleotide polymorphisms (SNPs) or haplotypes in multiple genes may improve the prediction of CAD. Despite the complexity of CAD genetics, steady progress is expected.

Introduction

Atherosclerotic coronary artery disease (CAD) is a major health problem worldwide. Intensive studies have been reported and are currently ongoing for appropriate risk assessment, prevention, and treatment of CAD. A small proportion of CAD cases may be attributed to rare, monogenic effects with high penetrance, but most CAD cases are multifactorial in etiology. Development of CAD is only roughly defined by traditional cardiovascular risk factors. In the Framingham Heart Study, a parental history of premature vascular disease was a strong risk factor (adjusted odds ratio=1.7-2.0).1) The INTERHEART study confirmed that a family history of CAD is an important risk factor (adjusted odds ratio=1.5).2) However, because the Framingham risk scoring, which is the most widely used risk assessment tool, does not include family history, thus refining a CAD prediction model is needed. At the same time, with the advances in genomic research, the potential for translating genomic information to clinical practice is greater now than ever before.

During the last 50 years, we have witnessed a remarkable progression of genomic studies, from the initial description of the deoxyribonucleic acid (DNA) double helix by Watson and Crick in 1953 to the completion of the Human Genome Project in 2003. Indeed, genomics is now being used to identify newly discovered pathogens in an effort to stratify patients according to the risk of a disease by genotyping, and to assess the effects and guide therapy. Technological breakthroughs in genetics and genomics have facilitated an understanding of disease pathogenesis.3) The Human Genome Project and International HapMap Project have provided a large amount of DNA-based information. In addition, current technology can assess the expression of thousands of genes from different tissues.

The term, personalized medicine, indicates disease prevention and treatment based on knowledge of individual genetic susceptibilities. By combining information gained through genomic methodologies with traditional methodologies, we expect to advance our ability to assess and treat CAD better. In particular, because DNA-based genetic markers remain unchanged throughout one's life, genetic tests for disease prediction could be performed at a young age to promote early intervention. This review considers basic concepts and approaches, recent genomic studies on CAD, and the clinical implications for personalized medicine.

Basic Concepts and Approaches in Genomic Research

The 20,000-25,000 protein coding genes that comprise the human genome represent only 30% of its sequence. The remainder is intergenic sequences that may contain elements for the regulation of gene expression. In a typical human gene, 5% of the sequence is coding exons that are partly translated into a protein, and the remainder is composed of introns and regulatory regions. The most common human sequence variations are differences in individual base pairs, termed single nucleotide polymorphisms (SNPs). Other variations consist of short or long repetitions of the same motif, such as mini- and micro-satellites, insertion or deletions, and variants that affect large chromosomal regions.

Mendelian versus complex inheritance

Genetic variation predisposing to CAD spans from rare and highly deleterious mutations related to Mendelian disease to common polymorphisms with weak effects, that alone or in combination, are associated with the risk of common diseases (common variants-weak effect-common disease model).4) Rare deleterious mutations (e.g., mutations that cause familial hypercholesterolemia) are related to an important risk for CAD; however, the impact at the population level is low. In contrast, frequent polymorphisms, such as the apolipoprotein E polymorphism, may have a population impact despite the weak effect at the individual level. Because polymorphisms have common alleles, numerous combinations of susceptibility alleles at several loci in an individual are possible. Furthermore, some of the combinations of susceptibility alleles can affect the risk for CAD in a way that cannot be predicted from the separate effect of each variant. Since this is the major problem when one characterizes the genetics of complex traits, it is rational and justified to explore systems of genes rather than a single gene.

Almost all polymorphisms identified by association studies have a modest effect, even when combined. Furthermore, there is a marked disparity between the extent of familial aggregation and those collectively accounted for by known variants in common diseases. One emerging hypothesis suggests that a significant proportion of this unknown area of heritability can be explained by low-frequency intermediate penetrance variants. To date, however, these variants have been unyielding to conventional gene-seeking approaches (Fig. 1).5)

Fig. 1

Variants with different frequencies, penetrance, and disease susceptibility. GWA: genome-wide association.

Genetic linkage and candidate gene association studies

Two basic approaches have been used to discover genetic influence on diseases:6) genetic linkage and candidate gene association studies. Genetic linkage studies use genetic and phenotypic data from families. Linkage analysis intends to localize genomic regions that might contain genes influencing a trait. On the contrary, candidate gene association studies compare genotypic frequencies between cases and control groups. A statistical difference in frequencies between the groups suggests that a genotype is associated with the trait. Genetic association studies rely on the existence of linkage disequilibrium (LD) among polymorphic sites which are physically close within the genome. It is referred to as LD when two alleles at different loci occur together on the same chromosome more often than would be predicted. Even if a causal involvement of a polymorphism in the disease is not proven, the association may be suggested by a measured proxy polymorphism in LD. Because of LD identified in most regions of the genome, the combination of alleles at neighboring SNPs (haplotypes) generates less diversity than would be expected. By re-sequencing the genome of 270 individuals from populations from African, Asian, and European ancestries, the HapMap Project has demonstrated a set of SNPs that tag most common haplotypes in humans. This resource can be used to search for polymorphisms associated with susceptibility to common diseases. Recent technological advances have allowed relatively low-cost dense genotyping arrays of 500,000 or more SNPs that cover most of the genome. For complex diseases, by typing hundreds of thousands of variants, genome-wide association studies provide excellent power. Specifically, genome-wide association studies are expected to help to resolve the questions regarding contribution of multiple variants to common diseases.

Genetic Susceptibility to Coronary Artery Disease

Until recently, most studies on the genetics of CAD were based on candidate genes. Accordingly, reports of genetic variations in CAD came from established risk factors, such as hyperlipoproteinemia, hypertension, and thrombosis. Nevertheless, relatively few polymorphisms in candidate genes were replicated for CAD.

Rare mutations causing premature coronary artery disease

Disease-causing mutations are defined as rare allelic variants that can markedly increase disease risk. Molecular genetic studies of rare, Mendelian forms of CAD have identified several mutations that cause premature CAD (Table 1). Many of the mutations affect the levels of low density lipoprotein (LDL) and high density lipoprotein (HDL)-cholesterol. Mutations in the LDL receptor gene (LDLR), which is associated with the majority of patients with familial hypercholesterolemia, is a classic example. Gain-of-function and loss-of-function variants of the proprotein convertase subtilisin/kexin 9 (PCSK9) gene, which modulates LDL receptor levels, has been reported to be associated with a reduction of LDL and CAD.7) Both variants are inherited in an autosomal dominant fashion with increased plasma LDL-cholesterol. Tangier disease is characterized by the absence of HDL-cholesterol and low plasma apoA-1. This disease results from mutations in the ATP-binding cassette transporter type 1 (ABCA1) gene and is associated with premature myocardial infarction.

Table 1

Rare Mendelian diseases that cause premature CAD

CAD: coronary artery disease, LDLR: low-density lipoprotein receptor, PCSK9: proprotein convertase subtilisin/kexin type 9, APO: apolipoprotein, LDL-C: LDL-cholesterol, ABCA 1: adenosine triphosphate (ATP)-binding cassette transporter type 1

Genetic linkage studies

Before the recent introduction of high-density SNP genotyping arrays, linkage analysis, also called positional cloning, using families with multiple affected individuals was considered the most useful way to identify susceptibility loci. Very few genes have been identified from genetic linkage studies, and the lack of replication emphasizes the difficulties in performing and interpreting these studies. The identification of the arachidonate 5-lipoxygenase-activating protein (ALOX5AP) gene encoding 5'-lipoxygenase activating protein (FLAP) represents one success in detecting a novel candidate from a linkage study.8) A few genome-wide linkage studies conducted for subclinical forms of atherosclerotic vascular disease have exhibited linkage evidence on chromosomes 109) and 12.10) Among the observed linkage regions on chromosomes 1, 2, 3, 13, 14, 16, 17, and X, only the linkage region detected by the PROCARDIS consortium on chromosome 17 was clearly replicated.11) It can be inferred from genome-wide linkage studies that there are no CAD loci with strong effects. In the future, putative loci that are implicated from genome-wide linkage scans are likely to help select regions for large-scale association studies.

Candidate gene association studies

Candidate gene studies have attempted to investigate variations in genes that are already implicated in the pathophysiology of disease. A large number of candidate gene association studies have been performed for myocardial infarction and atherosclerotic vascular disease, but few studies have been replicated. In addition, because risk ratios have largely been modest in association studies, meta-analyses often have been necessary to obtain statistical significance (Table 2). These analyses were shown in a large number of studies evaluating variants of the apolipoprotein E (APOE),19) plasminogen activator inhibitor 1 (PAI1),22) angiotensin-1 converting enzyme (ACE),20) and methylene tetrahydrofolate reductase (MTHFR) genes.12) Recently, high-throughput genotyping has been used in case-control studies. Variations in lymphotoxin-alpha (LTA),23) gap junction protein, a-4 (GJA4 or connexin 37),24) matrix metalloproteinase-3 (MMP3 or stromelysin-1),24) arachidonate 5-lipolygenase (ALOX5) genes,24) and thrombospondin (THBS) genes25) are associated with myocardial infarction (MI). Among many studies on subclinical atherosclerosis, the 5A/6A polymorphism of the matrix metalloproteinase (MMP3) gene26) revealed a consistent association with carotid intima-media thickness. Two candidate gene association studies have provided evidence for a role of the leukotriene pathway in CAD susceptibility. Promoter variants in the ALOX5AP gene are associated with increased carotid intima-media thickness.27) Another component of the leukotriene pathway which is considered a candidate gene for CAD is the leukotrine A4 hydrolase (LTA4H) gene. A recent study has shown an association between LTA4H and CAD and confirmed the importance of genetic variation in an inflammatory pathway gene in the development of CAD.

Table 2

Candidate gene variation for atherothrombotic cardiovascular disease reported in recent meta-analyses of association studies

MTHFR: methylene tetrahydrofolate reductase, CETP: cholesteryl ester transfer protein, eNOS: endothelial nitric oxide synthase, APOE: apolipoprotein E, ACE: angiotensin converting enzyme, PAI1: plasminogen activator inhibitor

Genome-wide association studies and the 9p21 locus

Genome-wide association studies that use large numbers of tagging SNPs have now come to replace linkage studies. In 2007, the first generation of genome-wide association studies using high-density, SNP genotyping arrays demonstrated several genetic variants that are significantly associated with CAD (Table 3).28-31) The design of these studies was quite simple. The frequencies of a large number of SNPs genotyped on arrays were compared between cases and controls, and the sites that showed significant differences were then validated in independent samples. There was no prior assumption of candidate genes, rather the search for associated genes throughout the genome was unbiased.32)

Table 3

Susceptibility loci identified in four genome-wide association studies of CAD

CAD: coronary artery disease, SNP: single nucleotide polymorphism

Currently, the most robust susceptibility locus for CAD is on chromosome 9p21.28-31) The association of this locus and CAD has been confirmed in independent Caucasian populations comprising greater than 55,000 individuals. In all of these studies, 9p21 was confined to a 58,000 bp region and the coronary risk was independent of traditional risk factors. As there is no known gene coding for a protein in the 58,000-bp region, the neighboring cyclin-dependent kinase inhibitor 2A (CDKN2A), cyclin-dependent kinase inhibitor 2B (CDNK2B), and methylthioadenosine pohosphorylase (MTAP) genes were initially suggested as candidates. The PROCARDIS consortium conducted an analysis of this CAD/type 2 diabetes susceptibility locus on chromosome 9 and demonstrated co-location of a high-risk haplotype with ANRIL, a large antisense non-coding RNA gene.33) It was shown to be expressed in tissues and cells affected by atherosclerosis. ANRIL has now been identified as the most probable candidate gene for the 9p21 locus. The 9p21 58,000 bp region is a target of major research and the mechanism whereby 9p21 causes CAD may imply a new pathway for cardiovascular risk independent of traditional risk factors.

Focusing on genes in biological pathways in coronary artery disease

A useful approach to search for candidate gene variants for CAD is to identify biological pathways in the development of CAD. Four interdependent pathways have been described:34) lipoprotein handling, endothelial integrity, arterial inflammation, and thrombosis (Fig. 2). First, as familial hypercholesterolemia is related to a LDL receptor deficiency, abnormal lipoprotein handling is known to promote atherosclerosis. In addition, protective and at-risk alleles of APOE correlate with the effects on LDL-cholesterol.19) PCSK9 loss-of-function mutations can reduce the LDL level and remarkably limit the development of CAD.7) Regarding HDL-cholesterol, rare variants of multiple genes are associated with low HDL levels. Specifically, non-synonymous variants in three genes (ABCA1, APOA1, and LCAT) were associated with the population variance of a low HDL. A series of human studies have shown how a genetic variation of apolipoprotein A-V (APOA5), in combination with apolipoprotein C-III (APOC3), influences triglyceride levels. Several recent studies support the link between the APOA1-APOC3-APOA4-APOA5 cluster and familial combined hyperlipidemia through a mechanism that involves alleles in the APOA5 and APOC3 loci.35) Another positional candidate gene, USF1 (upstream transcription factor 1), has shown evidence of an association with several familial combined hyperlipidemia phenotypes. This gene encodes a transcription factor that is implicated in the regulation of several apolipoproteins and other proteins.36) Secondly, endothelial integrity is an important factor that contributes to the pathophysiology of CAD. The thrombospondin family of extracellular matrix proteins, which affects endothelial cell adhesion and proliferation, was identified to be possibly associated with premature MIs25) and the association has been confirmed by a subsequent study.24) Connexin 37 is a gap junction protein; the C1019T gene variant has been identified to have an association with MIs. Wang et al.37) identified a variation in the myocyte enhancer factor 2a (MEF2A) gene in a family with MIs; the association with MIs has been confirmed by a subsequent study. Third, the arterial inflammation process has been implicated in several studies. LTA, lectin galactoside-binding soluble 2 (LGALS2),23) the leukotriene pathway involving ALOX5AP and LTA4,8) and the histocompatibility factor (MHC2TA) promoter, A168G SNP,38) are known to be related to inflammatory diseases, such as rheumatoid arthritis, multiple sclerosis, and MIs. Fourth, thrombosis has been implicated in CAD in previous studies. In a comprehensive meta-analysis, the 1691A variant of factor V (FV) and the 20210A variant of prothrombin were shown to be associated with CAD. The platelet glycoprotein receptor variants, vesicle-associated membrane protein 8 (VAMP-8) and PAI-1, have also been implicated as risk factors for CAD.39)

Fig. 2

Biological pathways implicated in CAD pathophysiology and examples of genes. PCSK9: proprotein convertase subtilisin/kexin 9, APOA5: apoliprotein A-V, USF1: upstream transcription factor 1, CAD: coronary artery disease, FLAP: 5'-lipoxygenase activating protein, OX40L: OX40 ligand, MEF2A: monocyte enhancer factor 2a, GP: glycoprotein, LDL: low-density lipoprotein-cholesterol, HDL: high-density lipoprotein-cholesterol, TG: triglyceride, EC: endothelial cell.

Studies performed recently in the Korean population

During the last decade in Korea, there has been remarkable progress in the field of genomic research. Many genomic studies on CAD were candidate gene association studies.

Several studies involving SNPs with CAD risk factors have been published. Park et al.40) showed a relationship between a cholesteryl ester transfer protein (CETP) Taq1B polymorphism and HDL-cholesterol levels and CAD. I405V, another CETP polymorphism, was reported to have a protective effect against CAD.41) The R219K variant of the ABCA1 gene was associated with high plasma HDL-cholesterol levels and a reduced severity of CAD.42) The APOA5 gene -1131T>C polymorphism has been reported to affect triglyceride levels, lipoproteins, and oxidative stress.43) A polymorphism of the adiponectin gene (AdipoQ), G276T, is associated with metabolic parameters that contribute to CAD.44) The risk for CAD based on specific folate levels required by the different MTHFR C677T genotype has been demonstrated.45) Recently, the association of the ADD1 G460W polymorphism and the risk of CAD was observed.46)

Evidence on the association between genes regulating inflammatory processes and CAD is accumulating. Associations between polymorphisms of endothelial nitric oxide synthase (eNOS),47-49) the NADH/NADPH oxidase p22phox gene,48) the C-1562T polymorphism of gelatinase B promoter, and CAD exist.50) The oxidative stress-related paraoxonase gene Q192R polymorphism is a risk factor for coronary spasm and stenosis.51) A variant of the lipoprotein-associated phospholipase A2 (Lp-PLA) gene52) and several polymorphisms of chemokine receptor 2 (CCR2) have also been reported to be associated with CAD.53) An association of polymorphisms with the receptor for advanced glycation end products (RAGE) gene and CAD have been shown.54) On the other hand, regulated upon activation, normally T cell-expressed and presumably secreted (RANTES) -403G>A promoter polymorphism have been shown to be related to a reduced risk for CAD.55) Reports on genes related to thrombosis or MIs include: G-33A polymorphism in the thrombomodulin (TM) gene,56) human platelet antigen-3 (HPA-3),57) prostacyclin synthase C1117A,58) prothrombin (PT) T165M, FV R485K polymorphism,59) and pregnancy-associated plasma protein-A (PAPP-A) IVS6+95C allele.60) In recent case-control studies, investigators expanded the association of 9p21 SNPs with CAD in Koreans beyond Caucasian populations.61)

Integration of data from genomic, molecular, and clinical medicine for the prediction of coronary artery disease

Recent successes in the identification of susceptibility variants have increased confidence that this information can be translated into clinical management. More and more genes and gene products are considered important in assessing patients at risk for CAD. Although single genetic polymorphisms increase individual risk, the integration of information from several polymorphisms can become clinically useful.62)

Markers of inflammation or plaque vulnerability may become part of clinical practice. However, most of the markers that have been discussed have not been validated in large-scale clinical studies. Thus, risk stratification and clinical decision-making may have the form of a combination of analysis of candidate genes, and multiple molecular and clinical factors. In addition, information regarding genomics and phenotypic markers must be considered in the context of an individual's medical history and environmental variables (Fig. 3).

Fig. 3

Combining information from genomic, molecular, and clinical data for personalized medicine in coronary artery disease.

Pharmacogenomics in Cardiovascular Diseases

The clinical goal of pharmacogenomics is to deliver the right drug to the right patient by accurate prediction of a therapeutic response and safety before prescription. In prior studies, commonly used cardiovascular medications, such as lipid-lowering agents, antihypertensives, and warfarin, have shown different effects on the basis of individual variation.

Associations between lipid response to HMG-CoA reductase inhibitors and variation in the HMG-CoA reductase gene have been found.63) In hypertension, variation in the alpha-adducin (ADD1) gene was associated with renal sodium reabsorption and salt-sensitive hypertension. Thus, variation in this gene might identify who would benefit more from diuretic therapy. Polymorphisms of the ACE pathway also show the possibility of clinical application of pharmacogenomics. Patients with the ACE DD genotype, which is associated with death or the need for transplantation in chronic heart failure, have shown the greatest benefit from beta-blocker or high-dose ACE inhibitor therapy. Variation in drug metabolism is related to implication for pharmacogenomics. Genetic variants of the enzyme that metabolizes warfarin, cytochrome P-450 2C9 (CYP2C9), and of a pharmacologic target of warfarin, vitamin K epoxide reductase (VKORC1), have been demonstrated to contribute to differences in patients' reponses to warfarin treatment.64)

Conclusions

In recent years, DNA sequence information, sophisticated genetic methods, and advances in statistics and bioinformatics have allowed researchers to identify some variation in genes that modify the risk for CAD. In particular, the consistently replicated locus on 9p21 related to non-coding RNA gene represents a region of high priority for further investigation.

Several limitations of current genomic approaches also need to be reviewed. First, before a personalized medicine for CAD becomes a reality, researchers need to validate novel markers in large independent cohorts, usually achieved by collaborative efforts. Second, unlike conditions such as age-related macular degeneration or diabetes, CAD has a wide range of clinical manifestations. Consequently, CAD cohorts usually have genotypic heterogeneity. Moreover, the lack of a clear definition for control subjects is also a significant problem. Third, although genomic study is to benefit an individual patient, because of the small effect size of most genetic risk factors, it may make only modest influence. Therefore, genetic information can be useful when it identifies candidates for intervention and guides more effective use of interventions.

In the coming years, we expect that the combination of genomic, clinical, and environmental risks will refine personalized approaches to treat potential CAD patients. Because DNA-based genetic markers remain static throughout life, predictive genetic tests could be performed at a young age to facilitate early intervention in high-risk individuals. Genetic tests will add to rather than replace the clinical risk stratification.