GENOMICS OF HEART FAILURE

Introduction

Cardiovascular disease is the leading cause of death in men and women, and heart Failure (HF), a cardiovascular disease condition characterized by dyspnea, fluid retention and decreased functional capacity, is associated with particularly high rates of morbidity and mortality. Apart from the few known Mendelian diseases that lead to HF, most common forms of HF are non-Mendelian and the evidence for heritability is modest. The substrates for development of HF are structural and/or functional changes in the myocardium that precede the onset of the clinical syndrome. Thus, a diverse group of etiologies (e.g., major etiologies being hypertrophic, dilated, restrictive, and ischemic cardiomyopathies) lead to HF, making it very challenging to undertake studies to discern the genomic factors underpinning this syndrome. Study of the genetic susceptibility to HF has until recently been limited to candidate gene association studies in patients with distinct sub-types of HF. However, with the completion of the human genome project¹ and the development of the HapMap template², new large scale genome wide association studies (GWAS) are possible and underway. We will review the status of these ongoing efforts and other important new directions in genomics, in particular genome-wide sequencing, as well as other “omics” technologies that will likely advance our understanding of HF etiologies.

State of knowledge pertaining to the genetics and genomics of HF

Disorders with Mendelian inheritance that lead to HF

Our understanding of the genetics and genomics of HF to date has been grounded in research in rare families with strong evidence for a genetic predisposition for HF. In these families, rare mutations underlie the predisposition to HF. Hypertrophic Cardiomyopathy (HCM), Dilated Cardiomyopathy (DCM), and arrhythmogenic right ventricular cardiomyopathy (ARVC) are disorders associated with HF that demonstrate classical Mendelian inheritance pattern. HCM has an estimated prevalence of 1 in 500 in the general population³ and familial DCM (which accounts for 30% of all DCM) has an estimated prevalence of 1 in 20,000.⁴ People with these syndromes exhibit a diverse array of phenotypic manifestations thus complicating the effort to identify them on the basis of clinical characteristics alone. Over 400 hundred mutations in several sarcomeric proteins have been identified, including MyHC⁵, MYH7, MYBP3⁶, TNNT2⁷ and TNNI⁸, encoding beta myosin heavy chain, myosin heavy chain, myosin binding protein, cardiac troponin I and troponin T respectively. Approximately 70% of all HCM cases are accounted for by mutations in the aforementioned 4 genes.

DCM has heterogeneous manifestations and its cause is unknown in the majority of cases. In selected cases of DCM that appear to be familial, mutations in genes coding for dystrophin⁹, emerin¹⁰ and tafazin¹¹ lead to X-linked inheritance pattern whereas mutations in genes coding for actin¹², lamin A/C ¹³ and desmin¹⁴ lead to an autosomal dominant pattern. In addition to these mutations, rare mutations in genes coding for laminin, troponin, titin and myosin have been reported in familial DCM cases. ARVC occurs approximately 1 in 5000 – 10000¹⁵. ARVC is a frequent cause of sudden cardiac death in the young, especially among athletes and is characterized by progressive degeneration, thinning and fibrosis of the right ventricle leading to left bundle branch block, arrhythmias and aneurysms. Mutations in the transforming growth factor-β¹⁶, ryanodine receptor¹⁷, plakophilin-2¹⁸, desmoplakin¹⁹ and plakoglobin²⁰ genes have been implicated; the first three lead to an autosomal recessive inheritance and the latter two to a dominant inheritance.

Genomics of HF risk factors

In contrast to rare Mendelian conditions underlying HF, common forms of HF account for the vast majority of HF cases that lead to very high rates of morbidity and mortality from HF. A substantial proportion of the etiology of common forms of HF is explained by the prior occurrence of well characterized, common risk factors including hypertension, left ventricular hypertrophy, type II diabetes mellitus, and prior myocardial infarction. Common HF appears to have a heritable component. Investigators from the Framingham Heart study have previously reported that offspring of people with HF are more likely to have systolic dysfunction, ventricular dilatation and increased left ventricular mass.²¹ In cross-sectional analyses, the odds of increased left ventricular mass, left ventricular internal dimension of left ventricular systolic dysfunction were 1.35, 1.29 and 2.37 in people with at least one parent with HF, compared to people with no HF in either parent. Indeed HF in one parent portended a 70% increment in risk and HF in both parents a 90% increment in risk, of HF prospectively in the offspring, compared to people with HF in either parent. The heritability (H²; percent of population variation attributable to genetic factors) of common HF is estimated to be 28%. While the minority of HF cases in the general population are explained by rare Mendelian causes, there is remarkably little reliable information on the actual proportion of risk for common HF that is conferred by rare mutations (eg, minor allele frequency [MAF] ≪0.1%) versus low frequency genetic variants (eg, MAF <0.1–1%) mutations versus high frequency variants (eg, MAF ≥ 1%).

Recently, investigators have conducted sequencing studies in general populations that lead to improved understanding about the role for of variants in genes known to underlie rare forms of HF. For example, in the Framingham Heart Study, genetic variation in exons of sarcomere protein genes and storage cardiomyopathy-causing genes was examined in association with increased left ventricular wall thickness in ~1800 persons. 7 mutations found in 5 sarcomere protein genes and 1 mutation in the alpha-galactosidase A (GLA) gene; one or more of these mutations was detected in only one percent of all persons in the community and in 18% of those with increased left ventricular wall thickness.²² In a separate sequencing study in the Framingham Heart Study, at least 1 out of 64 subjects was found to carry a functional mutation in one of three genes—NCCT, NKCC2 and ROMK—previously known to cause rare recessive diseases featuring large reductions in blood pressure.²³ These two studies underscore the potential role of low frequency variants in common conditions—left ventricular hypertrophy and hypertension—that are well established “risk factors” for HF. However, our understanding of the full role of low frequency variants in HF awaits the availability of much lower cost sequencing.

In the upcoming few years, the greatest opportunity for genomic discovery arises from the conduct of genome-wide association studies for HF and its risk factors. GWAS relate a large number of genomic sequence variants to a clinical disease or a continuous trait. To date, most GWAS have used a case-control study design. High-throughput genotyping technologies are used to assay hundreds of thousands of single-nucleotide polymorphisms (SNPs) in each subject, and agnostically compare genotype and allelic frequencies for each SNP between “cases” and “controls” to identify variants that are significantly associated with the disease or trait being studied. In the past three years, hundreds of new loci for many common diseases and traits have been identified and replicated in GWAS, leading to identification of many genes that have previously not been suspected to be related to the disease or trait under study. Of note, over a third of associated SNPs are not near protein coding regions of the genome. Although GWAS represent an important step in discovering new genome sequence variants associated with disease risk, they are also limited by their potential for false-positive and false-negative results and for biases related to selection and misclassification of participants, population substructure and genotyping errors. Therefore, replication studies in large, independent populations have become a mandatory part of the design of GWAS.²⁴

At present GWAS data for HF are very sparse, so we describe here the results of GWAS for the major risk factors that together explain a substantial proportion of HF risk—hypertension, type II diabetes mellitus, cardiac structural alterations, and prior myocardial infarction.

Hypertension

A substantial heritability of hypertension has long been observed and well characterized in the literature. In the St. Mary’s study, familial aggregation of blood pressures in families with hypertensive (defined at diastolic pressure >100) and normal (defined as diastolic pressure < 85) probands was noted.²⁵ A much higher correlation for systolic and diastolic blood pressures was noted between biological siblings versus adopted siblings in the Montreal Adoption study.²⁶ The H² for hypertension is 30%.

hypertension with a Mendelian form of inheritance is very rare and mostly associated with disorders that affect renal handling of sodium. Glucocorticoid-remediable aldosteronism²⁷ and Liddle’s syndrome²⁸ are two autosomal dominant disorders manifesting hypertension as part of their phenotype. The former is associated with increased aldosterone and sodium retention secondary to inappropriate control of aldosterone synthase by ACTH. The latter is associated with mutated sodium channel gene, leading to an increased number of sodium channels in the apical membrane and increased sodium reabsorption. Mineralocorticoid excess, secondary to 11-beta hydroxysteroid dehydrogenase mutations is an autosomal recessive condition with decreased turnover of cortisol to cortisone, thereby leading to salt retention.²⁹

The vast majority of people with hypertension have the common form, associated with polygenic influences. However, there has been little success in identifying putative genes. Candidates identified so have been poorly replicated, and most of the research has been focused on genes encoding renal sodium transport proteins (SLC4A5³⁰, DRD1³¹ etc) and elements of the renin angiotensin aldosterone system (ACE³², AGTR1³³, REN³⁴ etc). A major GWAS by the Welcome Trust Case Control Consortium failed to identify any significant hypertension loci.³⁵ However, two recent GWAS have for the first time identified a total of 19 loci associated with hypertension^{36, 37} including CYP17A1, CYP1A2, MTHFR, ATP2B1, PLEKHA7, CSK-ULK3, SH2B3 and TBX3-TBX5. Several loci have shown interesting pleiotropy with other phenotypes, which may potentially elevate the interest in these genes for HF. One example is SH2B3, which is associated with hypertension³⁶ as well as myocardial infarction.³⁸ These findings for the first time describe the genetic architecture of hypertension and suggest some loci that may be candidates for the study of HF.

GWAS for type 2 diabetes mellitus

A number of GWAS investigating the genetic susceptibility to both type 1 and type 2 diabetes have been published recently. Loci within, or in close proximity to several genes have been consistently identified in various GWAS as conferring increased risk for type 2 diabetes, and have been replicated in various racial and ethnic groups. The genes includeKCNQ1, IGF2BP2, CDKAL1, CDKN2A, CDKN2B, TCF7L2, SLC30A8, HHEX, FTO, PPARG2, and KCNJ11.^39–49 The functional significance of the products of these genes and the mechanism by which altered gene products lead to diabetes has yet to be elucidated. It is worth noting that knowledge of these variants does not provide incremental information over clinical factors alone in identifying people at increased risk for type 2 diabetes; 2 studies that evaluated the incremental value of SNP information over conventional risk factors observed that the c-statistic of the models containing clinical factors improved very little by addition of genetic information.^{50, 51}

Cardiac structural alterations that predispose to heart failure

Limited data exist on the genome wide associations of cardiac structural traits (assessed by imaging). A previous report by Vasan et al⁵² from the Framingham Heart Study reported several associations with echocardiographically assessed cardiac structural traits. The top SNP associations were: for LV diastolic dimension, a SNP associated with SLIT2 gene; for LV systolic dimension, a SNP associated with KCNB2 gene; and for LV mass, and left atrial size, SNPs associated with FAM5C gene. Several other large GWAS have been combined in a consortium of multiple community based epidemiological studies, including the Framingham Heart Study, have assessed the genomic determinants of LV structure and function assessed by echocardiography and publication of the results are anticipated later in 2009.

Myocardial infarction (MI)

MI is a leading cause of HF. Common occurrence of MI does not demonstrate a Mendelian inheritance pattern. However, there are several lines of evidence confirming the existence of familial risk. Data from twin studies show an exceptionally strong association between identical twins (relative risk of 8.1 and 15 for men and women respectively if sibling experienced cardiovascular death before age 55), with less strong, but significant association between nonidentical twins (relative risk of 3.8 and 2.6 for men and women respectively).⁵³ Two previous reports from the Framingham Heart study established the increased risk for MI in people with affected parents ⁵⁴ or siblings⁵⁵, with relative risks of the magnitude 1.7 to 2.0. Family history has proven to be a useful screening tool⁵⁶ to identify families at high risk for CVD.

A number of genome wide linkage and candidate gene studies reported both chromosomal loci and candidate genes with strong associations to MI and atherosclerotic CVD. Loci on chromosomes 1⁵⁷, 2^58–60, 3⁶¹, 5⁶², 13⁶³, 14⁶⁴, 16⁶⁵, 17⁶⁶ and X⁶⁰ have been implicated. However, loci identified through this approach have not been consistently replicated and its validity has been questioned.⁶⁷ A rare exception has been the study that identified the gene encoding the 5-lipoxygenase-activating protein (ALOX5AP)⁶³, which has been subsequently replicated in two other studies.^{68, 69}

A large number of candidate genes studies have been published that assessed the associations of individual genes to MI, CHD or subclinical measures of atherosclerosis. Similar to linkage analyses, candidate gene studies have suffered from poor replication, likely secondary to genetic heterogeneity, inadequate sample size, differing patterns of expression in different population groups, false-positivity and confounding by environmental factors or population stratification. Some of the main gene variants indentified by this approach are: APOB^{70, 71}, APOE^{72, 73}, ACE insertion/deletion^{74, 75}, PAI-1⁷⁶, MTHFR^77–79, CETP⁸⁰, eNOS⁸¹ and prothrombin.⁸² A comprehensive survey of candidate genes studies for MI is beyond the scope of this article.

The advent of GWAS has generated much excitement about the potential for identifying loci associated with a risk for MI. Several recent large scale GWAS reported associations with MI.^{35, 83–86} The Wellcome Trust Case Control Consortium (WTCCC) demonstrated several loci associated with MI, the principal locus being 9p21. This finding was replicated by investigators of the German MI study and in several other studies including the Framingham 100k dataset. Subsequently, the MIGen Consortium has confirmed and extended the findings of association with 9p21 SNPs in a discovery GWAS of 2,967 cases and 3,075 controls and replication in 9,746 cases and 9,746 controls.⁸⁷ Genome wide associations were found in a total of nine gene loci, including CELSR2, PSRC1, SORT1, MRPS6, KCNE2, MIA3, PHACTR1, LDLR, CXCL12, PCSK9, and WDR12; in addition a separate consortium reported an association of MRAS SNPs with MI.⁸⁸ Taken together, these loci identify genes/SNPs that are strongly and consistently associated with MI and further study is warranted to understand the role if any of these genetic variants in the etiology of HF.

Candidate gene associations with heart failure

Several genes that either influence the risk for HF⁸⁹ or modify the clinical course⁹⁰ (modifier genes) have been reported recently. Most of the genes relate to specific phenotypes described above viz. dilated, hypertrophic and restrictive cardiomyopathies, or polymorphisms in the genes encoding proteins of the rennin-angiotensin system that influence treatment effect (described below). As noted above, HF is the culmination of several types of injury and few single candidate genes that strictly influences the occurrence of the common form of HF has been identified.

GWAS for HF

Large GWAS that exclusively address the genetic susceptibility to HF are underway but have not yet been reported. The Framingham Heart Study 100K project, addressing genome wide associations for cardiovascular disease outcomes, reported the results of a relatively small GWAS on HF outcomes.⁸³ In this report, no single SNP met genome-wide significance; in additive regresion models, the lowest p-value was for a SNP associated with the gene KIAAI598 on chromosome 10 (p = 8.8 × 10⁻⁴), and in family based association testing models, a SNP on chromosome 5 (p = 4.72 × 10⁻⁵). A thought-provoking result from this report was the association with HF risk of a SNP in the ryanodine receptor [RYR2] gene (p = 3.6 × 10⁻⁴); as noted above, RYR2 is implicated in risk for ARVC.

Future Directions for HF Genomics

In the next few years, genomic studies of HF will likely include some or all of the following strategies: initial GWAS for HF; additional larger replication studies and larger GWAS using arrays in multiple ethnic groups; studies of copy number variations; targeted and whole-genome sequencing studies; and studies using other “omic” technologies to study HF. Ultimately, refinements in the HF phenotype due to advances in imaging and biomarker science will increase the quality and diagnostic accuracy of all future genomic studies of HF.

GWAS studies in larger sample sizes

A rising trend in GWAS studies is the formation of very large international consortia to allow study of ever larger numbers of disease cases and identify the totality of genetic variation underlying disease. For example, the GIANT Consortium has accumulated data on nearly 100,000 research participants for replicated GWAS studies of obesity.⁹¹ Similar consortia are under development for other HF risk factors including diabetes mellitus and myocardial infarction and will be valuable for pooling GWAS studies of HF as well. Due to increases in statistical power, large GWAS are able to detect significant associations with lower frequency alleles that may not be noted in smaller studies and the larger studies can more reliably exclude associations. An additional advantage of very large consortia is the possibility of studying gene by environment interactions and gene by gene interactions, for which smaller studies are statistically underpowered. Modifiers of HF that will be of interest in such large studies will include HF risk factors (HT, type II Diabetes, antecedent MI, LVH), as well as lifestyle and dietary factors such as physical activity and salt intake. Ultimately, dozens to hundreds of associated genetic variants may be detected for each clinically important phenotype as larger and larger studies are conducted.

Copy number variation

Copy number polymorphisms (CNPs), small insertions or deletions in various repeat patterns, are common in the human genome and have been hypothesized to underlie common diseases. To date, CNPs have not been shown to be associated with CVD in GWAS studies. For example, there was no evidence for association of CNPs with myocardial infarction in the MIGen Consortium.⁸⁷ However, available GWAS arrays only incompletely capture the total number and extent of CNPs in the human genome, so future more comprehensive studies will be needed to assess the role of CNPs in diseases including HF. Notably, a recent candidate gene study demonstrated the association of a deletion polymorphism in MYBPC3 with HF phenotypes.⁹² Participants with a 25-basepair deletion in this gene had a seven-fold higher risk of developing HF, and this mutation was present in 4% of populations of Indian descent. This finding represents a rare example of a consistent, replicated association with HF, and it remains to be determined whether this gene influences restrictive or dilated CM.

Targeted and whole genome sequencing

One strategy that may become mandatory for identification of the functional genetic variation underlying associations from GWAS is sequencing targeted to the chromosomal region of association. A number of the SNPs that are strongly associated with MI or other phenotypes in GWAS are in or near protein coding genes. For these loci, sequencing is often conducted in persons selected from the extremes of a phenotype (eg, with versus without HF) targeted at exons, introns and nearby regulatory regions to indentify differences in the prevalence of functional genetic variations in cases versus controls. As mentioned previously, this strategy has been successful in identification of gene variants underlying left ventricular hypertrophy and hypertension.^{22, 23} However, for the substantial proportion of GWAS associations tare far away from any known protein-coding gene, the conduct of sequencing studies will require more detailed and complex sequencing studies. Targeted sequencing studies have been successfully employed to identify rare variants underlying Mendelian HF conditions and may well be useful to identify the important genetic variation underlying GWAS associations with common forms of HF.

More generally, the emergence of “next generation” sequencing technology and decreases in the costs of sequencing have made possible for the first time the conduct of targeted sequencing in larger numbers of subjects as well as genomewide sequencing to characterize part or all of the personal genome sequence of thousands of persons at a time.⁹³ Technological advances have allowed the conduct of high throughput, genome-wide capture of exons and other parts of the human genome.⁹⁴ Such sequencing will allow characterization of the phenotype associations with the so-called “dark matter” of the genome, the large amount of currently unsequenced genome. While technological advances may enable the sequencing of the entire human genome sequence of over 3 billion base pairs in under $1000 in upcoming years, a more imminent and tractable sequencing goal will be the sequencing of the entirety of human exons (ie, the “exome”). Multiple exon capture technology is now available and large-scale sequencing studies are now being envisioned to conduct exome sequencing in large sample sizes, including populations containing persons with HF.⁹³

The Role of “omics” in HF research

In parallel with largescale genomic studies of HF, a number of other “omic” technologies including transcriptomics and metabolomics are now feasible and increasingly cost effective in large populations of research participants.⁹⁵ Integration of transcription information with GWAS and HF phenotypes may allow more sophisticated understanding of the specific functional genetic variation underlying HF.⁹⁶ When possible, actual heart tissue obtained via biopsy or at time of surgery may provide helpful signals when tissue is available for such uses. An additional challenge for “omic” studies is the difficulty of conducting direct studies of human heart tissue. Because actual heart tissue is usually inaccessible in healthy persons and “omic” profiling studies in diseased heart tissue may be difficult to interpret in the setting of multiple treatments.

Clinical applications of genomic information

A number of potential “personalized medicine” clinical applications may be considered once there are significant, replicated findings of common and/or lower frequency SNPs underlying HF. First, genetic risk scores may be constructed to assess the additive and incremental predictive effect of risk prediction for each associated SNP. While such a score cannot yet be constructed for the HF phenotype, initial attempts have been made to construct and assess genetic risk scores for type II diabetes mellitus⁹⁷ and for myocardial infarction.^{87, 98} For myocardial infarction, a risk score was constructed comprising nine SNPs associated in prior GWAS with adverse lipid levels; the age- and sex-adjusted genetic risk score was associated with incident MI but there was no incremental prediction over and above traditional risk factors.⁹⁸ Thus, to date there is no evidence that genetic risk scores comprised of limited numbers of SNPs will predict CVD risk over and above known clinically measurable risk factors, and a similar assessment will need to take place. Other personalized medicine applications of HF genomic studies might include the development of new HF drug targets from novel HF genes and pharmacogenetic applications, which are covered elsewhere.

Conclusion

The field of genomics has advanced rapidly since completion of the human genome sequence. While a number of novel, strongly replicated genetic discoveries have been made in the past few years using GWAS for myocardial infarction and other HF risk factors, the genomic studies of HF and left ventricular dysfunction are now underway. In the near future, it is anticipated that initial GWAS studies of HF and next step investigations, including replication studies in larger consortia and targeted resequencing studies of top loci, will be conducted in a similar manner as for GWAS studies already completed for myocardial infarction and other HF risk factors.