Exploring Predisposition and Treatment Response—the Promise of Genomics

Abstract

Spurred by large-scale public and private efforts as well as technological developments, the last few years have seen a major leap forward in our understanding of the genetic basis of cardiovascular disease. This revolution is in its infancy and will continue to alter the medical landscape for years to come. There is a need within the general cardiology community to develop a better understanding about how these developments may alter routine clinical care. In this review, we will provide an overview of the current state of genetics as pertains to rare cardiovascular diseases and then review advances in the discovery of the genetic basis of common disease with the potential for improved risk assessment and drug development. We will also outline a few recent examples of pharmacogenetic advances that are already starting to become a part of clinical management and finally discuss the promise as well as the challenges in using next-generation sequencing technologies to provide personalized cardiovascular care.

The last decade has brought a tremendous revolution in our understanding of cardiovascular genetics. This has been spurred by technological developments in genotyping and sequencing along with the increasing availability of population-level genetic information through efforts such as the Human Genome project¹; the International HapMap Consortium²; and, more recently, the 1000 Genomes Project³ and the National Heart, Lung, and Blood Institute Exome Sequencing Project.^4,5 This deluge of genetic information along with the increasing availability of patient-specific genetic information has led to a need for increased understanding regarding the promise and pitfalls of using genetic information in cardiovascular care.

In this review, we will provide an overview of the current state of genetics in cardiovascular disease (CVD), starting with a brief introduction to the current use of limited gene panels for genetic testing in rare CVDs. We will then discuss advances in the discovery of the genetic basis of common CVDs like hyperlipidemia and myocardial infarction (MI), which has led to potential clinical uses such as for risk assessment and for novel drug development. We will also discuss one of the most promising applications of genetic technology, pharmacogenomics, where genetic information is used to guide therapy with the potential to reduce the rate of adverse effects and enhance treatment response. Finally, we will conclude with a glimpse of how new sequencing technologies may begin to fulfill the promise of a more personalized delivery of medical care, albeit with several challenges that will need to be overcome.

Current genetic testing in rare CVD

Itmaybesurprisingtolearnthat1in17peopleintheworld has a “rare” inherited disease.⁶ Although each of these diseases has a low prevalence, cumulatively, they have a high societal burden, especially those with cardiovascular consequences such as hypertrophic cardiomyopathy (HCM), familial hypercholesterolemia (FH), and long QT syndrome (LQTS), which have prevalence estimates of 1:500,⁷ 1:500,⁸ and 1:2500.⁹ Patients with these conditions can often be unaware of their condition, with the first manifestation having tragic consequences like sudden cardiac death. Family history can be useful in identifying risk before this but is often difficult to obtain or unreliable, and can be complicated by variable phenotypic presentation even within families.

Genetically, most of these diseases have classically been thought of as having Mendelian inheritance patterns (eg, HCM as autosomal dominant, FH as autosomal co-dominant) with mutations in single genes having large effects. This inheritance pattern lends itself to a targeted strategy whereby most genetic testing is done by sequencing all (or part) of a limited number of genes that have been associated with a particular phenotype. Current panels for HCM, FH, and LQT include sequencing 12-18, 2-3, and 12-13 genes, respectively.

For many of these conditions, genetic testing has an integral role in confirming the clinical diagnosis and/or screening at risk relatives. Although relatively expensive, genetic testing can be cost effective especially by reducing the need for other tests or identifying at-risk individuals and initiating therapy before they present with a catastrophic outcome. Guideline documents are available from several governing organizations to assistin clinical decision-making regarding when to consider ordering these tests.^8,10-12

Although this information can be incredibly valuable, there are important caveats, the most important of which are incomplete yield and the fact that sequencing often identifies many “variants of unclear significance.” These are variants whose effects on the resulting gene product are unknown and thus do not rise to the level of evidence necessary to be considered pathogenic. In most cases, the strongest evidence of pathogenicity is derived from studying multiple affected families and demonstrating co-segregation of the variant with disease. Increasingly, in silico (computer based) prediction algorithms taking into account the type of variant, its location, how conserved the position within the gene is throughout other species, and how it might affect the resulting structure and activity of the gene product are used to provide evidence for or against pathogenicity, although these algorithms at this time are far from definitive. Given the complexity of these results and their potential consequences, interpretation of these tests often requires the assistance of a certified genetic counselor.

Although the standard gene panels can be very useful, the current yield of these tests is highly variable. For instance, the quoted yield is 20% to 30%, 40% to 60%, and 60% to 80% for familial dilated cardiomyopathy (DCM), HCM, and LQTS, respectively.⁹ Recent advances in sequencing technology (“next-generation sequencing” [NGS]) may significantly increase the yield of these tests and will be discussed in more depth below.

Inherited basis of common disease

Unlike the rare diseases mentioned above, common CVDs such as “garden variety” hyperlipidemia, hypertension, coronary artery disease (CAD)/MI, and atrial fibrillation are due to the complex interaction of many separate genetic and environmental effects. All of these conditions are moderately heritable (meaning a good proportion of the phenotypic variance can be explained by the genotypic variance), but the genetic susceptibility is due to small effects at many genetic loci rather than large effects in single genes. Until recently, the genetic basis of these diseases was very poorly understood because it had been hard to identify these relatively small genetic signals. However, there have been recent spectacular advances in this field largely due to genomewide association studies (GWASs). Genomewide association studies take advantage of the ability to simultaneously genotype hundreds of thousands to millions of common (defined as a frequency of >5%) single nucleotide polymorphisms (SNPs) across the genome. By comparing allele frequencies in large numbers of well-phenotyped cohorts, it is possible to robustly associate the presence (or absence) of an SNP with an increased risk of disease or phenotype. Typically, these studies require thousands of patients to reliably demonstrate associations. This is partly because the signals are usually very small (the odds ratio currently being identified for complex disease often range from ∼1.05 to 1.20) and because the tests for significance are very stringent to limit false positives. For instance, to correct for multiple testing, the generally accepted P value threshold for GWAS is P < 5 × 10⁻⁸. Nevertheless, GWASs have now identified more than 1500 common SNPs associated with complex diseases; and the list is growing daily,¹³ many of which have been identified for CVDs such as CAD/MI or risk factors such as dyslipidemia or hypertension.

Initially, these GWAS signals were identified by relatively small studies of several thousand patients.^14-19 However, more recently, larger meta-analyses have been performed to maximize the power to find small genetic signals. In one striking example, the Global Lipids Genetics Consortium in 2010 performed a large metaanalysis of 46 separate GWAS cohorts that in total comprised more than 100,000 persons of European descent collected from the United States, Europe, and Australia.²⁰ The authors were able to uncover 95 separate genetic loci associated with some aspect of the lipid profile, including 36 previously reported genetic loci and 59 novel ones, and were able to attribute 25% to 30% of the genetic variance (approximately 9%-12% of the entire variance) for each of the lipid traits (total cholesterol, low-density lipoprotein [LDL], high-density lipoprotein [HDL], and triglycerides). Although many of the signals identified were obvious candidates (such as the LDL receptor), others were completely unexpected, yielding unique insights into the pathogenesis of disease as well as offering additional therapeutic targets.

One such target that has since been explored further is sortilin 1 (SORT1). A noncoding variant in this gene was found in the above-mentioned GWAS of lipids with significant association with LDL levels. Subsequent work showed that these genetic changes modified the liver-specific expression of the SORT1 gene and that over-expression of SORT1 in the livers of mice results in drastically reduced LDL and very low density lipoprotein (VLDL) levels. SORT1 thus represents an exciting novel drug target for treatment of hyperlipidemia.²¹

Interestingly, genetics in hyperlipidemia is one field that has benefitted from both family studies using linkage and, more recently, exome sequencing strategies as well as a GWAS strategy looking for common variants in large cohorts. Two recent examples highlight this. Gain-of-function variants in proprotein convertase subtilisin/kexin 9 (PCSK9) were originally identified in a couple of families with autosomal dominant hypercholesterolemia.²² Subsequently, Cohen et al demonstrated by sequencing this gene in a cohort of African Americans that rare loss-of-function mutations in PCSK9 are associated with very low LDL levels and a reduced incidence of CAD.²³ Not surprisingly, common SNPs in this gene were also found to be associated with CAD/MI risk in large GWASs.²⁴ These discoveries have led to the development of a monoclonal antibody to PCSK9 that has been shown in Phase 1 trials to significantly reduce LDL levels in familial or nonfamilial hypercholesterolemia.²⁵

A similar story involves the ANGPTL3 gene, only in that case GWAS first identified a common variant near the gene with association with triglyceride levels,¹⁹ which was then solidified by a family study of familial hypobetalipoproteinemia (which manifests with very low triglyceride, LDL, and VLDL levels) using exome sequencing with causal variants found within the gene itself.²⁶ That same year, the above-mentioned Global Lipids Genetics Consortium association meta-analysis confirmed associations of variants in this gene locus with all 3 phenotypes (triglyceride, LDL, and VLDL levels), bringing this discovery full circle. Combinations of GWAS and family linkage/sequencing studies have been quite fruitful in bringing a number of very promising future therapeutic targets to bear and illustrate that both rare and common variants in the same gene may affect disease processes.

The study of the genetic predisposition to CAD/MI has also benefited greatly from GWAS approaches. In initial studies, 3 groups identified a genetic region in the 9p21 (chromosome 9, short arm band 21) locus that has consistently and reproducibly been associated with coronary heart disease.^15-17 Since this original discovery, multiple GWASs, including the recent CARDIoGRAM²⁴ and C4D²⁷ studies, have confirmed the association of approximately 25 new loci with an increased risk of CAD/MI. Some of these variants appear to increase risk through affecting traditional risk factors (particularly LDL levels). However, more than half of the CAD/MI risk variants seem to be operating independently of traditional risk factors, thereby shedding new light on the pathogenesis of atherosclerotic heart disease and opening the door to new therapeutics.

Improving CAD/MI risk prediction

Despite these phenomenal discoveries, much work still needs to be done to translate the findings to the clinic. For most of the variants that have been discovered, the mechanism of action is unclear; and scientists are working to uncover how these variants are acting at a molecular level to increase risk. However, although much progress has been made, the source of much of the estimated heritable risk of disease remains undiscovered. Despite very large studies, current known genetic loci explain less than 10% to 20% of the heritable risk for many common diseases such as MI/CAD.²⁸ The source of this “missing heritability” remains a field of intense investigation.²⁹

Nevertheless, in 2012, we know more than we ever have about the genetics of CVD. One looming question is whether we can use these markers to improve risk stratification. One early study evaluated the ability of a 13-SNP genotyping panel containing GWAS-identified CAD/MI signals for risk prediction in several large longitudinal studies.³⁰ Using the results of these genotypes in a discovery cohort, Ripatti et al were able to develop a genetic risk score for CAD that was significantly predictive of CAD, with scores in the top quintile predicting 1.7 times increased risk compared with scores in the bottom quintile, comparable to risk predictions afforded by the Framingham model. Although the ability to predict risk was not improved vs using traditional risk factors as defined by the c-index statistic, there was improvement in other reclassification indices. As more genetic markers are identified, the hope is that the power of genetic testing for improving risk stratification for common diseases such as CAD may improve.

Pharmacogenomics and treatment response

One field that is already seeing active clinical use of genetics is that of pharmacogenomics, which is the practice of using genetic data involving drug effect and metabolism to help guide treatment and potentially balance benefit vs adverse effects. One reason that this field has gained more traction than other areas is that the impact of pharmacogenetic variants has generally been much larger than that of variants associated with disease states, thus making discovery easier and genetic testing more tractable. The poster child for this has been warfarin,³¹ for which algorithms for genetics-based dosing have been developed and are in active use in some clinical centers and have been discussed elsewhere.³²

More recently, there has been much interest in investigating the pharmacogenetics of statins. There unfortunately remains a small subset of the population that is statin intolerant, usually because of myalgias. To identify the genetic basis of this effect, the Study of the Effectiveness of Additional Reductions in Cholesterol and Homocysteine included a GWAS of only 85 cases of European descent with statin-associated myopathy along with 90 matched controls and was successful in identifying a coding variant in SLCO1B1 that conferred an 18% cumulative risk of myopathy within the first year of therapy in homozygotes for that variant, as opposed to 3% in heterozygotes and 0.6% in people without the variant.³³ Of note, SLCO1B1 encodes an organic anion transporter that has a role in hepatic uptake of many drugs including statins. Studies are ongoing regarding the clinical benefit of using these data in prescribing practices.

Clopidogrel pharmacogenetics has also rapidly progressed into the clinic. It was first noted in 2004 that patients who received stenting after an ST-elevation MI with suspected clopidogrel resistance had higher rates of recurrent cardiac events.³⁴ More worrisome was that this population was estimated to compose up to 25% of all patients presenting with ST-elevation MIs. Since then, it has become apparent that inactivating variants in CYP2C19 reduce conversion of clopidogrel to the active form.^35,36 In terms of clinical outcomes, the PLATelet inhibition and patient Outcomes (PLATO) trial found that those on clopidogrel who had any loss-of-function variant in CYP2C19 were noted to have an event rate of the primary outcome (cardiovascular death, myocardial infarction, or stroke) of 5.7% vs 3.8% in those without.³⁷ Interestingly, a study of genotyped patients from the Clopidogrel in Unstable Angina to Prevent Recurrent Events (CURE) trial and the Atrial Fibrillation Clopidogrel Trial with Irbesartan for Prevention of Vascular Events A (ACTIVE-A) did not find a significant difference in the efficacy of clopidogrel in patients with and without loss-of-function mutations; but they did find that patients with gain-of-function mutations did better on clopidogrel, with a hazard ratio of 0.55 in patients with gain-of-function mutations as compared with 0.85 in normal genotype patients.³⁸

More recently, the Reassessment of Anti-Platelet Therapy Using An Individualized Strategy Based on Genetic Evaluation (RAPID-GENE) trial studied the feasibility of a point-of-care genetic test for the most common CYP2C19 loss-of-function allele (CYP2C19*2). In this trial, 200 patients undergoing percutaneous coronary intervention for acute coronary syndrome or stable angina were randomized to genotyping-based care involving the use of clopidogrel in those negative for the allele and prasugrel in those positive for the allele or standard care with clopidogrel without genetic testing. Patients were evaluated 1 week later to assess platelet function, and those in the standard-care group were genotyped at that time. The performance of the point-of-care test was noted to be excellent with 100% sensitivity and 99.4% specificity, with greater platelet inhibition in those in the genotype-based care arm.³⁹ Thirty percent of patients in the standard-care group were noted to still have high platelet reactivity despite antiplatelet therapy as compared with none in the genotype-based group. Although this study did demonstrate the feasibility of obtaining, synthesizing, and using genetic data in an efficient manner, the study was neither aimed nor powered to assess any differences in clinical events.

The future: personalized genomics

One barrier to the rapid incorporation of genetic information into clinical care has been the cost of DNA sequencing using traditional methods such as dideoxy chain termination (“Sanger”). However, recent advances in the technology behind sequencing that now takes advantage of large-scale parallelization and advanced computing resources to reconstruct whole genomes from millions of short sequences (50-150 base pairs) have made it possible to sequence whole genomes in days and soon possibly mere hours at a fraction of the cost of traditional sequencing. This technology, now known as NGS, has made it possible for hundreds of genomes and thousands of exomes to already be sequenced. Soon, it will be feasible to obtain the raw data for one's whole genomic sequence at a cost comparable to that of a magnetic resonance imaging scan.

NGS for rare disease discovery

Already, the power of NGS has been shown to impact mutation discovery. Herman et al applied NGS technology to sequence the very large gene TTN, encoding the protein titin, in several cohorts of patients with DCM.⁴⁰ Although TTN had already been considered a candidate gene for cardiomyopathy, its immense size (> 100 kilobases) had previously precluded routine sequencing. Their study found that variants in TTN predicted to result in truncated forms of the resulting protein were found in much greater proportion in the DCM cohorts than in cohorts with HCM or unaffected individuals, implying that mutations in TTN are likely contributory to the pathogenesis of inherited DCM. The data from this study support the expansion of clinical genetic testing for idiopathic DCM to include this gargantuan gene, which will only be practical using NGS methods.

Although this study demonstrates the power of NGS as applied to one (very large) gene, the full potential of this technology lies in its ability to give genetic data for the entire exome or genome. Already, several studies have applied the power of whole exome sequencing or genome sequencing to find disease-associated rare variants that could never be uncovered using other techniques. For example, one study from Iceland used whole genome sequencing to evaluate a locus at 14q11 with an association with sick sinus syndrome found by GWAS.⁴¹ By using data from 87 genomes including 7 cases of sick sinus syndrome, the study's authors were able to narrow down the cause of the GWAS association to a rare variant in the gene MYH6 that they verified in more than 800 other patients. Their discovery implicates what had previously been thought of as a purely structural gene in an arrhythmic pathology, suggesting a new mechanism for disease.

Another enticing use of this technology is in novel variant discovery in rare CVDs. Norton and colleagues⁴² used a combination of exome sequencing and comparative genomic hybridization, a method to detect large insertions and deletions in the genome, in a family with DCM to identify a deletion in the gene BAG3 as the cause of DCM in that family. Large-scale efforts funded by the National Institutes of Health and the National Heart, Lung, and Blood Institute are now ongoing with the express goal of using NGS technology in families with rare disease to define the genetic basis of these conditions. There already have been some remarkable stories using whole genome sequencing in the clinical setting to make an impactful change in patient outcomes.⁴³ Several centers have begun to construct pathways through which sequencing can be applied to “medical mystery” cases in hopes of guiding clinical decision making and improving patient outcomes.

NGS for the masses?

The rapidly dropping cost of genome sequencing, now within the $3000 to $5000 range on a research basis, has even prompted many to believe that it will one day be feasible for the majority of people to have their genomes sequenced. Although there are numerous challenges that will need to be overcome to make this a reality, it is clearly no longer in the realm of science fiction.

Within the cardiovascular realm, several test cases of applying whole genome sequencing in individual patients have been described. Ashley and colleagues⁴⁴ used information gleaned from the whole genome of a patient to develop an entire disease risk profile including a high genetic risk for CAD. This information in corroboration with pharmacogenomic data that indicated the patient would tolerate a statin without adverse effects led to a clinical decision to prescribe a statin in a case where traditional risk factors were borderline.

Dewey and colleagues⁴⁵ expanded on this kind of analysis by using family information using a quartet of genomes from a nuclear family not only to explain the father's hypercoagulable state but also to estimate the heritable risk of this condition in the children using phasing algorithms to identify how alleles were passed within the family. And even more recently, Chen and colleagues⁴⁶ used an approach of combined genomic, transcriptomic, proteomic, metabolomic, and autoantibody information to detect a high genetic risk for diabetes in an individual who was not suspected based on traditional risk factors. This led to increased monitoring for diabetes that was eventually diagnosed after a viral infection, leading to treatment with lifestyle modification and remission.

Challenges for high-throughput genotyping and sequencing approaches

Although the preceding examples provide credible evidence for the potential of genomics to transform the delivery of patient care, the road to successful implementation of personalized medicine is not without several hurdles that will have to be overcome. The analysis of the data produced from exome and genome sequencing is massive. For example, in one case of the use of whole genome sequencing to identify a novel genetic cause of acute promyelocytic leukemia in a patient, data processing, from DNA sample to validated mutation information delivered to the treating physician, took no less than 52 days.⁴⁷ Of course, this is expected to improve as computer hardware, algorithms, and work flows improve, although the storage and integration of such large amounts of data still represent a considerable challenge especially in the current electronic health record environment.

Even more challenging is the interpretation of these data. One recent study by MacArthur and colleagues⁴⁸ using data from the 1000 Genomes Project estimates that each person on average has approximately 100 loss-of-function mutations in their genome, challenging our ideas of what is benign and what is pathogenic. With estimates that the majority of genetic variants found in the population are extremely rare in frequency (many in only single individuals), how to determine the effects of such variants represents an unanswered question. Furthermore, although the cost of sequencing is rapidly plummeting, this does not include the costs of analysis and storage; and there are substantial regulatory hurdles (such as Clinical Laboratory Improvement Amendments (CLIA) certification) that must be faced before results of genomic sequencing can be used in clinical decision making.

Further challenges are extensions of challenges already faced by genetic counselors today. The ethical issues surrounding the sequencing of whole genomes are not trivial.⁴⁹ The concept of informed concept will need to be redefined in an era where a single test such as whole genome sequencing can give information about the risk of dozens to hundreds of possible conditions simultaneously, some of which will have no available therapy. Questions about how genomic information can affect discrimination, employment, and insurability abound, although the Genetic Information Nondiscrimination Act of 2008 affords some protections.

Finally, whereas the abundance of data continues to grow from the results of massive GWAS and sequencing projects, the translation of this information into new treatments and drugs has been less rapid. Genomewide association studies and sequencing studies produce hundreds of genetic associations with disease, but not all of these (or possibly even the majority) are necessarily causal. Further investigation is needed to solidify the pathways by which these genetic variants result in disease. Given the large complexity of these data, these investigations will require a shift moving away from the paradigm of reductionist biology to that of systems biology, with more sophisticated analytical and statistical methods to overcome the inherent challenges of this kind of framework.

In the end, will this deluge of genetic information (either from sequencing or genotyping) really have an impact on patient care? How do patients respond to getting such information? For sequencing, it is too early to know whether use of these data will routinely improve patient outcomes in a cost-effective manner. However, there has been some work exploring the use of less expensive and more analytically tractable approaches offered by genome-wide SNP genotyping.

In an interesting preliminary study, Bloss and colleagues⁵⁰ evaluated the impact of a direct-to-consumer genotyping test that included genetic risk assessments of heart disease on 2037 subjects. They found no significant differences in anxiety, diet, or exercise in general in those completing follow-up after receiving their risk assessment scores in an average of 5.6 months of follow-up. Although encouraging that subjects did not seem to suffer from increased distress from the results of these tests, it was also somewhat disappointing that these tests did not seem to increase the rate of patient lifestyle modifications for the better.

What is most lacking at this time are more targeted prospective trials to further evaluate how patients react to receiving genetic information and how clinical outcomes are affected using this information to guide management. This is especially important given the idea of genetic exclusivity, a term that describes the phenomenon that some people may react more to genetic data than other clinical information. To address at least one of these issues, we are currently conducting a randomized study within the Stanford Preventative Cardiology Clinic to determine whether we can (in a “real-world” setting) motivate patients to reduce cardiovascular risk profiles by providing them information about their inherited risk of CVD, as determined by a genotyping panel of 19 SNPs associated with an increased risk of CAD/MI in GWASs.⁵¹ The primary outcome of the trial will be to evaluate for any differences in LDL cholesterol levels between the 2 groups at 6 months, with secondary outcomes consisting of changes in other traditional risk factors, medication adherence, perceived risk, and psychological outcomes. A similar trial is under way in diabetes.⁵²

Conclusion

The genetic revolution spawned by public and private efforts along with technology development is still in the nascent stages. Although some have lamented that the discoveries have not led to major changes in clinical practice yet, there are now many examples in the areas of drug discovery, pharmacogenetics, and risk prediction that suggest that we are just beginning to harvest the fruits of our labor (Fig). As with many breakthrough technologies, reliable clinical trial data that show we can improve patient outcomes take some time to develop; and there remain other major hurdles to translate these discoveries to the general population. As these data become available over the next several years, it will become increasingly clear where we can best apply these resources. What is clear is that becoming familiar with the terminology, general principles, and pitfalls of genetic medicine will be increasingly important for all physicians.

Fig.

The current and potential future landscape of genetic diagnostic testing for CVD, with common examples of conditions or drug responses with known genetic components. Solid lines represent currently available diagnostics. Dashed lines represent potential genetic tests in development for clinical use and/or currently available for research purposes. Abbreviations: CAD = coronary artery disease, HTN = hypertension.

Abbreviations and Acronyms

AF

atrial fibrillation

BS

Brugada syndrome

CVD

cardiovascular disease

DCM

dilated cardiomyopathy

MI

myocardial infarction

NGS

next-generation sequencing

HCM

hypertrophic cardiomyopathy

LQTS

long QT syndrome

FH

familial hypercholesterolemia

GWAS

genomewide association study

TG

triglyceride

LDL

low-density lipoprotein

HDL

high-density lipoprotein

VLDL

very low density lipoprotein

SNP

single nucleotide polymorphism