Geoffrey S. Ginsburg
Personalized and Precision Cardiovascular Medicine
April 14, 2013 marked the 10th anniversary of the completion of the Human Genome Project. In just 10 years the field of genomics—the scientific study of genomes, their complete DNA sequences, and the functional interaction of their genes—has flourished as a result of high throughput technologies to generate, analyze, and interpret genome-derived data efficiently and cost-effectively. A broad aspiration of the Human Genome Project has been the concept of personalized medicine—a rapidly advancing field of health care that is informed by a person’s unique clinical, genetic, genomic, and environmental information.1 Personalized medicine seeks to couple established clinical-pathologic indices with state-of-the-art molecular profiling to create diagnostic, prognostic, and therapeutic strategies precisely tailored to each patient’s requirements—hence the term precision medicine. Although this concept is not entirely new, many patients and providers have had great expectations that the genome would enable the development of novel diagnostic and predictive tests as well as therapies based on an individual’s genetic information. This chapter presents a broad overview of the potential of personalized medicine. Subsequent chapters (Chapters 8 to 10, and 42) will elaborate specific approaches to various aspects of personalized medicine.
This decade also marks the 50th anniversary of the introduction of the term “factor of risk,” coined by William Kannel, principal investigator of the Framingham Heart Study (FHS).2 The risk factors for developing coronary artery disease (CAD)—male sex, hypertension, diabetes mellitus, increased low-density lipoprotein (LDL) cholesterol, tobacco use, and family history of heart disease—remain foundational for stratifying individuals to therapeutic strategies based on their risk of developing CAD. The FHS was among the first studies to illustrate the benefit of data integration to achieve refined risk classification. The massive and comprehensive collection of clinical and biologic data associated with the outcome of coronary disease enabled development of the Framingham predictive models3 and the resulting Framingham risk score (FRS).4 Today, it is anticipated that the inclusion of data that address the subtle distinctions in individuals revealed through genomic analyses might greatly enhance this prediction—a concept that has stimulated the development of genomic risk scores (GRSs) combined with the FRS (see later in this chapter) to enhance predictive accuracy. The opportunity for impact on clinical decision making offered by genome technologies lies in increased resolution: the potential to improve a person’s placement on the complex, multidimensional risk spectrum based on detailed, individual molecular characteristics on a genomic scale. The FHS example emphasizes the value of making use of the full spectrum of available clinical and demographic data; the genomic era simply expands this view toward integrated approaches that embrace and exploit genomic data in conjunction with other data.
Assessment of Disease Risk: Family Health History and Health Risk Assessments
Several approaches to risk assessment for cardiovascular disease have emerged that, if routinely used, might impact our ability to tailor chronic disease prevention strategies to the individual and promote improved cardiovascular public health. These include the FRS,4 the Reynolds risk score,5 and the European Society of Cardiology score.6 My colleague and I proposed a framework that includes family history assessment to identify high-risk persons for disease, thus enabling preventive and therapeutic interventions.7 Family health history (FHH) is a simple yet invaluable tool for the delivery of personal health risk information. Reflecting the complex combination of shared genetic, environmental, and lifestyle factors, a thorough FHH can approximate genetic/genomic risk information for integration into patient care. FHH assessments can identify persons at higher risk for common chronic diseases, enabling preemptive and preventive steps including lifestyle changes, health screenings, testing, and early treatment as appropriate.
Systematic collection of FHH for cardiovascular risk assessment was recently implemented in 24 family practices in the United Kingdom using a pragmatic cluster randomized controlled trial design, and demonstrated a highly significant (40%) increase in the identification of individuals at high risk.8 This was the first rigorously designed prospective study to show that the collection and use of FHH in a primary care setting can improve risk stratification for cardiovascular disease and health behaviors. Thus, ascertainment of FHH data is a feasible practice-level intervention that could improve cardiovascular risk assessment and help target patients who most need preventive interventions.
Family history and genomic testing are complementary techniques for evaluating health risks.9 Rather than choosing between the two, an approach that incorporates both types of information, in addition to nongenetic risk factors, promises the most accuracy. The combination of detailed family history, medical history, clinical evaluation, and genome sequence information, as exemplified by the ClinSeq Project at the National Human Genome Research Institute (NHGRI),10 may eventually provide the most accurate cardiac risk prediction.
A Genomic Toolbox for Personalized and Precision Medicine
Several genome-wide technology platforms are now routinely available for the exploration of the impact of the genome and its expressed products on health and disease states (see also Chapter 8). Concurrently, several cohort studies with longitudinal clinical data and biologic specimens sponsored by the National Heart, Lung, and Blood Institute (NHLBI) provide the opportunity for molecular analyses, disease classification, and predictive modeling. These include the FHS, the Coronary Artery Risk Development in Young Adults (CARDIA) study, the Atherosclerosis Risk in Communities (ARIC) study, the Jackson Heart Study (JHS), the Women’s Health Initiative (WHI) study, the Cardiovascular Health Study (CHS), and the Multi-Ethnic Study of Atherosclerosis (MESA). These powerful longitudinal studies and their clinical data and biospecimens can be accessed via the NHLBI’s BioLINCC program (https://biolincc.nhlbi.nih.gov), which contains a vast catalog of biospecimens resources that can be used to facilitate population genomics, using the tools outlined below. The discovery and development of genome-based biomarkers requires high-quality biospecimens linked to exquisitely defined phenotypes, assayed using one or more genome-based technologies. Their translation to clinical application forms the basis for personalized medical care.
DNA Variation
Genome-wide association studies (GWASs) emerged in 2005 as an unbiased strategy to provide information on common DNA variants associated with complex phenotypes. GWASs are predicated on the common disease–common variant hypothesis, which postulates that common diseases result from many disease-influencing alleles that occur at relatively high frequencies in the population, but individually have little predictive value. Nineteen published GWASs on CAD are recorded in the NHGRI Catalogue of GWASs (http://www.genome.gov/26525384), the largest being a meta-analysis of 63,746 CAD cases and 130,681 control cases.11 The total number of loci for CAD now exceeds 46. These loci encompass genes related to lipid metabolism and other CAD risk factors, but some novel loci—such as the region on chromosome 9 near the genes CDKN2A/CDKN2B—represent truly novel risk variants that will advance our understanding of the mechanisms underlying CAD. Together, these variants account for less than 10% of the heritability of CAD, suggesting the involvement of genetic factors beyond common variants.
Whole-Genome Sequencing
Advances in sequencing technologies have reduced costs such that a human genome can now be sequenced for less than $5000, and may be at the $1000 level in the coming year.12 At this cost, sequencing a patient’s genome will fall within the range of DNA-based diagnostic tests. More than 30,000 human genomes13 have now been sequenced and applied to elucidation of the biology and diagnosis of malignancies, rare genetic diseases, and microbial infections.14–16 Whole-genome sequencing also has advanced to the clinic, where it permits definitive diagnosis and even guides treatment.17–19 Although these approaches have yielded success when applied to mendelian disorders and cancer, methods for identifying rare variants for common diseases such as CAD are still nascent.20
Gene Expression
The genome-wide study of RNA expression levels includes a spectrum of molecules from mRNA to noncoding RNAs. Microarrays and RNA sequencing now can assay the entire complement of RNA expressed in a cell, tissue, or biologic fluid. Clustering of co-expressed genes using parametric or nonparametric methods provides the foundation for generating a “pattern” or “signature” of gene expression that is associated with a phenotype or physiologic state. These methods have been applied to classify a disease or to predict future disease states; the same data may also serve to generate molecular pathway information for the biologic mechanisms underlying disease. Two recent reviews nicely summarize the emerging gene expression–derived biomarkers for clinical applications in cardiovascular medicine.21,22
A surprising feature of the transcriptome is the significance of noncoding RNAs in the regulation of genes. Of particular interest are the expression patterns of small interfering RNAs (siRNAs) and microRNAs (miRNAs). Whereas siRNAs interfere with transcription through degradation of the message RNA, miRNAs work differently. The latter are usually 22 nucleotides in length, and through an miRNA-induced silencing complex, they inhibit gene expression on a post-transcriptional level by binding to complementary 3′ untranslated regions (UTRs) of target mRNA.23 The miRNAs play a role in several diseases and are advancing to clinical application in acute coronary syndromes,24 acute myocardial infarction (MI),25 cardiomyopathies,26 type 2 diabetes,27 hypertension,28 and heart failure.29 Most of these studies are small and require validation in larger populations.
Proteomics
Proteomics refers to the large-scale study of proteins, and the proteome often is considered to embody the full complement of proteins and their various derivatives (e.g., splice variants or post-translational modification) (see also Chapter 10). In the context of health and disease, proteomics seeks to define the full set of proteins associated with a particular physiologic state. Although this technology is relatively immature in its applications to human health and disease compared with RNA and metabolic profiling, application of these methods, combined with the development of mass spectroscopy technology, should advance proteomics to more routine use in disease classification and diagnosis, prognosis, and pharmacogenomics within the next several years.
Metabolomics
Metabolomics measures the approximately 5000 discrete small molecule metabolites and allows the identification of metabolic fingerprints for specific diseases. This technology may have practical use in the development of therapies, because metabolic changes immediately suggest enzymatic drug targets (see also Chapter 10). Similar to genomics and proteomics, metabolomics may be useful in disease diagnosis, prognosis, and drug development. Targeted mass spectroscopy–based metabolic profiling has been applied to cardiovascular disease to classify CAD and to predict ischemic events.30,31
Personalized and Precision Cardiovascular Medicine: Clinical Potential
Hypertension
Genetic variants associated with blood pressure (BP) that robustly replicate have finally emerged from GWASs. The single-nucleotide polymorphisms (SNPs) discovered have mainly been common variants (minor allele frequency [MAF] of ≥5%), with small effect sizes (mostly ≤1 mm Hg for systolic BP [SBP] and ≤0.5 mm Hg for diastolic BP [DBP]), and they collectively have explained only a small proportion (3% to 4%) of BP heritability. A recent GWAS investigated associations with SBP, DBP, mean arterial pressure (MAP), and pulse pressure (PP) by genotyping some 50,000 SNPs that capture variation in approximately 2100 candidate genes for cardiovascular phenotypes in 61,619 persons of European ancestry from cohort studies in the United States and Europe. Novel associations were identified for SBP (chromosomal locus 3p25.3 in an intron of HRH1; and 11p15 in an intron of SOX6, previously associated with MAP) and for DBP (1q32.1 in an intron of MDM4). Ten previously known loci associated with SBP, DBP, MAP, or PP were confirmed (ADRB1, ATP2B1, SH2B3/ATXN2, CSK, CYP17A1, FURIN, HFE, LSP1, MTHFR, SOX6; P < 2.4 × 10−6).32 These results represent a major advance in view of the fact that just a few years ago, almost no specific details were known about the genetic architecture of hypertension beyond the mendelian disorders. The results of ongoing fine-mapping studies of BP loci and sequence-based discovery of rare variants in extreme hypertensive cases and normotensive controls will provide further insights into the underlying genetic causes of BP, with the potential for improvement in the means for predicting and stratifying hypertension.
Coronary Artery Disease and Myocardial Infarction
As indicated previously, recent studies have identified a growing number of CAD-related and MI-related SNPs, and their results have stimulated additional studies to explore the value of these SNPs for risk prediction. Paynter and associates assessed the relationship of 101 SNPs to CAD in a cohort of 19,000 women, followed for 12 years, from the Women’s Genome Health Study.33 A GRS based on these 101 SNPs revealed a significant relationship between higher GRS and CAD, but failed to add incremental value to existing clinical models. Another GRS based on the counting of the number of “adverse” alleles influencing lipids has been shown to enhance risk prediction compared with measurement of lipids alone.34 Clinical adoption of GRS for CAD risk prediction will require unequivocal evidence that genotype predicts CAD, even after adjustment for plasma lipids and other known CAD risk factors.
Accompanying the transformative discoveries on genetic susceptibility variants just described are additional predictive CAD and MI biomarkers emerging from the expressed genome. Rosenberg and colleagues found that the gene expression signature of 23 genes obtained from the peripheral blood of nondiabetic patients undergoing coronary angiography for acute chest pain permitted reclassification of the risk of having CAD, at a rate of approximately 20% of that for traditional clinical models.35 The negative predictive value of 83% for the gene expression assay compared favorably with typically used clinical tests such as myocardial perfusion imaging. In addition, Voora and co-workers recently reported the development of an RNA signature associated with the platelet response to aspirin and the ability of that same signature to predict acute coronary syndromes in two cohorts.36
