Unraveling the Puzzle of the Role of Heritability in the Variability of the QT Interval Using Exome Array Analysis

Since the discovery the intricate structure of DNA by Watson and Crick¹, and the completion of sequencing of the entire human genome only a few decades later², genomic medicine has taken large strides in discovering novel disease causing genes and modifiers in an effort to better diagnose, manage and treat various diseases. The ultimate goal of all these studies is the one of personalized medicine catered to a person’s unique genomic fingerprint. Not only have the sequencing techniques become significantly cheaper - and even affordable to be utilized as mail-order kits -, genome and exome datasets have become large enough to perform large consortium studies to find novel genes or loci associated with common traits, such as QT prolongation of the surface ECG.

QT prolongation is associated with a risk for significant ventricular arrhythmias, torsades de pointes and sometimes ventricular fibrillation and sudden cardiac death (SCD). Not only is QT prolongation the pathognomonic feature of the rare, inherited condition known as congenital long QT syndrome (LQTS; estimated prevalence 1 in 2,000 individuals)³, it is often associated with far more common conditions, such as electrolyte imbalances (especially potassium and magnesium), certain comorbidities, or in combination with prescription medications known to effect the QT interval, such as ondansetron or amiodarone.⁴ In fact, attenuation of the QT interval is one of most common reasons of FDA ‘black box’ warnings or medications being halted in development because of the risk of SCD.⁵ And while the genetic causes for approximately 80% of the predominantly autosomal dominantly inherited LQTS have been established, the genetic fingerprint of QT-vulnerability for the general population, for example that what seems to make one person more susceptible to a drug-induced QT prolongation than another, remains to be determined. The fact that these effects are far more common than LQTS alone and the effect size on the QT interval is far less or only provoked by its circumstances, suggest they are likely by common variants or possibly a combination thereof. However, while monogenic diseases can be discovered using traditional linkage studies in pedigrees with penetrant disease or by candidate gene analyses in cohorts of patients, variants that only exert a moderate to small effect on the phenotype or clinical trait is a more complex puzzle to solve. Nevertheless, understanding the exact underpinnings of these polygenic diseases is an exciting area of research interest.

The first pieces to solving this puzzle in relation to the ECG-derived QT-interval were provided in studies by Arking et al.⁶ spearheaded by Dr. Newton-Cheh – both authors on the current article. Using a large collection of individuals of European ancestry (>76k) from a large consortium, they performed a genome wide association study (GWAS) meta-analysis across all human autosomes, and discovered 35 common variant loci that collectively explain approximately 8 –10% of QT interval variation. In their results, some of the loci were found in regions of known genes associated with LQTS (KCNQ1, KCNH2, SCN5A), but novel loci were identified as well, several of which associated with calcium regulation or cardiac repolarization.⁶ This study set the stage for several replication studies, but although these were able to identify loci that were significant with effects size attributable to each locus, they did not evaluate the intergenic and/or cumulative effect of these variants. To this end Strauss et al. introduced the genetic risk score (GRS) for the QT-interval to test the hypothesis that a weighted combination of the 61 common genetic variants previously associated with variability in the QT interval could predict individual response to multiple QT-prolonging drugs⁷. And, in a small, pilot study of 22 patients, they were in fact able to show that the cumulative QT-score explained the significant variability in QT response for all 3 tested QT prolonging drugs: 30% for dofetilide (p=0.02), 23% for quinidine (p=0.06) and 27% for ranolazine (p=0.03). Furthermore, the cumulative QT-GRS was a significant predictor of drug-induced torsades de pointes compared to controls (r² = 12%, p=1×10⁻⁷). Although such studies are promising, the rare and low frequency polygenic determinants of the QT interval remained unexplored that could account for the missing heritability in the variability of the QT interval, including such genetic determinants of the other intervals of the cardiac action potential.

In this issue of Circulation: Genomic and Precision Medicine, Bihlmeyer et al.⁸ have taken a step forward in understanding these complex interactions. And, while their approach seems similar to previous studies, there are two key differences as they i) include evaluation of the JT interval and, more interestingly, ii) choose to utilize an ExomeChip rather than the traditional GWAS. Although an exome array analysis focuses on only ~2% of the genome that would be covered by GWAS, it allows the results to more directly implicate gene and (abnormal) gene function in disease or the studied trait by targeting coding single nucleotide variants (SNVs) and allows the identification of significant low frequency and rare variants typically requiring large sample sizes. In their study, Bihlmeyer et al.⁸ performed a meta-analysis on over 95,000 individuals of predominantly European ancestry from 23 centers and analyzed 209, 449 common and rare variants in ~17,000 genes, and discovered 10 new loci that modulate the QT- (6 novel loci) and JT interval (4 novel loci), bringing the total number of repolarization-associated loci at 45. As expected, several previously described loci emerged and while the findings in the current study thereby only might seem incremental, the new approach has elucidated some interesting and important insights. First, by using an ExomeChip approach the results allowed for direct implication of identified variants in possible abnormal gene function. In fact of 6 novel QT-associated loci, 4 were nonsynonymous variants in 4 genes (PM20D1, SLC4A3, CASR and NRAP), while 3 novel coding variants were found to be associated with the JT interval (SENP2, SLC12A7 and NACA). This identification thereby not only identifies these novel loci for each of the intervals, it instantly upgrades these 7 genes for consideration as potential disease causing genes for LQTS, or one of the other repolarization disorders, like Brugada Syndrome. Secondly, by adding a Gene-Wide Significance (GWiS) analysis to their findings, the authors were able to implicate 17 genes (7 novel) which, in conjunction with their other exome array findings implicate known pathways involved in potassium, sodium and calcium ion regulation and autonomic control of the QT interval and pathways with proteins that encode for cytoskeletal and myocyte structure opening the door for new disease gene discovery as well as providing possible targets for anti-arrhythmic (gene) therapies. In using this previously published GWiS approach⁹, which accounts for the number of independent effects within a gene and is able to maintain power for studies assessing the association of low-frequency alleles, the authors are able to determine the independent effect of genes identified in the ExomeChip wide analysis. The authors by using this approach, for example, were able to attribute the significance of DLEC1 and SCN5A to non-coding variants and that of SCN10A to coding variants despite coding variants being significant in all 3 genes. This type of analysis could be especially useful to guide laboratory-based studies to elucidate the significance of a variant on gene function and protein expression. The interpretation of the results of this analysis however is clearly dependent on which genetic variants are used for the analysis and since not all variants implicated in prior GWASs were assayed using the ExomeChip, the GWiS analysis could not be conclusively applied to several QT associated genetic loci in the entire study population.

One of the limitations of this study is the lack of replication of the exome array findings although an argument could be made regarding the study’s legitimacy in its ability to replicate prior GWAS findings and that by performing a meta-analysis each cohort serves to validate the results from all the other cohorts. Another limitation is the potential relevance of these findings in other ethnic and racial groups due to relatively low representation of Chinese (n=750) and Hispanic (n=1,382) populations. Although correlation of the effect sizes between European-Americans and African-Americans was robust (r=0.801), it was less so in the Chinese (r=0.238) and Hispanic (r=0.227) populations. Although genetic effect sizes may be similar, risk allele frequencies maybe very different across various populations resulting in the inability to identify novel or previously identified significant genetic variants if sample sizes are small.

As compared to the QT interval the clinical significance of the JT interval is less clear, especially in the context of a normal QRS duration and the absence of left or right bundle branch block. The QT interval is an accurate and more comprehensive measure of cardiac repolarization in the absence of conduction abnormalities, since the inter-ventricular septum upon its early activation, already begins repolarization while other areas are getting activated, which ultimately is reflected within the QRS duration. The JT interval has been proposed as an alternative measure of repolarization to the QT interval in the presence of QRS prolongation as seen in bundle branch blocks, and is in fact predictive of cardiovascular events in men, and mortality in the general population in those with a wide QRS complex^10,11. The predictive value of JT and QT intervals for cardiac events and mortality is similar in individuals with LQTS, and the general population respectively^11,12. Therefore its clinical value over-and-above that of the QT interval is primarily seen in those with a QRS ≥ 120 ms, generally an exclusion criterion for this and other genetic studies. Therefore the electrophysiological and clinical implications of discovering novel genetic determinants of the JT interval remain unclear.

Despite some of these limitations, the results of this genetic study related to the QT interval may lead to exciting next steps. First of all, similar to the use of previously identified QT loci used for the derived GRS, the new loci identified in this study could be added to this model and be tested as the ‘QT-risk score 2.0’. The initial risk score was tested in a small, albeit prospective group of patients, but a subsequent large, prospective study of the QT GRS, including all novel loci, could shine light on risk markers for susceptibly for QT prolongation by QT prolonging drugs, increased risk for arrhythmias during electrolyte abnormalities or potentially even explain variable expressivity and penetrance in patients with congenital LQTS. Secondly, these studies may pave the way to fully understanding and potentially delivering the personalized genome, although caution and patience is imperative. From the well-characterized, but relatively straightforward QT GRS of 61 variants⁷, and other clinical variables that impact on the QT interval detected combined with the use of larger electronic health records datasets, machine learning and artificial intelligence, we can begin to imagine the development of much more complicated predictive scores, which would cover several traits or phenotypes that could aid in risk assessments and evaluate treatment outcomes. Studies such as these offer the start to solving the puzzle, and without these pieces, the full picture of not only the heritability of the QT interval,^l but other conditions that affect cardiac repolarization, such as LQTS and drug induced torsades des pointes would be impossible to see.