Common Coding Variants in SCN10A Are Associated With the Nav1.8 Late Current and Cardiac Conduction
Background: Genetic variants at the SCN5A/SCN10A locus are strongly associated with electrocardiographic PR and QRS intervals. While SCN5A is the canonical cardiac sodium channel gene, the role of SCN10A in cardiac conduction is less well characterized.
Methods: We sequenced the SCN10A locus in 3699 European-ancestry individuals to identify variants associated with cardiac conduction, and replicated our findings in 21,000 individuals of European ancestry. We examined association with expression in human atrial tissue. We explored the biophysical effect of variation on channel function using cellular electrophysiology.
Results: We identified 2 intronic single nucleotide polymorphisms in high linkage disequilibrium (r 2=0.86) with each other to be the strongest signals for PR (rs10428132, β=−4.74, P=1.52×10−14) and QRS intervals (rs6599251, QRS β=−0.73; P=1.2×10−4), respectively. Although these variants were not associated with SCN5A or SCN10A expression in human atrial tissue (n=490), they were in high linkage disequilibrium (r 2≥0.72) with a common SCN10A missense variant, rs6795970 (V1073A). In total, we identified 7 missense variants, 4 of which (I962V, P1045T, V1073A, and L1092P) were associated with cardiac conduction. These 4 missense variants cluster in the cytoplasmic linker of the second and third domains of the SCN10A protein and together form 6 common haplotypes. Using cellular electrophysiology, we found that haplotypes associated with shorter PR intervals had a significantly larger percentage of late current compared with wild-type (I962V+V1073A+L1092P, 20.2±3.3%, P=0.03, and I962V+V1073A, 22.4±0.8%, P=0.0004 versus wild-type 11.7±1.6%), and the haplotype associated with the longest PR interval had a significantly smaller late current percentage (P1045T, 6.4±1.2%, P=0.03).
Conclusions: Our findings suggest an association between genetic variation in SCN10A, the late sodium current, and alterations in cardiac conduction.
See Editorial by Maier et al
Genetic variants at the SCN5A/SCN10A locus are strongly associated with electrocardiographic PR and QRS intervals. In addition, variants in SCN10A have also been linked to atrial fibrillation. Although SCN5A is the canonical cardiac sodium channel gene responsible for the upstroke of the cardiac action potential, the role of SCN10A is less well characterized in cardiac conduction. In this study, we have characterized the electrophysiological function of several missense coding variants in SCN10A identified from large populations where the PR interval was known. We found that these missense variants all clustered in the cytoplasmic linker between 2 domains of the channel. When these missense mutations were generated and characterized via patch clamp, they were found to alter the amount of late sodium current present. Genetic haplotypes coding for channels that were associated with shorter PR intervals had a larger late sodium current, and therefore, our results suggest that there is an association between the amount of late sodium current and the PR interval. Clinically, this may be relevant as longer PR intervals are associated with a higher risk of atrial fibrillation, and therefore, further characterization of the role of SCN10A may provide a novel target for future therapies.
The PR and QRS intervals on the surface ECG reflect atrioventricular conduction and ventricular depolarization, respectively. PR interval prolongation is associated with increased risk of atrial fibrillation, the need for pacemaker implantation, and death.1 Similarly, QRS interval prolongation is associated with increased risk for heart failure and death.2,3 Genome-wide association studies (GWAS) have identified a locus on chromosome 3 associated with the PR and QRS intervals, with the most significant signal mapping to an intronic region of SCN10A.4–7 SCN10A encodes for the voltage-gated sodium channel, Nav1.8, and is flanked by SCN5A and SCN11A, both of which encode for voltage-gated sodium channels, Nav1.5 (cardiac) and Nav1.9 (neuronal), respectively.8,9
Until recently, Nav1.8 channels were thought to be restricted to the dorsal root ganglia and peripheral sensory neurons.10,11 However, we have previously shown that SCN10A is expressed in the myocardium and preferentially in the specialized Purkinje fibers of the cardiac conduction system.6 Further, blocking the Nav1.8 channel in mice with a selective antagonist, A-803467, prolongs the PR and QRS intervals, reduces the late sodium current (INa,L) in ventricular myocytes, and slows action potential firing in intracardiac neurons, suggesting a direct role for SCN10A in cardiac electrophysiology.6,12,13 In GWAS of PR and QRS intervals, 2 common SCN10A coding variants have been identified, both significantly associated with cardiac conduction.4–7 Recent work has also found that rs6801957, the most significant intronic single nucleotide polymorphism (SNP) in SCN10A associated with the QRS interval, alters a transcription factor–binding site for Tbx5 in mice that regulates the expression of SCN5A.14–16
To further define the relationship of variation in SCN10A with cardiac conduction, we sequenced the SCN10A gene in the CHARGE-TSS (Cohorts for Heart and Aging Research in Genomic Epidemiology Targeted Sequencing Study) to identify rare and common variants in SCN10A. We examined the association of these variants with atrioventricular conduction. We then replicated our findings among European and African-ancestry subjects sequenced as part of the Exome Sequencing Project (ESP) and by direct genotyping in a larger sample of CHARGE participants. Next, we examined whether coding variants are functionally related to cardiac conduction using cellular electrophysiology. Finally, we sought to determine whether the most significant SCN10A signal is associated with SCN5A and SCN10A transcript expression in humans.
Study Populations and Genetic Assays
Discovery Study Sample: CHARGE Targeted Sequencing Study
The CHARGE Consortium is a consortium of population- and community-based cohorts designed to facilitate studies of genetic epidemiology.17 CHARGE-TSS included a random sample, stratified by sex, of at least 1000 ARIC (Atherosclerosis Risk in Communities) Study, 500 CHS (Cardiovascular Health Study), and 500 FHS (Framingham Heart Study) participants of European ancestry. In each cohort, investigators selected an additional sample of individuals from the extremes of the phenotype distribution of several cardiometabolic traits, including PR and QRS intervals. Loci for targeted sequencing were identified by GWAS conducted by the CHARGE consortium for several phenotypes. The present analysis focuses on targeted sequencing of regions of the SCN10A gene for PR and QRS intervals. In total 3699 participants of European ancestry were used in this analysis, including individuals selected on the basis of extreme PR, QRS, or other cardiometabolic traits and a cohort random sample. All studies performed were approved by the Institutional Review Board of the participating institutions. Details of case and cohort selection and sequencing can be found in Methods in the Data Supplement.
Replication and Haplotype Study Sample: Exome Sequencing Project
The National Heart, Lung, and Blood Institute ESP, a parallel exome sequencing consortium, served as the replication cohort. In total, our replication sample included 607 ESP participants of European ancestry and 972 African American participants. Haplotypes were derived from all ESP participants, increasing the ESP–Haplotype sample to 4306 participants of European ancestry and 972 African American participants. Full details of the ESP cohort selection are available in Methods in the Data Supplement.
Extension Study Sample: CHARGE Exome Chip Cohorts
To increase the sample size, we examined directly genotyped data from the entire ARIC, CHS, and FHS cohorts. Participants with a QRS interval >120 ms were excluded from the QRS interval analyses. A total of 20 666 European ancestry participants (CHARGE-Exome sample) were analyzed. Full details of cohort selection are available in Methods in the Data Supplement.
Genotypes of both European and African-descent populations from ESP were phased separately by race using PHASE.18 PHASE estimates of haplotype frequency were reported. Full details are available in Methods in the Data Supplement.
Gene Expression Analysis
We performed an expression quantitative trait loci analysis using the most significant 5 SNPs at the SCN10A locus associated with PR or QRS intervals in this study or in prior GWAS efforts of PR or QRS intervals (rs10428132, rs6795970, rs6801957, rs6800541, and rs6599250). Please see Methods in the Data Supplement for details.
Human SCN10A Clone, Mutagenesis, Expression, and Characterization
The human SCN10A alpha subunit cDNA built in a pCMV6-XL5 vector was purchased from OriGene. The SCN10A DII–DIII linker variants/haplotypes (I962V [IV], P1045T [PT], V1073A [VA], L1092P [LP], VA+LP, IV+VA, and IV+VA+LP) were generated using appropriately designed mutagenic primers. The SCN10A constructs were expressed in a Neuroblastoma 2a cell line derived from mouse (ATCC, catalog number CCL-131). Neuroblastoma 2a cells were used because these neuronal cells expressed functional SCN10A channels after transfection. SCN10A expression was verified in freshly obtained human atrial tissue via immunohistochemistry. Wild-type (WT) and variant/haplotype Nav1.8 currents were measured using the whole-cell patch clamp recordings. Tetrodotoxin (150 nmol/L) was used to block endogenous sodium currents. The data are presented as mean±standard error, and a student’s t test was used to determine a significant difference (P<0.05). Full details of patch clamping methods and analysis are available in Methods in the Data Supplement.
Sequence Results in Discovery Samples
We sequenced an ≈117 kilobase region encompassing the SCN10A gene in a discovery sample of 3699 individuals of European ancestry from the ARIC study, the CHS, and the FHS (Table I in the Data Supplement). We identified 3005 noncoding (intronic/intergenic) and 204 coding variants (Tables II and IIIA in the Data Supplement). There was evidence of selection pressure against putatively functional variants; nonsynonymous variants tended to have a lower minor allele frequency than synonymous variants (P=0.046), and rare nonsynonymous variants were more likely to be predicted as having adverse effects compared with common nonsynonymous variants, P=0.014–0.000013 (Table IIIB in the Data Supplement).
Association of Common Variants With PR and QRS Intervals
We examined the association of cardiac conduction with common and rare SCN10A variants. The SNP with the most significant association with the PR (rs10428132, β=−4.74; P=1.5×10−14; n=3699; Figure 1A; Table IV in the Data Supplement) and QRS (rs6599251, β=−0.73; P=1.2×10−4; n=3699; Table IV in the Data Supplement) intervals were both intronic and in high linkage disequilibrium (LD) with each other (r2=0.86). These variants are also in high LD (r2≥0.72), with the most significant SNPs identified by prior GWAS of the PR or QRS intervals and, importantly, with the most significant nonsynonymous SCN10A coding SNP rs6795970, VA (Table 1; Tables IV and V in the Data Supplement).4–7 The most significant SCN10A SNPs associated with PR (rs10428132) and QRS (rs6599251) were not in LD, with common variants in 2 neighboring loci, SCN5A and SCN11A, which also encode for voltage-gated sodium channels (Figure I in the Data Supplement).
Expression Quantitative Trait Loci Analyses for SCN5A and SCN10A and Nav1.8 Channel Protein Expression
Examining transcripts from human atria (n=369 samples), ventricle (n=313 samples), and other tissues, we found no evidence that SCN5A transcript expression is associated with the PR and QRS index SNPs identified in the current study (rs10428132 or rs6599251), nor with the most significant SCN10A SNPs from prior studies of PR or QRS, including rs6801957 (Figure II, Methods, Results, and Table VI in the Data Supplement).14,15 Although we found expression of Nav1.8 channel protein in freshly obtained human atrial tissues using immunohistochemistry (Figures III and IV in the Data Supplement), expression of SCN10A transcript was low in atrial and ventricular tissue, therefore, hampering our ability to identify an SNP SCN10A expression quantitative trait loci association. Nonetheless, in an additional 121 samples from human left atrial appendage, we found no association by quantitative polymerase chain reaction between SCN5A and SCN10A transcript expression and the most significant SCN10A SNP (rs6800541) identified by prior GWAS of the PR interval (Figure V and Table VII in the Data Supplement).5,19
Although the most significant intronic PR and QRS SNPs were not associated with expression of the 2cardiac sodium channels, SCN5A and SCN10A, they were in high LD (r2=0.83 and 0.72, respectively) with a common SCN10A missense variant, rs6795970 (VA). We therefore examined common SCN10A missense variants for association with cardiac conduction. We identified 7 common missense SNPs in SCN10A among individuals of European descent, and 4—IV, PT, VA, and LP—were associated with PR interval duration (Table 1; Table II in the Data Supplement). The 4 SNPs associated with PR interval duration clustered within a 130 amino acid residue region of the cytoplasmic linker that connects the second and third domains (DII–DIII linker) of the Nav1.8 channel (Figure 1B). The VA and PT variants were associated with the largest effects on PR interval and were nominally associated with QRS duration (Table 1; Table IV in the Data Supplement).
We sought to replicate and extend our significant findings from the CHARGE-TSS in additional samples using 2 approaches. First, the coding regions of SCN10A were sequenced among those of European and African descent as part of ESP. Both the VA and PT variants were associated with PR interval in 607 European-descent individuals who were part of our ESP replication sample. The other 2 coding variants, IV and LP, although not reaching statistical significance in the smaller sample size of ESP, showed an equal or larger magnitude of effect on PR interval. We did not find significant associations among African Americans (n=972); however, power to identify a significant association was more limited because of lower minor allele frequencies and smaller sample sizes (Table 1).
In our second approach, we directly genotyped the 7 SCN10A common coding variants in 20 666 individuals from the CHARGE Exome Chip Project (FHS, ARIC, and CHS) using the Illumina HumanExome v1.0 array (Table 1; Table I in the Data Supplement). We found that all 4 common coding variants (IV, PT, VA, and LP) were significantly associated with both PR (P values ranging from 10–7 to 10–49; Table 1) and QRS (P values ranging from 10–4 to 10–14; Table IV in the Data Supplement) intervals.
To determine whether the 4 common coding SCN10A SNPs associated with the PR interval were independent from SNPs in neighboring sodium channel genes, SCN5A and SCN11A, we performed a conditional analysis of the PR interval using the ESP cohort and common SNPs in SCN5A and SCN11A. We found that both VA and PT remained significantly associated with PR interval duration and that the effect size was preserved among the 4 common coding SNPs (Table VIII in the Data Supplement).
Three of the 4 SNPs associated with the PR interval were in moderate to high LD, Table V in the Data Supplement, and 6 common haplotypes were present among individuals of European or African ancestry (Figure 1C). Haplotype analyses were performed using 5278 individuals of European and African ancestry sequenced in ESP. The most frequent haplotype among individuals of European ancestry (frequency of 36%) was designated the WT SCN10A haplotype and was associated with a mean PR interval of 161 ms among European Americans and 169 ms among African Americans. The VA haplotype and IV+VA+LP haplotype both had mean PR intervals that were shorter compared with WT in both ethnic groups (Figure 1C). By contrast, the PT haplotype had a significantly longer mean PR interval compared with the WT (Figure 1C). There was suggestive evidence that haplotype structure influenced PR interval given that the VA+LP haplotype had a longer mean PR interval compared with the VA only haplotype (P=0.02). Similar to the PR findings, the PT haplotype tended to have the longest QRS interval; however, no significant haplotype–QRS associations were detected, due in part to the smaller effect sizes with QRS (Table IX in the Data Supplement).
Biophysical Effects of SNPs and Haplotypes on Channel Function
Given the common location of the 4 nonsynonymous variants in the DII–DIII linker of SCN10A (Figure 1B), we sought to determine the biophysical effects of the haplotype combinations on channel function using cellular electrophysiology. SCN10A WT and haplotype whole-cell currents were acquired by heterologous expression in N2A cells using patch clamp electrophysiology (See Methods in the Data Supplement). Analyses were performed for all 6 common haplotype combinations observed, as well as for the 2 variants, IV and LP, which were not seen in isolation among European Americans and African Americans (Table 2; Figure VI in the Data Supplement).
Representative current traces of the most frequently observed haplotypes are illustrated in Figure 2A. Plots of current density versus voltage for the 4 most common haplotypes are shown in Figure 2B. The VA haplotype had a significantly larger current density than the WT at +10 mV, while the other haplotypes were not significantly different from WT (Table 2). Plots of conductance–voltage (G-V) relationships are shown in Figure 2C. We observed a subtle relationship in voltage-dependent opening (V1/2) and PR interval among the haplotypes (Figures VI and VII in the Data Supplement; Table 2). The midpoint of channel opening (V1/2) of the G-V relationship for PT was significantly right-shifted (+11.8±1.8 mV; P=0.001; n=7), and both VA (−6.3±1.5 mV; P=0.0002; n=6) and IV+VA+LP (−3.5±1.0 mV; P=0.0008; n=7) were significantly left-shifted compared with WT (+3.0±1.0 mV; n=7). There was no difference in the slope for voltage-dependent opening (Table 2). Furthermore, there was no relation between haplotype differences in the PR interval and voltage-dependent inactivation (Figure 2D and Table 2), the recovery from inactivation (Figure VIII in the Data Supplement; Table 2), or open-state inactivation (Figure IX in the Data Supplement).
A distinct feature of the Nav1.8 channel compared with other sodium channels is the persistence of a significant INa,L. A visual inspection of the recordings (Figure 2A) suggested a difference in the INa,L between haplotypes. To assess this difference systematically, we measured INa,L at the end of a test pulse (+20 mV at 475 ms) when the inactivated current reached steady state, which was then normalized to the peak current. We observed marked differences in INa,L between haplotypes ranging from 6.4% to 22.4%. Compared with WT INa,L (11.7±1.6%, n=7), smaller INa,L for the PT (6.4±1.2%; P=0.03; n=5) and larger INa,L for the IV+VA+LP (20.2±3.3%, P=0.03, n=5) haplotypes were observed (Figure 2E and Table 2). We next sought to determine if there was a relation between haplotype differences in the PR interval and INa,L. When the haplotypes were arranged from the shortest to the longest PR intervals, we observed an inverse linear relation between percent INa,L versus the PR interval, Figure 2F (percent INa,L, r2=0.83, P=0.001).
To comprehensively examine the associations at this locus with cardiac conduction, we secondarily examined association of rare SCN10A variants in aggregate (locus-wide or limited to coding variants) with either the PR or QRS intervals in ≈3699 individuals from CHARGE-TSS and 1579 individuals from ESP. Although no significant association was identified (P>0.05 for all associations tested), power was limited because of a low cumulative allele frequency.
By sequencing the SCN10A locus, we identified a cluster of common coding variants in the DII–DIII linker that influence cardiac depolarization and conduction. We have identified an inverse linear relationship between the percent INa,L and the PR interval duration. Haplotypes that reduce the Nav1.8 INa,L were associated with a longer PR interval, and conversely, haplotypes that increased the Nav1.8 INa,L were associated with a shorter PR interval compared with WT.
The functional role of SCN10A/Nav1.8 channel on cardiac conduction has only recently begun to be elucidated.20 Increasing experimental data suggests that the SCN10A/Nav1.8 channel may have a direct role in cardiac conduction. First, the specific SCN10A blocker, A-803467, lengthened the PR and QRS intervals in mice6 and reduced the SCN10A/Nav1.8 INa,L in isolated ventricular myocytes from mouse and rabbit.12 Second, the SCN10A/Nav1.8 channel has been identified in the specialized cardiac conduction fibers and atrial myocytes, as well as in cholinergic and, to a lesser extent, sympathetic intracardiac neurons (Table X in the Data Supplement). Third, SCN10A missense variants are associated with PR and QRS intervals. Taken together, these data provide evidence that the SCN10A/Nav1.8 channel participates in cardiac conduction.
Our results suggest a correlation between the amount of INa,L and the PR interval measured among our study cohorts. The PR interval is modulated by many factors, and the correlation observed in this study is not necessarily causative. SCN10A influence on atrioventricular conduction may be secondary to neuronal modulation, a direct effect on cardiomyocyte depolarization, or both.20 Supportive of the cardiomyocyte hypothesis, we found qualitative evidence of Nav1.8 expression in human atrial cardiomyocytes using immunohistochemistry, similar to prior findings with a different antibody.21 Although we acknowledge that there is no quantitative data on SCN10A expression in human cardiac conduction tissue, we have documented SCN10A expression in the murine Purkinje system previously.6 Alternatively, SCN10A has also been identified in intracardiac neurons and in cardiac ganglionated plexi.13 Recently, injection of A-803467 into cardiac ganglionated plexi was shown to be protective from atrial fibrillation, underscoring the potential for neuronal SCN10A expression to influence the electrophysiology of the myocardium.22 Future studies of SCN10A expression in the human conduction system and cardiac innervation may further elucidate the role of SCN10A in atrioventricular conduction.
Our findings highlight the role of the DII–DIII linker in Nav1.8 channel inactivation. Previous studies indicated that point mutations in the DII–DIII linker of cardiac (SCN5A) and brain (SCN1A) voltage-gated sodium channels both increased the INa,L in patients with type III long-QT syndrome and epilepsy, respectively.23,24 Crystal structures of sodium channels showed that the S6 transmembrane spanning region forms the inner lining of the pore.25 Because the DII–DIII linker connects the S6 of domain II with S1 of domain III, we hypothesize that this region of the channel has an important role in the process of prolonged inactivation. In support of that conclusion, several variants in this linker are associated with Brugada Syndrome,26 including VA, which was independently shown to cause a higher INa,L density with an A in position 1073 compared with V, similar to our findings.27 In that study, the higher current density allele (VA) was protective from Brugada Syndrome, signifying that haplotypes at this position may have multiple effects depending on relative expression in atrial, ventricular, and conduction system myocytes.
In this study, we did not find an association between the intronic SCN10A SNP and expression levels of SCN5A in human atrial or ventricular tissue. However, the relation between the SCN5A/SCN10A genomic region and cardiac conduction is highly complex, with at least 6 independent association signals for QRS interval that span both genes.6 The intronic SCN10A SNP associated with the QRS interval, rs6801957, modulates the function of a Tbx5 and Tbx3 enhancer.16 A recent study demonstrated that while no SCN10A transcript could be identified in ventricular cardiomyocytes in either WT or SCN10A−/− mice, Nav 1.8 contributes to late sodium current at slow rates, perhaps through modulation of SCN5A.28 In our study, we did not characterize Nav1.8 or Nav1.5 directly in the human ventricular myocardium or Purkinje system; however, these will be important future investigations to define the contributions and potential interactions of the Nav1.8 and Nav1.5 channels on ventricular conduction in humans.
Our study was subject to several limitations. First, although we examined individuals of both European and African descent, our results may not be generalizable to other ethnicities. Second, despite performing sequencing in 3699 and 1579 individuals in the discovery and replication samples, respectively, our power to detect modest associations, particularly with rare variants, was limited. Third, we did not examine whether these genetic variants were related to expression of SCN10A in the conduction system. Fourth, our genetic association and heterologous electrophysiology data regarding SCN10A are correlative and do not imply causality. Finally, in vitro cellular electrophysiology studies may not fully recapitulate the cellular milieu of the cardiomyocyte or the cardiac conduction system.
In conclusion, our sequencing and functional data provide new insights for the physiological role of SCN10A/Nav1.8 channel as a contributing factor to human cardiac conduction. As the SCN10A/Nav1.8 channel has only recently been identified as a novel modifier on cardiac conduction and electrophysiology, further studies are necessary to define the relative contributions of the Nav1.8 and Nav1.5 channels on cardiac conduction.
We acknowledge support from the National Heart, Lung, and Blood Institute (NHLBI) and the contributions of the research institutions, study investigators, field staff, and study participants in creating this resource for biomedical research. We thank the Lifeline of Ohio Organ Procurement Organization and the Division of Cardiac Surgery at The OSU Wexner Medical Center for providing the explanted human cardiac tissue for immunohistochemistry.
Sources of Funding
The current work was funded by National Institutes of Health (NIH) grants: 1R01HL088456, 1R01HL111089, 1R01HL116747, 2K24HL105780, R01HL128914, 1RO1HL092577, HL088577, HL10599, HL115580, HHSN268201100005C, HHSN268201100006C, HHSN268201100007C, HHSN268201100008C, HHSN268201100009C, HHSN2682011000010C, HHSN2682011000011C, and HHSN2682011000012C, R01HL087641, R01HL59367, R01HL086694, HL103010, HL102924, HL102925, HL102926, 5RC2HL102419, N01-HC-25195, N02-HL-6-4278, HHSN268201200036C, HHSN268200800007C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, N01HC85086; U01HL080295, R01HL087652, R01HL105756, R01HL103612, R01DK089256, R01HL120393, R01AG023629, K23HL114724 HHSN268201800001C, R01HL130114, UL1TR000124, DK063491, HHSN268201500001I, R01HL105993, R01HL090620 and R01HL111314, CTSA UL1-RR024989, Cleveland Clinic Department of Cardiovascular Medicine philanthropy research funds, and the Tomsich Atrial Fibrillation Research Fund. by American Heart Association (AHA) grants: 13EIA14220013, 12FTF11350014; by the Max Schaldach Fellowship in Cardiac Pacing and Electrophysiology and the MGH Fund for Medical Discovery; by the Laughlin Family; and by the Fondation Leducq (14CVD01).
Dr Ellinor is the PI on a grant from Bayer AG to the Broad Institute focused on the genetics and therapeutics of cardiovascular diseases. The other authors report no conflicts.
Guest Editor for this article was Christopher Semsarian, MBBS, PhD, MPH.
The Data Supplement is available at http://circgenetics.ahajournals.org/lookup/suppl/doi:10.1161/CIRCGEN.116.001663/-/DC1.
- Received November 17, 2016.
- Accepted March 2, 2018.
- © 2018 American Heart Association, Inc.
- Abrahamsen B,
- Zhao J,
- Asante CO,
- Cendan CM,
- Marsh S,
- Martinez-Barbera JP,
- et al
- Yang T,
- Atack TC,
- Stroud DM,
- Zhang W,
- Hall L,
- Roden DM
- Verkerk AO,
- Remme CA,
- Schumacher CA,
- Scicluna BP,
- Wolswinkel R,
- de Jonge B,
- et al
- Psaty BM,
- O’Donnell CJ,
- Gudnason V,
- Lunetta KL,
- Folsom AR,
- Rotter JI,
- et al
- Denny JC,
- Ritchie MD,
- Crawford DC,
- Schildcrout JS,
- Ramirez AH,
- Pulley JM,
- et al
- London B
- Chen X,
- Yu L,
- Shi S,
- Jiang H,
- Huang C,
- Desai M,
- et al
- Hu D,
- Barajas-Martínez H,
- Pfeiffer R,
- Dezi F,
- Pfeiffer J,
- Buch T,
- et al
- Behr ER,
- Savio-Galimberti E,
- Barc J,
- Holst AG,
- Petropoulou E,
- Prins BP,
- et al
- Stroud DM,
- Yang T,
- Bersell K,
- Kryshtal DO,
- Nagao S,
- Shaffer C,
- et al