|Home | About | Journals | Submit | Contact Us | Français|
The electrocardiographic PR interval reflects atrial and atrioventricular nodal conduction, disturbances of which increase risk of atrial fibrillation (AF). To identify underlying common genetic variation, we meta-analyzed genome-wide association results for PR interval from seven community-based studies of European-ancestry individuals in the CHARGE consortium: AGES, ARIC, CHS, FHS, KORA, Rotterdam Study, and SardiNIA (N=28,517). Statistically significant loci (P<5×10-8) were tested for association with AF (N=5,741 cases). We identified nine loci associated with PR interval. At chromosome 3p22.2, we observed two independent associations in voltage gated sodium channel genes SCN10A and SCN5A, while six loci were near cardiac developmental genes CAV1/CAV2, NKX2-5 (CSX1), SOX5, WNT11, MEIS1, and TBX5/TBX3. Another signal was at ARHGAP24, a locus without known relevance to the heart. Five of the nine loci, SCN5A, SCN10A, NKX2-5, CAV1/CAV2, and SOX5, were also associated with AF (P<0.0056). Common genetic variation, particularly in ion channel and developmental genes, contributes significantly to atrial and atrioventricular conduction and to AF risk.
In myocardial excitation, the delay between the excitation of the atria and ventricles is determined by the sum of atrial and atrioventricular nodal conduction. This delay, measured in milliseconds, is reflected on the standard 12-lead electrocardiogram (ECG) by the PR interval or PQ interval. The PR interval has a substantial heritable component, with heritability estimates ranging between 30 and 50% (1,2,3,4).
Atrial fibrillation (AF) is the most common sustained arrhythmia and is independently associated with increased risk of stroke, heart failure, dementia, and death (5). AF prevalence increases markedly with age, to nearly 9% in those 80-89 years of age, and is estimated to triple by the year 2050. (6). Common genetic risk factors for AF (7) include variants on chromosome 4q25 near the PITX2 gene (8), in 16q22.3 near the ZFHX3 (ATBF1) gene (9), in 1q21 in the KCNN3 gene (10) and the K897T variant in the KCNH2 gene on 7q36.1 (11).
The PR interval is an intermediate phenotype for AF, as alterations in atrial action potential duration and in atrioventricular conduction influence both PR interval and AF risk (12). Longitudinal data from the Framingham Heart Study (FHS) and the Atherosclerosis Risk in Communities Study (ARIC) demonstrate that PR interval prolongation is a predictor of increased AF risk (13, 14). In addition, PR interval prolongation has been shown in FHS to be an independent predictor in a multifactorial risk score for AF predisposition (15).
We undertook a meta-analysis of GWAS to investigate the genetic determinants of the PR interval and their relationship to AF risk. Our goal was to identify genes that can provide insights into atrial disease and lead to novel opportunities for AF prevention and therapy.
We studied individuals of European descent from seven community based studies: the Age, Gene/Environment Susceptibility-Reykjavik Study (AGES) (16), ARIC (17), the Cardiovascular Health Study (CHS) (18), FHS (19), the Kooperative Gesundheitsforschung in der Region Augsburg Study (KORA) (20), the Rotterdam Study (RS) (21), and the SardiNIA study (3) (Table 1 and Online Methods). Phenotypic data including resting 12-lead electrocardiography, height, weight, systolic blood pressure, and medication use were collected using standardized protocols in all studies. Exclusion criteria and covariates are described in Supplementary Table 1.
Study participants were genotyped using a variety of genome-wide SNP arrays. To facilitate comparison of results across studies, we imputed to the 2.5 million HapMap SNPs (22). A recent review supports the validity of combining results across statistical and genotyping platforms (23). Genotyping details, SNP quality control filters, and imputation methods for each study are summarized in Supplementary Table 2.
After exclusions, 28,517 individuals were available for study. The association of each SNP with the PR interval was adjusted for age, sex, RR interval, height, body mass index (BMI), systolic blood pressure, and study site in studies with multiple recruitment sites. Studies adjusted for or excluded individuals using drugs known to alter the PR interval including beta-blockers, diuretics and non-dihydropyridine calcium antagonists.
Due to restrictions imposed by Institutional Review Boards at several of the study sites on the sharing of individual genetic data, it was not possible to perform analyses based on combined individual-level data. Therefore, we conducted inverse variance-weighted fixed-effects meta-analysis of the beta estimates from linear regression of PR interval. The coefficients, generated for each SNP, estimate the difference in PR interval per additional copy of the minor allele, adjusted for the covariates in the model. The genome-wide significance threshold was 5×10-8.
To determine if there was an association between the PR-associated loci and AF risk, we meta-analyzed results from 4 studies of AF in subjects of European descent. The first was a meta-analytic study of 896 prevalent AF cases and 15,768 referents from the CHARGE cohorts (9). The second was a meta-analytic study of 2,517 incident AF cases and 21,337 referents from the CHARGE cohorts. The third and fourth were independent case-control studies of prevalent AF: the German Competence Network on Atrial Fibrillation (AFNET, 2,145 cases and 4,073 controls) (10); and the Cleveland Clinic AF study (CCAF, 183 cases and 164 controls) (24) (Table 3 and Online Methods). We performed an inverse-variance weighted meta-analysis of the logistic-regression results from the prevalent AF studies and the proportional hazards results from the incident AF study. The Bonferroni adjusted significance threshold was P = 0.05/9 = 0.0056.
The study was performed in accordance with the Helsinki declarations and was approved by the local medical ethics and institutional review boards. All participants gave signed informed consent to use their DNA for genetic analyses.
The distribution of results from the meta-analysis of PR GWAS is summarized in Figure 1. The Q-Q plot in Figure 2 shows a clear excess of extreme p-values. Overall, nine loci showed independent association signals with P<5×10-8. We determined the genomic control factor (λ) for the linear regression analysis of PR interval to be 1.076 and report overall analysis results unadjusted for this λ value (25). We did not observe evidence of heterogeneity in effect sizes for any of the nine loci (I2-statistic, all p>0.05, Table 2).
The strongest genome-wide association signal for PR interval was in chromosomal region 3p22.2. In this region we detected two association signals, one covering SCN10A (rs6800541, P = 2.1×10-74) and the other SCN5A (rs11708996, P = 6.0×10-26) (Figure 2 and Table 2). These variants are in low LD (r2 = 0.031). In a meta-analysis of linear regression results from models including both SNPs, these SNPs remained independently associated with PR interval (rs6800541, P= 9.7×10-82; rs11708996, P = 1.1×10-33), suggesting they represent independent association signals.
SCN10A encodes the voltage gated sodium channel Nav1.8, essential for cold perception in afferent nociceptive fibers of sensory dorsal root ganglia (26). Nav1.8 is expressed in the peripheral sensory nervous system but has not been identified in the heart (27). In SCN10A, two common nonsynonymous SNPs are in high to moderate linkage disequilibrium with the sentinel SNP: rs6795970 (V1073A, r2=0.933) and rs12632942 (L1092P, r2=0.220) making both SNPs good candidates to mechanistically explain the strongest PR interval association identified in the human genome.
The neighboring SCN5A gene encodes Nav1.5, the major cardiac sodium channel, with mutations resulting in Brugada Syndrome, long-QT Syndrome, dilated cardiomyopathy, cardiac conduction disease, idiopathic ventricular fibrillation and AF (28). SCN5A sentinel SNP rs11708996 is in weak LD with rs1805124 (H558R, r2=0.034) as well as with two non-coding variants, rs12053903 (r2=0.030) and rs11129795 (r2=0.058), recently reported to be associated with QT interval (29, 30). This suggests that the PR interval and QT interval modifying effects are distinct.
Six PR interval associations were identified in or near genes involved in human cardiac development (Figure 2 and Table 2). NKX2-5 (rs251253 P= 9.5×10-13) is the homolog of the Drosophila tinman gene and encodes the cardiac specific homeobox transcription factor Nkx2.5 (Csx). Mutations can cause atrium septum defect (ASD) with conduction defects (OMIM #108900), tetralogy of Fallot (OMIM #187500), and high degree AV block (31). In the NKX2-5 gene region the association extends over 200Kb and includes three other genes BNIP1, C5orf41, and ATP6V0E1. The signal at the TBX5/TBX3 locus is 230kb downstream of the paralog TBX5 and TBX3 genes (rs1896312, P= 3.1×10-17). Both encode T-box containing transcription factors important for cardiac conduction system formation in the developing heart (32). TBX5 is required for the patterning and maturation of the murine atrioventricular and bundle branch conduction system (33). Deletion of TBX5 results in longer PR intervals in mice (34). Mutations are seen in Holt-Oram syndrome (OMIM #142900) with atrial and ventricular septal defects, conduction disease, and occasionally AF (35). TBX3 controls formation of the sinus node and imposes pacemaker function on atrial cells (36). Mutations cause ulnar-mammary syndrome (OMIM #181450) with limb, mammary, tooth, genital and cardiac abnormalities (37).
The CAV1 and CAV2 genes (rs3807989, P = 3.7×10-28) encode caveolins necessary for the development of caveolae involved in signal transduction (38). CAV1 is expressed in atrial myocytes. Mice deficient in Cav1 develop dilated cardiomyopathy and pulmonary hypertension (39). SOX5 (rs11047543, P = 3.3×10-13) and the nearby C12orf67 encode transcription factors. SOX5 knockout mice die with heart failure marked by hepatic congestion and peripheral edema (40). MEIS1 (rs11897119, P = 4.6×10-11) encodes a homeobox transcription factor implicated in cardiac, hematopoietic and neural development. MEIS1 deficient mice have malformed cardiac outflow tracts with overriding aorta and ventricular septal defect (41).
WNT11 (rs4944092, P = 3.2×10-8) encodes a signaling protein inducing cardiogenesis in Xenopus and in mice by noncanonical WNT signaling (42). The nearest gene to a signal on chromosome 4 (rs7692808, P = 6.0×10-20) is ARHGAP24, which encodes a Rho-GTPase-activating protein and key angiogenic regulator involved in cell polarity, cell morphology, and cytoskeletal organisation (43), but without known relevance to the heart.
Of the nine identified PR loci, five were associated with AF risk (p<0.0056). These were at SCN10A (rs6800541, P=1.5 × 10-4) and SCN5A (rs11708996, P=7.0 × 10-4), as well as at three regions harboring developmental genes, NKX2-5 (P=2.3 × 10-3), CAV1/CAV2 (P=2.2 × 10-5), and SOX5 (P=2.1 × 10-4). In all instances the minor alleles were associated with a decrease in AF risk, irrespective of the direction of their association with PR interval (Table 4). Protective ratios against AF were between 0.93 and 0.88 for the minor alleles.
The observation that SNPs associated with both PR interval and AF risk did not exhibit consistent directions of effect may initially seem counterintuitive. However, PR interval is an amalgamated measure of atrial and atrioventricular nodal conduction, which independently affect AF risk. PR intervals at both high and low extremes may be associated with an increase in AF risk. Indeed, existing data from humans and animal models suggest that the effects of genetic variants on atrial repolarization and action potential duration, and their relationship with atrial arrhythmias, are complex (11). In analogy to the QT interval duration, where both long and short QT intervals are associated with increased ventricular tachycardia risk, assuming a linear association model for AF with PR interval and its underlying genetic variants may not capture the complexity of these relations.
Our study was subject to a number of potential limitations. False positive associations from multiple testing is a limitation of any GWAS, so we used a well-accepted genome-wide association significance threshold equivalent to a Bonferroni correction for 1 million independent tests to reduce false positive findings (44). Population stratification is also a concern, so we only included study subjects of European descent. The low genomic control inflation factor suggested there was no strong influence of population stratification on our results.
Our study also did not examine patterns of haplotype association. Thus complex haplotype associations may not have been captured. However, genome-wide meta-analysis of haplotypes is currently not feasible, and, in common with other GWAS, our use of imputation to the HapMap leverages available linkage disequilibrium information.
The identification of SCN10A was unexpected, as Nav1.8 was previously not thought to play a role in cardiac electrophysiology. In addition, SCN10A was the only locus where two common nonsynonymous variants were in high LD with a sentinel SNP and thus are likely causal candidates. The second key finding was that the majority of association signals we identified were in cardiac developmental genes including two, NKX2-5 and TBX5, in which mutations are known to cause well-defined cardiac malformations involving the atrial septum and the atrioventricular junction. The biological mechanisms by which the identified variants influence PR interval and AF remain speculative, and detailed functional investigation will be required to determine the potential contribution of each genomic region.
We gratefully acknowledge all of the participants in the studies.
AGES: NIH N01-AG-12100, NIA and NIH Intramural Research Programs, Hjartavernd (Icelandic Heart Association), Althingi (the Icelandic Parliament), NHLBI, NEI, and NIDCD.
ARIC: NHLBI N01-HC-55015, N01-HC-55016, N01-HC-55018 through N01-HC-55022, R01-HL-087641, R01-HL-59367 and R01-HL-086694, and R01-HL-054512, NHGRI U01-HG004402, NIH HHSN268200625226C, and the Donald W. Reynolds Cardiovascular Clinical Research Center. Infrastructure was supported by NIH UL1-RR025005.
CCAF: NHLBI R01-HL090620 and P50-HL077107, Intramural funding from the Heart and Vascular Institute, Department of Cardiovascular Medicine, Cleveland Clinic.
CHS: NHLBI N01-HC-85079 through N01-HC-85086, N01-HC-35129, N01-HC-15103, N01-HC-55222, N01-HC-75150, N01-HC-45133, U01-HL-080295 R01-HL-087652, R01-HL-088456, NCRR M01-RR-00425, NIDDK DK063491, NINDS, the Cedars-Sinai Board of Governors Chair in Medical Genetics.
FHS: NIH N01-HC-25195, HL-076784, AG-028321, N01-HC25195, HL-080025 and 6R01-NS-17950, NHLBI N01-HC-25195, BU School of Medicine and Boston Medical Center (LINGA-II), the Robert Dawson Evans Endowment, the Doris Duke Charitable Foundation, the SHARe project, DFG Fellowship SCHN 1149/1-1, Affymetrix contract for genotyping services (N02-HL-6-4278) and Pfizer.
KORA/AFNET: We thank Benno Pütz, Michael Putz and Guido Fischer for their contributions to genotyping and imputation. BMBF NGFN 01-GS-0499, 01-GR-0103, 01-GR-0803, AFNET 01-GI-0204 01-GS-0838, the Leducq Foundation 07-CVD 03, LMU FöFoLe 557/569, the LMU Excellence Initiative, MC Health as part of LMUinnovativ, the HMGU and the State of Bavaria.
Rotterdam Study: We thank Pascal Arp, Mila Jhamai, Dr Michael Moorhouse, Marijn Verkerk and Sander Bervoets for their help in creating the database, Karol Estrada for his help with the analyses and Maxim Struchalin for contributions to genotype imputation. NWO 175.010.2005.011, 911.03.012 and 050-060-810, RIDE, NGI, ZonMw, Netherlands Hartstichting, Ministry of Education Culture and Science, Ministry of Health Welfare and Sports; the European Commission; Erasmus Medical Center, Erasmus University Rotterdam and the Municipality of Rotterdam
SardiNIA: We thank Angelo Scuteri and Marco Orrù for longstanding continuous support of the project and for phenotype characterization. NIA NO1-AG-1-2109, 263-MA-410953, NIH and NIA Intramural Research Programs, NHGRI and NHLBI.
Role of the Sponsor: None of the funding organizations had any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Extended acknowledgements section: A more detailed acknowledgement section can be found in the supplementary material.
Conflict of Interest Statement: Aravinda Chakravarti is a paid member of the Scientific Advisory Board of Affymetrix, a role that is managed by the Committee on Conflict of Interest of the Johns Hopkins University School of Medicine.
Methods: Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturegenetics/.
Note: Supplementary information is available on the Nature Genetics website.
URL: Information about the CHARGE consortium is available at: http://depts.washington.edu/chargeco/wiki/Main_Page
Author Contributions: Study concept and design: AP, DEA, AVS, NS, JIR, AH, BHCS, CMVD, GE, AC, KLL, VG, PTE, SS, SK, JCMW, EJB, SRH; Acquisition of data: AP, CVN, DEA, KVT, MFS, JIR, FR, JAK, BHCS, AGU, BMB, WS, CG, CNC, TJW, MKC, JDS, DRVW, SSN, GBE, AC, EZS, SP, JCMW, AA, SRH; Analysis and interpretations of data: AP, CVN, KDM, DEA, AVS, MM, NS, GCV, ML, JIR, CNC, TJW, RSV, TA, SSN, GBE, AC, EZS, KLL, SP, VG, EJB, SRH; Drafting of the manuscript: AP, DEA, NS, PTE, SK, EJB, SRH; Critical revision of the manuscript: CVN, KDM, MGL, AVS, KVT, MM, MFS, GCV, WHLK, AK, JC, JCB, BMP, KR, JIR, FR, AH, JAK, BHCS, AGU, CMVD, BMB, CG, SAL, CNC, TJW, JWM, RBS, MKC, JB, JDS, DRVW, RSV, GE, LJL, TBH, EL, DS, MU, GRA, BMM, EB, EZS, KLL, HEW, TM, DL, VG, SS, JCMW, AA; Statistical analysis: AP, CVN, DEA, MGL, AVS, MM, GCV, ML, WHLK, JCB, KR, TA, KLL; Obtained funding: AP, MFS, BMP, JIR, FR, AH, AGU, MKC, JDS, RSV, GE, DS, MU, GRA, EB, AC, HEW, TM, DL, VG, JCMW, SRH; Study supervision: JIR, FR, AH, BHCS, AGU, CMVD, GE, AC, VG, JCMW, SRH; The following authors had full data access and take responsibility for analysis: AP, CVN, MM, JIR, AC, KLL, SRH.