Replication of SCN5A Associations with Electrocardio-graphic Traits in African Americans from Clinical and Epidemiologic Studies

Abstract

The NAv1.5 sodium channel α subunit is the predominant α-subunit expressed in the heart and is associated with cardiac arrhythmias. We tested five previously identified SCN5A variants (rs7374138, rs7637849, rs7637849, rs7629265, and rs11129796) for an association with PR interval and QRS duration in two unique study populations: the Third National Health and Nutrition Examination Survey (NHANES III, n= 552) accessed by the Epidemiologic Architecture for Genes Linked to Environment (EAGLE) and a combined dataset (n= 455) from two biobanks linked to electronic medical records from Vanderbilt University (BioVU) and Northwestern University (NUgene) as part of the electronic Medical Records & Genomics (eMERGE) network. A meta-analysis including all three study populations (n~4,000) suggests that eight SCN5A associations were significant for both QRS duration and PR interval (p<5.0E-3) with little evidence for heterogeneity across the study populations. These results suggest that published SCN5A associations replicate across different study designs in a meta-analysis and represent an important first step in utility of multiple study designs for genetic studies and the identification/characterization of genetic variants associated with ECG traits in African-descent populations.

1 Introduction

A necessary step to establish a robust genotype-phenotype relationship in the literature is statistical replication. There are multiple challenges with replication studies, one of which is the phenotypic heterogeneity that exists between the original study and the replication study population(s). Although the use of quantitative traits or intermediate phenotypes for both the discovery and replication study can alleviate some of the expected phenotypic heterogeneity observed for complex diseases, between study differences may not be eliminated. Therefore, it is ideal to include multiple independent studies to confirm a robust genotype-phenotype association.

The quantitative traits examined here are derived from electrocardiograms (ECGs). The ECG is a useful tool in assessing electrical conduction in the heart, and perturbations in the ECG are routinely used to diagnose cardiac arrhythmias, myocardial infarction, pericarditis, and other cardiac abnormalities by measuring and recording the electrical activity of the heart [1]. The ECG begins with the P wave that occurs during atrial depolarization, and represents the electrical impulse from the sinoatrial (SA) node towards the atrioventricular (AV) node that then spreads to the left and right atrium. The PR interval is the time the electrical impulse takes to go to the sinus node to the AV node then to the ventricles (lower chambers of the heart). It is measured from the P wave to the start of the QRS complex. The QRS marks the start of depolarization of the ventricles, and the ST segment represents the ventricle once they have fully depolarized. After depolarization, the left and right ventricles repolarize, represented as the T wave on the ECG. The QT interval represents the time it takes the ventricles to depolarize and repolarize and is measured in the ECG from the start of the QRS complex to the end of the T wave. In this study, we evaluate the associations with PR interval (representing atrioventricular conduction) and QRS duration (representing intraventricular conduction).

Slower cardiac conduction is thought to contribute to cardiac arrhythmias[2]. The PR interval is influenced primarily by body mass index, increased age, and height. QRS duration is influenced sex and somewhat by body mass index and height[3]. Clinical factors, such as hypertension, cardiac disease, and medications, in addition to genetic factors, can also cause abnormal electrical activity in the heart. Heritability studies suggest that >35% of the variation in ECG traits can be explained by genetics [4–6]. Several genetic association studies of ECG traits have been focused on genes that encode for proteins in voltage-gated ion channels [6–12]. The NA_v1.5 sodium channel α subunit is the predominant α-subunit expressed in myocytes and is encoded by the SCN5A gene, located on chromosome 3 [11,13]. Genetic association studies have identified variants in the SCN5A gene that are associated with long QT syndrome and Brugada syndrome [14–17]. Common variants in SCN5A that are associated with longer QRS are also associated with atrial fibrillation[18]. These associations have been reported to explain ~2% of the variation of ECG traits [19].

There have been several candidate gene studies performed in African Americans for SCN5A and ECG traits [19–21]. Although there are few genome-wide association studies in African descent populations for ECG traits, there have been at least three GWAS or fine-mapping studies on various ECG traits in African Americans [22–24]. These studies typically include African Americans ascertained from epidemiological longitudinal studies focused on cardiovascular diseases (CVD). In the present study we sought to replicate previously reported associations in SCN5A in two study populations of African Americans: the Third National Health and Nutrition Survey (NHANES III, n= 552) and a combined dataset (n= 455) from two biobanks linked to electronic medical records from Vanderbilt University (BioVU) and Northwestern University (NUgene) as part of the Electronic Medical Records and Genomics (eMERGE) network.

2 Methods

2.1 Study populations and ECG measurements

African Americans from two study populations were used for the present study (Table 1): the Third National Health and Nutrition Examination Survey (NHANES III) and participants from two biobanks, the Vanderbilt Genome-Electronic Records (VGER) and the Northwestern biobank (NUgene) as part of the eMERGE network [31]. All ECG traits followed a normal distribution and participants with QRS duration >120 m/sec were excluded from all analyses [32].

Table 1.

Population characteristics.

Trait	Original Study (n= 3,054 )*		NHANES III (n= 552)		eMERGE (n = 455)
	Mean	SD	Mean	SD	Mean	SD
Age (yrs)	56.5	±11.73	53.9	±11.61	46	±15
Sex (% female)	62%	n/a	67%	n/a	77%	n/a
Body mass index (BMI)	not reported		29	±6	34	±10
Type 2 diabetes (%)	19%	n/a	8%	n/a	20%	n/a
PR interval (msec)	171.6	±33.02	164.6	±25.56	159	±21
QRS duration (msec)	92.3	±10.12	95.5	±10.90	82	±8

^*Original Study data represent data from the Jackson Heart Study abstracted from Jeff et al 2011[19]. Mean and standard deviation was calculated for ECG traits, age, and sex for African Americans from the epidemiologic (NHANES III) and clinic-based (VGER/NUgene) study populations. Analysis of variance statistical test was performed to determine significant differences across study populations (bolded italicized data denote p<0.001 for all tests).

NHANES III was conducted from 1988–1994 as a complex survey that over-sampled minorities, the young, and the elderly [33]. Biospecimens for DNA extraction were collected in phase 2 of NHANES III (1991–1994). All NHANES participants were interviewed for demographic, socioeconomic, dietary, and health-related data. Additionally, all NHANES study participants undergo a detailed medical examination at a central location known as the Mobile Examination Center (MEC). Electrocardiograms (ECGs) were recorded on adult (40 years of age or greater) men and women in the mobile examination center (MEC) using a standard 12-lead resting ECG [33]. ECGs were recorded using the Marquette MAC 12 (Marquette Medical Systems, Inc, Milwaukee, Wisconsin) (U.S. DHHS, 1996). NHANES III 12-lead ECG data were recorded with eight independent components of the 12 standard leads simultaneously. ECG data were also sampled at 250 samples per second per channel; giving the availability of multiple simultaneous ECG leads for analysis. This study was limited to self-identified non-Hispanic blacks (referred to here as African Americans) in NHANES III with normal ECG measurements. All procedures were approved by the CDC Ethics Review Board and written informed consent was obtained from all participants. Because no identifying information was accessed by the investigators, Vanderbilt University’s Institutional Review Board determined that this study met the criteria of “non-human subjects.”

VGER and NUgene are study sites of the National Human Genome Research Institute’s electronic MEdical Records and GEnomics (eMERGE) Network [31]. The Vanderbilt study site (VGER) accesses BioVU, which is a collection of DNA samples extracted from discarded blood samples collected for routine clinical care linked to de-identified electronic medical records (EMRs) [34]. The Northwestern biobank, NUgene, combines DNA samples from consented participants with an enrollment questionnaire and longitudinal data from the EMR [31]. Study individuals from both sites were identified using a previously validated algorithm that used ECGs, laboratory data, medication exposures, and natural language processing of clinical notes [35]. Study participants included in this study had a normal ECG without evidence of cardiac disease (or abnormal ECG) before or within one month following the ECG, without concurrent use of medications that interfere with QRS duration, and who did not have abnormal electrolyte values at the time of the ECG. All ECGs had normal Bazett’s corrected QT intervals (<450ms), heart rates (between 50–100 bpm), and QRS duration (65–120 ms). All participants were African American indicated by either observer reported (VGER) or self-reported (NUgene) ancestry. Both biobanks were approved by Institutional Review Boards at their respective sites.

2.2 Genotyping and statistical analysis

DNA was extracted from crude cell lysates from lymphoblastoid cell lines established for NHANES III participants aged 12 over [36]. We chose five SNPs that were significant at p 1.0E-4 from the original study [19] for genotyping in NHANES III. All genotyping was performed in the Center for Human Genetics Research DNA Resources Core using either Sequenom’s iPLEX Gold assay on the MassARRAY platform (San Diego, CA) or Illumina’s BeadXpress. All genotype data reported here passed CDC quality control (QC) metrics and are available for secondary analysis through CDC. All statistical analyses in NHANES III were performed using the Statistical Analysis Software (SAS v.9.2; SAS Institute, Cary, NC) either locally or via the Analytic Data Research by Email (ANDRE) portal of the CDC Research Data Center (RDC) in Hyattsville, MD.

Genotyping for VGER and NUgene was performed by the Center for Inherited Disease Research (CIDR) and the Broad Institute. All individuals that met the inclusion criteria (n = 501) were genotyped for >1.1 million SNPs using the Illumina 1M BeadChip at the Broad Institute. Data were cleaned by the eMERGE QC pipeline [37]. There were 46 individuals that did not meet the QC thresholds and were removed from further analysis.

Using standard linear regression, assuming an additive model, we tested each SNP for an association with PR interval and QRS duration. We did not test SCN5A variants with heart rate and QT interval since these SNPs were not associated with these traits in the original analysis [19]. All tests were limited to African Americans and adjusted for age and sex. We declared significance at p<0.05 uncorrected for multiple testing. Using a fixed-effects inverse-variance weighted approach, we performed a meta-analysis using the effect sizes, standard errors, and p-values from each study populations using METAL [38]. Pairwise F_ST was calculated between the original study and each study population using the Platform for the Analysis, Translation, and Organization of large-scale data (PLATO) [39] (Table 2).

Table 2.

Comparison of minor allele frequencies across studies.

SNP	MA	Original Study* (n= 3,054)	NHANES III (n =552)		eMERGE (n= 455)
		MAF	MAF	FST	MAF	FST
rs7374138	G	0.23	0.15	<0.0001	0.23	<0.0001
rs7637849	A	0.19	0.20	0.014	0.20	<0.0001
rs11129796	T	0.15	0.08	0.019	0.11	0.005
rs7629265	T	0.08	0.09	<0.0001	0.08	<0.0001
rs6768664	G	0.36	0.36	<0.0001	0.38	<0.0001

^*Original Study data represent data from the Jackson Heart Study abstracted from Jeff et al 2011[19]. We calculated three-way F_ST for all SCN5A SNPs to test for differences between study populations.

Abbreviations: minor allele (MA) and minor allele frequency (MAF).

3 Results

We abstracted data from the previously published Jackson Heart Study [19] referred to here as the “original study” for comparison with our population-based (NHANES III) and clinical (eMERGE) collections. We compared the three study populations and observed differences across study populations for age, sex, and ECG measurements (Table 1). On average, eMERGE (46 years) and NHANES III (54 years) participants were younger compared with the original study (57 years). Both eMERGE and NHANES III had more female participants (77% and 67%, respectively) compared with the original study (62%; Table 1). Additionally, the measurements QRS duration and PR interval in eMERGE were shorter compared to the other studies, which is a reflection of the more stringent selection criteria within eMERGE to select subjects without any prior heart disease or abnormalities on their ECG.

To further characterize similarities and differences between the original study and the other two study populations, we first calculated the minor allele frequency and compared these estimates across study populations. Though not statistically significant, NHANES III had a lower minor allele frequency for SCN5A rs7374138 (0.15) and rs11129796 (0.08) compared with the original study and eMERGE (Table 2). To further characterize study population differences at these loci, we calculated F_ST using the Weir and Cockerham algorithm [25] between the original study and each study site separately for each SNP (Table 2). The fixation index F_ST is a measure of population differentiation, and an F statistic >0.15 is indicative as a significant difference between populations. As might be expected, there were no significant differences between studies for any of the SNPs tested at this stringent threshold for population differentiation (F_ST<0.019, Table 2).

We performed single SNP tests of association for SCN5A SNPs identified in the original study with PR interval and QRS duration in eMERGE and NHANES III [19]. There were no significant associations (p<0.05) observed between SCN5A SNPs for any ECG traits in African Americans from NHANES III (Table 3, Figure). Three SNPs (rs7374138, rs7629265, and rs6768664) have a consistent direction of effect compared to the original study, despite not being statistically significant for PR interval. Likewise for QRS duration three SNPs (rs7637849, rs7374138, and rs6768664) have a consistent direction of effect compared to the original study in NHANES III.

Table 3.

Association results across three independent African American study populations and meta-analysis results across all study populations.

SNP	CA	Original study (n = 3,054)	NHANES III (n = 552)	eMERGE (n = 455)	Meta Analysis (n~4,000)
		β (p)	β (p)	β (p)	β (p)	Q P
PR Interval
rs7374138	G	−4.00 (2.4E-5)	−1.65 (0.40)	−1.94 (0.21)	−3.17 (2.2E-5)	0.37
rs11129796	T	−3.50 (2.4E-4)	0.12 (0.96)	−3.89 (0.05)	−3.14 (1.0E-4)	0.36
rs7637849	A	−4.20 (2.3E-5)	0.11 (0.96)	−3.11 (0.04)	−3.11 (2.0E-4)	0.17
rs7629265	T	−7.80 (2.4E-7)	−4.13 (0.17)	−4.00 (0.08)	−6.32 (5.3E-8)	0.30
rs6768664	G	3.00 (2.4E-4)	0.16 (0.92)	0.16 (0.90)	1.80 (5.0E-3)	0.13
QRS duration
rs7374138	G	−1.30 (6.2E-6)	−0.70 (0.41)	−1.15 (0.07)	−1.22 (1.0E-3)	0.79
rs11129796	T	−1.10 (2.5E-3)	0.09 (0.94)	−1.40 (0.11)	−1.03 (1.0E-3)	0.52
rs7637849	A	−0.76 (1.2E-2)	−0.39 (0.66)	−1.55 (0.02)	−0.85 (1.3E-6)	0.48
rs7629265	T	−1.50 (1.8E-3)	1.07 (0.82)	−2.08 (0.03)	−0.15 (0.77)	0.01
rs6768664	G	0.67 (9.5E-3)	1.93 (0.73)	−0.20 (0.71)	0.21 (0.45)	0.11

For each test of association, SNP rs number, coded allele (CA), beta, and p-value are given. For the meta-analysis, the Q p-value is also given.

^*Original Study data represent data from the Jackson Heart Study abstracted from Jeff et al 2011[19].

Figure. Association results for SCN5A SNPs by study site.

We performed single SNP tests of association for five previously identified SCN5A SNPs with PR interval (red) and QRS duration (blue). The direction of effect is indicated by the direction of the arrow and the −log p-value is plotted in the y-axis. SNPs are sorted based on location on chromosome 3. SNPs that met our liberal significance threshold p<0.05 are above the line.

To better understand the impact SCN5A variation has on these ECG traits, we performed a meta-analysis using the effect estimates from all three study populations (Table 3). All SCN5A SNPs were significant (p<0.05) and the direction of effect was consistent with the original study for PR interval. For QRS duration, all SCN5A SNPs have a consistent direction of effect; however, rs7629265 and rs6768664 did not meet our liberal significance threshold in NHANES III samples (Table 3). In NHANES III, having two copies of the risk allele “T” for rs7629265 increased QRS duration, whereas in both the original study and in eMERGE; having two copies of the “T” allele decreased the QRS duration.

In contrast, having two copies of the risk allele “G” at rs6768664 is associated with increased QRS duration in both the original study and NHANES III but with decreased QRS duration in eMERGE samples.

4 Discussion

In the present study we replicate and or generalize previously identified SCN5A associations in African Americans ascertained from clinical and population-based collections. Associated variants in the clinical-based study population, eMERGE, had a consistent direction of effect with previous studies in African Americans for four of the five SNPs between the two traits. However, in African Americans from the US population-based cohort, NHANES III, several associations had opposing direction of effects, and none of the SNPs tested reached our liberal significance threshold (p=0.05) for any ECG trait.

Here we tested five SNPs: rs7374138, rs11129796, rs7637849, rs7629265, and rs6768664, all located in various SCN5A introns. Consistent with published data, three of the five tested SNPs are specific to African-descent populations and are monomorphic or rare in European and Asian-descent populations [19,20,22,24,26]. While there is not a direct biological correlation with the role intronic SNPs have on protein function, intronic regions are known to play an important role in splicing, which could possibly affect protein function. Previous studies report that in African-descent populations one SNP, rs7629265, is in high linkage disequilibrium (LD, r² = 0.87) with coding non-synonymous variant, rs7626962 (S1103Y), and has been consistently associated with long QT syndrome [26]. This SNP is associated with PR interval in the combined analysis and is not associated with QRS duration, which is also consistent with the literature [19] (Table 3).

We did not test all SCN5A SNPs reported in the literature; as a consequence, there are several important associations we did not test. In spite of this, there are several trends we observed that are consistent with previous studies which suggest testing more SNPs in these regions will likely yield the same results [19,22,24]. Another limitation to this study is sample size. Compared to the previous reports for SCN5A, both eMERGE and NHANES III were limited in sample size: n= 455 and n=520, respectively. However, despite this limitation we were able to observe a consistent direction of effect compared to original reports for most SCN5A SNPs in African Americans from eMERGE.

In addition to small sample size, differences in study population may explain the lack of replication in NHANES III African Americans. Indeed, there are several similarities and differences across study populations. We sought to replicate associations originally detected in African Americans from the Jackson Heart Study (JHS). The Jackson Heart Study is a longitudinal study collected with the primary objective to identify and explain the disparity of cardiovascular diseases in African Americans [27–29]. As a result, the JHS has more samples with in-depth phenotype information for cardiovascular diseases and related traits compared to the other studies and can possibly include individuals with CVD. Another difference between the NHANES III population compared to the other study populations could be explained by geography. Unlike NHANES III, which is representative of the US population, the JHS is limited to African Americans from the southeastern United States. Similarly, most samples from eMERGE are limited to African Americans that visit the clinic or hospital in the Nashville metropolitan areas, also in the southeastern United States. The prevalence of CVD and related environmental risk factors are disproportionately higher in the southeastern US, which likely makes up majority of the eMERGE samples tested here and all of the JHS samples[30]. Lack of replication in NHANES III might be explained by this difference. Another possibility could be the time in which the study was conducted. Both eMERGE and JHS samples were collected fairly recently (in the past 10 years) compared to NHANES III, which was collected over 20 years ago. There are several environmental factors that have changed since then, such as diet, that might be interacting with these SNPs and thus have an impact on our study and explain our failure to replicate in NHANES III. Other non-genetic risk factors such as age, sex, type 2 diabetes mellitus (T2DM) status, and BMI are associated with ECG traits [3]. Indeed, we did observe significant differences across study populations for age, sex and BMI (Table 1). Study individuals from eMERGE were more likely to be female and were significantly younger compared to the original study and NHANES III. Also, eMERGE study individuals had a higher mean BMI and higher proportion of T2DM cases compared with NHANES III. It is possible that associations between SCN5A and ECG traits are modified by environmental factors (such as poor diet and BMI). Additional analyses are needed to statistically assess the impact of gene-environment interactions on this complex trait.

Perhaps the main reason for lack of replication in NHANES III is phenotype definition. It is important to note that despite small sample sizes in the replication cohorts, SCN5A associations have a much more consistent direction of effect in eMERGE than do the NHANES III samples compared to the original study (Table 3). One of the main goals of the eMERGE consortium was to successfully replicate GWAS associations in clinical populations, thus accurately identifying phenotypes from electronic medical record (EMR) data was critical. As previously mentioned, eMERGE study participants are limited to ECG measurements in normal range without any evidence of pre-existing heart conditions, laboratory values, or medications that may alter their ECG results. The algorithm was validated by blinded physician review with a positive predictive value of 97% at both eMERGE sites participating in this study[18]. This stringent phenotype definition was not used in either the original study or NHANES III, and both had longer PR interval and QRS duration compared to eMERGE (Table 1). The broad phenotype definition might have resulted in a loss of power in NHANES III and thus explain why SCN5A variants did not replicate. While the original study also had a broad phenotype definition, the loss of power is not as significant given the large sample size, which is almost seven times larger than NHANES III.

Despite these study population differences, most tests of association were significant at p<0.05 after meta-analysis for both PR interval and QRS duration. And, tests of heterogeneity suggested little detectable differences across the studies in this meta-analysis. Overall, our data validate SCN5A associations with QRS duration and PR interval in African Americans by meta-analysis. Most importantly, this work high-lights the challenge of conducting and interpreting genetic association studies in multiple study populations with differing ascertainment strategies. This latter finding is of particular important in this era of meta-analysis of genetic association studies where study design and phenotypic precision are traded in favor of larger sample sizes for greater power.