A Common Variant in SLC8A1 Is Associated with the Duration of the Electrocardiographic QT Interval

Abstract

Prolongation of the electrocardiographic QT interval, a measure of cardiac repolarization, predisposes one to ventricular arrhythmias and sudden cardiac death. Since NOS1AP, a regulator of neuronal nitric oxide synthase, was discovered in a genome-wide association study (GWAS) as a novel target that modulates cardiac repolarization, several loci have been linked to the QT interval in studies (QTGEN and QTSCD) of European descendents. However, there has been no GWAS of the QT interval in Asian populations. We conducted a GWAS with regard to the QT interval in Korea Association Resource (KARE [n = 6,805]) cohorts. Replication studies in independent populations of Korean (n = 4,686) and Japanese (n = 2,687) groups validated the association between a SNP, rs13017846, which maps to near SLC8A1 (sodium/calcium exchanger 1 precursor, overall p = 8.0 × 10⁻¹⁴), and the QT interval. The minor allele frequency (MAF) of rs13017846 varies widely between ethnicities—0.053 in Europeans (HapMap CEU [Utah residents with ancestry from northern and western Europe from the Centre d′Étude du Polymorphisme Humain collection] samples) versus 0.080 in Africans (HapMap YRI [Yoruba in Ibadan, Nigeria] samples)—whereas a MAF of 0.500 has been reported in Asians (HapMap HCB [Han Chinese in Beijing, China] and JPT [Japanese in Tokyo, Japan] samples). This might explain why this locus has not been identified in Europeans in previous studies.

Main Text

Sudden cardiac death by acute ventricular arrhythmia is a common cause of mortality in developed countries.^1,2 Prolongation of ventricular repolarization increases the risk of ventricular arrhythmia, as observed in congenital long QT syndrome (LQTS), a rare Mendelian disorder.³ The QT interval—the span between the start of the Q wave and the end of the T wave of the cardiac electrical cycle—is the traditional measure of the duration of ventricular depolarization and repolarization⁴ and is genetically influenced (it has approximately 50% heritability).^5,6

Since NOS1AP (MIM 610141), a regulator of neuronal nitric oxide synthase, was identified in a genome-wide association study (GWAS) as a novel target that modulates cardiac repolarization,⁷ several loci have been linked to the QT interval in studies (QTGEN and QTSCD) of European descendents.^8,9 However, no GWAS on the QT interval has been performed on Asian populations.

To understand the genetic architecture of QT intervals in Asians, we conducted a GWAS by using data from Korea Association Resource (KARE [phase 1, n = 6,805])¹⁰ during the discovery phase and two subsequent replication studies (phase 2) in independent Korean (n = 4,686) and Japanese (n = 2,687) populations. The studies from which this report collected data acquired informed consent from all participants and were approved by the appropriate ethics committees at the respective institutions and countries.

The overall study design is shown in Figure S1, available online. Of the 8,842 original KARE discovery-phase participants, those who had missing electrocardiographic data were excluded (n = 1,713). Also, participants with a history of ischemic heart disease, stroke, coronary-artery bypass surgery, or electrocardiographic findings, such as atrial fibrillation, QRS duration > 120 ms, right or left bundle-branch block, and frequent ventricular premature beats >10% of the beats, were excluded (n = 324). Participants with electrocardiogram (ECG) abnormalities, as defined by the presence of Minnesota codes 1.1, 1.2 (major Q wave), 4 or 5 (ST and T wave changes), 7.1–7.6 (ventricular-conduction defect), or 8.1 (frequent atrial or ventricular premature contraction) to 8.3 (atrial fibrillation or flutter),¹¹ were excluded from the analysis. We also excluded subjects who were being treated with medications that were likely to modify the duration of the QT interval (n = 34).⁴

The final number of subjects for the GWAS was 6,805. The same exclusion criteria were applied for the phase 2 replication studies, although the population characteristics and genotyping platforms varied between studies (Tables S1 and S2 and Figure S2).

To measure QT intervals in the phase 1 study, we obtained a supine 12-lead ECG by using a MAC 5000 (GE Medical System, CT, USA). On the basis of the 12-lead ECG, the QT interval was measured from the earliest onset of the Q wave to the latest offset of the T wave, and the heart-rate corrected QT interval (QTc interval) was calculated by the Bazett method with the 12SL program.¹²

The majority of genomic DNAs that were genotyped on the Affymetrix Genome-Wide Human SNP array 5.0 were isolated from peripheral blood that was drawn from KARE participants. We sequentially discarded 38,364 markers with a Hardy-Weinberg equilibrium p value < 10⁻⁶, 17,926 markers with genotype call rates below 95%, and 92,050 markers with a minor allele frequency (MAF) < 0.01; we were left with 352,228 SNPs for subsequent analysis. The genotyping quality-control protocol is described in a previous report by Cho et al.¹⁰

Population stratification was assessed by multidimensional scaling and principal-component analyses, which demonstrated that the characteristics of the phase 1 KARE participants were similar to those of the Japanese and Chinese (HCB [Han Chinese in Beijing, China] and JPT [Japanese in Tokyo, Japan]) components of HapMap.¹⁰ The genomic inflation factor of the phase 1 participants was 1.013.

Most association analyses were performed with PLINK (version 1.07) and SAS (version 9.1). SNPs were tested for association by linear regression with an additive model (1 degree of freedom) after adjustments for age, sex, systolic blood pressure, and recruitment area (Figure S3). We combined KARE GWAS and replication-study results by inverse-variance meta-analysis under the assumption of fixed effects by using Cochran's Q test to assess between-study heterogeneity.¹³ All meta-analysis calculations were implemented in METAL (see Web Resources).

We performed an association analysis in the phase 1 study by using Bazett QTc intervals and a subsequent meta-analysis by using inverse normal transformed (INT) values of the QTc interval (INT QTc interval). The INT method permits p values and the direction of the effect to be combined independently of beta estimates, allowing for differences in QTc distribution between studies (Figure S2). Genome-wide significance was inferred at p < 5 × 10⁻⁸.

Genotyping of the 4,686 subjects in the Korean replication study was performed with the GoldenGate assay¹⁴ (Illumina). The accuracy of the GoldenGate assay was tested by duplicate genotyping of 1%–2.5% of the samples as quality control. All replicate SNPs showed concordance rates in the duplicates of over 99% and genotype success rates of over 98%. On the basis of these results, genotyping errors could be disregarded for these SNPs in subsequent association analyses.

Genotyping of the 2,687 subjects in the Japanese replication study (Millenium Genome Project) was performed with the TaqMan assay. The genotype success rates of the genotyped SNPs exceeded 98%. This project has been detailed elsewhere.^15,16

In phase 1, eight SNPs, which lie near genes that have been linked to the QT interval (such genes are NOS1AP [MIM 610141], KCNQ1 [MIM 192500], KCNH2 [MIM 613688], NDRG4 [MIM 604917], PLN [MIM 172405], and SCN5A [MIM 603830]), had a p value < 0.01 (Table 1). Of the lead SNPs (pairwise linkage-disequilibrium statistics r² < 0.2 and MAF > 0.05) that had not been reported, we selected the seven most strongly associated with nonintergenic SNPs (p < 10⁻⁴) and four additional SNPs near replication-related genes in 4,686 Koreans and 2,687 Japanese (Figure 1 and Table S3).

Table 1.

Previously Reported Loci Showing Associations with QTc Interval in the Phase 1 GWAS

Gene	GWAS SNP	Effect Size (β) in ms	95% Confidence Interval		GWAS p	Reported SNP	Effect Size (β) in ms	95% Confidence Interval		Reported p	Study Population	Distance from GWAS SNP (bp)	LD States
			Lower	Upper				Lower	Upper				r2	D′
NOS1AP	rs4657175	1.709	1.156	2.261	1.4 × 10−9	rs4657178	2.19	1.76	2.62	7 × 10−33	QTSCD	14,873	0.53	0.97
NDRG4-CNOT1	rs37060	1.52	0.957	2.084	1.3 × 10−7	rs37060	1.75	1.41	2.09	3 × 10−25	QTGEN	0	1	1
KCNQ1	rs16928297	1.105	0.5582	1.653	7.6 × 10−5	rs12296050	1.49	0.95	1.93	3 × 10−17	QTSCD	3,222	0.91	1
PLN	rs7764093	0.869	0.2894	1.449	0.00331	rs11970286	1.64	1.25	2.03	2 × 10−24	QTSCD	−11,930	0.57	0.96
SCN5A	rs2268755	0.805	0.2663	1.344	0.00341	rs12053903	1.23	0.88	1.57	1 × 10−14	QTGEN	91,477	0.597	0.79
KCNH2	rs2968854	1.632	0.416	2.848	0.00853	rs2968863	1.35	0.98	1.8	2 × 10−15	QTSCD	−14,697	0.55	0.74

The following abbreviations are used: GWAS, genome-wide association study, and LD, linkage disequilibrium.

Figure 1.

Manhattan Plot of Genome-wide Association Signals from Phase 1 Study

The −log₁₀(p) values are plotted against chromosomal base-pair positions. Pink labeling indicates previously reported loci for the QT interval, and green labeling indicates previously unreported loci showing associations with the QT interval in the phase 1 study and tested for replication. The red line represents the genome-wide significance level (p = 5 × 10⁻⁸).

Of the 11 SNPs that were examined in the phase 2 study, rs13017846, located approximately 18 kb upstream of SLC8A1 (MIM 182305), correlated significantly with the QTc interval in both replication analyses (Table 2). In the meta-analysis of the phase 1 and 2 studies, we observed a highly significant p value for rs13017846, wherein the G allele was linked to shorter QTc intervals (combined beta = −0.08 by INT QTc interval; standard error of the mean [SEM] = 0.01; p = 8.0 × 10⁻¹⁴; Table 2 and Figure 2).

Table 2.

Previously Unreported Top-Ranking SNPs of INT QTc Interval in the Phase 1 and 2 Studies and the Combined Results

Genea	SNP ID	Type	Chr	Minor Allele	MAF	Phase 1 (n = 6,805)			Phase 2						Combined
									Japanese (n = 2,687)			Korean (n = 4,686)
						β	SEM	p	β	SEM	p	β	SEM	p	Combined p	Q	I2
KCNS3	rs6748095	downstream	2	G	0.31	0.072	0.017	1.80 × 10−5	−0.038	0.026	1.53 × 10−1	0.017	0.020	3.86 × 10−1	4.47 × 10−3	0.003	82.79
NCOA2	rs11993276	intron	8	A	0.48	0.075	0.016	3.50 × 10−6	−0.024	0.024	3.27 × 10−1	0.038	0.019	5.33 × 10−2	5.52 × 10−5	0.0027	83.14
DAB1	rs1213770	intron	1	A	0.39	0.062	0.016	9.80 × 10−5	−0.037	0.025	1.30 × 10−1	−0.004	0.019	8.53 × 10−1	5.13 × 10−2	0.0011	85.38
AJAP1	rs2071995	intron	1	A	0.35	0.067	0.017	9.90 × 10−5	0.042	0.026	1.04 × 10−1	0.001	0.020	9.69 × 10−1	4.88 × 10−4	0.3887	0
TMEM20	rs10882364	downstream	10	T	0.45	−0.095	0.024	1.10 × 10−4	−0.017	0.037	6.55 × 10−1	0.031	0.029	2.71 × 10−1	1.97 × 10−2	0.0422	68.41
KCNS3	rs4832537	downstream	2	A	0.46	−0.058	0.016	3.10 × 10−4	0.015	0.024	5.34 × 10−1	0.003	0.020	8.85 × 10−1	3.07 × 10−2	0.0112	77.72
CORIN	rs6447580	intron	4	C	0.26	0.055	0.016	6.50 × 10−4	0.012	0.024	6.15 × 10−1	0.009	0.019	6.36 × 10−1	3.73 × 10−3	0.0119	77.45
SLC8A1	rs13017846	upstream	2	G	0.42	−0.069	0.016	1.70 × 10−5	−0.087	0.026	8.83 × 10−4	−0.104	0.020	1.36 × 10−7	8.00 × 10−14	0.0904	58.4
SLC6A5	rs4922798	intron	11	G	0.41	−0.066	0.018	2.90 × 10−4	−0.010	0.028	7.14 × 10−1	0.016	0.022	4.71 × 10−1	2.34 × 10−2	0.1214	52.58
MAPRE1	rs6119294	downstream	20	A	0.34	−0.053	0.017	1.50 × 10−3	−0.003	0.025	9.09 × 10−1	−0.002	0.020	9.29 × 10−1	2.06 × 10−2	0.0013	85.01
SLC9A9	rs986376	intron	3	G	0.39	−0.064	0.017	1.30 × 10−4	0.053	0.025	3.50 × 10−2	0.002	0.020	9.13 × 10−1	8.24 × 10−2	0.0003	87.79

The effect size of β is on a standard-deviation scale of the inverse normal transformed (INT) QTc (Bazett) interval. Minor alleles for which the effect is estimated refer to the positive strand and are based on NCBI Build 36. The following abbreviations are used: Chr, chromosome; MAF, minor allele frequency; and SEM, standard error of the mean.

^aNearby genes are defined as the closest genes to the SNP within the signal boundary or a 200 kb window.

Figure 2.

Signal Plot for rs13017846 across an 800 kb Window

Association of individual SNPs in the phase 1 study is plotted as −log₁₀(p) against the chromosomal base-pair positions. The y axis on the right shows the recombination rate estimated from the HapMap CHB and JPT populations. All p values are from the discovery phase. The purple diamond represents the result of the meta-analysis of the phase 1 and 2 studies.

SLC8A1 (sodium/calcium exchanger 1 precursor) is a strong candidate for the observed association. SLC8A1 extrudes calcium from cardiac myocytes during relaxation and returns the myocardium to its resting state after excitation.¹⁷ Targeted disruption of SLC8A1 causes heartbeat defects—SLC8A1^−/− mouse embryos experience slow and arrhythmic heart contractions.¹⁸

Notably, the MAF of rs13017846 varies widely between ethnicities—0.053 in Europeans (HapMap CEU [Utah residents with ancestry from northern and western Europe from the Centre d′Étude du Polymorphisme Humain collection] samples) versus 0.080 in Africans (HapMap YRI [Yoruba in Ibadan, Nigeria] samples)—whereas a MAF of 0.500 was reported in Asians (HapMap HCB and JPT samples); this might explain why this locus was not identified in earlier studies in Europeans. The eight reported SNPs that also showed a significant association in our phase 1 study had similar effects between ethnicities; when an overlap of 95% confidence intervals between ethnicities was considered, their effect sizes were also similar.

In general, the Bazett formula overcorrects QT intervals at high heart rates and undercorrects them at low rates.¹⁹ Because the QT interval is heavily influenced by heart rate, there is concern that the common variant of SLC8A1 governs heart rate rather than repolarization. Thus, we used several methods, including the Hodges,²⁰ Fridericia,²¹ and Framingham²² formulas, to correct heart rate (Table S4), which showed that our finding was robust to various correction methods. In the phase 1 and phase 2 Korean replication studies, the correlation between the RR interval and rs13017846 was insignificant.

Limitations of our study include the method by which QT intervals were measured. Because a digital caliper was not used, the measurements might have been inconsistent, especially if U waves or nonspecific ST-T changes were present. However, we excluded subjects whose endpoints of T waves were unclear. Also, QT-interval covariates, such as serum potassium and calcium levels, were not taken into account because they were not uniformly available across studies.

In summary, we report an association between a common genetic variant in SLC8A1 and the electrocardiographic QT interval. This locus appears to be involved in myocardial relaxation, and our findings provide insight into the electrophysiological mechanisms of myocardial repolarization. Further evaluation of these pathways might facilitate the discovery of genetic biomarkers and the development of new strategies for the prevention of sudden cardiac death in high-risk individuals.

Web Resources

The URLs for the data presented herein are as follows:

• Ensembl, http://www.ensembl.org/index.html

• LocusZoom, http://csg.sph.umich.edu/locuszoom/

• Metal, http://www.sph.umich.edu/csg/abecasis/Metal/

• Online Mendelian Inheritance in Man (OMIM), http://www.omim.org/

• Plink, http://pngu.mgh.harvard.edu/∼purcell/plink/