Two Further Blood Pressure Loci Identified in Ion Channel Genes With a Genecentric ApproachCLINICAL PERSPECTIVE
Background—Blood pressure (BP) is highly heritable, but our understanding of the genetic causes underlying variations in BP is incomplete. In this study, we explored whether novel loci associated with BP could be identified using a genecentric approach in 3 community-based cohorts with accurate BP measurements.
Methods and Results—Genotyping of 1857 single nucleotide polymorphisms (SNPs) in 91 ion channel genes was performed in a discovery cohort (n=358). Thirty-four SNPs associated with BP traits (P≤0.01) were followed up in an independent population (n=387); significant SNPs from this analysis were looked up in another independent population (n=1010) and meta-analyzed. Repeated clinic and ambulatory measurements were available for all but the discovery cohort (clinic only). Association analyses were performed, with systolic, diastolic, and pulse pressures as quantitative traits, adjusting for age and sex. Quantile–quantile plots indicated that the genecentric approach resulted in an inflation of association signals. Of the 29 SNPs taken forward from the discovery cohort, 2 SNPs were associated with BP phenotypes with the same direction of effect, with experiment-wide significance, in follow-up cohort I. These were rs2228291, in the chloride channel gene CLCN2, and rs10513488, in the potassium channel gene KCNAB1. Both associations were subsequently replicated in follow-up cohort II.
Conclusions—Using a genecentric design and 3 well-phenotyped populations, this study identified 2 previously unreported, biologically plausible, genetic associations with BP. These results suggest that dense genotyping of genes, in pathways known to influence BP, could add to candidate-gene and Genome Wide Association studies in further explaining BP heritability.
Heritability estimates for blood pressure (BP) are 15% to 40% and 15% to 30% for clinic systolic BP (SBP) and diastolic BP (DBP), respectively,1,2 and 69% and 51% for the more accurate measures of ambulatory nighttime SBP and DBP.1 Candidate gene, targeted gene array, and genome-wide association studies have discovered >50 common genetic variants robustly associated with clinic BP. As each of these variants has only a modest effect on BP (<1 mm Hg systolic), in aggregate, these common variants explain <3% of the variance of BP. In this study, we have adopted a targeted genotyping approach in an attempt to find additional variants associated with ambulatory BP measurements.
Clinical Perspective on p 879
Ion channels are known to influence BP level through multiple mechanisms: renal epithelia ion channels are instrumental in the control of fluid volume,3 ion channels additionally influence vascular tone,4,5 and cardiac contractility.6 Twin studies have confirmed genetic influences on the renal excretion of sodium and potassium.7 Mutations in several ion channel genes which alter renal salt reabsorption have long been known to cause rare Mendelian BP disorders.8,9 More recently, common variants in these same genes have been shown to influence BP variation in the general population.10,11
Our targeted approach in this study involved interrogating multiple ion channel genes on the basis that they were likely to play a role in BP variation. We adopted a 3-stage experimental design. We first conducted extensive genotyping across 91 sodium, potassium, chloride, and calcium channel genes and tested for associations with clinic BP measurements in a discovery cohort (n=358). We then followed up promising associations in a second, independent cohort, follow-up cohort I (n=387), with clinic and ambulatory BP measurements. Both of these cohorts were from Irish community dwelling populations. Finally, significant associations from follow-up cohort I were looked up in a third cohort, follow-up cohort II (n=1,010), from The GRAPHIC (Genetic Regulation of Arterial Pressure of Humans in the Community) study consortium, a community-based, UK study with ambulatory BP measurements.
All subjects provided written informed consent, and studies were approved by Institutional Research Ethics Committees.
This cohort consisted of 358 individuals from a study of healthy bank employees and their spouses. All participants were free of diagnosed hypertension and vasoactive drugs when recruited. Age, sex, smoking habit, alcohol consumption, salt intake, past medical history, current drug treatments, height, weight, and clinic BP were recorded. Sitting clinic BP was measured 3 times at 5-minute intervals, from the right brachial artery, using a regularly calibrated validated automated sphygmomanometer (Omron HEM-705CP). The mean of the last 2 measurements was recorded as representative of clinic BP.
Follow-Up Cohort I
This cohort was independent of the discovery cohort and was recruited separately. Initially, 815 current and retired bank employees and their spouses, who were free of diagnosed hypertension and vasoactive drugs at baseline, were recruited.12 At the baseline assessment (phase I), age, sex, smoking habit, alcohol consumption, salt intake, past medical history, current drug treatments, height, weight, clinic, and 24-hour ambulatory BPs were recorded. After a mean interval of 8.4 years, 441 subjects responded to a written invitation to undergo a repeated assessment (phase II), including collection of blood for DNA extraction. At this time, 4 subjects were taking antihypertensive medication, but this therapy was discontinued 1 week before the phase II assessments. Sitting clinic BP was measured 3 times at 5-minute intervals from the right brachial artery arm using a mercury sphygmomanometer. The mean of the last 2 measurements was recorded as representative of clinic BP. Ambulatory BP measurements were made every half-hour throughout a 24-hour period using validated oscillometric 90202 or 90207 SpaceLabs recorders. Mean daytime (mean of all measurements between 0900 and 2100 hours) and nighttime (mean of all measurements between 0100 and 0600 hours) SBP and DBP were calculated for each individual for each phase. Of the 441 subjects who attended the phase II assessments, 9 and 45 participants were excluded because of technically unsatisfactory ambulatory blood pressure recordings and unsuccessful DNA extractions, respectively. Hence, there were 387 participants in the follow-up cohort I.
Follow-Up Cohort II
This cohort comprised the parental generation in The GRAPHIC (Genetic Regulation of Arterial Pressure of Humans in the Community) study (n=1010 unrelated subjects). Details of recruitment and phenotyping of The GRAPHIC (Genetic Regulation of Arterial Pressure of Humans in the Community) study subjects are described elsewhere.10,13 Briefly, nuclear families (all of white European ancestry) with both parents (aged 40–60 years) and 2 adult offspring were identified through general practices in Leicestershire. Participants had a detailed history taken and were examined by research nurses following standard protocol. The protocols of clinic BP and 24-hour ambulatory BP monitoring in The GRAPHIC (Genetic Regulation of Arterial Pressure of Humans in the Community) study have been reported10,13 and were similar to those described for follow-up cohort I above.
The ion channel genes analyzed in this study were originally selected as targets for a multicentre search for genetic susceptibility loci for epilepsy.14 All known members of the voltage-gated sodium and calcium channels and a subset of members of the chloride and potassium channels were included. Single nucleotide polymorphisms (SNPs) were selected for genotyping such that all common variation across the set of 91 candidate genes was captured. Preference was given for putative functional SNPs, including exonic SNPs, those in promoter and other regulatory regions and transcription factor binding sites. Full details of the genotyped SNPs can be found in tables in the Data Supplement of Cavalleri et al.14 In the discovery population, 1857 SNPs were genotyped in total using a custom Illumina GoldenGate genotyping assay by the Duke Genomic Analysis Facility in Durham, North Carolina.
SNPs were considered for genotyping in follow-up cohort I if they had an association with SBP, DBP, or pulse pressure (PP) with a P value of ≤0.01 in the discovery cohort. We deliberately used a relaxed P value threshold to take as many likely candidates into the follow-up cohort as possible, with the assumption that any false-positive associations would fail to replicate. Forty SNPs qualified for genotyping in follow-up cohort I, but 6 were in complete linkage disequilibrium with another SNP in the same data set (r2=1 value in Hapmap release 27), and hence were excluded. One SNP was replaced with a surrogate (r2=1, Hapmap release 27) because of a low genotyping designability score. A full list of the 34 SNPs, from 20 genes, genotyped in follow-up cohort I is available in Table I in the Data Supplement. Genotyping was performed using the Illumina VeraCode technology by the Duke Genomic Analysis Facility in Durham, North Carolina.
Quality control measures were applied for the discovery and follow-up I cohorts as described in Table II in the Data Supplement. Briefly, no samples had to be excluded because of low genotyping call rate (<95%), X-chromosome heterozygosity inconsistent with phenotypic sex, or cryptic relatedness. SNPs were excluded on the basis of low call rate (<95%), minor allele frequency <1%, or Hardy–Weinberg Equilibrium P value of <1×10−07. Of the 1857 SNPs genotyped in the discovery cohort, 1485 SNPs passed quality control. Of the 34 follow-up SNPs, 29 passed quality control.
Subjects in follow-up cohort II were genotyped using the 730 525 SNP Infinium HumanOmniExpress-12v1_h beadchip (Illumina Inc). Genotypes were clustered and called using Genotyping Module v1.9 within GenomeStudio™ Software v2011.1 (Illumina, Inc). Imputation of the untyped genotypes (with the 1000 Genome Project as a reference) was performed using IMPUTE-v2.2 through the SNPTEST-v2.3.0 program.
A linear regression, using genotype as an additive covariate and BP—clinic and ambulatory SBP, DBP, or PP—as the response, was fitted for each SNP individually. All analyses were adjusted for age and sex. P values were calculated from the test statistics using a 2-tailed test for the discovery cohort and a 1-tailed test for the follow-up cohorts.
For follow-up cohort I, the phase I and phase II, clinic, daytime, and nighttime BP measurements were included in the analyses models. As data for this cohort were collected in 2 phases (mean interval 8.4 years), a mixed model analysis was performed in addition to the linear regression analysis. The mixed model analysis included an additional random-effect term, which included in the model the intraindividual variation in BP measurements after adjustment for time (age) and was performed using the R package nlme. The findings were in keeping with those of the linear regression analysis (results shown in Table III in the Data Supplement).
Population stratification was not anticipated to be a problem, as all study participants were white Europeans recruited within Southern Ireland (discovery cohort and follow-up cohort I) or the English Midlands (follow-up cohort II). Genomic control lambda values were ≈1 (range 0.99–1.02), indicating that any effect of population stratification in this study was indeed minimal.
Meta-analysis of the association results from the screening and follow-up cohorts was performed using PLINK.15 Effect sizes and P values were calculated using both fixed and random-effects meta-analysis for each association, along with Cochrane’s Q statistic for inconsistency across studies. Where Cochrane’s Q statistic was nonsignificant (P>0.05), results of the fixed-effects meta-analysis are reported. Otherwise, results from the random-effects meta-analysis are shown.
Correction for multiple testing in follow-up cohort I was achieved by obtaining empirical thresholds for rejecting the null model of no association using permutation tests. This approach is robust and has the advantage of drawing the threshold directly from the experimental data.16 Permutation was performed in PLINK using the MaxT permutation procedure and performing 1000 permutations, the number needed to estimate an α=0.05 threshold.16 In this permutation test, the phenotypic data (all 3 BP phenotypes) were randomly shuffled relative to the genotypic data (all SNPs) and association tests were repeated 1000 times on the shuffled data, simulating results expected under the null hypothesis. The correction for multiple testing can then be made empirically, based on where an association ranks compared with the results of the associations randomly generated from the data.
Only follow-up cohort II included individuals on antihypertensive treatment. BP values of subjects on antihypertensive treatment in this cohort were adjusted for the BP-lowering effect of therapy using a semiparametric algorithm as described elsewhere.13
All cohorts in this study comprised unrelated individuals. Spousal pairs, however, were included. Previous reports of spousal correlations in BP have indicated that correlation between spouses is modest;17 however, it is possible that there might be some spousal correlation in BP because of shared environment and/or assortative mating, which would lower the overall variability of BP in this study. Because spousal pairing information was not available for these cohorts, it was not possible to account for this possible source of variability in the analysis. However, it is reasonable to assume that if correlation does exist, it is unlikely to be causing spurious SNP associations because there is no reason to think that the spouses in these cohorts were any more related to each other (and therefore likely to share SNP genotypes) than to the other members of the cohort.
The characteristics of the discovery cohort and follow-up cohort I are summarized in Table 1.
Quantile–quantile plots of the P values for association of the 1485 SNPs with BP phenotypes in the discovery cohort demonstrated enrichment for significant associations, indicating that the genecentric approach was successful in enriching for polymorphisms associated with BP variation (Figure 1). The null hypothesis of no influence of these variants on BP was therefore rejected. The quantile–quantile plots for the 29 SNPs which passed quality control in follow-up cohort I illustrated further enrichment for significant associations (Figure 1). Observed P values clearly exceeded the 95% probability envelopes for association with both clinic and ambulatory (daytime and nighttime), systolic, and diastolic pressures.
Nine SNPs were consistently associated with variation in BP in the discovery cohort and follow-up cohort I (P≤0.05 in both cohorts, with the same direction of effect), shown in Table IV in the Data Supplement. Of these, 2 variants had statistically significant P values in follow-up cohort I after permutation testing.
The first, rs2228291, is a synonymous (coding) SNP located in exon 16 of the CLCN2 gene, which encodes chloride channel 2. Carriership of 1 minor allele at this locus was significantly associated with ≈2 mm Hg increases in nighttime SBP and in both daytime and nighttime DBP (Table 2). The other SNP, rs10513488, is located in intron 2 of the KCNAB1 gene, which encodes potassium voltage-gated channel, shaker-related subfamily, beta member 1. Carriership of 1 minor allele at this locus was significantly associated with a 7 mm Hg increase in clinic SNP and a 2.5 mm Hg increase in nighttime PP (Table 2).
The regional association plot for CLCN2 (Figure 2, upper panel) shows that only 1 SNP other than rs2228291 in this region was genotyped in follow-up cohort I. The minor allele of this SNP, rs6770808, was associated with decreased SBP in both the discovery cohort and in follow-up cohort I. It is located ≈10 kb upstream of rs2228291, there is little linkage disequilibrium between the SNPs, and analysis of the association at 1 SNP conditioning on the other did not markedly change these results. The regional association plot for KCNAB1 (Figure 2, lower panel) shows that 4 SNPs other than rs10513488 in the region were genotyped in follow-up cohort I, none of which was in high linkage disequilibrium with rs10513488 (r2<0.3). One SNP, rs3821686, located ≈100 kb downstream of rs10513488, was also associated with increased SBP in follow-up cohort I (each minor allele of which was associated with an increased daytime SBP of 2.0 mm Hg [P=0.004]; increased nighttime SBP of 2.0 mm Hg [P=0.002]; and increased clinic SBP of 3.1 mm Hg [P=0.001]). However, the minor allele of this same SNP was associated with decreased SBP in the discovery cohort and therefore probably represents a false association.
An in silico look-up of the 2 significantly associated SNPs in follow-up cohort II, in which rs2228291 was directly genotyped and rs10513488 was imputed (imputation quality score 0.92), confirmed the associations between the SNPs and BP. Each copy of the minor allele of rs10513488 was associated with an increase of 2.2 mm Hg in ambulatory nighttime SBP (P=0.05), 1.8 mm Hg in ambulatory nighttime PP (P=0.01) and daytime PP (P=0.02), and 2.2 mm Hg in clinic PP (P=0.05). The minor allele was also associated with a nonsignificant increase in other ambulatory SBP and DBP and clinic SBP measures (Table 2). Each copy of the minor allele of rs2228291 was associated with an increase of 1.3 mm Hg in ambulatory nighttime SBP (P=0.03) and 0.6 mm Hg in ambulatory nighttime PP (P=0.04) in follow-up cohort II. The minor allele was also associated with a nonsignificant increase in daytime SBP and PP and nighttime DBP (Table 2).
Meta-analysis of the associations of rs10513488 and rs2228291 with BP phenotypes across the screening and replication studies (Table 2) showed a significant (experiment-wide) association between rs10513488 and all SBP and PP phenotypes. Heterogeneity across studies was higher for rs2228291; however, meta-analysis showed a significant association with nighttime SBP and a nominal association with daytime and clinic SBP.
This study has identified 2 novel loci which were consistently associated with BP variation in 3 independent, community-based cohorts. Both SNPs are located on the long arm of chromosome 3, ≈28 million base pairs apart.
CLCN2 is widely expressed in human tissues, including in the vascular smooth muscle and endothelial cells of the heart and in the kidney.19,20 In a human embryonic kidney cell line, the protein encoded by this gene has been shown to be activated in response to oxidative and metabolic stress and its activity modulated by membrane lipid environment, especially by cholesterol.21 Mammalian studies have implicated CLCN2 channels in the regulation of cardiac electric activity and cell volume control under physiological and pathological conditions.22
KCNAB1 is widely expressed in human tissues, including in the ventricle and the atrium of both healthy and diseased human hearts.23 KCNAB1 encodes a subunit of the shaker-related subfamily of voltage-gated potassium channels, which are known to be important modulators of cardiac action potential. β-subunits are thought to modulate the inactivation of the pore-forming α subunits.24 In keeping with this observation, KCNAB1 transcripts have been shown to be upregulated in the mouse heart after treatment with the antiarrhythmic drug amiodarone,25 suggesting a role in cardiac electrophysiology.
There are no Hapmap3 (release 2) SNPs in high linkage disequilibrium (r2>0.6) with either rs2228291 or rs10513488, functional or otherwise. Functional annotations from the University of California, Santa Cruz genome browser (http://genome.ucsc.edu/) indicate that although neither of these SNPs causes an amino-acid change in the encoded protein, strong evidence exists for their regulatory functions.
rs2228291 lies within a digital DNase 1 hypersensitivity cluster, indicating that the region is likely to be regulatory. The position of rs2228291, close to an exon/intron boundary, led us to hypothesize that this mutation may affect splicing of this gene. Analysis in Human Splicing Finder version 2.4.126 indicated that the minor allele variant of rs2228291 may have major implications for the correct splicing of the exon. The variant destroys important exon-identity elements, which are enriched in exons and important for exon recognition. It also alters as a binding site for the pre-mRNA binding protein, hnRNP A1, and results in the creation of a silencer element, believed to repress splicing of pseudoexons. Although several splice variants of human CLCN2 have been described, none of those reported to date affects exon 16.
Human Splicing Finder analysis suggested that rs10513488 may also affect splicing, although the evidence was not as strong as it was for rs2228291.
Modest population sizes are a limitation of this study—the study had low power to detect associations of rare variants, or of those with modest effect sizes, with BP level. However, the smaller population sizes studied also brings several strengths. The targeted design meant that each SNP had a higher a priori chance of having a true association with BP variation. Furthermore, unlike GWAS, the study was not encumbered with a high multiple testing burden, inflicted largely by variants which are unrelated to BP variation. Finally, the size of the study meant that permutation testing was computationally feasible, providing empirical P values for association and avoiding the common problem of overcorrection for multiple testing.
A further strength of this study was the quality of the phenotypic data available. Unusually for a genetic association study of BP, ambulatory BP measurements were available for both follow-up cohorts. Ambulatory BP offers the ability to measure BP in real-life settings, track BP at night, and avoid the white- coat phenomenon, making it a more accurate measure of true BP than clinic measurements. The added value of ambulatory BP measurements is borne out by its substantially higher heritability and by studies showing that ambulatory BP is a better predictor of target organ damage and cardiovascular morbidity and mortality than BP measured in the clinic. Many review articles published in recent years have emphasized the potential of using ambulatory BP measurements in genetic studies of BP to find new variants.27,28 In addition, unlike many cohorts used in BP studies, all the participants in the discovery cohort and in follow-up cohort I were healthy volunteers free of antihypertensive medication. The most common method of correction for antihypertensive medication in other studies is the addition of 10 mm Hg to treated SBP measurements and 5 mm Hg to treated DBP measurements.29 Although this approach has been shown to be superior to ignoring antihypertensive treatment or to excluding individuals on therapy, factors such as medication number and dosage make this adjustment scheme an oversimplification, leading to a substantial reduction in the power to detect the small (often <2 mm Hg) changes in BP typically associated with genetic variants.30
The effect sizes of the variants that were associated with BP in this study were clinically meaningful—a minimum of 2 mm Hg BP change per minor allele carried was observed at both loci. Observational data indicate that a prolonged reduction in SBP of 2 mm Hg corresponds to 6% reduction in stroke and 5% reduction in coronary artery disease.31
Despite recent candidate-gene and genome-wide association studies conducted by large consortia, BP variation still suffers from a large amount of missing heritability. This deficit is common to many complex disease phenotypes, which are likely to be influenced by a large number of genetic variants. Any GWAS or large-scale candidate gene study will only identify a subset of all truly associated genetic variants, because their power to detect such variants is low.32 We think it likely that studies that use a genecentric approach, such as this study, may contribute importantly toward the elucidation of the complete genetic architecture of BP level and hypertension. Ultimately, identifying variants associated with BP will provide new insights into the pathophysiology of BP regulation and may point to further drug targets for the treatment of hypertension.
We acknowledge the participants, collaborating general practitioners, research nurses, and laboratory staff who contributed to the Allied Irish Bank Studies and The GRAPHIC (Genetic Regulation of Arterial Pressure of Humans in the Community) study. This work forms part of the portfolio of translational research of the National Institute of Health Research Biomedical Research Unit at Barts.
Part of the work contained in this article has been presented at the following conferences:
‘Association of Variants in Candidate Genes Influencing Electrolyte Transport and the Autonomic Nervous System with Blood Pressure Variation.’
European Society of Hypertension, Milan, June 2009: Oral presentation.
‘Gene-centric Study Identifies Two Novel Genes, CLCN2 (a Voltage-gated Chloride Channel) and KCNAB1 (a Voltage-gated Potassium Channel) Associated with Blood Pressure in Two Independent Irish Populations.’
American Society of Human Genetics, Washington DC, November 2010: Poster presentation.
European Society of Hypertension, Oslo, June 2010: Poster presentation.
Sources of Funding
The Allied Irish Bank Study and genotyping was funded in part by the Health Research Board, Ireland (Grant number PHD 2007/11), the Higher Education Authority (Ireland), and the Charitable Infirmary Charitable Trust (Ireland). The GRAPHIC (Genetic Regulation of Arterial Pressure of Humans in the Community) study was funded by the British Heart Foundation. Genotyping was funded by Medisearch medical research charity and the Wellcome Trust Functional Genomics Initiative in Cardiovascular Genetics.
The Data Supplement is available at http://circgenetics.ahajournals.org/lookup/suppl/doi:10.1161/CIRCGENETICS.113.000190/-/DC1.
- Received April 24, 2013.
- Accepted July 16, 2014.
- © 2014 American Heart Association, Inc.
- Levy D,
- DeStefano AL,
- Larson MG,
- O’Donnell CJ,
- Lifton RP,
- Gavras H,
- et al
- Guyton AC
- Davis MJ,
- Hill MA
- Haddy FJ,
- Vanhoutte PM,
- Feletou M
- Churchill GA,
- Doerge RW
- DeMarco MM,
- Coresh J,
- Woodward M,
- Butler KR,
- Kao WHL,
- Mosley TH,
- et al
- Pruim RJ,
- Welch RP,
- Sanna S,
- Teslovich TM,
- Chines PS,
- Gliedt TP,
- et al
- Cid LP,
- Montrose-Rafizadeh C,
- Smith DI,
- Guggino WB,
- Cutting GR
- Hinzpeter A,
- Fritsch J,
- Borot F,
- Trudel S,
- Vieu DL,
- Brouillard F,
- et al
- England SK,
- Uebele VN,
- Shear H,
- Kodali J,
- Bennett PB,
- Tamkun MM
- Le Bouter S,
- El Harchi A,
- Marionneau C,
- Bellocq C,
- Chambellan A,
- van Veen T,
- et al
- Desmet FO,
- Hamroun D,
- Lalande M,
- Collod-Béroud G,
- Claustres M,
- Béroud C
The genetic architecture of blood pressure variation is complex, and identifying all of its elements will require many complimentary study designs. The quality of the data used in association studies is as important as the quantity. Here we have used a gene-centric approach in 3 modestly sized populations with high-quality phenotypic data. This resulted in the identification of 2 novel variants, in biologically plausible genes, a chloride channel gene and a potassium channel gene. These variants seem to have clinically relevant effects on blood pressure variation in healthy individuals. The finding of these 2 variants adds to the current list of ≈50 common genetic variants robustly associated with blood pressure and provides new insights into the pathophysiology of blood pressure regulation. Furthermore, new drug targets for the treatment of hypertension have been identified.