THE EPITHELIAL SODIUM CHANNEL γ-SUBUNIT GENE (SCNN1G) AND BLOOD PRESSURE: FAMILY-BASED ASSOCIATION, RENAL GENE EXPRESSION AND PHYSIOLOGICAL ANALYSES

Abstract

Variants in the gene encoding the γ-subunit of the epithelial sodium channel (SCNN1G) are associated with both Mendelian and quantitative effects on blood pressure. Here, in four cohorts of 1611 white European families comprising a total of 8199 individuals, we undertook staged testing of candidate SNPs for SCNN1G (supplemented with imputation based on data from the 1000 Genomes Project) followed by a meta-analysis in all families of the strongest candidate. We also examined relationships between the genotypes and relevant intermediate renal phenotypes as well as expression of SCNN1G in human kidneys. We found that an intronic SNP of SCNN1G (rs13331086) was significantly associated with age-, sex- and BMI-adjusted blood pressure in each of the four populations (P < 0.05). In an inverse variance-weighted meta-analysis of this SNP in all four populations each additional minor allele copy was associated with a 1 mmHg increase in systolic blood pressure and 0.52 mmHg increase in diastolic blood pressure (SE = 0.33, P = 0.002 for SBP; SE = 0.21, P = 0.011 for DBP). The same allele was also associated with higher 12-h overnight urinary potassium excretion (P = 0.04), consistent with increased epithelial sodium channel activity. Renal samples from hypertensive subjects showed a non-significant (P = 0.07) 1.7-fold higher expression of SCNN1G compared with normotensive controls. These data provide genetic and phenotypic evidence in support of a role for a common genetic variant of SCNN1G in blood pressure determination.

The epithelial sodium channel (ENaC) plays a vital role in blood pressure (BP) regulation. Mutations in the gene encoding the γ-subunit of ENaC, SCNN1G, have been identified as causes of Mendelian forms of hypertension and hypotension. A number of independent studies including our own have consistently shown an association between the SCNN1G locus and blood pressure variation in the general population.^1-7 More recently in the Victorian Family Heart Study (VFHS) we used a selective genotyping approach by comparing genotypes in unrelated individuals in the highest and lowest systolic blood pressure (SBP) deciles to identify associated SNPs in a gene-centred analysis of SCNN1G.⁸ This approach identified six SNPs with association to BP. In the present study we extend our observations by undertaking much larger and more comprehensive analyses in the VFHS and in three other independent populations. Our strategy was to initially test the six SCNN1G candidate SNPs in all available participants of the VFHS, and also in the Utah High Risk Pedigree Study (UPS).⁹ From these we selected the two most strongly associated SNPs for testing in two additional independent family studies, the Silesian Hypertension Study (SHS)¹⁰ and the Genetic Regulation of Arterial Pressure of Humans in the Community (GRAPHIC)⁷ cohorts. We then combined data from all family studies in a meta-analysis of the highest ranked SNP and examined its association to several available biochemical intermediate phenotypes. Finally, using the human kidney samples from Silesian Renal Tissue Bank we have investigated the correlation between hypertension, the genotype of SCNN1G and its renal expression at the transcript level.¹¹

Methods

Study Populations

8199 individuals of white European ancestry from 1611 families recruited into four cohorts: the Victorian Family Heart Study (VFHS),¹² the Utah High Risk Pedigree Study (UPS)⁹, the Silesian Hypertension Study (SHS)¹⁰ and the Genetic Regulation of Arterial Pressure of Humans in the Community (GRAPHIC)⁷were included in the association analysis. Details of the recruitment and phenotypic measures for each of the cohorts have been published previously and are summarised in supplementary Table S1. In brief, the VFHS and GRAPHIC families were recruited from the general population and the SHS and the UPS were recruited through probands with high cardiovascular risk. BP/hypertension was a primary phenotype of interest in all four studies. Systolic (SBP) and diastolic blood pressures (DBP) were measured using standard sphygmomanometry in the VFHS and SHS, and using automated devices in the UPS (Infrasonde SR-2 Automatic Blood Pressure Recorder, Sphygmetrics Inc., Woodlands, CA, USA) and GRAPHIC (Omron HEM-705CP digital BP monitor) cohorts. In all cohorts, subjects had three blood pressure measurements taken. Representative measures used in this study were those used originally by the specific cohorts: for the SHS this was the average of the three measurements and in the VFHS, UPS and GRAPHIC this was the average of the last two measurements.

The UPS cohort is a prospective study with data available at baseline and 25 year follow up time points. Only the recent 25 year follow up time point was used in the present study. In 648 of the UPS participants, urinary concentrations of sodium, potassium and aldosterone were obtained from a 12-hour overnight urine sample at the same 25 year follow up time point. Plasma sodium and potassium from fasting blood samples were obtained for 1292 of the UPS participants.^13-14

The Silesian Renal Tissue Bank is a collection of human kidneys obtained from adult Polish patients (all of white European ancestry) with non-invasive renal cancer who underwent elective unilateral nephrectomy.¹¹ Blood pressure measurements were conducted (in triplicate) in each subject according to the protocol used in the SHS.¹⁰ Hypertension was defined as SBP and/or DBP ≥140/90 mmHg on at least three separate occasions and/or treatment with anti-hypertensive medication.

Genotyping and Quality Control

Genotyping was first performed in all subjects from the VFHS and UPS cohorts on six SNPs identified previously in the VFHS based on evidence of suggestive association with SBP (P < 0.1) in our earlier selective genotyping case-control study (rs13331086, rs11074553, rs4299163, rs5740, rs4281710 and rs4470152).⁸ Genotyping was performed using the LightScanner High Resolution Melting (HRM) platform (Idaho Technology) which resolves genotype based upon melting temperature of the DNA strands. Genotyping of the two SNPs with strongest evidence of association with BP in the VFHS and UPS (rs13331086 and rs11074553) was conducted using commercially available TaqMan® assays on the ABI Prism 7900HT Sequence Detection System (Applied Biosystems) in the SHS. Genotypes of these two SNPs in the GRAPHIC Study were retrieved from large-scale genotyping conducted using 50K Illumina Human CVD BeadChip array.¹⁵ Genotypes for all SNPs passed our quality-control thresholds (call rate > 90%) and were in Hardy-Weinberg equilibrium (P > 0.05).

Statistical Analysis

Means or proportions for BP and covariates were calculated for each population. Genotypes were tested for Mendelian inconsistencies using family-based association test software (http://www.biostat.harvard.edu/~fbat/fbat.htm). Prior to statistical analysis, BP values were adjusted for treatment by adding 10mmHg to SBP and 5mmHg to DBP according to previously described and validated methods.^16-17 Initial investigation of association with BP phenotypes was performed in the VFHS population using quantitative trait linkage disequilibrium (QTLD) analysis¹⁸ in SOLAR¹⁹ using a variance components model specific to the VFHS family structure including additive genetic, shared environment and non-shared environmental components of variance (described in detail in the online supplement). In order to adjust for multiple testing we used the Single Nucleotide Polymorphism Spectral Decomposition (SNPSpD) method of correcting for non-independence of SNPs in LD with each other.²⁰ We did not correct for the number of phenotypes tested because of the strong linear correlation between SBP and DBP (r = 0.74).

Associations between SCNN1G SNPs and SBP and DBP in the UPS, GRAPHIC and SHS were examined using GEE analyses with an exchangeable correlation structure to adjust for intra-familial correlations and using age, sex and BMI as additional covariates. In addition, because of the age structure in the GRAPHIC Study (paternal generation - 40-60 years; offspring –18-40 years) we included age² as an additional covariate in the association analysis. In the UPS (with three generations) we included age³.

We then combined the inverse variance weighted averages of β-coefficients and SEs from VFHS, UPS, GRAPHIC and SCS in a fixed effects meta-analysis. A summary effect size and overall P value were calculated for the combined sample under an additive model of inheritance using METAL software (http://www.sph.umich.edu/csg/abecasis/metal/index.html). The between-study heterogeneity was evaluated using a χ² test.

Genotype imputation analyses

In order to evaluate whether the association signal could be attributed to other SNPs in the region we performed imputation of genotypes in the GRAPHIC cohort. A total of eleven SCNN1G SNPs genotyped using the 50K IBC array in the GRAPHIC cohort ²¹ and listed in Supplementary Table S2 were used for imputation. All of these SNPs were in Hardy-Weinberg equilibrium (in the parental generation; P > 0.05) and had very low missingness rate (<2%). The genotyped SNPs provide approximately 90% coverage for common SCNN1G SNPs (minor allele frequency > 5%).

We used IMPUTE version 2^22-23 (http://mathgen.stats.ox.ac.uk/impute/impute_v2.html) to impute genotypes for SNPs in the chromosome 16 interval 23,110,000 to 23,320,000 using reference haplotypes from the 1,000 Genomes Project and HapMap Phase 3. We then performed GEE analysis on the imputed genotypes as described above. Only genotypes with ≥ 80% probability after imputation were utilised in the analyses.

SCNN1G mRNA expression analysis

In brief, renal tissue was sampled immediately after surgery from the pole of the kidney that was unaffected by cancer and was secured in RNAlater (Ambion, TX) for further mRNA analysis. Total RNA was extracted from 43 human renal samples (24 hypertensive, 19 normotensive; 25 males, 18 females) with the use of a commercially available assay (RNeasy, Qiagen) according to the manufacturer’s protocol. First-strand cDNA was synthesized with the use of 500 ng of total RNA and random hexamer primers (High Capacity cDNA Reverse Transcription Kit, Applied Biosystems) and analysed by qPCR using Power SYBR Green chemistry and the Eppendorf Realplex system (Eppendorf, Germany). Pre-designed QuantiTect primers (Qiagen, USA) were used (QT00063217) and normalized to expression of the cyclophilin control gene. Each assay was performed in triplicate using 20 μl reaction mixtures containing 10 μl of Power SYBR Green PCR Master Mix (2X) (Applied Biosystems, USA) and 200 nM of forward and reverse primers. Amplification was performed according to the following conditions: one cycle at 95 °C for 10 min, 50 cycles at 95 °C for 15 s and 60 °C for 1 min. Melting curve analyses were performed to check PCR product specificity. Data, expressed as cycle threshold (Ct), were used to determine dCt values [Ct(SCNN1G) – Ct(Control)]. The fold difference in SCNN1G expression between hypertensive and normotensive subjects was calculated according to the following formula:

$$\text{fold difference}=2^{-\text{difference between dCt of hypertensive and normotensive samples}}.$$

²⁴ Genotyping of rs13331086 was completed in the Silesian Renal Tissue Bank (SRTB) using TaqMan® assays on the ABI Prism 7900HT sequence detection system (Applied Biosystems). Due to low minor allele frequency, individuals with minor homozygous genotype were grouped with heterozygous genotype subjects in the data analysis. A two tailed t test was used to compare differences in renal expression of SCNN1G between both groups.

Additional information on materials and methods is provided in the online Data Supplement (please see http://hyper.ahajournals.org).

Results

Association analyses

SCNN1G and blood pressure in the Victorian Family Heart Study

Of the 2959 participants in the VFHS cohort, DNA was available for 2876 individuals and this was included in the present study. Six SNPs were genotyped in this cohort, rs13331086, rs11074553, rs4299163, rs5740, rs4281710 and rs4470152. After adjustment for treatment effect and demographic variables in QTLD analysis we identified an association between SBP and one SCNN1G SNP (rs13331086) (QTLD P = 0.014) (Table 1). This result passed the significance threshold of P < 0.017 as determined by the SNPSpD method to account for multiple testing. Our estimations showed that each additional minor allele was associated with a 0.81 mmHg (SE ± 0.46 mmHg) increase in SBP (Table 1). The direction of the allele effect on DBP was consistent with SBP but the association was not significant (P = 0.58) (Table 1). SNP rs11074553 showed borderline association with SBP (P = 0.08). No other SNPs displayed any evidence of association with these phenotypes (all other P values > 0.1).

Table 1.

QTLD variance components analysis of SCNN1G SNPs and blood pressure in the VFHS

SNP	Minor allele	MAF	SBP			DBP
			β	SE	P	β	SE	P
rs13331086	G	0.23	0.81	0.46	0.01	0.37	0.29	0.58
rs11074553	G	0.42	0.47	0.37	0.08	0.23	0.26	0.70
rs4299163	C	0.19	−0.07	0.47	0.87	−0.01	0.31	0.62
rs5740	A	0.19	−0.41	0.48	0.70	−0.09	0.31	0.44
rs4281710	A	0.18	−0.21	0.48	0.99	−0.10	0.32	0.25
rs4470152	T	0.18	−0.23	0.49	1.00	−0.08	0.32	0.31

MAF, minor allele frequency; SBP, systolic blood pressure; DBP, diastolic blood pressure.

Utah Pedigree Study

Evidence of association in the UPS cohort was detected between the SCNN1G SNP rs13331086 and DBP - each additional copy of the minor allele was associated with a 1 mmHg (± 0.48 mmHg, P = 0.04) increase in DBP (Supplementary Table S3). The association between SBP and rs13331086 was not observed in this population (P = 0.49). Other SNPs were not associated with BP (all other P values > 0.1).

GRAPHIC Study

There was a borderline association between rs13331086 and SBP (β = 1.15, SE = 0.6, P =0.06) in the GRAPHIC Study. No association was observed with rs11074553 (all P values > 0.1, see Supplementary Table S4).

Silesian Hypertension Study

In the SHS cohort, there was association between rs13331086 and both SBP (β = 4.25, SE = 1.74, P = 0.02) and DBP (β = 2.05, SE = 1, P = 0.04). No association was observed with rs11074553 (P > 0.1, see Supplementary Table S5).

Meta-Analysis

In the combined inverse variance fixed effects meta-analysis of the data from the 6753 informative subjects from the VFHS, UPS, SHS and GRAPHIC cohorts, each copy of the minor allele of rs13331086 was associated with 1.01 mmHg higher SBP (P = 0.002) and 0.52 mmHg higher DBP (P = 0.01) after adjustment for age, sex and BMI (Table 2). There was no evidence of heterogeneity in effect sizes across the four populations included in the meta-analysis (P = 0.19).

Table 2.

Association of SCNN1G SNPs rs13331086 and blood pressure by inverse variance weighted fixed effects meta-analysis

Cohort	n	SBP			DBP
		β	SE	P	β	SE	P
VFHS	2876	0.81	0.46	0.01	0.37	0.29	0.58
UPS	1274	0.69	0.86	0.42	1.09	0.57	0.05
GRAPHIC	2020	1.12	0.60	0.06	0.27	0.36	0.45
SHS	583	4.25	1.74	0.02	2.53	1.00	0.04
Meta-Analysis	6753	1.01	0.33	0.002	0.52	0.21	0.01

VFHS, Victorian Family Heart Study; UPS, Utah Pedigree Study; GRAPHIC, Genetic Regulation of Arterial Pressure of Humans in the Community cohort; SHS, Silesian Hypertension Study. SBP, systolic blood pressure; DBP, diastolic blood pressure. Individual population β-coefficients (β), Standard Errors (SE), and P values were obtained from GEE-based association tests for quantitative traits for UPS, GRAPHIC and SHS cohorts and from QTLD analysis for VFHS. Combined β, SE and P were obtained using inverse variance meta-analysis in METAL. All P values were obtained after adjustment for hypertension treatment, age, (and age², age³, where appropriate), sex and BMI. n- number of subjects with complete genotypic and phenotypic information.

Imputation analyses

A total of 43 SCNN1G SNPs were included in the analysis of clinic SBP and DBP after all post-imputation quality filters in the GRAPHIC Study. GEE-based analysis was adjusted for treatment effect as well as age, age², sex and BMI as per the original analyses. None of the imputed SCNN1G SNPs showed association with clinic BP at the nominal level of statistical significance; however the magnitude of association with SBP for a total of 26 imputed SNPs (linked to rs13331086, LD r² > 0.9, Supplementary Table S6) was similar to that with rs13331086.

Intermediate biochemical phenotypes

Given the associations described above, and knowing the function of the epithelial channel to promote renal sodium reabsorption and potassium secretion under the influence of aldosterone, we performed further analyses of data available from the Utah Pedigree Study. rs13331086 was associated with urinary potassium (P = 0.04), such that each additional minor allele was associated with an estimated 2.21 mmol (SE = 1.15) increase in 12-hour overnight urinary excretion of potassium above the mean. No significant evidence of association was detected between rs13331086 and 12-hour urinary excretion of sodium (P = 0.24) or aldosterone (P = 0.24). No association was observed with the urinary sodium/potassium ratio (P = 0.99), nor with plasma sodium or potassium levels (P = 0.30 and 0.57, respectively).

SCNN1G and Hypertension: Renal Expression Analysis

Our preliminary analyses of SCNN1G expression in renal tissue revealed that compared with normotensive individuals, hypertensive patients had a 1.7-fold higher expression of SCNN1G in kidneys (dCt for normotensives = 2.5 ± 0.33 vs 1.7 ± 0.22 for hypertensives; P = 0.07). The genotype distribution of rs13331086 in these individuals was in HWE (P = 0.41) and we observed that there was no difference in renal expression of SCNN1G between contrasting genotypes of rs13331086 (P = 0.36; minor allele carriers n = 20, dCt = 1.87 ± 0.17; major allele homozygotes n = 23, dCt = 2.22 ± 0.21).

Discussion

In these independent family-based analyses of the association of BP with genetic variation in SCNN1G, our attention was focused on the SNP rs13331086, which we had detected as the most significant variant in our previous study.⁸ In the present study we undertook family-based association analyses in the entire VFHS cohort and obtained corroborative evidence for association between rs13331086 and SBP. Meta-analysis in the four populations indicated that the minor allele of this SNP (G) was associated with a 1.01 mmHg increase in SBP and 0.52 mmHg increase in DBP. In an individual homozygous for the minor allele of the SNP this equates to a BP on average 2/1 mmHg higher than those homozygous for the common allele. Although modest in clinical terms, the magnitude of association between blood pressure and rs13331086 corresponds well with expected effect of a single SNP on this complex, polygenic trait. Similar estimates were documented in recently published GWAS for BP.²⁵

The SNP rs13331086 is located in an intronic region of SCNN1G that is not evolutionarily conserved and has no obvious functional annotation, however according to the UCSC Genome browser database (http://genome.ucsc.edu/), the SNP rs13331086 lies within a human expressed sequence tag and it therefore has a potential role as a regulator of SCNN1G expression. It is possible that a functional variant in strong LD with rs13331086 could be a driver for the observed association. Indeed, 26 SCNN1G SNPs are in strong LD with rs13331086 (Figure 1). Of these, at least 3 may affect mRNA structure/protein function when introduced into in silico analyses. The SNP rs5723 is a synonymous coding variant, SNP rs5727 is located in the 3′ untranslated region and is predicted in silico by the SNP Function Annotation Portal (http://brainarray.mbni.med.umich.edu/Brainarray/default.asp) to be a potential microRNA binding site for hsa-miR-29a and hsa-miR-29c, while SNP rs13306653 is predicted to be in a functional region of the gene (splice site). Similar to rs13331086 these 3 potentially functional SNPs showed borderline association with clinic SBP in the imputation analysis.

Figure 1.

1000 Genomes Project SNPs (CEU) in LD with rs13331086. Plot derived from the SNP Annotation and Proxy Search (http://www.broadinstitute.org/mpg/snap/). SNPs with r²>0.8 to rs13331086 are indicated as large diamonds above the dotted line.

This predominantly genetic analysis was not designed to investigate potential phenotypic correlates in any detailed fashion. The availability of data from the UPS cohort regarding relevant urinary and plasma measures related to the renal function of the epithelial sodium channel offered an opportunity for preliminary analysis. We found that the G allele of rs13331086, which is associated with higher BP, was also associated with increased urinary potassium excretion measured over a 12-hour overnight period. One could postulate increased expression of ENaC in the distal nephron of the kidney that would favour increased sodium reabsorption, which as a result of subsequent electrochemical gradients would favour the excretion of potassium via the potassium channel ROMK.²⁶ We did not observe any significant association of the SNP with either urinary Na or the Na/K ratio. A previous study examining hypertensive patients compared to normotensive controls found that variants in the β- and γ-subunit ENaC genes were associated with the urinary potassium to plasma renin ratio, though they did not detect association with either variable alone.²⁷ They suggested that even without strong effects on the phenotypes, the variants could act as subtle genes conferring predisposition to sodium retention and hypertension acting continuously over a lifetime. Additionally, a recently published study has shown that SCNN1G variants are associated with the BP response to dietary sodium intake, indicating that SCNN1G may be involved in short term response to sodium.²⁸

Accordingly, the expression study of human renal tissue indicates a trend towards increased expression of SCNN1G mRNA in hypertensive patients compared to controls. Although not significant, and while it may not be directly linked to the SNP rs13331086, the fact that this trend emerges from the diverse sources of confounding variables that might obscure potentially subtle expression changes in the kidney indicates a potentially important lead that requires more detailed studies. It is clear that an increase in ENaC expression is associated with the development of hypertension while the inverse is related to decreased BP, as is evident in the phenotypes of Liddle’s syndrome and Pseudohypoaldosteronism Type I, respectively.²⁹ Moreover, decreased ENaC expression/activity has been shown to prevent the development of hypertension in a mouse model of activation of the thiazide sensitive sodium chloride cotransporter.³⁰ To our knowledge, this is the first study to compare the expression of ENaC between the hypertensive and normotensive kidney in humans.

It would be fair to acknowledge that although traditionally considered as a renal determinant of BP, the epithelial sodium channel is also expressed in other tissues. For example, ENaC has been located on smooth muscle cells in arteries where it appears to serve as an initiator of pressure dependent myogenic responses in juxtamedullary afferent arterioles.³¹ Other vascular autoregulatory roles of this channel have been attributed to endothelial mechanisms that mediate the arterial flow-dependent vasodilatation. Evidence exists that ENaC mediates sodium entry into endothelial cells, thereby altering their deformability and sensitivity to shear stress that results in the release of nitric oxide and consequently vasodilatation.³² These channels are also known to exist in central nervous regions involved in BP homeostasis and have been identified as important mediators of aldosterone production in the adrenal cortex.^33-34

Perspectives

Identifying the genes responsible for population variation in BP is a complex challenge. While the large GWAS have been successful in identifying novel genes for multiple complex disorders, in the field of hypertension they have had limited success. Using a candidate gene approach with a strong physiological justification, we have identified a common genetic variant in the SCNN1G gene associated with increased SBP and DBP in a meta-analysis of over 6000 individuals. Additionally, our preliminary experiments indicate increased SCNN1G mRNA expression in hypertensive patients and hint at increased ENaC activity in subjects carrying the minor allele. This common variant with relatively small phenotypic effects on a population-wide basis adds to the role for SCNN1G in BP, better known for the well-characterized rare SCNN1G mutations associated with hypertensive Mendelian diseases. It is clear that a thorough investigation of the physiological actions associated with the rs13331086 SNP and those in strong LD should be a focus for further research.