|Home | About | Journals | Submit | Contact Us | Français|
Essential hypertension is a multi-factorial disorder and is the main risk factor for renal and cardiovascular complications. The research on the genetics of hypertension has been frustrated by the small predictive value of the discovered genetic variants. The HYPERGENES Project investigated associations between genetic variants and essential hypertension pursuing a two-stage study by recruiting cases and controls from extensively characterized cohorts recruited over many years in different European regions.
The discovery phase consisted of 1,865 cases and 1,750 controls genotyped with 1M Illumina array. Best hits were followed up in a validation panel of 1,385 cases and 1,246 controls that were genotyped with a custom array of 14,055 markers. We identified a new hypertension susceptibility locus (rs3918226) in the promoter region of the endothelial nitric oxide synthase (eNOS) gene (odds ratio 1.54; 95% CI 1.37-1.73; combined p=2.58·10−13). A meta-analysis, using other in-silico/de novo genotyping data for a total of 21714 subjects, resulted in an overall odds ratio of 1.34 (95% CI 1.25-1.44, p=1.032·10−14). The quantitative analysis on a population-based sample revealed an effect size of 1.91 (95% CI 0.16-3.66) for systolic and 1.40 (95% CI 0.25-2.55) for diastolic blood pressure. We identified in-silico a potential binding site for ETS transcription-factors directly next to rs3918226, suggesting a potential modulation of eNOS expression. Biological evidence links eNOS with hypertension, as it is a critical mediator of cardiovascular homeostasis and blood pressure control via vascular tone regulation. This finding supports the hypothesis that there may be a causal genetic variation at this locus.
Essential hypertension (EH) is a clinical condition affecting a large proportion (25-30%) of the adult population and is a major risk factor for cardiovascular and renal diseases.1,2 It is a complex trait influenced by multiple susceptibility genes, environmental and lifestyle factors and their interactions. 3 In the last years, huge efforts have been performed in recruiting and genotyping tens of thousands of individuals and meta-analysing dozens of cross-sectional population-based studies. In spite of this, the research on the genetics of EH has been frustrated by the small predictive value of the discovered genetic variants and by the fact that these variants explain a small proportion of the phenotypic variation. 4-13 EH is a late-onset disease and therefore the small discovered effect sizes could in part be due to the effect of misclassification, sample selection bias and inappropriate phenotyping of cases and controls.9,14,15 The selection of cases and controls may have important effects on the results as misclassification bias can lead to loss of power. For common traits, such as EH, this bias can be remedied by defining more stringent selection criteria, by recruiting hyper-normal controls and adopting a more stringent case definition.14,15
The HYPERGENES Project pursued a two-stage study to investigate novel genetic determinants of essential hypertension. Cases and controls were recruited from extensively characterized cohorts over many years in different European regions using standardized clinical ascertainment. Particular care was devoted to control selection. A large proportion of the sample has been followed for 5–10 years after DNA collection, allowing for the exclusion of controls that developed hypertension at a later age, thereby defining the hyper-normal controls.
Cases and controls were recruited from extensively characterized cohorts using standardized clinical ascertainment, collected over many years in different European regions (balanced within North Europe, Continental Italy and Sardinia). The inclusion criteria are described in Methods S1 (available online at http://hyper.ahajournals.org). In order to perform a genetic association with continuous BP phenotypes, we considered two additional cohorts (FLEMENGHO-EPOGH, n=1514 and WHSS, n=306, see methods S2 (please see http://hyper.ahajournals.org) that provided population-based data. Description of the different samples is reported in the methods S2 (http://hyper.ahajournals.org).
Genotyping details are shown in methods S3-S6 (please see http://hyper.ahajournals.org). Briefly, in the Discovery phase, the samples were genotyped using the Illumina 1M-duo array and the imputation was performed with MACH 16 using as reference the 1000 Genomes haplotypes (release June 2010) (method S3, please see http://hyper.ahajournals.org). To validate and fine map the genes found associated with EH in discovery phase an Illumina custom chip of 14,055 markers was created starting from the list of best-associated and of candidate SNPs based on a priori biological knowledge (methods S4-S5, please see http://hyper.ahajournals.org). For the replication stage, we used the in-silico data of rs3918226 from ASCOT/AIBIII/NBS, BRIGHT, EPIC Turin, HYPEST and NORDIL/MDC studies (methods S6, please see http://hyper.ahajournals.org).
All quality controls and statistical analyses were performed in accordance with the protocols written by C.A Anderson 17 and G.M. Clarke 18 (methods S7-S9, http://hyper.ahajournals.org). We tested each SNP for association with hypertension using a logistic regression under an additive model with adjustment for sex and for the first 10 principal components. Combined analysis for discovery, validation and replication results was conducted using METAL. 19 The quantitative effect of rs3918226 on SBP and DBP was tested on two additional population-based cohorts (methods S2, please see http://hyper.ahajournals.org). Moreover, we tested for multiplicative interaction between rs3918226 and the most plausible interactive partners of eNOS gene: actin genes and Heat Shock Protein-90 genes (methods S9, please see http://hyper.ahajournals.org). The quantitative effect of rs3918226 on SBP and DBP has been tested on two additional population-based cohorts (EPOGH-FLEMENGO and WHSS, see methods S2 http://hyper.ahajournals.org).
A classical two-stage case-control strategy was employed with a discovery phase of 1,865 cases and 1,750 controls (2,294 males, 1,321 females), all genotyped on the Illumina 1M Duo chip. The sample consisted of an ethnically diverse population (25.06% North Europeans, 38.70 % Sardinians and 36.24% Continental Italy subjects). The discovery phase was followed by a validation phase of additional 1,385 cases and 1246 controls (1,417 males and 1,214 females). According to ethnicity, the validation sample was comprised of 1262 North Europeans (47.97%), 788 Sardinians (29.95 %) and 581 Continental Italians (22.08%). Tables S1 and S2 (please see http://hyper.ahajournals.org ) show the demographic characteristics and baseline measures.
Principal Component Analysis (PCA) of the genotype data was carried out to find the major axes of variation used as covariates to correct for population stratification. 23 The discovery samples in the principal component map showed three (roughly) equal-sized distinct clusters corresponding to the three main ethnic groups, as expected from the study design (Figure S1, please see http://hyper.ahajournals.org). All association analyses were adjusted for the ancestry principal components and sex by including them as covariates in the logistic regression model. In addition genomic control (GC) correction was applied (since genomic inflation factor was 1.04). In the discovery phase, 90 SNPs (57% intragenic) with p-value < 1·10−4 were identified after GC (Figure S2, Table S4, please see http://hyper.ahajournals.org). The most promising SNPs were genotyped in the validation samples using an Illumina Infinium Custom chip. The meta analysis of the discovery and validation data revealed SNP rs3918226 to be associated with EH in Caucasians, reaching a Pcombined of 2.58·10−13 and OR of 1.54 per T allele (95% CI, 1.37-1.73) under an additive model (Figure 1, Table 1 and Figure S4 at http://hyper.ahajournals.org). Estimated odds ratios in the Discovery and Validation samples were consistent across the different Caucasian populations of the HYPERGENES sample (Figure S5, please see http://hyper.ahajournals.org).
The polymorphism rs3918226 maps to the promoter region of the eNOS gene (−665 C>T, NOS3). 24,25 The T allele frequencies in the present study are 13.8% in cases and 8.9% in controls. SNP rs3918226 is monomorphic in the non-Caucasian HYPERGENES samples (Wandsworth Heart & Stroke Study cohort, WHSS) and African and Asian HapMap samples. The second best hit chr7:150,314,954 (G/A SNP, MAF of A allele= 3%) with P value 2.46·10−6 and OR 2.25 was imputed based on the 1000 Genomes haplotypes (release June 2010), its imputation quality was very high (r2-hat = 0.94). Unfortunately we couldn’t replicate the observation in validation due to low imputation quality. Further 7 SNPs within eNOS gene showed significant p-values (1·10−3 < p-values < 1·10−5): rs2853792 (intronic, Pcombined = 7.76 ·10−5), rs1549758 (coding, Pcombined = 3.32 ·10−4), rs1800779 (intronic, Pcombined = 1.16 ·10−3), rs6951150 (intergenic, Pcombined = 1.64 ·10−3), rs743507 (intronic, Pcombined = 1.76 ·10−3), rs1800780 (intronic, Pcombined = 1.96 ·10−3), rs1800783 (intronic, Pcombined = 2.89 ·10−3) (Figure 1).
Table 1 shows also other significant SNPs with p-values between 1·10−3 and 1·10−5 mapping different genes as calcium-activated potassium channel subunit alpha-1 (KCNMA1), plasminogen (PLG), retinoid-related orphan receptor alpha (RORA) and WW domain-containing protein 1 (WWC1).
Moreover, the signals of SNPs previously presented in literature are in our study in the same direction as the original studies 5,6,8 showing evidence of a marginally significant association in HYPERGENES (Table S5, please see http://hyper.ahajournals.org).
We meta-analyzed rs3918226 using in silico data from ASCOT/AIBIII/NBS, BRIGHT, EPIC-Turin, HYPEST and NORDIL/MDC samples (methods S2, S6, please see http://hyper.ahajournals.org) resulting in an overall OR of 1.34 per T allele (95% CI 1.25-1.44, Pcombined = 1.032·10−14) (Table 2 and Figure 2) for a total of 21,714 subjects. Since case and control definitions differed between HYPERGENES and the in-silico replication samples, the ORs are not directly comparable. In our study, the p value of heterogeneity calculated for HYPERGENES samples is 0.13. It decreased slightly, but remained non-significant, as expected, when also EPIC-Turin was considered together in the meta-analysis (p=0.092) since the recruitment criteria for cases and controls were identical. Conversely, the heterogeneity increased significantly (p=0.005) when HYPERGENES samples were meta-analyzed with all the other samples (ASCOT/AIBIII/NBS, BRIGHT, HYPEST and NORDIL/MDC).
Moreover we tested for epistatic multiplicative interactions between eNOS rs3918226 and all available polymorphisms in genes known to be involved in targeting and regulating the overall availability of eNOS at the cell membrane 26,27,28 actin genes (ACTA1, ACTA2, ACTB, ACTG1, ACTG2) 29,30 and HSP90 genes (HSP90AA1, HSP90AA2, HSP90AB1).26 Nominally significant interactions were observed between rs3918226 and rs13447427 (p-value =1.34·10−3) in actin beta gene (ACTB), rs7503750 (p-value =1.57·10−3) in actin gamma 1 (ACTG1) and rs4922796 and rs17309979 (p-value =3.47·10−3, p-value =4.88·10−3) in heat shock protein alpha 2 (HSP90AA2) (Table S6, please see http://hyper.ahajournals.org). When controlling for multiple testing these interactions remained significant at a False Discovery Rate of 20%.
The quantitative analysis confirmed the qualitative observation. In fact, the β coefficient of the regression between SBP or DBP with rs3918226 is respectively 1.91 (95% CI 0.16-3.66) and 1.40 (95% CI 0.25-2.55) per T allele. The coefficient is the effect size on blood pressure in mm Hg per coded allele based on an additive genetic model. The BP distribution according to rs3918226 genotype is shown in Table S7 (please see http://hyper.ahajournals.org).
Since rs3918226 maps to the promoter region of eNOS, we tested whether it may fall into a regulatory binding site. Using the PATCH algorithm of TRANSFAC database 21 we characterized a putative binding site for transcription-factors of ETS family directly next to rs3918226. The ETS family members are present in endothelial cells and participated in activation of the eNOS promoter. 31 Using the TFSEARCH tool 22 we confirmed this finding with a score of 87.3.
We also tested the degree of evolutionary conservation of rs3918226 locus in primates and placental mammals using the conservation track of UCSC genome browser. Figure S6 (please see http://hyper.ahajournals.org ) shows that the region in which rs3918226 lies is conserved from placental mammals to primates.
Essential hypertension (EH) is a complex clinical condition representing the main risk factor responsible for renal and cardiovascular complications. The HYPERGENES Project investigated undiscovered associations between genetic variants and EH pursuing a two-stage study by recruiting cases and controls from extensively characterized cohorts recruited in different European regions.
We discovered rs3918226 in the promoter region of the eNOS gene (endothelial nitric oxide synthase) to be significantly associated with hypertension (OR, 1.54; 95% CI, 1.37 to 1.73; p-value = 2.58·10−13). The result was confirmed by meta-analyzing in-silico data for a total of 21714 subjects (OR, 1.34; 95% CI 1.25-1.44; p-value = 1.032·10−14). We observed heterogeneity in the findings of meta-analysis (p=0.005 for Q-test of heterogeneity) that could be due to both different sample sizes and recruitment criteria not directly comparable between HYPERGENES and the in-silico replication samples (Figure 2).
The quantitative effect of rs3918226 was also estimated in continuous BP phenotypes, resulting in a β coefficient of 1.91 for SBP and 1.40 for DBP, despite the low p-values of the regression probably due to the low sample size. This finding reinforces the observation on the qualitative phenotype.
We identified a potential transcription-factor binding site for the ETS-family domain directly next to rs3918226. The members of ETS family, as ETS-1 and ELF-1, are essential factors for the activation of eNOS promoter.31 This suggests that, by affecting transcription factor-binding affinity, rs3918226 might modulate the transcription of eNOS gene.
It is also worth noting that the region in which rs3918226 lies is conserved from placental mammals to primates.
We propose rs3918226 as a novel susceptibility SNP since among the GWASs so far published this is the first that points to eNOS: the novelty of the rs39118226 finding is that the association between eNOS and Hypertension has been found in Caucasians using a GWAS approach.
The use of the Illumina 1M array and Human CVD BeadArray was crucial in detecting the association since rs3918226 is not present on other commercial arrays.32 Besides being poorly covered by other genotyping platforms, the region has a relatively high recombination rate towards the coding region (Figure 1). This has resulted in low linkage disequilibrium (LD) with markers present on older platforms (e.g. rsq-hat < 0.2 for Affy500K platform). These facts largely limited the potential to replicate our finding using data from other GWAS samples, almost all of which relied on older platforms.
Indeed eNos has been found inconsistently associated to hypertension with several underpowered candidate gene studies, many of which only focused on a few variants with relatively small numbers of cases and controls, compared to the large sample sizes of GWAS. Positive studies were substantially on Asian cohorts 33,34,35, whereas the majority were negative in Caucasians, as summarised in a recent meta-analysis.36 The polymorphisms studied in our Caucasian sample, G894T (rs1799983) and T-786C (rs2070744) did not reach genome-wide significant association with hypertension. If looked with candidate gene threshold, the p-value and the sample size of the present study by far outnumber all the other published so far. rs1799983 was associated with EH with a p value = 2.63×10−3 (OR =1.038) and rs2070744 with a p value = 6.42×10−4 (OR = 1.04), as shown in Table S8 (please see http://hyper.ahajournals.org). To summarize, the ORs are clinically irrelevant. We would like to underline the low LD between rs3918226 and rs1799983 (R2=0.16) and rs2070744 (R2=0.17) suggesting that these two SNPs are independent from rs3918226 and do not have any additional effect on the phenotype.
There is considerable biological evidence linking eNOS with hypertension and hypertension-associated cardiovascular target organ damage. 37 eNOS, which catalyses the synthesis of nitric oxide (NO) by vascular endothelium, is responsible for the vasodilator tone that is fundamental for the regulation of blood pressure. Furthermore, eNOS is a critical mediator of cardiovascular homeostasis through regulation of blood vessels diameter and of the maintenance of an anti-proliferative and anti-apoptotic environment.
As NO is highly active, it cannot be stored inside producing cells. Indeed, eNOS signalling capacity must be controlled, at least in part, by regulating its targeting from Golgi apparatus to plasma membrane, by its compartmentalization within the plasma membrane and by its later internalization from the plasma membrane to the cytoplasm. eNOS is a dually acylated peripheral membrane protein that is targeted to endothelial plasmalemmal caveolae through an interaction with the caveolae structural protein, Caveolin-1 (Cav1). 26,27 Cav1 inhibition of eNOS is lessened by Calmodulin (Calm) causing dissociation of eNOS from Caveolin. This regulatory mechanism is further altered by Heat Shock Protein-90 (HSP90) 27 which binds to eNOS and facilitates displacement of Cav1 by Calm. Moreover, eNOS directly interacts with actin cytoskeleton. 29 Recently, Kondrikov added that beta-actin is associated with eNOS oxygenase domain increasing eNOS activity and NO production.30 To explore such pathway we tested the interaction between the discovered eNOS SNP and its most plausible interactive partners. We observed nominally significant interactions between rs3918226 and rs13447427 in actin beta (ACTB), rs7503750 in actin gamma 1 (ACTG1) and rs4922796 and rs17309979 in Heat Shock Protein-90 alpha 2 (HSP90AA2) gene.
In conclusion, with a stringent case-control design and a population based study, we identified a novel hypertension susceptibility locus in the promoter region of eNOS with a relatively high effect size. Our finding could provide new insights into the mechanism of vascular regulation and could help in better understanding the genetics of EH. Furthermore, we believe that this indication can be useful to guide fine-mapping or sequencing efforts to single out causal variants.
Further investigations and high-throughput sequencing of region of interest will help to identify the real causal variant and to clarify the functional role of eNOS in essential hypertension.
The complete list is reported in supplemental material.
Sources of Funding This work was supported by the HYPERGENES project (European Network for Genetic-Epidemiological Studies: building a method to dissect complex genetic traits, using essential hypertension as a disease model), grant HEALTH-2007-201550, funded by EU within the FP7.
Disclosures Toby Johnson was supported by the Wellcome Trust (grant number 093078/Z/10/Z). John Chalmers has received research grants from Servier International and from the NHMRC (Australia), administered through the University of Sydney, for the PROGRESS and ADVANCE Trials and for the ADVANCE-Post-trial follow-up study. He has also received honoraria for speaking about these studies at scientific meetings.
Mark Caulfield was supported by British Heart Foundation grant for developing CVD BeadArray and KASPAr assay.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.