These longitudinal data demonstrate that the combination GRS associated with higher BP levels in youth and early adulthood. Individuals with the highest combination GRS had significantly higher diastolic BP at the age of 9 years and the effect was persistent from childhood through adult age. Common variants generally have a small effect size, but in adults an effect size of 1 mmHg could translate to a 10% higher mortality risk (22
). The original GWA studies were conducted in older populations, and the extent to which their findings can be validated in the early age groups has not been clarified. Therefore we conducted individual replications for the effects on BP of the GRSs and each of the 13 SNPs in young individuals between 3 and 18 years of age. In older adults systolic blood pressure may be directly associated with cardiovascular risk and mortality (23
), but in a recent analysis of blood pressure data collected from Swedish conscripts, elevated diastolic blood pressure in late adolescence contributed more to subsequent mortality in middle age compared with systolic blood pressure (24
Thanassoulis et al. regarded family history as the best marker for the genetic risk of complex traits (25
). In our cohort, individuals with family history of premature hypertension had 2.46-fold increased odds for hypertension (p<0.0001), demonstrating the importance of both lifestyle and unknown gene-environment interactions. In our study, the GRSs were independent risk factors for hypertension in adulthood, even when adjusted for family history of premature hypertension. A similar finding was reported recently for GRS based on coronary heart disease loci (26
). These results show that even though effects of individual SNPs are small, they jointly add precision to the risk profiling of individuals over family history.
Several important genomic regions are connected to the SNPs used for calculation of the GRSs in this study. The individual SNPs with nominally significant effect sizes in this study were rs11191548, rs16948048, rs17367504, rs9815354 and rs2681492. In secondary analyses two SNPs, rs11014166 (OR=1.21, CI=1.01–1.45, p=0.04) and rs11191548 (OR=1.39, CI=1.01–1.93, p=0.046) were associated with an increased risk of adult hypertension. We also found a sex difference in the effect of the rs11191548, where the effect estimate was 0.9 mmHg per risk allele in women but there was no significant effect in men in this sample. The rs11191548 is located near the gene CYP17A1
encoding cytosolic purine 5′-nucleotidase, an enzyme in the cytochrome P450 family involved in catalyzing the synthesis of cholesterol, steroids and other lipids. However, this locus contains many genes other than CYP17A1,
such as the CYP17
gene where mutations cause congenital adrenal hyperplasia, a rare (Mendelian) cause of hypertension. Little else is known about how this locus might influence BP in humans. Therefore, we calculated the level of linkage disequilibrium (LD) of rs11191548 with other SNPs located within CYP17A1
in our population (please see Table S3 in the online Data Supplement
). rs11191548 was in strong LD with two other SNPs located within CYP17A1. However, as none of the SNPs in this study have a known functional role and there are no missense variants, no eQTL mapping was available to further support the possible involvement of CYP17A1
and therefore variants in other nearby genes could underlie the observed association between rs11191548 and adult hypertension.
The SNP rs16948048 was also associated with diastolic BP in women in our study cohort. The exact mechanism by which this variant can increase BP remains as yet unknown. The rs16948048 is a variant of the ZNF652
gene encoding the zinc finger protein 652, which belongs to a family of transcription-modulating proteins, able to switch genes on and off (27
). Zinc finger proteins are widespread in nature and comprise some 3% of the human genome, and as they are uniquely suited to specifically recognizing large sequences of the genome they are currently a target of intensive research for genetic engineering and gene therapy. Interestingly, the same variant (rs16948048) was associated with a significantly lower risk of hypertension in a recent Chinese Han population-based study by Niu et al, highlighting the need for more research on the pathophysiological mechanisms by which minor genetic differences may act upon the variation of BP both between individuals and populations (28
The SNP rs17367504 near the gene encoding natriuretic peptide B was significantly associated with BP in this study. Common genetic variants at the NPPA-NPPB
loci (natriuretic peptide precursor A and B) have previously been associated with inter-individual variation in plasma natriuretic peptides and BP (9
). The natriuretic peptides may lower BP by promoting the urinary excretion of sodium or through vasodilation, and they are synthesized in response to high BP or volume overload. The SNPs rs9815354 near the gene ULK4
(Unc-51-like kinase 4; serine/threonine protein kinase) and rs2681492 near ATP2B1
transporting, plasma membrane 1) were also associated with BP traits in the cross-sectional analyses.
A limitation of this study is that the combination of risk alleles was based on statistical analysis, and the biological significance of the genetic variants need further clarification before assigning a clinical value to a genetic risk score. Not all of the individual SNPs proved to be significantly associated with blood pressure in our data and the effect of GRS may be “diluted” by the inclusion of SNPs which have little or no effect. Because the GRSs in this study included SNPs which have been discovered in very large international consortia, no correction for multiple testing was performed, and the significance of individual SNPs should be interpreted with caution. Although they were statistically not significant, the smaller effect sizes at ages 27 and 42 possibly arose from cohort effects. We did not include the potential confounders or effect modifiers such as smoking or alcohol consumption in adulthood in this study, but antihypertensive treatment was accounted for by a substitution method. Although the combination GRS was statistically significantly associated with hypertension in multivariable models, it did not significantly improve AUC in addition to age, sex, and BMI as judged by the C-statistics.
Several gene-environment interactions and physiological pathways through which the genetic effects on BP could be mediated are currently under investigation, including salt sensitivity, the renin-angiotensin-aldosterone system as well as endothelial dysfunction (29
). We had cross-sectional data on sodium intakes collected with food frequency questionnaires in adulthood but modification of the association between the genetic markers and blood pressure by salt intake was not detectable. However, the used food frequency questionnaires were not designed to precisely capture salt intake or cumulative salt exposure, and with a more valid assessment method such as repeated 24-h urine collection the intake of salt could possibly have shown to play a more significant role in these analyses (33
The disease probability of complex polygenic traits such as BP can be studied with whole-genome prediction methods, as more SNPs will probably be recognized in future (34
). Recently, Ho et al. used a novel approach combining both GWAS-derived SNPs and a gene expression-guided approach in the identification of a novel locus associated with BP in a large cohort of middle-aged women (38
). Their strategy highlights the importance of identifying biologically functional SNPs and looking for associations beyond the conventional threshold for genome-wide significance (P<5*10−8