In this sample of African Americans, none of the SNPs evaluated were convincingly associated with hypertension as a binary trait or with blood pressure level as a quantitative trait. Thus, we were unable to confirm previously reported associations between PMS1, SLC24A4, YWHA7, IPO7, and CACANA1H with systolic and diastolic blood pressure. However, among normotensives only, one SNP (rs12748299) showed a significant association with systolic blood pressure even after adjusting for multiple comparisons, but was no longer significant after adjustment for age, sex, BMI and serum creatinine. This SNP was previously marginally associated with hypertension as a binary trait in the study by Adeyemo et al, but had not reached genome-wide significance in their study either. It is located on chromosome 1 in the intergenic region and its biological plausibility is unknown at this time.
We observed several suggestive, but not significant associations. SNP rs2146204, located on chromosome 1 in the intergenic region, was associated with systolic blood pressure in all subjects (treated and untreated). Similarly, rs12757682 which is also located on chromosome 1 in the intergenic region, showed borderline association with systolic blood pressure among normotensives. Among hypertensives, borderline associations (p > 0.002) were noted between rs1550576 (in the intergenic region near the gene aldehyde dehydrogenase 1 family, member A2, ALDH1A2
) and systolic blood pressure, and rs17365948 (in the intronic region of the gene tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide, YWHAZ
) and systolic and diastolic blood pressure. The biological plausibility of these associations is currently unknown (Additional file 1
: Table S4).
Similarly, Ehret et al
attempted to replicate six top-associated SNPs (rs2820037 (1q43), rs6997709 (8q24), rs7961152 (12p12), rs11110912 (12q23), rs1937506 (13q21) and rs2398162 (15q26)) from WTCCC in the Family Blood Pressure Program (FBPP) cohort [10
]. DNA was genotyped on 11,433 participants (39% African Americans, 39% European Americans, 21% Hispanic Americans) and the results showed that only one of these six SNPs (rs1937506 on 13q21) was negatively
associated with systolic and diastolic blood pressure among European Americans. The same SNP showed a significant but opposite effect with systolic blood pressure in Hispanic Americans and was not associated in African Americans [19
]. No replication could be shown for hypertension status. The associations that were closest to being suggestive could not be replicated, even in a well characterized sample such as the FBPP suggesting the difficulty in replicating the findings of GWAS [19
]. The most significant criticism of the WTCCC was the use of young controls that could potentially become cases at a later age.
Inconsistent results among several small-scale GWAS [11
] suggested that hypertension is polygenic disease with multiple low frequency genes acting in harmony to result in elevated blood pressure. To overcome some of these limiting factors, a large meta-analysis of CHARGE and BPgen Consortium results was conducted recently and several novel loci have been identified for blood pressure. Although these associations have not been replicated in independent populations, these results may be more robust than previous GWAS as the meta-analysis involved more than 60,000 subjects [15
]. We did not choose to replicate these associations as these studies involved subjects primarily from European and Asian Indian ancestry.
On a similar note, a large GWAS for blood pressure among African Americans, the Candidate Gene Association Resource (CARe) Study, has recently been published. In a meta-analysis across five community-based cohorts, two novel loci were identified that reached statistical significance: rs2258119 on chromosome 21 with systolic blood pressure and rs10474346 on chromosome 5 with diastolic blood pressure [21
]. However, neither of these associations were replicated in independent African American samples, again highlighting the difficulty in extending the findings of GWAS to independent populations. In addition, this study did not identify any of the loci identified by Adeyemo et al
Even though the evidence for a genetic contribution remains robust, the effect of environment may be a major modifier, and this may be the reason for inconsistent data. Many common everyday activities can profoundly affect blood pressure, such as salt intake, exercise, stress, etc., and many of these are not quantitated in population-based studies. To minimize environmental and genetic variability, Wang et al
carried out a GWAS of systolic and diastolic blood pressure in Amish subjects [11
]. Strong association signals with several common variants in a serine/threonine kinase gene (STK39)
were found and they confirmed these associations in an independent Amish and 4 non-Amish Caucasian samples [11
]. Adeyemo et al
found that several SNPs in this gene were associated with the blood pressure (9/136 for systolic blood pressure and 33/136 for diastolic blood pressure) in African Americans. Variants in STK39
may influence blood pressure by increasing STK39
expression and consequently altering renal sodium excretion [11
]. We were unable to confirm the associations of 2 top-associated SNPs in this gene for systolic and diastolic blood pressure in our cohort despite having adequate power. These associations were not tested in CARe study by Fox et al
Another novel locus contributing to blood pressure was identified by Org et al
in a GWAS of a German population and is known as Cadherin 13-Heart
is a calcium-dependent cell adhesion glycoprotein and may mediate interaction between cells in heart. Several SNPs in this locus were also found to be associated with systolic and diastolic blood pressures among African Americans by Adeyemo et al.
We tested the association of 4 of these SNPs with hypertension as a binary trait and blood pressure as a quantitative trait in our sample. We confirmed a borderline association (insignificant when considered with multiple tests) of rs17177428 and rs7200009 with blood pressure as quantitative trait among hypertensive and normotensive subjects. Again, this association was not tested in CARe study by Fox et al
As evidenced by these studies, identification of genetic contributors to hypertension remains challenging for several reasons. Population heterogeneity, environmental effects, and sample size all contribute to these failures. Our study is also subject to these criticisms. In addition, as already eluded to, inheritance of blood pressure is polygenetic where in a single gene or combination of genes act in concert with above mentioned environmental exposures to contribute only a modest effect on blood pressure. Indeed, genetic variants identified by GWAS contribute to only a small effect on blood pressure (1-2 mmHg). Although this may have a population-based impact, this effect may not be discernible among individuals. Moreover, the difficulty of obtaining a standardized blood pressure phenotype cannot be over emphasized.
We also acknowledge the limitations of using commercially available tag SNPs. While it is possible to identify genetic variation with tag SNPs without genotyping every SNP in the chromosomal region and are widely used in genome-wide association studies, it may make replication of target SNPs difficult due to differences in linkage disequilibrium between two samples. For this reason, many studies have now adopted the technique of "local replication" in which SNPs surrounding tag SNP are also targeted, especially for SNPs that are not directly replicated from previous studies [22
Apart from inherent difficulties in identifying genetic determinants of hypertension due to the phenotype itself as well as methods, we acknowledge that the differences between the populations investigated by our study and Adeyemo et. al
. may account for non-replication of the GWAS. Since African Americans as an ethnic group are quite diverse with different origins, we recruited only the individuals whose parents were both born in the United States, with English as their first language, to achieve some uniformity in ethnicity. However, this may not have been sufficiently homogeneous population and the differences in the results may reflect different population admixture rates [23
]. Since we do not know the admixture rate in our population, we were unable to adjust for that while Adeyemo et. al used genome-wide markers to compute principal components and used these in their models. These baseline differences in the African American population composition is also reflected in the allele frequencies of the various SNPs, e.g. CACNA1H (0.015 vs 0.109), IPO7 (0.013 vs 0.123), YWHAZ (0.012 vs 0.113), SLC24A4 (0.067 vs 0.178), PMS1 (0.352 vs 0.148) (Additional file 1
: TableS 5).
There were several strengths to our study. Cases and controls were clearly defined with normotensives having blood pressures in the lower third of the population distribution of blood pressure (~ 3 SD below the untreated hypertensive means) to eliminate the bias of cases masquerading as controls or vice versa. Subjects with secondary hypertension were excluded including those with renal insufficiency. In addition, the mean age of the cases and controls are in mid-forties, which is sufficiently young to capture genetically determined blood pressure increases and old enough to capture most primary hypertensives. However, it is still possible some mis-classification of cases and controls might have occurred.