We have completed the largest and highest density HTN admixture scan to date and failed to identify any loci that are significantly, or even suggestively, associated with HTN after correcting for multiple hypothesis testing. Our large sample size also allows us to exclude > 95% of the genome as harboring risk loci of > 1.3 due to African or European ancestry. Although we cannot rule out weak ancestry effects, the negative results suggest that there may be no common variants with a strong effect accounting for differences in HTN prevalence between African and European Americans. These results thus increase the weight of evidence that non-genetic causes (diet and environment) contribute to the different epidemiology across populations.
An intriguing aspect of this study is the analysis of four loci from a previously reported admixture scan of HTN [11
]. We observed nominal replication of the admixture association at one of these loci (p
= 0.016 correcting for four hypotheses tested), and the direction of the association (increased HTN with increased African ancestry) is the same as previously reported. We caution that these results could represent statistical fluctuations, as numerous other loci in our scan scored more strongly. Another concern is that the estimate for the increased risk for HTN (1.19) arising from the inheritance of one African allele at this locus is sufficiently small that it would have been very surprising to observe genome-wide significant peaks in a scan with the sample size and map density studied by Zhu et al. [11
]. Further follow-up studies will be necessary to properly test these loci for association. A possible way to reconcile the results from the two studies is that the samples in the Zhu et al. [11
] study had a somewhat different phenotype than the ones we studied. Their definition of HTN, with multiple affected family members, may have been more genetically heritable, and thus their phenotype may have been more likely to yield association signals.
We also specifically evaluated risk at 40 biologically plausible HTN candidate gene loci as well as eight previously identified linkage peaks. For a complex disease, such as HTN, in which effects are expected to be weak, combining the wealth of prior genetic and biochemical data with whole genome scans may be essential for uncovering genes. Although our data highlight a small number of candidate loci, including the angiotensin Type I receptor and CYP4A11
, independent studies will be needed for corroboration. We also examined five candidate genes that were highlighted by Young et al. [3
], who proposed that HTN may arise from the interaction of salt-availability in humans populations with heat-adapted alleles that vary widely in frequency across populations. None of the five genes produced an admixture signal, suggesting that the underlying alleles do not explain a substantial amount of differential HTN risk across these populations.
A potential pitfall for our analysis is that we combined samples from two different studies, each with a different definition of HTN. The GCI study is based on physician-diagnosis, while the MEC, with its questionnaire-based data collection, obtains most information from patient self-report. To increase comparability across the two studies, we restricted our analysis of the MEC samples to individuals who reported using HTN-specific medications, guaranteeing that both the GCI and MEC samples were physician-treated individuals with HTN. We also performed additional analyses with non-diabetics only (since the percentage of participants with diabetes differed significantly in the two studies). We recognize that the difference in the phenotypes may reduce power for some analyses, as genetic determinants of HTN may not have the same effect in the two populations. In general, in our admixture mapping studies—not only for HTN, but also for prostate cancer [13
] and multiple sclerosis [12
]—we have taken an inclusive approach, analyzing as many individuals as possible that fit a loose definition of the phenotype, and following up marginal peaks by exploratory analysis across different subgroups. This was successful in identifying a locus for prostate cancer [13
]; however, since it can also reduce power in some contexts, here we also present analyses of more homogeneous subgroups.
We conclude by noting that other factors may have contributed to our inability to identify HTN genetic variants by admixture mapping. It is possible that the phenotype definition we focused on was not sufficiently strong. Past successes at finding genetic risk factors for HTN have focused on families with extreme, familial forms of HTN, and here we aimed to find common variants affecting more commonly observed HTN in the community and the clinic [21
]. For HTN, which is a classic complex trait, there are also a number of covariates that we did not consider, and that may have contributed to reduced power for detection of genetic determinants. In future admixture mapping and whole genome scans for HTN genes in African Americans, it will be particularly important to study samples that have been assessed not only for presence or absence of HTN, but also for differences in covariates that are known to differ across populations such as plasma renin activity, urinary kallikrein, and dopamine levels [4
]. This may also offer insights into the differences in blood pressure and salt handling known to exist between African and European Americans [15