Genomic Exploration of Essential Hypertension in African‐Brazilian Quilombo Populations: A Comprehensive Approach With Pedigree Analysis and Family‐Based Association Studies

Anonymized summary statistics data have been made publicly available at the GWAS Catalog (study GCST90454187) and can be accessed at https://www.ebi.ac.uk/gwas/studies/GCST90454187. We use the term “African American” to refer to individuals of African ancestry with historical and cultural backgrounds in the Americas, aligning with guidelines (PMID: 34402850) for precise and contextually appropriate terminology.

Significant differences in mean ages were observed between affected and unaffected individuals across all constructed pedigrees (details in Data S6), as shown in Figure 3 (details in Table S1): Abobral: 35.9 versus 49.7 years, P=2.112×10⁻³; André Lopes and Nhunguara populations: 36.4 versus 57 years, P=1.291×10⁻⁷; Galvão and São Pedro populations: 37 versus 54.3 years, P=4.771×10⁻⁷; Pedro Cubas population: 38.4 versus 60.6 years, P=1.026×10⁻⁴; Ivaporunduva population: 33.2 versus 64.3 years, P=4.963×10⁻⁷; and Sapatu population: 38.2 versus 53.8 years, P=8.866×10⁻⁴. Overall, the combined data showed a mean age of 36.73±15.4 years for unaffected individuals versus 55.11±17.3 years for affected individuals (P=6.855×10⁻²⁵). One‐way ANOVA was performed with 5 degrees of freedom for pedigrees and 161 for residuals, and the sum of squares for pedigrees was 2783 and for residuals was 46 617, resulting in mean squares of 556.6 and 289.5, respectively. Comparing the average age of affected individuals across pedigrees for both sexes, this analysis yielded an F value of 1.922 with a P value of 0.093, indicating nonsignificant differences in average age across pedigrees. Two‐way ANOVA highlighted a significant effect of affected status on average age (P<0.001). However, neither sex nor the interaction between sex and affected status showed significant effects on average age, with P values of 0.601 and 0.671, respectively. This suggests that although affected status significantly influences average age and sex, the interaction between sex and affected status does not have a significant impact. The multiple regression model showed that affected status significantly influenced average age (P<0.001). The coefficients for sex (P=0.197), pedigree (P=0.551), body mass index (P=0.328), as well as for SBP (P=0.278) and DBP (P=0.440) are not statistically significant. The overall model's performance is high (strong explanatory power), as indicated by the R² value of 0.761, suggesting that about 76.11% of the variability in average age can be explained by the predictors in the model. Additionally, the F statistic is significant (P=9.109×10⁻⁵), indicating that the model is significant in predicting average age.

We calculated ancestry proportions for individual admixture segments (local ancestry) using data from 145 467 SNPs considering all of the 431 samples (Figure 4A) and individually per pedigree (Figure 4B). The 3 top PCs, PC1, PC2, and PC3, explain 12.42%, 6.17%, and 1.68% of the variance, respectively (Figure 5A), allowing to visualize the genetic distance between the inference and the reference data sets (Figure 5B). All 6 quilombo remnants pedigrees studied have a high degree of admixture. Among all 431 individuals, the estimates are 47.4%, 36.3%, and 16.1%, respectively, for African, European, and Native American ancestries (Table 3). The pedigrees from Abobral and Pedro Cubas had the highest (51.5%) estimates of African ancestral contribution. In comparison, Sapatu had the highest (50.8%) estimate of European ancestral contribution, and Abobral had the highest estimate (16.9%) of Native American ancestral contribution. For comparison, the ancestry proportions for the general Brazilian population were estimated as 19.6%, 68.1%, and 11.6% for African, European, and Native American ancestral contribution,⁴⁹ respectively, with vivid contrasts among each Brazilian region (Tables S2 and S3).

The MCMC runs diagnoses were conducted, within each pedigree, on the smallest chromosome exhibiting a positive LOD score (Figures S2 through S7). Through pedigree analysis, we identified 30 ROIs (Table S4), of which 8 were discarded due to a lack of corroboration between the 3 subpanels of SNPs. The remaining 22 ROIs contain 2363 genes (Table S5) (example in Figure 6A).

We used FBAS (Figure 6B and 6C) to analyze a total of 1 612 754 SNPs, derived from both pedigree‐ and population‐based imputation strategies, across the 22 ROIs. The summary statistics for the extended SNP set are publicly accessible in the GWAS Catalog (GCST90454187). From this analysis, we identified 117 variants (Table S6) with suggestive or significant associations with EH.

Furthermore, our investigation of EH‐related genes, using publicly available databases and supported by the literature, narrowed down the initial set of 2363 genes to 60 promising candidate genes (Table 4) located within the 22 ROIs.

Among the 22 remaining ROIs, 20 of them were single‐supported by the investigation of EH‐related genes as a fine‐mapping strategy, and 17 of them were additionally supported by FBAS as a fine mapping strategy (double supported). Highlighting the common ground between the double‐supported results (by both of the fine‐mapping strategies), 14 genes (highlighted in Table 4) were identified within the mapped regions with compelling evidence of association with the phenotype. These genes include PHGDH and S100A10 (ROI 1); MFN2 (ROI 2); RYR2, EDARADD, and MTR (ROI 3); SERTAD2 (ROI 4); LPP (ROI 5); KCNT1 (ROI 11); TENM4 (ROI 13); P2RX1, ZZEF1, and RPA1 (ROI 18); and ALPK2 (ROI 20). All of these genes are broadly described in the supplementary section (SS7: double‐supported EH genes). These genes harbor 29 SNPs (highlighted in Table S6 implicated by our FBAS.

From our analysis, the estimated genetic correlation between our study and GCST90436066 was 0.427, with a genetic covariance of 0.0105 (P=0.743). Heritability estimates for GCST90436066 showed a total observed scale heritability of 0.0023 (Z=3.15, P=7×10⁻⁴), whereas for our study, scale heritability was 0.2623 (Z=0.271), reflecting a significant difference between the data sets. These results are likely influenced by the fact that GCST90436066 represents a European cohort (there are no available African or admixed EH summary statistics in public repositories such as the GWAS Catalog), in contrast to the admixed nature of our data set, which contains a high proportion of African ancestry. Despite these limitations, the moderate genetic correlation observed provides some level of validation, indicating some level of shared genetic architecture between the 2 cohorts for this trait.

Despite significant financial investment and the substantial number of samples analyzed in genome‐wide linkage analysis and GWAS, much of the heritability of complex phenotypes such as EH remains elusive. Moreover, due to the underrepresentation of African and Native American populations in global genetic studies, much of what is currently known about complex phenotypes may not be fully applicable to admixed populations, such as the descendants of quilombo populations and, more broadly, the Afro‐descendant Brazilians. By studying EH in semi‐isolated quilombo remnants, we reduce biases from population and clinical–biological heterogeneity, improve the signal‐to‐noise ratio (and consequently the statistical power), and increase the representation of admixed populations (like the Brazilian population) in genomic studies.

In our complementary approach, we combined 2 main strategies, linkage analysis and FBAS. Together, these strategies minimize the need for multiple testing, reduce population stratification, and leverage chromosomal location information provided by meiotic events. This combined approach resulted in 22 filtered ROIs with 60 single‐supported genes (harboring 117 suggestive/significant associated variants) by the EH‐genes investigation, and 14 double‐supported genes by both the fine‐mapping (investigation of EH‐genes and FBAS) harboring 29 suggestive/significant associated variants.

The 14 double‐supported genes imply EH through their roles in vascular, cardiac, and metabolic biological pathways. For instance, LPP harbors BP‐associated SNPs,⁹⁸ and plays a role in maintaining vascular smooth muscle cell contractility,⁹⁹ which helps prevent hypertension‐induced arterial remodeling.¹⁰⁰ Similarly, P2RX1, identified in both African ancestry populations and our Brazilian Afro‐descendant cohort, plays a key role in vasoconstriction in small arteries¹⁰¹ and renal autoregulation, both of which are critical for maintaining BP homeostasis.¹⁰² Apparently, this gene controls glomerular hemodynamics in angiotensin II hypertension.¹⁰³ RYR2 harbors several BP and EH‐associated SNPs,^{104, 105, 106} and is key to calcium handling in cardiac muscle,¹⁰¹ directly affecting heart function and BP regulation, whereas MFN2 is involved in mitochondrial function,¹⁰¹ with mutations linked to vascular smooth muscle proliferation, which may contribute to hypertension.^{107, 108, 109} Collectively, these genes are involved in mechanisms affecting vascular tone, muscle contraction, and energy metabolism, all central to BP regulation.

Several genes related to oxidative stress, endothelial dysfunction, and NO regulation, were associated with EH in this study and supported by the literature. RPA1 influences the repression of endothelial nitric oxide synthase,¹¹⁰ thus reducing NO levels and leading to endothelial dysfunction,¹¹¹ a key player in hypertension. EDARADD has been linked to pulse pressure regulation,^{101, 112} whereas S100A10 is involved in cellular processes such as inflammation,^{113, 114, 115} which can contribute to vascular dysfunction and elevated BP. MTR, responsible for methionine biosynthesis,¹⁰¹ has been associated with DBP,¹¹⁶ indicating its possible role in homocysteine metabolism and cardiovascular risk in patients with EH,¹¹⁷ and has been suggested as a genetic marker to assess the action of antihypertensive drugs.¹¹⁸ We have identified a BP‐related SNP in MTR (rs1805087) in linkage disequilibrium (r²≥0.97) with the suggestive SNP rs10925261 (P=2.5×10⁻³) identified in this study. This set of genes collectively suggests that endothelial health, inflammation, and metabolic regulation converge on pathways influencing BP.

Lastly, genes such as ZZEF1,^{119, 120} PHGDH,^{121, 122} and TENM4^{123, 124} were associated with BP regulation via several genetic polymorphisms. PHGDH has been linked to BP through epigenetic regulation, with methylation changes influencing systolic and diastolic pressure.¹²² KCNT1 may influence vascular tone indirectly, although it is mainly involved in potassium channel regulation.^{125, 126} Although SERTAD2¹⁰⁹ and ALPK2¹²⁷ present BP regulation‐associated polymorphisms, they have less direct evidence connecting them to EH, but their involvement in cellular signaling and protein interactions suggest possible contributions to BP regulation.

SNPs linked to the EH phenotype identified in this study were genotyped using genomic arrays and are therefore common variants. Many of these variants are situated in noncoding or intergenic regions and may or may not have recognized functional or regulatory effects. Although these SNPs are not expected to affect the phenotype directly, several of these variants are in LD with variants of more significant impact. Notably, among these, some may be rare and remain ungenotyped. Specifically, we identified unique tag SNPs: 77 noncoding SNPs (3′ untranslated region, 5′ untranslated region, or transcription factor binding), 196 regulatory, and 15 missense SNPs (Table S7).

Recent GWAS studies have highlighted significant findings in African‐derived populations, revealing genetic variants associated with BP traits that differ from those identified in European populations. For example, a large meta‐analysis involving 80 950 individuals of African ancestry identified 10 variants for SBP and 9 for DBP, including a novel variant, rs562545 in the MOBP gene, associated with DBP.¹²⁸ The African Wits‐INDEPTH Genomic Study (AWI‐Gen), which focused on Sub‐Saharan African populations, identified 2 genome‐wide significant signals for BP traits: 1 near P2RY1 for SBP and another near LINC01256 for pulse pressure.¹²⁹ These findings, absent in European ancestry cohorts, demonstrate the critical need for African‐derived specific studies. Similarly, large‐scale GWAS efforts, such as those using the UK Biobank, discovered that only a fraction of loci identified in European populations replicate in African‐derived populations, demonstrating that ancestry‐specific genetic variants play a crucial role in hypertension susceptibility.^{128, 129} African‐derived populations often exhibit unique genetic associations due to differences in genetic architecture. For example, variants in genes such as CACNA1D and KCNK3 that have been found to significantly contribute to BP regulation in African populations are less prominent in other populations.

Admixture mapping has provided valuable insights into loci with potential effects on BP and EH, identifying genomic regions particularly relevant in populations with African ancestry. For instance, NPR3, associated with BP regulation in both African and European populations, was uncovered through admixture mapping in African American individiuals.¹³⁰ Additional regions such as 1q21.2–21.3, 4p15.1, 19q12, and 20p13 have demonstrated significant associations with DBP (P≤2.07×10⁻⁴), whereas regions 1q21.2–21.3 and 19q12 were also associated (P≤5.32×10⁻⁴) with mean arterial pressure.¹³¹

Several regions identified in the literature have been corroborated by findings from our study. The 3q27.3 region (ROI 5), containing the AGT gene, is strongly linked to EH, especially in African ancestry populations.¹³² Likewise, the 17q23.2 region (ROI 19) has been implicated in BP traits, particularly through its role in regulating the renin‐angiotensin‐aldosterone system.¹³³ Furthermore, 6p24.1 (ROI 9) has shown previous associations with BP regulation.¹³² Linkage studies have also implicated 1q43 (ROI 3) in essential hypertension,¹³⁴ whereas 8p23.1 (ROI 10) has been recognized for its relevance in BP regulation, particularly in populations of African descent.¹³³

These findings underscore the value of including diverse populations in genetic studies to uncover novel loci and better understand the genetic basis of complex traits like hypertension. In this study, our approach identified 22 ROIs, with 14 relevant genes, from which we highlight PHGDH, S100A10, and RYR2 implicated in hypertension susceptibility. The overlap between findings in African ancestry populations and those in Brazilian Afro‐descendants, such as P2RX1 highlighted in our study, further supports the role of GWAS and admixture mapping in identifying population‐specific loci relevant to hypertension risk.

To prioritize our results, we developed a comprehensive score based on the weights assigned to each strategy (Table S8), including linkage analysis, EH genes investigation, and association studies. The scores allowed us to classify the ROIs into 3 tiers based on their priority level (Table 5): high (top 20% of the ROIs), intermediate (30% of the ROIs), and low (50% of the ROIs).

As presented, our study has some limitations. One such limitation is that the investigation of EH‐related genes, as part of the fine‐mapping strategy, was focused on genes previously supported by the literature as being related to BP regulation. Consequently, it is possible that genes that have never been implicated in EH may contain rare or novel variants relevant to the origin of hypertension in quilombo remnant populations.

On the other hand, the strength of this study comes from the improvement and application of a unique multilevel computational approach that combined mapping strategies to deal with large family data, which provided reliable results. By using genome‐wide linkage analyses based on MCMC methods and adjusted for admixture, association studies such as the primary fine‐mapping strategy, and limiting analyses to candidate genomic regions, this study took advantage of meiotic information provided by pedigrees while simultaneously reducing the need for multiple tests and avoiding population stratification. Therefore, the ROIs identified in this study are credible and provide valuable insights into the genetic basis of essential hypertension in the quilombo remnant populations.

Conducting analyses by merging all 6 pedigrees into only 1 would be a formidable challenge, and probably not feasible. However, the prospect of replicating these analyses using alternative computational packages is exciting and not an impossible task. Our study has demonstrated that BP and hypertension in the quilombo remnant populations are likely influenced by multiple genes, in a polygenic or oligogenic mechanism of inheritance. We have identified several loci across different chromosomes that contain genes and variants involved in the development of hypertension. Additionally, we have identified genomic ROIs not previously associated with EH and will therefore be important targets for future research.

To overcome the limitations, future steps of this investigation will involve the use of whole‐genome sequencing and whole‐exome sequencing from this data set. Whole‐genome sequencing and whole‐exome sequencing will enable the investigation of coding and noncoding variants within all ROIs, with a focus on rare variants that may have a higher impact on gene functioning. To optimize this process, the prioritization of ROIs as performed in Table 5 is essential, because high and intermediate ROIs will be addressed first when filtering variants detected after whole‐genome sequencing and whole‐exome sequencing, and the low‐priority ROIs afterward.

Furthermore, our study has the potential to shed light on important genomic regions, genes, and variants that are specific to African‐derived populations. We have provided insights into the genetic factors that contribute to hypertension in a group that has been often underrepresented in genetic studies and databases.

The authors acknowledge funding from CEPID‐FAPESP (Research Center on the Human Genome and Stem Cells–São Paulo Research Foundation, grants 1998/14254‐2 and 2013/08028‐1), and FAPESP/INCT‐CNPq (São Paulo Research Foundation/National Institutes of Science and Technology–National Council for Scientific and Technological Development) 2014/50931‐3 led by (Dr Zatz). This research also received support from FAPESP (São Paulo Research Foundation) grant 2012/18010‐0, led by Dr Meyer (Institute of Biosciences, University of Sao Paulo‐Brazil). Additionally, the authors are grateful to CAPES (Brazilian Federal Agency for Support and Evaluation) for the sandwich PhD fellowship number 88887.371219/2019‐00 and CNPq (National Council for Scientific and Technological Development) for the PhD fellowship number 142193/2017‐8. The authors also acknowledge funding from the Marshall University Joan C. Edwards School of Medicine, Marshall University Data Science Core, WV‐INBRE (West Virginia–Idea Networks of Biomedical Research Excellence) grant (National Institutes of Health P20GM103434), Bench‐to‐Bedside Pilot grant (Dr Nato) under the West Virginia Clinical and Translational Science Institute (WV‐CTSI) grant (National Institutes of Health 5U54GM104942), and Dr Nato startup fund.

The authors are thankful for the support of Dr Meyer (Institute of Biosciences, University of Sao Paulo‐Brazil), Dr Pavan Soler (Institute of Mathematics and Statistics, University of Sao Paulo‐Brazil), Dr Ruiz Giolo (Federal University of Parana‐Brazil), Dr Otto (Institute of Biosciences, University of Sao Paulo‐Brazil), and Dr da Costa Pereira (Heart Institute, University of Sao Paulo‐Brazil) for valuable suggestions. The authors also thank Dr Kubisch (University Medical Center Hamburg/Eppendorf‐Germany) for the many ideas that stimulated the development of this project.