|Home | About | Journals | Submit | Contact Us | Français|
The angiotensinogen gene locus has been associated with essential hypertension in most populations analyzed to date. Increased plasma angiotensinogen levels have been proposed as an underlying cause of essential hypertension in whites; however, differences in the genetic regulation of plasma angiotensinogen levels have also been reported for other populations. The aim of this study was to analyze the relationship between angiotensinogen gene polymorphisms and haplotypes with plasma angiotensinogen levels and the risk of essential hypertension in the Mexican population. We genotyped 9 angiotensinogen gene polymorphisms in 706 individuals. Four polymorphisms, A-6, C4072, C6309, and G12775, were associated with increased risk, and the strongest association was found for the C6309 allele (χ2 = 23.9; P = 0.0000009), which resulted in an odds ratio of 3.0 (95% CI: 1.8–4.9; P = 0.000006) in the recessive model. Two polymorphisms, A-20C (P = 0.003) and C3389T (P = 0.0001), were associated with increased plasma angiotensinogen levels but did not show association with essential hypertension. The haplotypes H1 (χ2 = 8.1; P = 0.004) and H5 (χ2 = 5.1; P = 0.02) were associated with essential hypertension. Using phylogenetic analysis, we found that haplotypes 1 and 5 are the human ancestral haplotypes. Our results suggest that the positive association between angiotensinogen gene polymorphisms and haplotypes with essential hypertension is not simply explained by an increase in plasma angiotensinogen concentration. Complex interactions between risk alleles suggest that these haplotypes act as “superalleles.”
Essential hypertension (EH) is a complex disease that results from an interaction between genes and environmental factors. It has been stated that 40% of hypertension is attributable to genetics.1 The genes involved in the renin-angiotensin system have been suggested as candidate genes for EH because they play an important role in the regulation of blood pressure (BP).2 Genetic linkage studies have demonstrated that there is a relationship between the angiotensinogen gene (AGT) locus and EH in white, African-Caribbean, Asian, and Mexican-American populations.3–6 However, association studies were unable to replicate the influence of these genetic polymorphisms in the risk to EH across other populations. For example, the polymorphisms C4072T (T235M) has been associated with EH in white populations,3,7–9 but it has been reported as borderline in a Japanese population10 and was not associated in Arabian and African-derived populations.4,11,12 Increased plasma levels of angiotensinogen (AGT) produced by increased levels of A-6G–induced transcription, which is in almost complete linkage disequilibrium with C4072T (T235M), has been proposed as a causal mechanism.13 Nevertheless, the C4072T (T235M) polymorphism was not significantly associated with plasma AGT levels or with EH risk in either African-Caribbean or Mexican-American families.4,6 Moreover, an analysis of the AGT polymorphisms that affect plasma AGT levels in a Japanese population showed that C4072T (T235M) did not alter the trait.14 These facts suggest that population genetic diversity plays an important role in the control of intermediate traits, such as plasma AGT levels, and the risk for developing EH. In the case of the AGT, differences in the linkage disequilibrium pattern, number, and frequencies of haplotype blocks among populations with distinct ancestry have been described, further supporting the idea that population differences influence the association results.15,16
Mexican Mestizos, like other Latino populations, are a recently admixed population conformed by Amerindian, European, and, to a lesser extent, African ancestries. Previously, we evaluated genetic diversity, linkage disequilibrium patterns, and extent of haplotype sharing using genome-wide data from Mexicans Mestizos. Ancestry was evaluated by comparing 1 Mexican Amerindian group and data from the HapMap.17
The aim of this study was to analyze the effect of 9 polymorphisms across AGT on plasma AGT levels, and the risk of EH in an urban sample of the Mexican Mestizo population. To address this issue, we performed an association study and haplotype analysis. We decided to sample individuals >65 years of age as controls (hypercontrols) to limit false-negative subjects in the analysis.
Patients were recruited from the outpatient clinic of the Central North Hospital and Central Military Hospital in Mexico City from February 2008 through July 2010. The patients mainly resided in the Mexico City metropolitan area, and they were self-defined as Mexican Mestizo. The diagnosis of EH was established in patients who lacked a secondary cause of hypertension after a complete clinical and biochemical examination. The hypertensive subjects included in this study met the following criteria, ≥30 years old, have a previous diagnosis of hypertension by a physician, and use a prescription antihypertensive medication and/or have a systolic BP ≥145 mm Hg and/or a mean diastolic BP ≥95 mm Hg during interview. For further information on exclusion criteria, BP measurement methods, and power statistics, see the online-only Data Supplement.
Genomic DNA was obtained from peripheral blood using the Qiagen Maxi-kit. We genotyped the following 9 single nucleotide polymorphisms (SNPs) at the AGT locus using 5′ nuclease TaqMan assays (ABI Prism 7900HT Fast RT-PCR System; Applied Biosystems, Foster City, CA): C-532T (rs5046), G-217A (rs5049), A-20C (rs5050), A-6G (rs5051), C3389T-T174M (rs4762), C4072T-T235M (rs699), C6309T (rs2493132), C11535A (rs7079), and G12775A (rs943l580). These SNPs are distributed over 13 kb across the AGT locus. All of the samples included had genotype call rates >95% per SNP.
Plasma AGT levels were quantified using a sandwich ELISA method as described previously.18
For single-marker analysis, the allele frequencies of the 9 SNPs were estimated using direct allele counting. The Hardy-Weinberg equilibrium was calculated for each SNP using χ2 with 1 degree of freedom. Comparisons between cases and controls for each SNP were tested using a χ2 test with the Pearson P value under the allelic and dominant models. Odds ratios (ORs) and 95% CIs were calculated to estimate the relative hypertension risk associated with the AGT polymorphisms.
Single SNP association analysis was performed via a pairwise analysis between genotypes and plasma AGT levels and reported as the mean and SD; we used the AA genotype as a reference group and the other genotypes, AB and BB, for comparison. We used the Student t test to identify differences between genotypes, and a 2-tailed P value <0.005 was considered to be statistically significant after adjusting for 9 comparisons (Bonferroni correction). Potential associations between the set of haplotypes and EH status were examined by using an omnibus likelihood ratio test with permutation-based hypothesis testing procedure as implemented in HAPLOVIEW software.19
The haplotype inference and its effects on plasma AGT levels were evaluated using the maximum-likelihood model with the Stochastic-EM algorithm as implemented in the THESIAS software.20 Also, Bonferroni correction was used to control for multiple testing. The genetic distance between haplotypes was estimated by the Kimura method, and a neighbor-joining network was constructed using the PHYLIP 3.6 software.21
The power to detect association between SNPs and EH was estimated using the Genetic Power Calculator.22 The case-control study with 706 subjects reached 85% and 89% statistical power at = 0.05 for the allelic and dominant models, respectively. The statistical power calculation for trait association was established based on the differences in means of plasma AGT levels between genotypes, using Stata 10.1 (Stata Corp). This data set achieves 99% and 86% power at 2-tailed 0.05 for C3389T (T174M) and T-20G, respectively. For further information on statistical analysis detail, see the online-only Data Supplement.
The demographic information and the distribution of characteristics related to obesity and hypertension in the sample population are summarized in Table S1, available in the online-only Data Supplement. There was a difference in age between cases and controls, because the controls were deliberately selected to include subjects >65 years of age, with the hypothesis that these healthy controls have minimal genetic risk for developing EH. Sex proportions were 0.66 females and 0.34 males for cases and 0.47 females and 0.53 males for controls. Genotype distribution analysis showed no significant differences by sex (data not shown). This is consistent with the logistic regression analysis using sex as a covariate, where risk to EH by associated alleles was not affected. We analyzed a potential population stratification using 8 ancestry informative markers to distinguish between Amerindian and European ancestral components and used the principal component analysis implement in Eigensoft software to detect significant deviation in cases and controls (Figure S1, available in the online-only Data Supplement). Our results showed Fst (measure of genetic structure) between cases and controls at 0.006, with SE = 0.003. This led us to conclude that there is no evidence of stratification between cases and controls in the population studied.
An association analysis was performed with 502 cases and 204 hypercontrols. No deviation from Hardy-Weinberg equilibrium in the 9 SNPs was observed. Four AGT alleles, A-6, C4072, C6309, and G12775, were found to be associated with increased EH risk (Table S2). The strongest association was found for the C6309 allele (χ2 = 23.9; P = 0.0000009). This allele produced the maximum OR in the recessive model (OR: 3.0; 95% CI: 1.8–4.9; P = 0.000006) with allele positivity (OR: 2.1; 95% CI: 1.4 –3.2; P = 0.0008; Table S3). After 10000 permutations, the associations remained significant, at P = 0.0003, P = 0.001, P = 0.002, and P = 0.002 for C6309, A-6, C4072, and G12775, respectively. In addition, the logistic regression analysis showed that the alleles A-6, C4072, C6309, and G12775 were independent risk factors for EH after adjustment for age, sex, body mass index (BMI), and plasma AGT (Figure S2).
We tested the potential association between SNPs and plasma AGT levels in 506 individuals included in the sample (386 cases and 120 controls) and identified 2 SNPs, A-20C and C3389T, associated with this trait (Table S4). The genotype association for SNP A-20C was AA = 26.7 ± 8.9 versus AC = 24.1 ± 8.2 µg/mL (P = 0.003) and for AA = 26.7 ± 8.9 versus CC = 22.9 ± 8.1 µg/mL (P = 0.04). For SNP C3389T, CC = 26.8 ± 8.8 versus CT = 22.1 ± 8.0 µg/mL (P = 0.000001) and CC = 26.8 ± 8.8 versus TT = 21.5 ± 6.9 µg/mL (P = 0.0000009). Also, the analysis with 1-way ANOVA between groups shows 2 SNPs significantly associated with plasma AGT levels, A-20C (P = 0.003) and C3389T (P = 0.0001). None of the alleles associated with the risk to EH (A-6, C4072, C6309, and G12775) were associated with the regulation of the plasma AGT levels.
Nine haplotypes accounted for 93% of all potential combinations. Association between haplotypes and plasma AGT levels is shown in Table 1. In the control group, 2 haplotypes showed association with plasma AGT level: haplotype 3 (H3: CGCATCCCG) was associated with decreased plasma AGT level (−4.3 µg/mL [95% CI: −7.6 to −1.0]; P = 0.01), and haplotype 5 (H5: CGCACCCCG) was significantly associated with increased plasma AGT level (7.2 µg/mL [95% CI: 2.9 –11.5]; P = 0.001). In the cases, H3 was also associated with decreased plasma AGT level (−3.6 µg/mL [95% CI: −6.3 to −0.9]; P = 0.008), and haplotype 6 (H6: TAAACCTCG) was significantly associated with increased plasma AGT level (3.2 µg/mL [95% CI: 0.42–5.90]; P = 0.02). This association is reported after full adjustment for covariates, that is, age, sex, and BMI. The total contribution of the haplotypes and covariates on plasma AGT level variance was 14%. A likelihood ratio test for haplotype-phenotype association, adjusted for covariates, that is, age, sex, and BMI, resulted in a χ2 of 36.8 (P = 0.00001). The haplotypes H1 and H5 showed association with EH with χ2 = 8.1 (P = 0.004) and χ2 = 5.1 (P = 0.02), respectively. In addition, haplotypes H2 and H4 showed protective effect with χ2 = 5.1 (P = 0.02) and χ2 = 8.0 (P = 0.004), respectively. However, only 2 remained significant after 10000 permutations, H1 (P = 0.04) and H4 (P = 0.04; Table 2).
Age, sex, and 2 polymorphisms, A-20C and C3389T, were the independent variables that affected plasma AGT levels in the Mexican-Mestizo population. However, the plasma AGT level alone is not associated with BP-related traits after adjustment for age, sex, and BMI covariates (Table S5). Furthermore, the plasma AGT level was associated with a marginally increased EH risk (OR: 1.04 [95% CI: 1.01–1.07]; P = 0.01), but this association was decreased after an adjustment for sex, age, BMI, and rs2493132.
A summary of the haplotype effects on quantitative and qualitative traits and the haplotype SNP compositions is shown in Table 3. The H1, with a population frequency of 0.39 in our study, contains the EH risk alleles and increased plasma AGT levels. The H5, which was associated with increased plasma levels and with the EH risk, contains alleles associated with EH risk and C3389 but not the A-20 allele. The H2 and H4, with population frequencies of 0.15 and 0.10, respectively, displayed protective effects and contain alleles associated with increased plasma AGT levels and diastolic BP but do not contain EH risk alleles. The H3, which is associated with decreased plasma AGT levels, contains the EH risk alleles but not those associated with increased plasma AGT levels. Interestingly, H3 itself did not increase the EH risk although it contains hypertensive alleles; however, if the haplotype gained the hypertensive properties of C3389 (H5), the EH risk was increased, suggesting that there is a synergistic effect of alleles. To test the combined effect of associated alleles (C3389 and C6309), we used a logistic regression analysis to assess the effect of combined genotypes in the risk to EH (Tables 4 and S6). Allele risks showed an additive effect on the development of EH. This findings support the notion of a combined effect of individual alleles in the risk for EH.
The unrooted haplotype phylogenetic tree constructed using the neighbor-joining method is shown in the Figure. The haplotype in primates (H0) was considered as a proxy for the ancestral haplotype. Using Kimura’s method21 to infer the distance between haplotypes, we noted that the risk haplotypes H5 and H1 have a shorter distance from H0 (H0 to H5 = 0.12; H0 to H1 = 0.27) and may be considered as human ancestral haplotypes. H2 and H4 are more recent haplotypes in the evolutionary history of the AGT gene because they show the longest distances from H0 (H0 to H2 = 1.68 and H0 to H4 = 1.14). Figure S3 shows the phylogenetic tree using the Amerindian population (Zapotecs). We observed a decreased in haplotype diversity in this population; however, the risk haplotypes H5 and H1 were also considered as human ancestral in this population, with similar distances as the haplotypes from the Mestizo population (H0 to H5 = 0.13 and H0 to H1 = 0.31).
The AGT gene is the only one for which association with EH has been consistently replicated in multiple populations.13 Several polymorphisms across the AGT gene have been associated with both EH and plasma AGT levels. Increased plasma AGT levels have been postulated as a causal factor for increased risk to EH. However, the lack of replication across populations has prevented the translation of this association to clinical applications. Differences in the genomic structure between human populations, including that of the Mexican population,18 have been described in the last decade, and these differences might be the source for this lack of replication. Thus, analysis of the risk-associated alleles in multiple interethnic populations may become an important tool to determine the role of genetic predisposition in the development of the disease.
This study identified 4 alleles associated with EH risk and were found to be independent risk factors in a logistic regression model. We were able to replicate the 2 most important associations, A-6 and C4072 (T235), and we identified 2 new markers associated with higher risk, C6309 and G12775. The association signal at intron 2 of the AGT (C6309) is particularly interesting because another polymorphism was found recently in intron 2 (C6233T) associated with a major effect on the risk to EH in whites.8 An important point in our study is that we did not find any association between these SNPs and plasma AGT levels in the Mexican population, which reproduces the negative association between C4072T (T235M) and plasma AGT reported previously in a Mexican-derived population.6,23 This finding suggests that mechanisms other than an increased level of plasma AGT contribute to EH risk in the Mexican-Mestizo population. Interethnic variability in the genetic control of plasma AGT levels has been reported for Japanese and African-American children. Sato et al14 found no association between C4072T (T235M) and plasma AGT levels, and the SNP G-1074T associated with AGT levels did not increase the EH risk. In another study, Bloem et al24 were also unable to find an association between plasma AGT levels and C4072T (T235M).
In the Mexican-Mestizo population, the A-20C and C3389T (T174M) were found to be associated with plasma AGT levels, and the C3389 (T174) allele showed the strongest association. However, these SNPs were not associated with increased risk for EH. Both SNPs have been associated with EH and plasma AGT levels in other populations. For example, A-20C located in the promoter region of AGT is associated with both in vitro changes in AGT transcription and plasma AGT levels in the Japanese population.25,26 In a meta-analysis including 11079 subjects, an association between C3389T (T174M) and EH was identified in Asian and multiethnic populations but not in a European population.15 The functional mechanism by which C3389T (T174M) increases BP among carriers is currently unknown, and several studies have failed to show association between this SNP and plasma AGT levels.3,27,28 In our sample, levels of plasma AGT by itself did not increase the risk of EH nor showed association with any BP-related traits in our covariate-controlled linear regression analysis. In addition, a recent study failed to associate plasma AGT levels and BP-related traits in a family cohort of white ancestry.29 These results differ from those in an initial report that found a positive correlation between plasma AGT levels and diastolic BP.30 The observations that plasma AGT levels are not associated with any BP-related traits and are not independent risk factors for EH support the notion that, at least in our sample, the increased EH risk from the AGT locus involves mechanisms other than an increase in plasma AGT levels alone.
To summarize these results, individual SNP analysis identified polymorphisms that were associated with EH risk but not were associated with plasma AGT levels; we also identified SNPs associated with plasma AGT levels that were not associated with EH risk. Plasma AGT is not an independent risk factor for EH and does not impact the BP-related trait.
If we consider that each SNP in a genomic region is fixed to another SNP by evolutionary forces, it would be anticipated that this haplotype background could be more informative than individual SNPs alone. Several publications have described the higher informative value of haplotype analysis as compared with individual SNP analysis.8,31–33 Our haplotype analysis found that the H1 and H5 are associated with risk to EH, and the H2 and H4 are associated with protection against EH. The lack of association between haplotypes H2 and H5 with hypertension after a permutation test could be influenced by sample size and modest haplotype frequency differences between cases and controls. For this reason, a replication study with a larger sample size could be useful to strengthen this analysis and contribute to clarifying the effect of these haplotypes on the risk of EH.
The SNP composition of these haplotypes shows that H1 contains the EH risk alleles and plasma AGT levels. The second risk-associated haplotype H5 also contains the EH risk alleles but lacks the A-20 allele in the promoter, similar to one of the ancestral haplotypes. Interestingly, H3, which includes the EH risk alleles but lacks the plasma AGT levels, did not increase the risk for EH. The protective haplotypes, H2 and H4, include plasma AGT level–associated alleles but not those from the EH risk alleles. Our analysis suggests a more complex model than a single polymorphism effect that involves a combination of variants within the AGT gene, which modulate the risk for EH and plasma AGT levels. There are multiple indications that several polymorphisms within a gene position interact to affect quantitative trait variation. Thus, multiple locus interactions create a major locus that has a large effect on the observed phenotype (superallele).34 Quantitative trait mapping in Drosophila melanogaster has shown that major gene effects are not necessarily attributed to single site polymorphisms but are the result of the combined effects of multiple associated polymorphisms.35 This phenomenon has also been described in human traits and diseases; for example, the apolipoprotein gene (APOB) affects plasma low-density lipoprotein and high-density lipoprotein cholesterol,36 the LCT gene influences intestinal lactase activity,37 and ADRB2 influences the actions of catecholamines on bronchodilation and risk to asthma.38 Similarly, the AGT gene has a demonstrated additive effect on the risk to EH, acting as a superallele, when specific SNPs are present.39,40
Genetic distance analysis on these haplotypes showed that H5 and H1 have the shortest distances from the ancestral chimpanzee haplotype. Considering that haplotypes H1 and H5 contain the major alleles and have the shortest distances with respect to the ancestral haplotype H0, we named these as human ancestral. The difference between H1 and H5 haplotypes is the presence of the A-20 allele in H5 and H0. The H2 and H4 have the longest distances from H0, suggesting a recent expansion of haplotypes. The Zapotecs, an ancestral Amerindian population, contain a similar genetic distance pattern for H1 and H5, supporting the ancestral character of these haplotypes and also the important genetic contribution of the Amerindian population on the modern Mexican-Mestizo population. These observations are in agreement with the ancestral-susceptibility model for common diseases in which the ancestral alleles reflect risk in the modern lifestyle, whereas in human ancestors these same ancestral alleles provided adaptive advantages to both a low-salt intake and vegetable-based diets.41
In conclusion, our results show heterogeneity in the effects of AGT polymorphisms on EH risk and plasma AGT levels. Two haplotypes act as superalleles for the risk to EH, containing EH risk alleles and plasma AGT levels. Two SNPs were associated with plasma AGT levels, but no association was identified between plasma AGT variation and risk to EH and BP-related traits. These findings suggest that population genetic diversity plays an important role in the control of intermediate traits and helps to elucidate interethnic variability in plasma AGT levels and the role of the AGT locus in EH in the Mexican-Mestizo population.
Our results contribute to understanding the influence of the AGT locus to the risk to EH in a population with a unique genomic ancestry as the Mexican ancestry. Expression analysis of intron 2 and its interactions with functional SNPs across the AGT may be interesting to determine its role in disease. Furthermore, cohort follow-up of genotyped individuals in the general population would be valuable to understand the effect of haplotypes on BP continuous traits and also the role of haplotypes in the clinical setting.
We thank Trinidad Gil and Sara del Carmen Barrios for excellent clinical coordination.
Sources of Funding
This work was supported by the National Institute of Genomic Medicine (INMEGEN), PEMEX Central North Hospital, SEDENA Central Military Hospital, and the Tulane University Health Sciences Center (grants R01DK072408 from the National Institute of Diabetes and Digestive and Kidney Diseases, and P20RR017659 from the National Center for Research Resources).
The online-only Data Supplement is available with this article at http://hyper.ahajournals.org/lookup/suppl/doi:10.1161/HYPERTENSIONAHA.111.176453/-/DC1.