|Home | About | Journals | Submit | Contact Us | Français|
To better understand the relationship between angiotensinogen (AGT) genetic variation and essential hypertension, AGT genotypes and haplotypes were tested for association with hypertensive endophenotypes and essential hypertension.
256 HyperPATH/SCOR cases and 126 controls were genotyped for 24 SNPs in the AGT gene. SNPs and AGT haplotypes were tested for association with plasma AGT, renal plasma flow, and essential hypertension.
New associations between essential hypertension, plasma AGT, and renal plasma flow (RPF) are reported for alleles −1178G, 6066A, 6152A, 6233C, and 12822C. The maximum odds ratio for association of hypertension and AGT genetic variation was 2.3 (95% CI 1.5 – 3.8; p < 0.0003) for allele 6233C. Previous associations for −1074T, −532T, −217A, −6A, and 4072C are confirmed (p < 0.05). Sodium depletion enhances associations between AGT SNPs and plasma AGT. Most individually associated SNPs, including −6A and 4072C, are found on a common complete AGT haplotype, H4 (frequency = 0.09). Individuals with haplotype H4 have significantly higher plasma AGT and reduced renal plasma flow (p < 0.003 and p < 0.0002, respectively). Other common haplotypes are not associated with plasma AGT levels in this data set despite the presence of the −6A and 4072C alleles, suggesting that AGT haplotype H4 is more predictive of elevated plasma AGT than is −6A or 4072C.
This study demonstrates the importance of analyzing haplotypes in addition to single genotypes in association studies. By demonstrating the dependence of AGT associations on sodium depletion status, it helps to explain previous conflicting association results.
Linkage studies, animal models, and pharmacological data show that the Renin-Angiotensin-Aldosterone System (RAAS) plays a key role in the regulation of blood pressure [1–3]. Angiotensinogen (AGT) is initially cleaved by renin to produce angiotensin I, which is rapidly processed by angiotensin converting enzyme (ACE) to produce the powerful octapeptide vasopressor, angiotensin II. Blood pressure and volume are normally regulated through changes in circulating renin concentrations . Plasma AGT concentrations, however, are lower than the Km for renin, implying that increases in plasma AGT can increase plasma angiotensin under physiological conditions . Factors, including genetic variation, that elevate plasma AGT may contribute to elevated blood pressure, while blood pressure can be reduced and renal blood flow increased by blocking components of the RAAS pathway with ACE inhibitors and angiotensin II (type-1) receptor blockers [6–8].
Linkage between essential hypertension and the AGT locus was initially demonstrated in European sib-pairs, and affected individuals homozygous for amino acid 235T had, on average, ~20% higher plasma AGT than did 235M homozygotes . In transgenic mice, adding up to four additional copies of the AGT gene can increase plasma AGT levels and blood pressure by ~8 mmHg per copy . A meta-analysis of more than 45,000 individuals has provided summary evidence that the AGT locus (marker T235M) is associated with plasma AGT levels in Europeans and with hypertension in Europeans and Asians in a dose-dependent manner . Although many studies show a critical role for AGT in essential hypertension, linkage to this gene has not been replicated in some studies [10, 11]. The −6A promoter allele is in linkage disequilibrium with 235T, and the presence of the −6A leads to increased transcription . Other alleles in the AGT promoter, including −20A, −217A, −532T and −793A, also affect plasma AGT levels and blood pressure [12–19], but these effects can vary by ethnicity and sex [17, 19, 20].
To provide further resolution of the contribution of AGT genetic variation to hypertension, we have performed high-density genotyping and created haplotypes across the AGT promoter and gene region in 256 cases and 126 controls from the HyperPATH / Specialized Center of Research (SCOR) study group. New associations between AGT haplotypes and plasma AGT levels, renal plasma flow, and hypertension provide insight into the complex contribution of angiotensinogen to blood pressure regulation. Our results show that haplotypes can provide more robust associations with angiotensinogen levels and hypertension than individual SNPs (e.g., A-6G and T235M), and we show that these associations are strongly influenced by physiological sodium status. These findings help to explain previous inconsistencies in linkage and association results between AGT and essential hypertension.
Hypertensive patients and controls were recruited and phenotyped using inpatient procedures at the General Clinical Research Centers at Brigham and Women's Hospital (Boston, MA), University of Utah (Salt Lake City, UT), Institut National de la Santé et de la Recherche Médicale (Paris, France), Vanderbilt University (Nashville, TN), and Sapienza-Università di Roma (Rome, Italy). All individuals were of European or European-American ancestry and were participants in the HyperPATH / Specialized Center of Research (SCOR) collaborative studies. Recruitment details have been previously described [21–23]. Individuals with evidence of secondary hypertension, diabetes, ethanol consumption exceeding 250g per week, or renal insufficiency (serum creatinine > 1.9 mg/dL) were excluded from the analyses. All study subjects had BMI values < 40 kg/m2. Hypertensives were withdrawn from medications over a 2-week period prior to phenotype evaluation. The final study set was created from the multicenter samples by selecting unrelated cases and controls (from independent pedigrees) for which DNA samples were also available for genotyping (Boston (145), Italy (24), Paris (47), Utah (136), Vanderbilt (30). All patient recruitment and protocols were performed with Institutional Review Board approval, and all participants provided informed written consent.
Genomic regions containing each polymorphic variant were amplified by PCR (20 mM Tris pH 8.8, 10 mM KCl, 10 mM (NH4)2SO4, 2 mM MgSO4, 50 μM each dNTP, 1 μM each primer, and 1 U Taq DNA polymerase). Primers for PCR were selected using the Primer3 software package. PCR amplicons for multiplex SNaPshot reactions were combined in roughly equal molar amounts. Primers and unincorporated dNTPs were removed from the PCR reactions by adding shrimp alkaline phosphatase (SAP) (2U) and ExoI nuclease (1U) to the pooled amplicons and incubating 90 minutes at 37°C followed by heat inactivation at 75°C for 15 minutes. A detection primer adjacent to each SNP was annealed and extended by 1 base with fluorescently labeled ddNTPs using SNaPshot reagents and conditions recommended by the manufacturer (Applied Biosystems (ABI)). Multiple SNPs were assayed in a single reaction by using detection primers of varying length (20 – 62 bp with 8 bp separation). SNaPshot extension reactions were dephosphorylated with 1U of SAP. Extension products were assayed by capillary electrophoresis on an ABI 3100 genetic analyzer. Genotypes were called using the Genescan and Genotyper software packages (ABI). All SNPs conformed to Hardy-Weinberg expectations in the study populations.
Phenotypic data were obtained through the HyperPATH / SCOR collaborative research group. Variables, including blood pressure, body mass index (BMI), plasma AGT, renal plasma flow (RPF), and infusion of angiotensin II, have been previously described [21, 22, 24, 25]. Briefly, blood pressure was measured by a standard auscultatory method using a mercury sphygmomanometer with patients in the seated position. Hypertensive status was assigned for patients with blood pressure exceeding 140 mmHg systolic and 90 mmHg diastolic on three separate visits. The average blood pressure from the three measurements was used for quantitative analyses. Plasma AGT was measured by radioimmunoassay of angiotensin I after complete cleavage by renin using standard methods . Angiotensin II was administered by intravenous injection at 3ng/kg/min for 45 minutes. Following a loading dose, steady-state plasma PAH was 2 mg/dl prior to experimental protocol. PAH clearance, measured by spectrophotometry, was then used to calculate baseline and angiotensin-stimulated effective renal plasma flow under low-salt (10 mEq) and high-salt (200 mEq) daily dietary conditions . Average trait values for the study population are shown in Table 1.
Associations between single SNPs and hypertension as a categorical variable were tested by chi-square analysis. We calculated the power to detect associations between variation at the AGT locus and hypertension as a discrete trait in our case-control data set . Assuming a prevalence of hypertension of 30% in the general population, relative risk for the risk allele of 1.5, and a risk allele frequency of 0.1, this data set achieves 78% power at α = 0.05 for an additive genetic model. AGT haplotypes were tested for association with hypertension using a randomization test (10,000 permutations).
Each of the 24 SNPs was tested for association with plasma AGT and renal plasma flow under high- and low-sodium conditions. Because standard association and regression analyses are sensitive to phenotypic outliers, we identified and removed outlier values that exceeded the mean +/− three times the inter-quartile range in the raw data set for each quantitative trait. Genotype-phenotype associations for quantitative variables were tested using linear regression as implemented in the PLINK software package . Specifically, we tested whether the response variable (e.g., plasma AGT, renal plasma flow) was correlated with AGT genotype. All single-SNP regression analyses were performed with age, sex, BMI, and sampling location as covariates. Plasma AGT and renal plasma flow were not adjusted by blood pressure status because such an adjustment could reduce or eliminate any potential association between AGT genotype and the hypertensive phenotypes. Using the SNPs individually for the regression analyses maximized the number of samples available for testing. Significant p-values were obtained for the additive and dominant models but, as the quantitative trait means suggested an allele-dosage effect on plasma AGT, p-values are reported for the additive model. We report p-values at the 0.05 level. Due to linkage disequilibrium in the AGT region, test results for individual SNPs are correlated, making the Bonferroni correction for multiple comparisons very conservative. Nonetheless, we also indicate which associations are significant at the Bonferroni-corrected level of 0.002.
AGT haplotypes across the entire ~14 kb AGT region were constructed from genotype data using the FAST_PHASE program . Six major haplotypes, labeled H1 through H6, were inferred. Haplotype frequencies were evaluated for stratification by sampling location using a permutation test. Only haplotype H5, in one population, showed a significant difference due to location (frequency of 0.100 in Utah vs. 0.044 in all other locations, p = 0.003). ANOVA was used to test for significant differences in average plasma AGT among haplotype groups. Assuming equal group weighting, this data set has 80%, 50%, or 25% power to detect an effect size of ~25%, ~18%, or ~13%, respectively, at α = 0.025. A linear regression analysis was performed in which the dependent variable was plasma AGT and the independent variables were haplotype groups H4 and H1/H2 and sex. Genetic distances between haplotypes were estimated, and a neighbor-joining network was constructed using the PHYLIP package .
Twenty-four SNPs spanning the AGT locus and selected for allele frequencies of >5% in the general population were genotyped and tested for association with hypertension. Eleven of the 24 SNP alleles, including −6A and 4072C (aa. 235T), showed positive associations with hypertensive status (p < 0.05) (Figure 1). Four SNP alleles (1164A, 6066A, 6233C, and 12822C) had significance levels < 0.01, and alleles 1164A and 6233C remained significantly associated with hypertension after a Bonferroni correction for 24 comparisons. Allele 6233C (rs3789669), which was found in 21.8% of hypertensives and 10.6% of normotensive controls, produced the maximum odds ratio (2.3, 95% CI 1.5 – 3.8, p < 0.0003).
We tested the hypothesis that AGT genotype is correlated with plasma AGT levels using the 113 subjects for whom plasma AGT values were available (Paris (46), Boston (36) and Utah (31)). Under low-salt conditions (10 mEq), 16 of 24 SNPs are associated with elevated plasma AGT after controlling for age, sex, BMI, and sampling location as covariates in a linear regression analysis (Table 2). Notably, all but one of the SNPs that showed significant association with the trait of essential hypertension were also significantly associated with elevated plasma AGT under physiological conditions of low-sodium. All 16 SNPs remained associated with elevated low-salt plasma AGT using the Paris samples alone, and 7 SNPs (including 4 promoter SNPs) were associated in the Utah sample alone (p < 0.05), demonstrating that the observed associations are not due to stratification among collection sites. As plasma AGT levels increased, there was a modest, but non-significant, increase in low-salt diastolic and systolic blood pressure (r = 0.08, p > 0.4; r = 0.09, p > 0.3, respectively).
Under high-sodium conditions (200 mEq), only 7 of the 16 SNPs remained associated with plasma AGT (p < 0.05), none of the 16 were associated at p < 0.01, and most associations were much less significant than for the low-sodium conditions (Figure 2). This result demonstrates that the association between AGT SNP variation and elevated plasma AGT is enhanced under physiological conditions of sodium depletion.
56 unique haplotypes were inferred in the entire case-control set, including 33 singletons and 17 low-frequency (< 0.05) haplotypes. Six major haplotypes each exceeded a population frequency of 0.05 and account for 84% of all haplotypes in the study population (Table 3). Among the six major haplotypes, the configuration of haplotype H4 is unusual: it contains 15 of the 16 SNP alleles associated with elevated plasma AGT and, uniquely among the common haplotypes, contains 4 promoter alleles associated with elevated plasma AGT. Haplotype H4 contains the −6A promoter allele, and it is uniquely identified by SNP alleles −1178G, −1074T, −532T, −217A, and 6152A. These five markers are in high LD over all samples (r2 ≥ 0.85), and each effectively tags the H4 haplotype. However, r2 values between H4 tagging SNPs and the A-6G polymorphism are low (r2<0.16) (see Figure 2). In a comparison of the four major haplotypes that contain the −6A promoter allele, 78% of those who have the H4 haplotype are hypertensive, while only 69% of those who have any of the other three haplotypes are hypertensive. Over all cases and controls, H4 is the only haplotype positively associated with hypertension at a level that approaches statistical significance (p = 0.053).
For the 113 individuals in whom plasma AGT was measured, we divided the sample into those that carry H4 and those without H4. Under low-sodium conditions, the mean plasma AGT level for the H4 group (1304 ng/ml, n = 26) is 18% higher than for the group containing all other haplotypes (1101 ng/ml, n = 87; p < 0.002) and 15% higher than the group containing one or more copies of non-H4 −6A-bearing haplotypes (1136 ng/ml, n = 42; p <0.036). Plasma AGT was also significantly higher in the H4 group for the Paris samples alone (p < 0.003). Under low-salt conditions, plasma renin activity was higher in the H4 group than in non-H4 group (2.54 vs. 2.32 ng/mL/hr), but this difference was not statistically significant. Under high-salt conditions and consistent with the single-marker results, the mean value for plasma AGT in the H4-bearing group is nearly identical to that of the non-H4 bearing group (1160 ng/ml vs. 1162 ng/ml).
The association results suggest three primary AGT haplotype classes. H4 is associated with hypertension and elevated plasma AGT, H1 and H2 are protective, and others haplotypes have minimal effects. To test this hypothesis formally, we estimated the additive effects of H4 and H1/H2 haplotype copies on plasma AGT values. In the best-model linear regression, with sex as a significant covariate (p < 0.027), low-salt plasma AGT values were significantly predicted by these haplotype classes. H4 was positively associated with plasma AGT (p < 0.013), while H1/H2 was negatively associated (p < 0.025; overall p < 0.001, adjusted R2 = 0.14).
Because previous studies have demonstrated an association of the promoter variant rs5051 (-6A) with a renal vascular response , we tested whether this association could be further refined with additional AGT SNPs and haplotypes. Association statistics were calculated for each of the 24 SNPs and renal plasma flow (RPF) under high- and low-sodium conditions, using sex, age, and BMI as covariates in those samples measured for RPF. Study location was also added to the regression to control for possible between-location differences. Two patterns are apparent: promoter alleles −1178G, −1074T, −532T, −217A and several intronic SNP alleles are associated with reduced renal plasma flow. The associations are present under low-salt conditions and are less significant under high-salt conditions unless the patient was infused with angiotensin II (Table 4). SNP alleles −1178G, −1074T, −217A, and 6152A remained associated with RPF when hypertensive and normotensive groups were analyzed separately (p < 0.05). However, the difference between baseline RPF and angiotensin II stimulated RPF was not significantly associated with AGT genetic variation under either high- or low-salt conditions. These findings suggest that associations between AGT genotype and RPF are increased by conditions that systemically activate the RAAS.
The AGT alleles showing significant associations with reduced RPF are found on haplotype H4, and the associated promoter alleles occur uniquely on haplotype H4. An approximate 11% reduction in high-salt RPF following angiotensinogen II infusion occurs in heterozygotes for those SNPs showing an association (Table 5). Reduced RPF for associated alleles was also observed in hypertensives or normotensives separately. As expected, there is a significant negative correlation between plasma AGT and low-sodium RFP in the subset of samples having both measurements (r = −0.38, p < 0.02; n = 37).
To refine the relationship between AGT and RPF, we again split the data set into those samples with H4 (n = 44) and all others (n = 201). The average high-salt RPF after angiotensin II infusion is significantly lower in the H4 group (388.9 pmol/L vs. 440.6 pmol/L; p < 0.00017). To further evaluate this result, all 24 SNPs were retested for association in individuals lacking the H4 haplotype. The removal of the 44 H4 samples completely eliminates significant associations between AGT SNP genotypes and RPF. These results suggest that haplotype H4 is the major contributor to the association between AGT genotype and renal plasma flow in this study cohort.
We tested each of the 24 AGT SNPs for association with diastolic and systolic blood pressure. No significant associations were found after correcting for the effects of age, sex, and BMI. Thus, although significant associations are found between AGT variation and the categorical variable of hypertension, associations between AGT genotypes and the continuous variables of diastolic and systolic blood pressure did not reach significance in this data set (p > 0.05).
Previous assessments of the contribution of the AGT gene to essential hypertension have obtained variable results, depending in part on the geographic origin of the study population. Since haplotype H4 captures much of the association signal for the hypertension-related traits of increased plasma AGT and reduced renal plasma flow in Europeans, we analyzed the distribution of haplotype H4 relative to the A-6G SNP in worldwide populations using previously published AGT sequence data . Consistent with ascertainment on hypertensive status, haplotype H4 occurs at a higher frequency in our European hypertensive study cohort than in randomly selected western Europeans (9% vs. 6%).
The six major haplotypes identified and characterized by 24 SNPs in this study cohort make up at least 74% of all AGT haplotypes in major Eurasian population groups (Figure 3). In contrast, these six haplotypes represent only 17% of all haplotypes in sub-Saharan Africans (n = 74), where AGT haplotype diversity is high . The frequency of H4 in Africans is 7%. Another haplotype, nearly identical to H4 but having cytosine at position −532, has a frequency of 6% in Africans, making H4 and this nearly identical haplotype frequent in Africans relative to other haplotypes.
Haplotype H4 harbors the −6A and 4072C alleles often used in association studies. Due to the relatively low frequency of H4 and the high frequency of the −6A promoter allele, haplotype H4 captures only a portion of the −6A alleles in a given population. Worldwide, the frequency of haplotype H4 is not well-correlated with the population frequency of the −6A promoter variant (r = 0.36, p > 0.25) (Table 6).
Many previous studies have provided evidence that common SNPs at the AGT locus influence transcription rates, plasma AGT levels, or blood pressure [9, 12, 13, 18–20, 31–34]. Variants −1074T, −532T, −217A, −20C, and −6A within the promoter have been implicated in essential hypertension, individually or in combination, and replicated in a number of recent functional assays or association studies [13, 15–17]. Consistent with these studies, our analysis detects significant associations between these AGT alleles and increased plasma AGT levels and essential hypertension.
Our analysis identifies a complete AGT haplotype, H4, which contains the SNP alleles individually associated with essential hypertension and elevated plasma AGT. This is the fourth most common haplotype in the analysis cohort and has a frequency of 6–13% in European populations. Individuals with H4, primarily heterozygotes (e.g., H4/H1), account for a large proportion of the association signal in the data even though most copies of the −6A and 4072C are found on other haplotypes. Of the common Eurasian AGT haplotypes, H4 uniquely contains promoter alleles −1178G, −1074T, −532T and −217A. Consistent with these results, Brand-Herrmann et al. showed that the −532T/-6A AGT haplotype is associated with elevated blood pressure (systolic, diastolic, ambulatory) in untreated hypertensive subjects . In addition, Jain et al.  have recently shown that a human promoter haplotype containing −217A confers an allele-dose-specific increase in AGT transcription in liver and kidney tissues compared to the −217G-containing promoter haplotype in transgenic mice.
A key factor affecting the strength of the association between AGT genotype and plasma AGT levels was sodium intake. Associations were more numerous and more significant under physiological conditions of sodium depletion. It is well-established that sodium loading has an inhibitory effect on the RAAS system and renin expression in humans and model organisms [36–39]. Our results suggest that differences in plasma AGT levels associated with AGT SNP variation are detectable under low-sodium conditions, but that RAAS down-regulation under sodium loading reduces these allelic differences and associations. Thus, these results suggest that future AGT-blood pressure association studies should account for physiological sodium status.
Several previous studies have found an influence of AGT genetic variation on renal traits. Individuals homozygous for the −6A allele are over-represented among hypertensive patients with low renal plasma flow on high-salt diets . Younger hypertensive 235T homozygotes are characterized by a prematurely blunted renal vascular response to increased salt intake . The non-modulator hypertensive phenotype, characterized by blunted renal response to angiotensin II infusion under a high-sodium diet, is correlated with genotype 235TT, but this effect is enhanced by adding other correlated genotypes from ACE and CYP11B2 [41, 42].
In our study, baseline associations AGT variation and renal plasma flow (RPF) were stronger under low-salt than high-salt conditions. More significant associations, however, occurred following angiotensin II infusion under high-salt conditions. While several interpretations are possible, we suggest that allelic variants found on the H4 haplotype background lead to increased responsiveness of AGT expression, and that H4 identifies a subset of the −6A–bearing individuals with an enhanced angiotensin II response and lower RPF. Conditions that activate RAAS systemically, such as sodium depletion and angiotensin II infusion, may have a greater effect on AGT H4 than on other haplotypes, thus producing detectable differences among haplotype classes. This model permits H4 to maintain higher angiotensinogen levels under RAAS-activating conditions, possibly leading to a higher set point for pressure natriuresis and blood pressure over time.
Increasing molecular evidence now suggests that several common diseases involving metabolism, sodium homeostasis, and hypertension may be the result of evolutionarily adapted alleles that are deleterious in modern environments [3, 43, 44]. The ancestral AGT −6A allele frequency decreases with latitude and, along with −217A, may confer positive heat adaptation . The derived −6G allele appears to be under positive selection in some non-African populations . Additionally, the frequency of the AGT 4072C allele (aa. 235T) is correlated with the unlinked CYP3A5*3 allele, which is also associated with salt and water retention . At these loci, alleles favoring salt retention, vasoconstriction, and a predisposition to hypertension are more common in African populations .
A comparison of the common Eurasian AGT haplotypes in this study shows that H4 (and H5) are more divergent from the ancestral form of AGT than are the H3 and H6 haplotypes, even though the latter also contain the ancestral −6A promoter allele (Figure 4). AGT haplotype H4 contains several human-specific alleles associated with elevated AGT levels (-1178G, −217A, 1164A, 6066A, 9597C). As humans underwent range expansion into more arid environments within Africa, specific mutations and recombination events may have produced AGT haplotype configurations (i.e., H4) that favored vasoconstriction, sodium retention, and limited volume depletion via RAAS activation and the effect of this activation in the kidney.
Haplotype H4, along with a closely related haplotype that differs from H4 at position −532T only, accounts for 13% of our sample of sub-Saharan haplotypes. These two haplotypes are common in Africans, relative to other haplotypes. An independent assessment of 57 Nigerians found that one of seven common AGT haplotypes was significantly associated with elevated plasma AGT levels . For nine shared markers located between −1074 and 4072, all alleles reported for this Nigerian haplotype are identical to haplotype H4. Consistent with the potential adaptive features described above, the frequency of haplotype H4 is high (23%) in tropical South Indians.
We suggest that haplotype H4 may be a predisposing factor for essential hypertension via elevated AGT levels. In this study, activation of the RAAS system, either through sodium depletion or angiotensin II infusion, produced stronger associations between haplotype variation and hypertension-associated phenotypes. Replication of these results in larger cohorts is feasible and necessary. In particular, the relatively low frequency of the H4 haplotype limited the present study to those effects seen in heterozygotes. Larger studies can assess whether haplotype H4 has significant additive effects in homozygotes. Additional analyses of the molecular mechanisms influencing transcriptional regulation of the human AGT gene by SNP and haplotype class are also needed. The distributions of hypertensive trait values for H4 and non-H4 groups overlap extensively, so caution is warranted in using this haplotype to predict plasma AGT, RPF, or hypertensive predisposition of a single individual. Nevertheless, identification of hypertensive patients with AGT haplotypes that increase plasma AGT levels could help to optimize pharmacological approaches to hypertension management.
We thank the many patients that participated in the HyperPATH / SCOR study and the collaborators and personnel who produced and provided the phenotype data used in the analyses. We thank David Witherspoon and Jackie Ying for helpful discussions. We thank Teena Varvil for sample management. We also thank Mark Leppert and Nancy Brown for their contributions to the HyperPATH / SCOR collection.
Sources of Funding This work was supported by National Institutes of Health grants HL070048, HL55000, and HL086907.
Conflict of Interests The authors have no conflicts of interest to declare.
Supplemental files Genotype and phenotype data are provided in Supplementary File 1.