Genetic Variants Associated With Uncontrolled Blood Pressure on Thiazide Diuretic/β‐Blocker Combination Therapy in the PEAR (Pharmacogenomic Evaluation of Antihypertensive Responses) and INVEST (International Verapamil‐SR Trandolapril Study) Trials
Background The majority of hypertensive individuals require combination antihypertensive therapy to achieve adequate blood pressure (BP) control. This study aimed to identify genetic variants associated with uncontrolled BP on combination therapy with a thiazide diuretic and a β‐blocker.
Methods and Results A genome‐wide association study of uncontrolled BP on combination therapy was conducted among 314 white participants of the PEAR (Pharmacogenomic Evaluation of Antihypertensive Responses) trial. Multivariable logistic regression analysis was used. Genetic variants meeting a suggestive level of significance (P<1.0E‐05) were tested for replication in an external cohort, INVEST (International Verapamil‐SR Trandolapril study). We also examined genome‐wide variant associations with systolic and diastolic BP response on combination therapy and tested for replication. We discovered a single nucleotide polymorphism, the rs261316 major allele, at chromosome 15 in the gene ALDH1A2 associated with an increased odds of having uncontrolled BP on combination therapy (odds ratio: 2.56, 95% confidence interval, 1.69–3.88, P=8.64E‐06). This single nucleotide polymorphism replicated (odds ratio: 1.86, 95% confidence interval, 1.35–2.57, P=0.001) and approached genome‐wide significance in the meta‐analysis between discovery and replication cohorts (odds ratio: 2.16, 95% confidence interval, 1.63–2.86, P=8.60E‐08). Other genes in the region surrounding rs261316 (ALDH1A2) include AQP9 and LIPC.
Conclusions A single nucleotide polymorphism in the gene ALDH1A2 may be associated with uncontrolled BP following treatment with a thiazide diuretic/β‐blocker combination.
Clinical Trial Registration URL: https://www.clinicaltrials.gov. Unique identifier: NCT00246519.
- combination therapy
- high blood pressure
- thiazide diuretics
What Is New?
We discovered a variant in the gene ALDH1A2 associated with uncontrolled blood pressure on combination therapy with a thiazide diuretic/β‐blocker in the PEAR (Pharmacogenomic Evaluation of Antihypertensive Responses) trial and replicated this variant in an independent cohort, INVEST (International Verapamil‐SR Trandolapril Study).
We discovered a potential biomarker for identifying patients at risk of uncontrolled blood pressure on thiazide diuretic/β‐blocker combination therapy.
What Are the Clinical Implications?
Clinicians may use the information to prescribe more appropriate alternative antihypertensive medications for patients carrying the risk allele; however, further replication of the identified variant in other external cohorts is needed to confirm the finding.
Hypertension is a leading risk factor for cardiovascular morbidity and mortality, affecting 1.4 billion people worldwide1 and 1 of every 3 adults in the United States.2 While many effective antihypertensive medications are available, 75% of hypertensive patients require at least 2 drugs to achieve blood pressure (BP) control, 25% require at least 3 drugs, and a subset, defined as resistant hypertensives, requires at least 4 agents of different drug classes to achieve BP control.3 Given these data, the current US hypertension recommendations for adult BP management include suggestions for initiation of combination therapy in those with stage II hypertension or systolic BP (SBP) ≥20 mm Hg or diastolic BP (DBP) ≥10 mm Hg above goal.4 However, whether all patients with high baseline BP would benefit from starting on combination therapy and on which drug combinations is unclear. Among patients on dual therapy, combination of an angiotensin‐converting enzyme (ACE) inhibitor and calcium‐channel blocker (CCB) was associated with better cardiovascular outcomes than the ACE inhibitor/hydrochlorothiazide or β‐blocker/hydrochlorothiazide combinations in the ACCOMPLISH (Avoiding Cardiovascular Events in Combination Therapy in Patients Living with Systolic Hypertension) and the ASCOT‐BPLA (Anglo‐Scandinavian Cardiac Outcomes Trial‐Blood Pressure Lowering Arm) trials, respectively.5, 6 However, in the INVEST study (International Verapamil‐SR Trandolapril Study), both ACE inhibitor/CCB and β‐blocker/hydrochlorothiazide combinations showed comparable reductions in BP and adverse cardiovascular outcomes.7 Because most hypertensive patients need at least 2‐drug combinations, gaining a better understanding of which patients would most benefit from initiation on different combination therapies would not only be clinically useful but may also lead to the discovery of novel BP regulatory mechanisms and drugs that would work in different subsets of hypertensive patients.
The widespread use of combination therapy also underscores the need for more effective targeted therapies. While adding optimized doses of additional drugs from different drug classes increases the likelihood of BP control in most patients, poor adherence to antihypertensive therapy is an already prevalent issue that increases as drug regimens become more complex.8 Indeed, developing targeted therapies that achieve the desired effect while optimizing as well as minimizing the number of drugs needed will be essential to the successful delivery of precision medicine.
There has been a long‐standing interest in understanding the genetic underpinnings of differential response to common antihypertensive agents.9 A few candidate gene studies have examined the associations of genes in the renin–angiotensin–aldosterone system or sodium/volume regulation pathways with BP response to dual combination therapy10, 11, 12; however, to our knowledge, no study has taken a non‐hypothesis‐driven genome‐wide association study (GWAS) approach to examine BP response to 2‐drug antihypertensive combination therapy. A few studies have examined the genetics of a more serious form of drug nonresponse, resistant hypertension, including recent GWAS and candidate gene studies.13, 14, 15 Additionally, several BP or hypertension GWAS and large candidate gene studies have been conducted and have included participants treated with antihypertensive therapy, typically with an adjustment of the BP values to account for the effects of medication16, 17; however, these studies were not specifically designed to evaluate the genetics of antihypertensive drug response.
In this study, using data from the PEAR (Pharmacogenomic Evaluation of Antihypertensive Responses) trial, we sought to identify genetic variants associated with uncontrolled BP in participants taking a thiazide diuretic and a β‐blocker combination. Findings from this study may contribute to the discovery of novel BP regulatory mechanisms and drugs that would work for different subsets of hypertensive patients with different genetic backgrounds.
Study Design and Study Population
PEAR was a prospective, randomized controlled clinical trial with a parallel‐group design that aimed to identify the genetic associations of the effects of the thiazide diuretic hydrochlorothiazide, the β‐blocker atenolol, and their combination, on BP and adverse metabolic responses (clinicaltrials.gov: NCT00246519).18 Institutional Review Boards at the University of Florida, the Mayo Clinic, and Emory University, where the participants were enrolled, approved the study, and voluntary written informed consent was obtained from all participants before enrollment. Additional details on the PEAR study have been published previously.18 Briefly, male and female participants aged 17 to 65 years with mild‐to‐moderate essential hypertension (DBP 90–110 mm Hg) and no major comorbidities were enrolled and randomized, following a washout period of ≈4 weeks duration, as necessary, to hydrochlorothiazide 12.5 mg or atenolol 50 mg once‐daily monotherapy. After 3 weeks on monotherapy, participants with BP >120/70 mm Hg underwent dose titration to hydrochlorothiazide 25 mg or atenolol 100 mg once daily for an additional 6 weeks of treatment. If BP remained >120/70 mm Hg at the end of monotherapy treatment, the alternate agent was added, followed by a similar dose‐titration step for another 6 to 9 weeks. BP and laboratory measurements were obtained at baseline and after monotherapy and combination therapy. Participants with severe or secondary hypertension, diabetes mellitus, cardiovascular disease, or renal disease were excluded. A nested case–control study was created within PEAR that comprised participants with uncontrolled office BP following 6 to 9 weeks of combination therapy (cases) and controlled office BP following 6 to 9 weeks of monotherapy or combination therapy (controls). Uncontrolled office BP was defined as having SBP ≥140 mm Hg and/or DBP ≥90 mm Hg, and controlled office BP was defined as having SBP <130 mm Hg and DBP <85 mm Hg to account for elevated daytime BP and “white coat” or isolated office hypertension. A total of 314 genetically defined white participants or individuals of European ancestry (123 cases, 191 controls) met these criteria and were included in the study. Black participants were not included because of insufficient numbers of cases and controls for discovery and replication studies.
The top genetic variants discovered in PEAR were tested for replication in INVEST (clinicaltrials.gov: NCT00133692).19 Details of the INVEST study have been published elsewhere.19 Briefly, INVEST was an international, multicenter, outcomes‐based randomized controlled clinical trial in which male and female participants of multiple race groups aged ≥50 years with a history of hypertension and coronary artery disease were randomized to treatment with a CCB (verapamil) or a β‐blocker (atenolol), followed by add‐on therapy with an ACE inhibitor (trandolapril) in the CCB arm or a thiazide diuretic (hydrochlorothiazide) in the β‐blocker arm. Only participants of the INVEST‐GENES (INVEST‐Genetic Substudy) were studied,20 and only participants who were randomized to the β‐blocker arm were included in order to create a replication cohort for the PEAR discovery analysis. Additionally, only BP measurements taken at the end of β‐blocker and thiazide diuretic combination therapy, before a third drug add‐on, were used to assess BP control status. The replication cohort included 414 participants, including genetically defined 221 whites (85 cases, 136 controls) and 193 Hispanics (81 cases, 112 controls). We considered INVEST Hispanics as a valid replication cohort for PEAR whites because INVEST Hispanics were recruited from Puerto Rico where Hispanics have more European genetic ancestry.21
Genotyping, Imputation, and Quality Control
Genome‐wide single nucleotide polymorphism (SNP) genotype data on PEAR participants were obtained through genotyping on the Illumina Human 1M‐duo beadchip (Illumina, San Diego, CA). All genome‐wide genotypes, including autosomal and sex chromosome variants, underwent quality control steps in PLINK.22 To confirm participants' self‐identified race, Principal Components Analysis for genetic ancestry was performed on a linkage disequilibrium pruned data set using the EIGENSTRAT method.23 Genotypes of variants that passed quality control underwent haplotype phasing in Markov Chain Haplotyping (MaCH) software24 followed by imputation to 1000 Genomes phase III reference panels using minimac/minimac2.25, 26 Variants with imputation quality r2<0.30 and minor allele frequency <3% were excluded.
INVEST DNA samples were genotyped on the Illumina OmniExpressExome Beadchip (Illumina, San Diego, CA). The 1000 Genomes phase III imputed genome‐wide variants data were used for replication analysis. Imputation was performed using a similar strategy as in PEAR. Principal Components Analysis for genetic ancestry was performed on a data set of linkage disequilibrium–pruned high‐quality SNPs using EIGENSTRAT.23 Race/ethnicity was genetically classified as white or Hispanic based on Principal Components Analysis.
Discovery analysis in the PEAR study.
Descriptive statistics were obtained using the Student t test for continuous variables and the χ2 or Fisher exact test for categorical variables. Continuous variables are shown as means±SDs and categorical variables as numbers and percentages. Multivariable logistic regression modeling was used for the GWAS analyses. The dependent variable or phenotype was a dichotomous variable: uncontrolled versus controlled BP on combination therapy, as described previously. In all models, covariates that met P<0.20 in the univariate analyses were selected for inclusion in a stepwise logistic regression model building procedure, with P<0.05 used as the criterion for a variable to stay in the model. The list of tested covariates is shown in Table. The final model included the covariates age, sex, ancestry‐specific principal components (PCs) 1 and 2, randomization assignment, baseline SBP, whether a higher dose of the add‐on drug was used, and smoking status. An additive genetic model was assumed, and variants with minor allele frequency <3% were excluded. All regression analyses were conducted using 1000 Genomes imputed expected genotype dosage files implemented in ProbABEL.27 Genetic race was defined using Principal Components Analysis as described in the Methods section. In all analyses, variants meeting P<5.0E‐08 were considered as genome‐wide significant markers of uncontrolled BP on combination therapy, and variants with P<1.0E‐05 were considered as having a suggestive level of evidence. Additionally, in all analyses, the effect of the allele associated with uncontrolled BP was reported. Regional plots of the top variants were visualized in LocusZoom28 and variant function was assessed using HaploReg and GTEx browsers.29, 30
We also conducted additional genome‐wide association analyses on the whole sample with the phenotypes SBP or DBP response on combination therapy, defined as the difference in BP between baseline and end of combination therapy. Multivariable linear regression modeling was used, with a similar strategy for covariate selection as previously described.
Replication analysis in the INVEST study.
Phenotypes were defined similarly to PEAR: uncontrolled versus controlled BP on combination therapy (dichotomous), SBP or DBP response on combination therapy (continuous). In order to ensure that participants were on the maximal needed or tolerated dose of atenolol and hydrochlorothiazide, we used BP measurements taken at the last visit at which INVEST participants were on atenolol and hydrochlorothiazide combination therapy before a third drug add‐on. Participants with controlled BP on atenolol or hydrochlorothiazide monotherapy were included among controls based on the assumption that if a participant is already controlled on 1 drug, they would be controlled on 2 drugs; and only BP measurements taken before a cardiovascular event were used for participants with an event. Further, BP values used to define “controlled BP” were the same as in PEAR and were also the same as those used to define “controlled BP” among diabetic participants in INVEST, in accordance with Sixth Report of the Joint National Commission (JNC VI) guidelines at the time of the trial.
Multivariable logistic or linear regression modeling was used depending on the phenotype. Covariates were selected using a similar strategy as in PEAR and included age, sex, ancestry‐specific PCs, atenolol dose, and hydrochlorothiazide dose. Atenolol and hydrochlorothiazide dose categories were collapsed into 3 groups based on clinically meaningful dosing cut points (hydrochlorothiazide ≥25 mg once daily or atenolol ≥100 mg once daily; hydrochlorothiazide <25 mg once daily or atenolol <100 mg once daily; none or “no dose” for the alternate drug, if controlled on 1 drug). Models with the dichotomous phenotype were further adjusted for history of left ventricular hypertrophy, and models with the continuous phenotypes were further adjusted for history of stroke and peripheral vascular disease as indicated by the stepwise covariate selection process.
The top variant in each gene region with lowest P value was selected for replication analysis in INVEST. Variants with consistent direction of effect between PEAR whites and INVEST whites, or between PEAR whites and INVEST white‐Hispanic meta‐analysis (but not INVEST Hispanics only) that met P<0.05/# of variants tested were considered as validated genetic markers of uncontrolled BP on combination therapy. A 1‐sided P value threshold was used, consistent with a 1‐sided hypothesis. Further, in all analyses, the effect of the allele associated with uncontrolled BP was reported.
We tested the robustness of the primary genome‐wide association analyses in PEAR with further model adjustment for baseline lipid levels. We also conducted sensitivity replication analysis in INVEST excluding individuals with uncontrolled BP on a low dose of both combination therapy drugs despite having a normal heart rate. Further, we performed sensitivity analysis in INVEST with additional adjustment for history of diabetes mellitus, heart failure, and renal insufficiency to account for the fact that INVEST included participants with these comorbidities while PEAR did not. Finally, we tested whether the top findings from PEAR validate among INVEST participants treated with CCB/ACE inhibitor combination therapy.
Study Population Demographic and Clinical Characteristics
Table shows characteristics of PEAR study participants by case–control status (uncontrolled or controlled BP on combination therapy). In general, participant characteristics were similar between cases and controls, with the exception of sex, baseline SBP and DBP, use of higher dose of add‐on drug, and smoking status.
Characteristics of INVEST participants by case–control status within each race/ethnicity group are summarized in Table S1. In total, there were 221 white and 193 Hispanic participants. Participant characteristics did not differ between cases and controls in each race/ethnicity group, with the exception of baseline SBP and DBP, hydrochlorothiazide and atenolol dose, and history of left ventricular hypertrophy (Hispanics).
Results of Discovery GWAS and Replication Analysis
No signals met the threshold of genome‐wide significance in PEAR; however, 24 variants were associated with uncontrolled BP on combination therapy at a suggestive level of significance (Table S2); all 24 variants imputed with good quality (r2≥0.80). These 24 variants arose from 5 distinct gene regions, and the top SNP in each region was tested for replication in INVEST (Table S2). Figure 1 summarizes the results for the top SNP. As shown, rs261316 in ALDH1A2 (aldehyde dehydrogenase 1 family member A2) was associated with uncontrolled BP on combination therapy with suggestive significance in PEAR whites, with the major allele (T) showing increased odds of having uncontrolled BP (odds ratio [OR]: 2.56, 95% confidence interval [CI], 1.69–3.88, P=8.64E‐06). This SNP showed consistent direction of association in INVEST whites (OR: 1.94, 95% CI, 1.23–3.06, P=0.008) and Hispanics (OR: 1.79, 95% CI, 1.13–2.82, P=0.018), replicated in white–Hispanic meta‐analysis (OR: 1.86, 95% CI, 1.35–2.57, P=0.001), and approached genome‐wide significance in the meta‐analysis between PEAR whites and INVEST whites and Hispanics (OR: 2.16, 95% CI, 1.63–2.86, P=8.60E‐08). The major allele frequency of rs261316 was 0.59–0.63, consistent with frequencies reported in 1000 genomes for individuals of European and Hispanic (Puerto Rico) ancestry.21 A plot of the genomic region surrounding the top variant rs261316 is shown in Figure 2. Other genes in the region include AQP9 and LIPC.
We also identified a variant rs35123024 near OR5H14 that was associated with the phenotype at a suggestive level of significance in PEAR, with the minor allele (C) associated with an increased odds of having uncontrolled BP on combination therapy (OR: 3.71, 95% CI, 2.13–6.45, P=3.46E‐06). rs35123024 replicated with nominal significance in the INVEST white‐Hispanic meta‐analysis (OR: 1.72, 95% CI, 1.06–2.79, P=0.033) and showed consistent direction of associations across all cohorts. This SNP did not reach genome‐wide significance in the meta‐analysis between PEAR whites and INVEST whites and Hispanics (OR: 2.56, 95% CI, 1.72–3.82, P=3.77E‐06). Minor allele frequencies of rs35123024 ranged from 0.12 to 0.18, consistent with those reported in 1000 genomes for individuals of European and Hispanic ancestry.21 A regional plot is provided in Figure S1. Results on the top variants from the additional GWAS of SBP and DBP response on combination therapy are summarized in Tables S3 and S4, respectively. In these additional analyses, none of the variants that met a suggestive level of significance in PEAR replicated in INVEST.
Results of Sensitivity Analysis
The top replicated variant, rs261316, in the primary analysis identified a region containing the gene LIPC, which has well‐known associations with lipid levels. We therefore conducted sensitivity GWAS analyses with additional models separately adjusted for baseline level of serum low‐density lipoprotein, high‐density lipoprotein, triglycerides, or total cholesterol in PEAR. rs261316 remained associated with uncontrolled BP on combination therapy in a consistent direction (adjusted for low‐density lipoprotein: P=1.79E‐06; high‐density lipoprotein: P=2.42E‐06; triglycerides: P=1.88E‐06; total cholesterol: P=1.95E‐06). Second, to ensure that most INVEST participants were on the maximum needed or tolerated doses of combination therapy, we identified and excluded the few individuals in INVEST (n=6) who were on a low dose of both combination therapy drugs despite having a normal heart rate. The top signal remained associated with the primary phenotype in the replication cohort (INVEST white–Hispanic meta‐analysis P=0.005). This issue did not arise in PEAR because the protocol required dose optimization on the first drug before addition of the second drug. Furthermore, because participants with diabetes mellitus were included in INVEST but not in PEAR, we performed sensitivity replication analysis in INVEST, with further model adjustment for history of diabetes mellitus. The top signal remained associated with the primary phenotype (INVEST white–Hispanic meta‐analysis P=0.002). The association also held after concurrently adjusting for history of diabetes mellitus, heart failure, and renal insufficiency (INVEST white–Hispanic meta‐analysis P=0.005). Lastly, we tested whether the top variant, rs261316, was associated with uncontrolled BP on combination therapy with a CCB and an ACE inhibitor in INVEST. rs261316 showed a trend toward significance in the INVEST white–Hispanic meta‐analysis (OR: 1.22, 95% CI, 0.94–1.59, P=0.104), with the association driven by INVEST Hispanics only (OR: 1.56, 95% CI, 1.05–2.32, P=0.030). The direction of association was consistent with PEAR discovery results.
To our knowledge, this is the first pharmacogenomic GWAS to investigate the genome‐wide associations of uncontrolled BP in people taking a combination of a thiazide diuretic and a β‐blocker. We identified a region on chromosome 15 containing the genes ALDH1A2, AQP9, and LIPC associated with uncontrolled BP on combination therapy, and replicated this finding in an external cohort. ALDH1A2, also known as RALDH2, is a highly conserved gene located at chromosome 15q21.3 that encodes the protein aldehyde dehydrogenase 1 family member A2. This enzyme catalyzes the irreversible oxidation of retinaldehyde from vitamin A to produce retinoic acid, which is essential for normal mammalian development of the kidneys and heart.31, 32 SNPs in ALDH1A2 have previously been associated with kidney size.33 Smaller kidneys may indicate having fewer nephrons, which has been associated with hypertension in animal and human studies.34 A GWAS among black participants identified an intergenic SNP near ALDH1A2 that was associated with hypertension (rs1550576, OR: 0.52, P=1.03E‐05).35 It is also notable that a recent meta‐analysis of pharmacogenomics GWAS by Hiltunen et al identified an intronic SNP in ALDH1A3, also important for retinoic acid synthesis, associated with BP response to hydrochlorothiazide (rs3825926, β: 6.7 mm Hg, P=5.6E‐06).36
The primary function of ALDH1A2 is to synthesize retinoic acid from vitamin A. Retinoic acid has been extensively studied in relation to kidney disease and has been considered a potential therapeutic agent because of its anti‐inflammatory, antifibrotic, and cell differentiation properties.31 In regard to BP regulation, retinoic acid administration has been shown to modify the expression of renin–angiotensin–aldosterone system components37, 38 and leads to decreased BP in a rat model.39, 40 Studies have also shown that the retinoic acid receptor RXR dimerizes with peroxisome‐proliferator‐activated receptor‐γ to upregulate renin gene expression.41 This may be of significance given that peroxisome‐proliferator‐activated receptor‐γ has well‐established roles in many cardiovascular and metabolic syndrome phenotypes, including hypertension.42 Other mechanisms through which retinoic acid may modulate BP include epigenetic alterations43 and endothelium‐dependent NO‐cGMP signaling.44 Retinoic acid may also inhibit the development of atherosclerosis45, 46 and reduce the risk of cardiovascular mortality.47
The region surrounding the top variant also includes the genes LIPC and AQP9. LIPC encodes hepatic lipase, which has well‐known associations with lipids48 and coronary artery disease.49, 50 LIPC has also been linked to BP and hypertension in some studies.51, 52 Of note, the top variant was in moderate linkage disequilibrium (r2: 0.51) with rs4775032, an expression quantitative trait loci (eQTL) in ALDH1A2 for LIPC expression.29 The third gene in the region, AQP9, encodes a member of a subset of aquaporins involved in transferring water, glycerol, and other small noncharged solutes across membranes. Association of AQP9 with BP regulation is unclear; however, 1 study found that it was upregulated in preeclamptic placenta.53 AQP9, in coordination with AQP7, may regulate adiposity and glucose homeostasis.54 Notably, other members of the aquaporin family (AQP1, AQP2, and AQP4) have been associated with BP in animal studies.55, 56
Retinoic acid–induced signaling, and the enzymes involved in its biosynthesis, may regulate BP through a variety of molecular mechanisms and pathophysiological processes. Further, retinoic acid may have BP‐lowering effects. Because the top variant in ALDH1A2 is intronic, it may be an eQTL, though this is not yet known, or it may be in linkage disequilibrium with the causal or functional variant. How that functional variant may affect retinoic acid is unclear, but it is possible that the T allele alters ALDH1A2 function such that levels of retinoic acid are decreased, leading to downstream effects on the renin–angiotensin–aldosterone system and peroxisome‐proliferator‐activated receptor‐γ pathways that may ultimately result in higher BP.
We also identified a top variant near OR5H14 that replicated with nominal significance and showed consistent direction of effect across discovery and replication cohorts. With additional replication studies, it is possible that this signal may be validated. OR5H14 has not previously been linked to BP; however, there is a growing body of literature on the role of sensory receptors, including olfactory receptors, in modulating BP.57, 58 Further, in additional analysis with SBP and DBP response phenotypes, a few variants showed nominal significance in the replication analysis, including a variant in the genes SETD7, MIR720, and RP11‐461F11.3 associated with DBP response. SETD7 may play a role in mediating high glucose‐induced vascular dysfunction.59 Less is known about the other genes, which are micro‐ and long non‐coding RNAs, in regard to the cardiovascular system.
The strengths of the study include the use of data from a randomized controlled clinical trial with a robust design for evaluating the pharmacogenomics of antihypertensive drug response. We used a non‐hypothesis‐driven GWAS approach to identify the genetic correlates of uncontrolled BP on combination therapy with the 2 well‐known antihypertensive drug classes, thiazide diuretics and β‐blockers. Through this approach, we identified genetic variants corresponding to genes with plausible connections to the cardiovascular system and BP regulation. The top variant replicated in an external cohort of complicated hypertensives, indicating that the finding may have broader validity beyond uncomplicated hypertensives. More research is needed to understand the role of the identified genes in the cardiovascular system.
Our study also adds to the growing literature on the pharmacogenomics of antihypertensive drug nonresponse to combination therapy. A few GWAS and candidate gene studies have studied a more extreme form of drug nonresponse, namely, resistant hypertension,13, 14, 15 defined as the use of at least 4 drugs to achieve BP control. Further, many BP or hypertension GWAS and large candidate gene studies have been conducted that have included treated individuals,16, 17 including 1 study that identified BP associations with the gene ALDH2, which encodes aldehyde dehydrogenase.60 It must be noted, however, that these studies were not specifically designed to investigate the genetics of drug response, and therefore, the genes identified may not always overlap with those of hypertension pharmacogenomic studies.
Lastly, another strength of the study is that both PEAR and INVEST had well‐documented drug and dose optimization protocols, with good adherence in PEAR (>85%),18 and while INVEST did not directly assess adherence, study drugs were delivered to participants' homes (with receipt confirmations), and we observed expected reductions in BP and heart rate in both treatment arms, consistent with drug utilization.61, 62
The limitations of the study are acknowledged. We focused on the discovery and replication of variants associated with uncontrolled BP on thiazide diuretic and β‐blocker combination, and did not examine other drug class combinations. While β‐blocker combinations are necessary in many patients, particularly those with ischemic heart disease or other indications for a β‐blocker, we acknowledge that other combinations, particularly CCB/ACE inhibitor combinations, are now more widely used for uncomplicated hypertensives in light of the results of landmark studies.5, 6 As such, we also tested whether the top variant, rs261316, was associated with uncontrolled BP on CCB/ACE inhibitor combination therapy in INVEST, which showed that rs261316 trended toward association in INVEST white–Hispanic meta‐analysis, with consistent direction across INVEST cohorts and with PEAR. Although this would need to be studied in another, larger cohort, these results suggest that rs261316 might also be a marker of uncontrolled BP among patients treated with CCB/ACE inhibitor combination therapy.
Other study limitations are noted. We did not identify variants that reached Bonferroni‐corrected level of genome‐wide significance in the discovery analyses; however, we selected variants that met the suggestive level of significance and focused on those that replicated or trended toward replication in a secondary cohort. Further, with the exception of the top variant in ALDH1A2, other top variants from the discovery analysis did not meet the multiple‐comparisons corrected P value threshold for replication and did not approach genome‐wide significance in the meta‐analysis across discovery and replication cohorts. The relatively small sample size of the study is likely a contributing factor. Despite the reduced power, we had moderate‐to‐large effect sizes, which is consistent with findings from pharmacogenomics GWAS.63 Differences in participants' baseline characteristics as well as differences in treatment protocol between the discovery and replication cohorts may also have precluded identifying additional significant signals. We tried to overcome these differences by adjusting for participant and study‐related differences. Given that we were able to identify a replicated signal in or close to genes with plausible connections to various cardiovascular phenotypes suggests that our findings may not be spurious. Lastly, our study focused on white and Hispanic participants only. Further replication of the findings in other race/ethnic groups is warranted.
In summary, our study identified a SNP in the gene ALDH1A2 where the major allele T was associated with an increased odds of uncontrolled BP on a thiazide diuretic and a β‐blocker combination therapy. ALDH1A2 and its byproduct, retinoic acid, may influence BP through a variety of alternative pathways. Additionally, we found suggestive evidence that the top SNP may also be associated with better BP control among individuals treated with a CCB/ACE inhibitor combination. More research on the genes corresponding to the top SNP, particularly ALDH1A2, and their pathways, are needed, which may illuminate novel or understudied mechanisms of BP regulation. Additionally, replication of the study's findings among participants treated with the same or different drug combinations is needed to determine their clinical utility for identifying patients who would benefit from initiation of combination therapy.
Sources of Funding
Funding for PEAR comes from the National Institutes of Health (NIH) Pharmacogenetics Research Network grant U01 GM074492, the National Center for Advancing Translational Sciences under award numbers UL1 TR000064 (University of Florida), UL1 TR000454 (Emory University), and UL1 TR000135 (Mayo Clinic), the Mayo Foundation, and the National Heart Lung and Blood Institute grant R01 HL74730. INVEST was supported by the University of Florida and grants from BASF Pharma and Abbott Laboratories. Additional support for this project comes from NIH grants T32 DK104721 (Magvanjav) and KL2 TR001429 (McDonough).
Table S1. Characteristics of INVEST‐GENES Study Participants
Table S2. Association of Top Genetic Variants With Uncontrolled BP on Combination Therapy in PEAR With Replication Analysis of Top Variant Per Region in INVEST
Table S3. Association of Top Genetic Variants With SBP Response on Combination Therapy in PEAR With Replication Analysis in INVEST
Table S4. Association of Top Genetic Variants With DBP Response on Combination Therapy in PEAR With Replication Analysis in INVEST
Figure S1. Regional plot of top genetic variant associated with uncontrolled BP on combination therapy in PEAR
We are grateful to the participants, site investigators, physicians, and support staff of the PEAR and INVEST studies.
- ↵Mills KT, Bundy JD, Kelly TN, Reed JE, Kearney PM, Reynolds K, Chen J, He J. Global disparities of hypertension prevalence and control: a systematic analysis of population‐based studies from 90 countries. Circulation. 2016;134:441–450.
- ↵Writing Group Members , Mozaffarian D, Benjamin EJ, Go AS, Arnett DK, Blaha MJ, Cushman M, Das SR, de Ferranti S, Després JP, Fullerton HJ, Howard VJ, Huffman MD, Isasi CR, Jiménez MC, Judd SE, Kissela BM, Lichtman JH, Lisabeth LD, Liu S, Mackey RH, Magid DJ, McGuire DK, Mohler ER, Moy CS, Muntner P, Mussolino ME, Nasir K, Neumar RW, Nichol G, Palaniappan L, Pandey DK, Reeves MJ, Rodriguez CJ, Rosamond W, Sorlie PD, Stein J, Towfighi A, Turan TN, Virani SS, Woo D, Yeh RW, Turner MB, American Heart Association Statistics Committee; Stroke Statistics Subcommittee. Heart Disease and Stroke Statistics‐2016 Update: a Report From the American Heart Association. Circulation. 2016;133:e38–e60.
- ↵James PA, Oparil S, Carter BL, Cushman WC, Dennison‐Himmelfarb C, Handler J, Lackland DT, LeFevre ML, MacKenzie TD, Ogedegbe O, Smith SC, Svetkey LP, Taler SJ, Townsend RR, Wright JT, Narva AS, Ortiz E. 2014 evidence‐based guideline for the management of high blood pressure in adults: report from the panel members appointed to the Eighth Joint National Committee (JNC 8). JAMA. 2014;311:507–520.
- ↵Dahlöf B, Sever PS, Poulter NR, Wedel H, Beevers DG, Caulfield M, Collins R, Kjeldsen SE, Kristinsson A, McInnes GT, Mehlsen J, Nieminen M, O'Brien E, Ostergren J; ASCOT Investigators . Prevention of cardiovascular events with an antihypertensive regimen of amlodipine adding perindopril as required versus atenolol adding bendroflumethiazide as required, in the Anglo‐Scandinavian Cardiac Outcomes Trial‐Blood Pressure Lowering Arm (ASCOT‐BPLA): a multicentre randomised controlled trial. Lancet. 2005;366:895–906.
- ↵Brunner M, Cooper‐DeHoff RM, Gong Y, Karnes JH, Langaee TY, Pepine CJ, Johnson JA; IINVEST Investigators . Factors influencing blood pressure response to trandolapril add‐on therapy in patients taking verapamil SR (from the International Verapamil SR/Trandolapril [INVEST] Study). Am J Cardiol. 2007;99:1549–1554.
- ↵Dumitrescu L, Ritchie MD, Denny JC, El Rouby NM, McDonough CW, Bradford Y, Ramirez AH, Bielinski SJ, Basford MA, Chai HS, Peissig P, Carrell D, Pathak J, Rasmussen LV, Wang X, Pacheco JA, Kho AN, Hayes MG, Matsumoto M, Smith ME, Li R, Cooper‐DeHoff RM, Kullo IJ, Chute CG, Chisholm RL, Jarvik GP, Larson EB, Carey D, McCarty CA, Williams MS, Roden DM, Bottinger E, Johnson JA, de Andrade M, Crawford DC. Genome‐wide study of resistant hypertension identified from electronic health records. PLoS One. 2017;12:e0171745.
- ↵Gong Y, McDonough CW, Beitelshees AL, El Rouby N, Hiltunen TP, O'Connell JR, Padmanabhan S, Langaee TY, Hall K, Schmidt SO, Curry RW, Gums JG, Donner KM, Kontula KK, Bailey KR, Boerwinkle E, Takahashi A, Tanaka T, Kubo M, Chapman AB, Turner ST, Pepine CJ, Cooper‐DeHoff RM, Johnson JA. PTPRD gene associated with blood pressure response to atenolol and resistant hypertension. J Hypertens. 2015;33:2278–2285.
- ↵Fontana V, McDonough CW, Gong Y, El Rouby NM, Sá AC, Taylor KD, Chen YD, Gums JG, Chapman AB, Turner ST, Pepine CJ, Johnson JA, Cooper‐DeHoff RM. Large‐scale gene‐centric analysis identifies polymorphisms for resistant hypertension. J Am Heart Assoc. 2014;3:e001398. DOI: 10.1161/JAHA.114.001398.
- ↵Padmanabhan S, Aman A, Dominiczak AF. Genomics of hypertension. Pharmacol Res. 2017;121:219–229.
- ↵Padmanabhan S, Caulfield M, Dominiczak AF. Genetic and molecular aspects of hypertension. Circ Res. 2015;116:937–959.
- ↵Johnson JA, Boerwinkle E, Zineh I, Chapman AB, Bailey K, Cooper‐DeHoff RM, Gums J, Curry RW, Gong Y, Beitelshees AL, Schwartz G, Turner ST. Pharmacogenomics of antihypertensive drugs: rationale and design of the Pharmacogenomic Evaluation of Antihypertensive Responses (PEAR) study. Am Heart J. 2009;157:442–449.
- ↵Pepine CJ, Handberg‐Thurmond E, Marks RG, Conlon M, Cooper‐DeHoff R, Volkers P, Zellig P. Rationale and design of the International Verapamil SR/Trandolapril Study (INVEST): an Internet‐based randomized trial in coronary artery disease patients with hypertension. J Am Coll Cardiol. 1998;32:1228–1237.
- ↵Beitelshees AL, Gong Y, Wang D, Schork NJ, Cooper‐Dehoff RM, Langaee TY, Shriver MD, Sadee W, Knot HJ, Pepine CJ, Johnson JA; INVEST Investigators I . KCNMB1 genotype influences response to verapamil SR and adverse outcomes in the INternational VErapamil SR/Trandolapril STudy (INVEST). Pharmacogenet Genomics. 2007;17:719–729.
- ↵Consortium G. Human genomics. The Genotype‐Tissue Expression (GTEx) pilot analysis: multitissue gene regulation in humans. Science. 2015;348:648–660.
- ↵Xavier‐Neto J, Sousa Costa AM, Figueira AC, Caiaffa CD, Amaral FN, Peres LM, da Silva BS, Santos LN, Moise AR, Castillo HA. Signaling through retinoic acid receptors in cardiac development: doing the right things at the right times. Biochem Biophys Acta. 2015;1849:94–111.
- ↵Kanzaki G, Tsuboi N, Haruhara K, Koike K, Ogura M, Shimizu A, Yokoo T. Factors associated with a vicious cycle involving a low nephron number, hypertension and chronic kidney disease. Hypertens Res. 2015;38:633–641.
- ↵Hiltunen TP, Donner KM, Sarin AP, Saarela J, Ripatti S, Chapman AB, Gums JG, Gong Y, Cooper‐DeHoff RM, Frau F, Glorioso V, Zaninello R, Salvi E, Glorioso N, Boerwinkle E, Turner ST, Johnson JA, Kontula KK. Pharmacogenomics of hypertension: a genome‐wide, placebo‐controlled cross‐over study, using four classes of antihypertensive drugs. J Am Heart Assoc. 2015;4:e001521. DOI: 10.1161/JAHA.114.001521.
- ↵Dechow C, Morath C, Peters J, Lehrke I, Waldherr R, Haxsen V, Ritz E, Wagner J. Effects of all‐trans retinoic acid on renin‐angiotensin system in rats with experimental nephritis. Am J Physiol Renal Physiol. 2001;281:F909–F919.
- ↵Todorov VT. PPARgamma‐dependent control of renin expression: molecular mechanisms and pathophysiological relevance. PPAR Res. 2013;2013:451016.
- ↵Stump M, Mukohda M, Hu C, Sigmund CD. PPARγ regulation in hypertension and metabolic syndrome. Curr Hypertens Rep. 2015;17:89.
- ↵Kumar P, Periyasamy R, Das S, Neerukonda S, Mani I, Pandey KN. All‐trans retinoic acid and sodium butyrate enhance natriuretic peptide receptor a gene transcription: role of histone modification. Mol Pharmacol. 2014;85:946–957.
- ↵Wang Y, Han Y, Yang J, Wang Z, Liu L, Wang W, Zhou L, Wang D, Tan X, Fu C, Jose PA, Zeng C. Relaxant effect of all‐trans‐retinoic acid via NO‐sGC‐cGMP pathway and calcium‐activated potassium channels in rat mesenteric artery. Am J Physiol Heart Circ Physiol. 2013;304:H51–H57.
- ↵Bechor S, Zolberg, Relevy N, Harari A, Almog T, Kamari Y, Ben‐Amotz A, Harats D, Shaish A. 9‐cis β‐carotene increased cholesterol efflux to HDL in macrophages. Nutrients. 2016;8: 435.
- ↵Liu Y, Chen H, Mu D, Li D, Zhong Y, Jiang N, Zhang Y, Xia M. Association of Serum Retinoic Acid With Risk of Mortality in Patients With Coronary Artery Disease. Circ Res. 2016;119:557–563.
- ↵Kobayashi J, Miyashita K, Nakajima K, Mabuchi H. Hepatic lipase: a comprehensive view of its role on plasma lipid and lipoprotein metabolism. J Atheroscler Thromb. 2015;22:1001–1011.
- ↵Aggarwal A, Srivastava S, Velmurugan M. Newer perspectives of coronary artery disease in young. World J Cardiol. 2016;8:728–734.
- ↵Ríos‐González BE, Ibarra‐Cortés B, Ramírez‐López G, Sánchez‐Corona J, Magaña‐Torres MT. Association of polymorphisms of genes involved in lipid metabolism with blood pressure and lipid values in mexican hypertensive individuals. Dis Markers. 2014;2014:150358.
- ↵Montiel V, Leon Gomez E, Bouzin C, Esfahani H, Romero Perez M, Lobysheva I, Devuyst O, Dessy C, Balligand JL. Genetic deletion of aquaporin‐1 results in microcardia and low blood pressure in mouse with intact nitric oxide‐dependent relaxation, but enhanced prostanoids‐dependent relaxation. Pflugers Arch. 2014;466:237–251.
- ↵Pluznick JL. Renal and cardiovascular sensory receptors and blood pressure regulation. Am J Physiol Renal Physiol. 2013;305:F439–F444.
- ↵Miyamoto J, Kasubuchi M, Nakajima A, Irie J, Itoh H, Kimura I. The role of short‐chain fatty acid on blood pressure regulation. Curr Opin Nephrol Hypertens. 2016;25:379–383.
- ↵Paneni F, Costantino S, Battista R, Castello L, Capretti G, Chiandotto S, Scavone G, Villano A, Pitocco D, Lanza G, Volpe M, Lüscher TF, Cosentino F. Adverse epigenetic signatures by histone methyltransferase Set7 contribute to vascular dysfunction in patients with type 2 diabetes mellitus. Circ Cardiovasc Genet. 2015;8:150–158.
- ↵Tragante V, Barnes MR, Ganesh SK, Lanktree MB, Guo W, Franceschini N, Smith EN, Johnson T, Holmes MV, Padmanabhan S, Karczewski KJ, Almoguera B, Barnard J, Baumert J, Chang YP, Elbers CC, Farrall M, Fischer ME, Gaunt TR, Gho JM, Gieger C, Goel A, Gong Y, Isaacs A, Kleber ME, Mateo Leach I, McDonough CW, Meijs MF, Melander O, Nelson CP, Nolte IM, Pankratz N, Price TS, Shaffer J, Shah S, Tomaszewski M, van der Most PJ, VanIperen EP, Vonk JM, Witkowska K, Wong CO, Zhang L, Beitelshees AL, Berenson GS, Bhatt DL, Brown M, Burt A, Cooper‐DeHoff RM, Connell JM, Cruickshanks KJ, Curtis SP, Davey‐Smith G, Delles C, Gansevoort RT, Guo X, Haiqing S, Hastie CE, Hofker MH, Hovingh GK, Kim DS, Kirkland SA, Klein BE, Klein R, Li YR, Maiwald S, Newton‐Cheh C, O'Brien ET, Onland‐Moret NC, Palmas W, Parsa A, Penninx BW, Pettinger M, Vasan RS, Ranchalis JE, M Ridker P, Rose LM, Sever P, Shimbo D, Steele L, Stolk RP, Thorand B, Trip MD, van Duijn CM, Verschuren WM, Wijmenga C, Wyatt S, Young JH, Zwinderman AH, Bezzina CR, Boerwinkle E, Casas JP, Caulfield MJ, Chakravarti A, Chasman DI, Davidson KW, Doevendans PA, Dominiczak AF, FitzGerald GA, Gums JG, Fornage M, Hakonarson H, Halder I, Hillege HL, Illig T, Jarvik GP, Johnson JA, Kastelein JJ, Koenig W, Kumari M, März W, Murray SS, O'Connell JR, Oldehinkel AJ, Pankow JS, Rader DJ, Redline S, Reilly MP, Schadt EE, Kottke‐Marchant K, Snieder H, Snyder M, Stanton AV, Tobin MD, Uitterlinden AG, van der Harst P, van der Schouw YT, Samani NJ, Watkins H, Johnson AD, Reiner AP, Zhu X, de Bakker PI, Levy D, Asselbergs FW, Munroe PB, Keating BJ. Gene‐centric meta‐analysis in 87,736 individuals of European ancestry identifies multiple blood‐pressure‐related loci. Am J Hum Genet. 2014;94:349–360.
- ↵Pepine CJ, Handberg EM, Cooper‐DeHoff RM, Marks RG, Kowey P, Messerli FH, Mancia G, Cangiano JL, Garcia‐Barreto D, Keltai M, Erdine S, Bristol HA, Kolb HR, Bakris GL, Cohen JD, Parmley WW; Investigators I . A calcium antagonist vs a non‐calcium antagonist hypertension treatment strategy for patients with coronary artery disease. The International Verapamil‐Trandolapril Study (INVEST): a randomized controlled trial. JAMA. 2003;290:2805–2816.
- ↵Maranville JC, Cox NJ. Pharmacogenomic variants have larger effect sizes than genetic variants associated with other dichotomous complex traits. Pharmacogenomics J. 2016;16:388–392.