Introduction

Hypertension is a major risk factor for cardiovascular disease (CVD), which is the leading cause of morbidity and mortality in the United States [1]. Over 76 million Americans have hypertension, of which an estimated 71% are taking antihypertensive medications [1]. Thiazide diuretics, β-blockers, angiotensin converting enzyme inhibitors, angiotensin receptor blockers, and calcium channel blockers are all acceptable first-line treatments for hypertension [2]. However, there has been debate about the first-line role of β-blockers and diuretics due to their adverse metabolic effects on glucose, insulin, and lipid levels [3]. For example, β-blockers have been shown to significantly decrease high-density lipoprotein cholesterol (HDL-C) levels, but the underlying mechanism is unclear [4]. These metabolic effects, including HDL-C reduction, are also strong contributors to the risk of coronary heart disease and stroke [1].

Data from four large, prospective studies, including The Framingham Heart Study, suggest that every 1 mg/dL (0.026 mmol/L) decrease in HDL-C is associated with a 2-3% increase in risk for coronary heart disease [5]. Additionally, post hoc analysis of the Treating to New Targets (TNT) trial demonstrated that low levels of HDL-C were predictive of higher rates of CVD, even at very low levels of low-density lipoprotein cholesterol (LDL-C) (< 70 mg/dL) [6] and low HDL-C levels were also predictive of major cardiovascular events in patients treated with statins [7]. However, randomized controlled trials designed to determine if increases in HDL-C prevent coronary heart disease and major cardiovascular events demonstrate inconsistent results [8,9]. Additionally, it remains unclear if decreases in HDL-C in response to β-blocker therapy are harmful in a similar way as low HDL-C observed in epidemiological studies.

Clear data exist regarding the influence of genetics on HDL-C levels. In a study of the sex specific genetic architecture of many quantitative traits, the heritability of HDL-C was estimated at 52% in males and 70% in females [10]. HDL-C is a sexually dimorphic trait, with higher HDL-C levels typically observed in females [10]. There have been multiple genome-wide association studies (GWAS) investigating the effect of genetic variants on blood lipid levels. In a large-scale meta-analysis including many of these GWAS, 45 loci were found to influence HDL-C [11]. When examining the entire set of single nucleotide polymorphisms (SNPs) associated with HDL-C, 12.1% of the total variance in HDL-C was explained, and ~25-30% of the genetic variance was explained [11].

While there have been many studies investigating genetic influences on blood lipid levels, including HDL-C [11], the genetics of HDL-C response to β-blockers remains unexplored. In order to evaluate potential genetic effects, we assessed SNPs across the genome from the Illumina HumanCVD chip [12] for association with atenolol induced HDL-C change in the Pharmacogenomic Evaluation of Antihypertensive Responses (PEAR) study [13].

Methods

Results

The results presented are for PEAR study participants on atenolol monotherapy (i.e. the atenolol arm). Baseline characteristics and demographics are shown in Table 1. Among a total of 384 study participants in the atenolol group, 58.6% were white and 38.3% were African American. Participants were overweight (average BMI = 30.8 kg/m²), and 12.8% were current smokers. The average baseline HDL-C was 49.7 mg/dL.

The Manhattan plots and corresponding Q-Q plots for HDL-C response to atenolol are shown in Figure S1 and Figure S2, respectively. No SNPs achieved a Bonferroni correct P-value. Overall, 25 SNPs met the screening P-value (P<1x10^-3) in whites and 42 SNPs met the screening P-value in the African Americans; however, no SNP replicated in the other race group, in the same direction. After examining the identified gene regions in the other race for evidence of consistent association, thirteen regions met our threshold for association (P<0.005 or P<0.05 with prior HDL-C or lipid association, Table 2). The allele frequencies, genotype counts, and Hardy-Weinberg equilibrium P-values for these SNPs are shown in Table S1. All SNPs identified in these gene regions were in Hardy-Weinberg Equilibrium (Hardy-Weinberg P-value > 0.01).

In whites, the strongest association was rs10157410 in PLA2G4A (P=6.66x10^-5, beta=3.46, Table 2), with consistent regional association at rs4648287 in the African Americans (P=0.0040, beta=3.21). PLA2G4A is located at chromosome 1q31.1 and encodes the cytosolic phospholipase A2. In African Americans, the strongest association was seen at rs3213619 on chromosome 7q21.12, with a P-value of 2.74x10^-4 and a beta of 5.22 (Table 2). rs3213619 lies in ABCB1, ATP-binding cassette sub-family B member 1, and consistent regional association was also observed at rs10267099 in whites (P=0.0059, beta=1.62).

Two other regions are of particular note. First, GALNT2, polypeptide N-acetylgalactosaminyltransferase 2, which has been previously associated in GWAS studies with circulating HDL-C levels [11], was associated with atenolol induced HDL-C change. In whites the G allele of rs2144300, one of the same SNPs identified from GWAS studies [23], was associated with an atenolol induced decrease in HDL-C (P=2.29x10^-4, Table 2, Figure 1A), with a mean HDL-C change of -4.18 mg/dL, -3.03 mg/dL, and -0.63 mg/dL for the GG, GA, and AA genotypes, respectively (Figure 1B). In African Americans, a signal was observed in GALNT2 with the A allele of rs2144297 and a decrease in HDL-C in response to atenolol (P=0.0233, Table 2, Figure 1C). The mean change in HDL-C was -5.39 mg/dL, -3.61 mg/dL, and -1.63 mg/dL for the AA, AG, and GG genotypes, respectively (Figure 1D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. GALNT2 Regional Plots and adjusted HDL-C response by genotype.

(A) Regional Plot in PEAR whites. (B) HDL-C response in PEAR whites by rs2144300 genotype. (C) Regional Plot in PEAR African Americans. (D) HDL-C response in PEAR African Americans by rs2144297 genotype. Error bars represent standard error.

https://doi.org/10.1371/journal.pone.0076984.g001

The second gene, FTO, the fat mass and obesity associated gene, has been previously associated with BMI, obesity and type 2 diabetes in GWAS [24], and was recently associated with HDL-C in a large, gene-centric analysis of the CVD array used herein [25]. In African Americans, the A allele of rs12595985 was associated with an increase in HDL-C in response to atenolol treatment, with an increase of 13.03 mg/dL in the AA homozygote, an average decrease of 0.07 mg/dL in AC heterozygotes, and an average decrease of 3.97 mg/dL in CC homozygotes (P=2.90x10^-4, Table 2, Figures 2A and 2B). The observed signal in whites was with rs9940629, with the G allele associated with a greater reduction in HDL-C after atenolol therapy (P=0.0041, Table 2, Figure 2C). The mean change in HDL-C was -1.09 mg/dL, -1.90 mg/dL, and -4.04 mg/dL for the AA, AG, and GG genotypes, respectively (Figure 2D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. FTO Regional Plots and adjusted HDL-C response by genotype.

(A) Regional Plot in PEAR African Americans. (B) HDL-C response in PEAR African Americans by rs12595985 genotype. (C) Regional Plot in PEAR whites. (D) HDL-C response in PEAR whites by rs9940629 genotype. Error bars represent standard error.

https://doi.org/10.1371/journal.pone.0076984.g002

Top initial signals with low minor allele frequencies were further investigated to determine if single study participants with an extreme HDL-C response to atenolol were driving the observed signals. One white participant with an HDL-C change of -36.4 mg/dL was identified and the individual was removed in a sensitivity analysis (Table S2). All of the initial signals in whites remained except for LIPC rs9652472 (sensitivity P-value=0.3998). Additionally, sensitivity analyses were conducted in African Americans at rs12595985 in FTO (Table S3). The signal remained under both a dominant genetic model, collapsing the AA homozygote with the AC heterozygotes (sensitivity P-value=9.71x10^-4) and an additive genetic model, after removing the single AA homozygote, (sensitivity P-value=3.18x10^-3).

The putative functional implications of the lead SNPs from the initial signals in the 13 gene regions are shown in Table S4. rs3736228, in LRP5 (low density lipoprotein receptor-related protein 5), is a non-synonymous SNP causing a conservative amino acid change of an alanine to valine at amino acid position 1330. Additionally, rs3736228 was predicted by FastSNP to affect splicing regulation. All of the other SNPs identified were located in intronic regions, or upstream or downstream of the nearest gene.

In order to examine the possible function of two the top SNPs, we measured gene expression at baseline (before atenolol treatment) of GALNT2 by rs2144300 genotype in 34 whites and of FTO by rs12595985 genotype in 35 African Americans. For GALNT2, study participants with one or two copies of the G allele had a 1.59 and 1.75 fold increase in relative expression at baseline, respectively, compared to AA homozygotes (P=0.0279, Figure 3). There was no difference in baseline expression between genotype groups for FTO (data not shown).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Plot of relative gene expression of GALNT2 by rs2144300 genotype.

Gene expression analysis performed in 34 PEAR whites, normalized to β-2-microglobulin. Error bars represent standard error.

https://doi.org/10.1371/journal.pone.0076984.g003

Discussion

We report, for the first time to our knowledge, a genome-spanning analysis using the HumanCVD Beadchip, which contains cardiovascular, metabolic and inflammatory candidate genes, to identify genetic variants associated with atenolol induced changes in HDL-C in hypertensive study participants. While no SNPs achieved a Bonferroni corrected P-value, using a staged analysis approach that required consistent regional association in white and African American participants, we identified 7 SNPs associated with HDL-C response to atenolol in whites that showed evidence of consistent regional association in African Americans. Additionally, 6 SNPs were found in African Americans with consistent regional association in whites. Most notable were seven gene regions with prior associations with HDL-C or lipid metabolism/trafficking: GALNT2, FTO, ABCB1, LRP5, STARD3NL, ESR1, and LIPC [11,25-29].

GALNT2, located at 1q42.13, is a member of the GalNAc-transferase family, which transfers N-acetyl galactosamine to the hydroxyl group of a serine or threonine residue in the first step of O-linked oligosaccharide biosynthesis [30]. Common variants in GALNT2 have previously been associated with both HDL-C and triglyceride levels in numerous GWAS analyses, and subsequently validated [11,23]. Additionally, altered expression of Galnt2 in mouse models, both knockdown and over expression, has been inversely correlated with altered HDL-C levels [11]. This is similar to what we observed; where those individuals with higher gene expression at baseline (G carriers at rs2144300), had greater decreases in HDL-C in response to atenolol, resulting in lower HDL-C levels. These prior data, the association of common SNPs in this gene with HDL-C response to atenolol in whites and African Americans, and the difference in GALNT2 expression by rs2144300 genotype in whites suggest GALNT2 may be an important mediator of the observed atenolol associated HDL-C response.

FTO, located at 16q12.2, is thought to be involved in the regulation of food intake and to affect lipolysis in adipose tissue [31]. Other SNPs in FTO have been previously associated with BMI, obesity and type 2 diabetes [24]. Additionally, FTO was recently associated with HDL-C in a large, gene centric analysis [25], and it was suggested that this association with FTO and HDL-C may be mediated through FTO’ s association with BMI, since the observed association with FTO and HDL-C lost significance after BMI adjustment [25]. However, in our data, when we add BMI to the model, rs12595985 remains associated with HDL-C response to atenolol (P=3x10^-4), suggesting the association may be through a different mechanism. Additionally, while we did not observe a significant difference in FTO expression by rs12595985 genotype at baseline, this does not preclude that expression differences might exist in more relevant tissues (e.g. liver).

Other top genes are also of note. ABCB1, also known as MDR1 (multi-drug resistance protein 1), belongs to the MDR/TAP subfamily responsible for transporting various molecules across cell membranes. Evidence indicates a link of ABCB1 variants with circulating lipid profiles, mainly LDL-C, and the efficacy of statins [26]. Furthermore, ABCB1 3435C>T (rs1045642) has been associated with efavirenz induced changes in HDL-C [32]. We did not observe association with HDL-C response to atenolol at this SNP in our study; however, because we did observe association in this gene, ABCB1 and drug associated HDL-C change warrants further investigation. Next, LRP5, low-density lipoprotein receptor-related protein 5, encodes a transmembrane low-density lipoprotein receptor. Mice lacking the production of LPR5 had increased total plasma cholesterol levels and impaired glucose tolerance [27]. STARD3 N-terminal L, STARD3NL, encodes a late-endosomal protein that binds cholesterol and may assist in endosomal cholesterol transport [28], a process important in maintaining HDL cholesterol levels [33]. Also, ESR1, the estrogen receptor 1 gene, was shown associated with HDL-C response to atenolol. Variants in ESR1 have been associated with many traits including type 2 diabetes in African Americans [34], and HDL-C response to atorvastatin in women [29]. The last gene of particular interest is LIPC, hepatic lipase, which is involved in lipoprotein metabolism, and other SNPs in the gene have also been associated with HDL-C in GWAS [11].

Many of the top gene regions identified have shown prior evidence of association with circulating HDL-C levels, HDL-C levels in response to drug therapy, or other metabolic traits or are involved in lipid metabolism. While this analysis was based on a candidate gene chip, the chip includes many genes involved in cardiovascular and inflammatory processes as well as genes involved in metabolic processes [12]. Thus, it is particularly interesting that a connection to HDL-C could be made for many of the gene regions meeting the consistent regional association requirement.

Like any study, the present investigation is not without limitations. For example, we had a relatively small sample size and did not have an external replication cohort. However, to our knowledge, there is not another suitable replication population with data on HDL-C response to a β-blocker at the present time. Also, due to our study design, we only focused on gene regions that showed a signal in both race groups. While true functional variants should be relevant to all race groups, it is very possible that we missed regions of interest due to our consistent regional association requirement and the differences in LD between the two race groups. Additionally, while we have identified gene regions of interest, we have likely not identified the functional variants in these regions. It is more likely we have identified SNPs in LD with the functional variant.

β-blockers are appropriate first-line therapy for hypertension [2]; however, there is growing concern regarding use in patients without compelling indications (i.e., heart failure or myocardial infarction), as the benefits associated with use of β-blockers may not outweigh the potential metabolic and other risks, including a decrease in HDL-C levels [3,4]. Our study investigated the possible genetic contributions to atenolol induced change in HDL-C in hypertensive study participants. We found 13 gene regions that showed initial evidence of association in one race (whites or African Americans), and consistent regional association in the other. Of particular interest were seven gene regions with prior associations with HDL-C or other metabolic traits, or functional implications in the lipid pathway: GALNT2, FTO, ABCB1, LRP5, STARD3NL, ESR1, and LIPC. We also observed differences in GALNT2 baseline expression by rs2144300 genotype. These results require further investigation in order to confirm the associations, identify potential functional variants, elucidate their role, and establish the mechanism through which they act. However, our findings provide initial insight into genes that may influence atenolol induced change in HDL-C.

Supporting Information

Acknowledgments

We acknowledge and thank the participants in the PEAR study. We also acknowledge and thank the PEAR support staff, and study physicians: Drs. George Baramidze, R. Whit Curry, Karen Hall, Frederic Rabari-Oskoui, Dan Rubin, Siegfried Schmidt, and Michel Diab.