Gene–Education Interactions Identify Novel Blood Pressure Loci in the Framingham Heart Study

Abstract

BACKGROUND

Blood pressure (BP) variability has a genetic component, most of which has yet to be attributed to specific variants. One promising strategy for gene discovery is analysis of interactions between single-nucleotide polymorphisms (SNPs) and BP-related factors, including age, sex, and body mass index (BMI). Educational attainment, a marker for socioeconomic status, has effects on both BP and BMI.

METHODS

We investigated SNP–education interaction effects on BP in genome-wide data on 3,836 subjects in families from the Framingham Heart Study. The ABEL suite was used to adjust for age, sex, BMI, medication use, and kinship and to perform 1 degree-of-freedrom (df) and 2 df SNP–education interaction tests.

RESULTS

An SNP in PTN was associated with increased systolic BP (5.4mm Hg per minor allele) in those without a bachelor’s degree but decreased systolic BP (1.6mm Hg per allele) in those with a bachelor’s degree (2 df; P = 2.08×10^–8). An SNP in TOX2 was associated with increased diastolic BP (DBP; 4.1mm Hg per minor allele) in those with no more educational attainment than high school but decreased DBP in those with education past high school (−0.7; 1 df; P = 3.74×10^–8). Three suggestive associations were also found: in MYO16 (pulse pressure: 2 df; P = 2.89×10^–7), in HAS2 (DBP: 1 df; P = 1.41×10^–7), and in DLEU2 (DBP: 2 df; P = 1.93×10^–7). All 5 genes are related to BP, including roles in vasodilation and angiogenesis for PTN and TOX2.

CONCLUSIONS

PTN and TOX2 are associated with BP. Analyzing SNP–education interactions may detect novel associations. Education may be a surrogate for unmeasured exposures and behaviors modifying SNP effects on BP.

Elevated blood pressure (BP) is a major risk factor for cardiovascular disease and renal failure and was the leading global health burden in 2010,¹ annually costing tens of billions of dollars in the United States alone. Evidence suggests that systolic BP (SBP) and diastolic BP (DBP) have a heritability of at least 30%, but genetic variants identified thus far explain <3% of the total phenotypic variability.² Several explanations have been offered for this so-called missing heritability,³ and diverse research strategies addressing these explanations have emerged to pursue it. In addition to meta-analyses in enormous genome-wide association study consortia,² investigators have focused on sequence data and the role of rare variants,⁴ as well as subphenotypic distinctions within hypertension⁵ (e.g., degree of salt sensitivity and of sympathetic nervous activation). Another recently expanding field explores the role of interactions amongst single nucleotide polymorphisms (SNPs)⁶ and between SNPs and other factors known to influence BP such as age,⁷ sex,⁸ and body mass index (BMI).⁸ Motivated by the observation that these variables are directly correlated with BP themselves, these studies of SNP–covariable interactions test the hypothesis that these covariables also modulate the effect of SNPs on BP and have reported compelling evidence supporting that hypothesis.

One such variable just beginning to receive attention for its role in BP is educational attainment. Noting that educational attainment is inversely associated with cardiovascular disease in developed countries,⁹ Loucks et al. tested BP as an explanatory mechanism in the Framingham Heart Study. They reported that educational attainment (≤12 years vs. ≥17 years) was also inversely correlated with BP, both as a baseline effect and as a change in SBP over a 30-year life span (the latter particularly in female subjects).¹⁰ The contrast between those with limited literacy and those with adequate literacy showed a complex effect on SBP, with the direction of effect varying depending on the health-care system used: those with limited literacy had lower SBP in the Veterans Affairs healthcare system but higher in the university healthcare system.¹¹ A study of low-income rural Mexican women showed no association between BP and educational attainment, although the study reported a negative association between BP and 3 other socioeconomic indicators: household income, housing, and assets.¹² Variance component modeling in the Strong Heart Family Study also indicated that educational attainment has a modifying effect on the contribution of SNPs to BP variance, with evidence suggesting distinct genetic effects on DBP among those with less than high school education as compared with those with at least high school education.¹³ Although these studies provide evidence of an association between educational attainment and BP, to our knowledge this study is the first to report associations arising from SNP–education interactions on BP.

METHODS

Subjects

The data were drawn from the Framingham SNP Health Association Resource (SHARe) from the database of Genotypes and Phenotypes (dbGaP; http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000342.v8.p8). The Framingham Heart Study is a longitudinal study with 3 cohorts: the Original cohort, the Offspring cohort, and the Third Generation cohort. The Original cohort was recruited in 1948. The Offspring cohort was recruited in 1971, consisting of the offspring of the Original cohort, as well as the spouses and children of those offspring. The biological and adopted offspring of the Offspring cohort form the Third Generation cohort, recruited in 2002. The analysis data were taken from the Third Generation cohort’s single visit because only this cohort provided the educational attainment data used in our analysis.

Phenotypes and genotypes

Three measurements (1 nurse/technician reading and 2 physician readings) were taken for each subject for SBP and DBP, and the averages of these 3 values were used for the analysis; mean arterial pressure (MAP; calculated as two-thirds of average DBP plus one-third of average SBP) and pulse pressure (PP; calculated as average SBP minus average DBP) were derived.

Subjects were genotyped on the GeneChip Human Mapping 500k Array Set (Affymetrix, Santa Clara, CA) containing 487,998 SNPs. Eight thousand four hundred seventy-seven subjects (across all 3 cohorts) had an additional 2.5 million SNPs imputed using MACH.¹⁴ These SNPs were filtered by the absence of Mendelian errors, minor allele count ≥30 among all 3 cohorts in the most recent visit, Hardy–Weinberg equilibrium (P ≥ 10^–6), and good quality for imputed SNPs (r ² ≥ 0.3).

All subjects had nonmissing genotype and imputed data, as well as complete data for SBP, DBP, both education covariables, age, sex, BMI, and hypertension medication use; this limited the analysis dataset to 3,386 Third Generation cohort subjects. Two education covariables were derived from the raw Framingham Heart Study education variable. This raw variable contains 9 categories: 0 = no schooling; 1 = grades 1–8; 2 = grades 9–12; 3 = high school graduate; 4 = some college; 5 = technical school; 6 = associate’s degree; 7 = bachelor’s degree; 8 = graduate/professional degree. We derived Comp_HS+ as a dichotomous variable distinguishing those with no more than high school education (categories 0–3) from those with at least some college (categories 4–8), and Comp_BA as a dichotomous variable distinguishing those without a bachelor’s degree (categories 0–6) from those with a bachelor’s degree (categories 7–8). Descriptive statistics for this sample are shown in Table 1.

Table 1.

Study sample descriptive statistics

Trait	Value
Sample size	3,836
% Comp_HS+	84.7%
% Comp_BA	53.2%
% Male sex	47%
Age	40.2±8.8
BMI	26.9±5.5
% Taking antihypertensive meds	8.6%
SBP	116.3±13.7
DBP	75.7±9.3
MAP	89.2±10.1
PP	40.6±8.9

Data are mean ± SD unless otherwise stated.

Abbreviations: BMI, body mass index; Comp_BA, dichotomous variable distinguishing those without a bachelor’s degree from those with a bachelor’s degree; Comp_HS+, dichotomous variable distinguishing those with no more than high school education from those with at least some college; DBP, diastolic blood pressure; MAP, mean arterial pressure; PP, pulse pressure; SBP, systolic blood pressure.

Analysis

For each combination of BP trait and education covariable, we used the following linear mixed model:

where X_cov refers to the covariables age, age squared, sex, BMI, and use of hypertension medication, and z_polygene is a random effect to account for family relationships using the kinship matrix. Using this model, we performed the 1 degree-of-freedom (df) test of the SNP–education interaction (β₃), and the 2 df joint test of the SNP main effect and SNP–education interaction (β₂ and β₃).¹⁵ These analyses were performed using the ABEL suite, including GenABEL (c 1.6–6), ProbABEL (v 0.1-9e), and MixABEL (v 0.0–9).^16,17 MixABEL is one of few packages that can provide the 2 df joint test for family data; we also found the ABEL suite to be faster than other options, which was an important practical consideration for this analysis of approximately 2.5 million imputed SNPs on multiple combinations of phenotype and education covariable.

For both the 1 df and 2 df SNP–education interaction tests, the genomic inflation factor λ was calculated as the ratio of the observed median χ² value to the expected median χ² value. None of the analyses yielding significant or suggestive loci exhibited substantial evidence of inflation, with λ values in the range 1.02–1.08 (Table 2), as expected for highly polygenic traits.¹⁸ Figure 1a–d show QQ plots for these analyses.

Table 2.

Genomic control λ values

Covariable (test)	SBP	DBP	MAP	PP
Comp_HS+ (1 df)	1.990	1.052	1.098	1.076
Comp_HS+ (2 df)	1.132	1.071	1.079	1.070
Comp_BA (1 df)	1.018	1.032	1.014	1.018
Comp_BA (2 df)	1.043	1.061	1.036	1.038

Abbreviations: Comp_BA, dichotomous variable distinguishing those without a bachelor’s degree from those with a bachelor’s degree; Comp_HS+, dichotomous variable distinguishing those with no more than high school education from those with at least some college; DBP, diastolic blood pressure; df, degrees of freedom; MAP, mean arterial pressure; PP, pulse pressure; SBP, systolic blood pressure.

Figure 1.

QQ plots for selected combinations of blood pressure phenotype, education covariable, and interaction test. (a) QQ plot for joint 2 degrees-of-freedom (df) test of single nucleotide polymorphism (SNP) main effect and interaction of SNP with a dichotomous variable distinguishing those without a bachelor’s degree from those with a bachelor’s degree (Comp_BA) for systolic blood pressure (SBP). (b) QQ plot for 1 df test of interaction of SNP with a dichotomous variable distinguishing those with no more than high school education from those with at least some college (Comp_HS+) for diastolic blood pressure (DBP). (c) QQ plot for 1 df test of SNP-Comp_HS+ interaction for pulse pressure (PP). (d) QQ plot for 2 df test of SNP main effect and SNP-Comp_BA interactions for DBP.

RESULTS

Figure 2a–d show Manhattan plots for the indicated combination of BP trait (SBP, DBP, or PP), educational attainment covariable (Comp_HS+ or Comp_BA), and test (1 df interaction or 2 df joint test). There were 2 loci reaching significance at α = 5×10^–8 and 3 suggestive loci with P < 5×10^–7. These SNPs are summarized in Table 3 (significant loci) and Table 4 (suggestive loci).

Figure 2.

Manhattan plots for selected combinations of blood pressure phenotype, education covariable, and interaction test. (a) Manhattan plot of the joint 2 degrees-of-freedom (df) test of the singlue nucleotide polymorphism (SNP) main effect and interaction of SNP with a dichotomous variable distinguishing those without a bachelor’s degree from those with a bachelor’s degree (Comp_BA) for systolic blood pressure (SBP). The –log₁₀(P value) is plotted by genomic position. One novel locus reaching significance is located on chromosome 7 (PTN). (b) Manhattan plot of the 1 df test of interaction of SNP with a dichotomous variable distinguishing those with no more than high school education from those with at least some college (Comp_HS+) for diastolic blood pressure (DBP). The –log₁₀(P value) is plotted by genomic position. One novel locus reaching significance is located on chromosome 20 (TOX2), and a suggestive locus is located on chromosome 8 (HAS2). (c) Manhattan plot of joint 1 df test of the SNP main effect and SNP-Comp_HS+ interaction for pulse pressure (PP). The –log₁₀(P value) is plotted by genomic position. A suggestive locus is located on chromosome 13 (MYO16). (d) Manhattan plot of joint 2 df test of the SNP main effect and SNP-Comp_BA interaction for DBP. The –log₁₀(P value) is plotted by genomic position. A suggestive locus is located on chromosome 13 (DLEU2).

Table 3.

Significant single nucleotide polymorphism–education interactions

SNP rs no.	Chr	Location	Type	MAF	Trait	Covariable	Beta (Main)	SE (Main)	Beta (Int)	SE (Int)	P value 1 df interaction	P value 2 df interaction
rs12113527	7	PTN intron	Imp	0.044	SBP	Comp_BA	−5.40	0.98	7.05	1.29	4.28E-08	2.86E-08
rs12536916	7	PTN intron	Imp	0.044	SBP	Comp_BA	−5.30	0.98	6.97	1.29	5.83E-08	4.61E-08
rs6979228	7	PTN intron	Imp	0.044	SBP	Comp_BA	−5.30	0.98	6.97	1.28	5.81E-08	4.58E-08
rs7792298	7	PTN intron	Gen	0.047	SBP	Comp_BA	5.50	0.99	−7.18	1.29	2.90E-08	2.08E-08
rs11086907	20	TOX2 intron	Gen	0.102	DBP	Comp_HS+	4.12	0.81	−4.81	0.87	3.74E-08	2.61E-07

Bolded values are those with P < 5.00E-08.

Abbreviations: Comp_BA, dichotomous variable distinguishing those without a bachelor’s degree from those with a bachelor’s degree; Comp_HS+, dichotomous variable distinguishing those with no more than high school education from those with at least some college; DBP, diastolic blood pressure; df, degrees of freedom; Gen, genotyped; Int, interaction effect; Imp, imputed; MAF, minor allele frequency; Main, main effect; SBP, systolic blood pressure; SNP, single nucleotide polymorphism.

Table 4.

Suggestive single nucleotide polymorphism–education interactions

SNP rs no.	Chromosome	Location	Type	MAF	Trait	Covariable	Beta (Main)	SE (Main)	Beta (Int)	SE (Int)	P value 1 df interaction	P value 2 df interaction
rs4771625	13	MYO6 intron	Imp	0.37	PP	Comp_HS+	0.23	0.05	−0.29	0.05	7.61E-08	2.89E-07
rs13274659	8	Near HAS2	Imp	0.49	DBP	Comp_HS+	−2.65	0.54	3.01	0.58	2.04E-07	1.22E-06
rs1994404	8	Near HAS2	Imp	0.49	DBP	Comp_HS+	2.65	0.54	−3.05	0.58	1.41E-07	9.11E-07
rs2171348	8	Near HAS2	Imp	0.5	DBP	Comp_HS+	−2.63	0.55	3.03	0.59	2.84E-07	1.79E-06
rs6990539	8	Near HAS2	Imp	0.49	DBP	Comp_HS+	−2.65	0.54	3.05	0.58	1.43E-07	9.21E-07
rs6991390	8	Near HAS2	Imp	0.49	DBP	Comp_HS+	−2.65	0.54	3.05	0.58	1.43E-07	9.19E-07
rs7839349	8	Near HAS2	Imp	0.49	DBP	Comp_HS+	−2.72	0.55	3.09	0.60	2.16E-07	1.28E-06
rs12100048	13	DLEU2 intron	Imp	0.12	DBP	Comp_BA	−2.36	0.43	2.01	0.58	5.60E-04	1.93E-07
rs17073826	13	DLEU2 intron	Imp	0.13	DBP	Comp_BA	−2.30	0.42	1.88	0.57	1.04E-03	2.41E-07
rs17073834	13	DLEU2 intron	Imp	0.13	DBP	Comp_BA	−2.27	0.42	−1.82	0.57	1.45E-03	2.86E-07
rs6561575	13	DLEU2 intron	Imp	0.13	DBP	Comp_BA	2.31	0.42	−1.92	0.58	8.53E-04	2.26E-07
rs7981648	13	DLEU2 intron	Imp	0.13	DBP	Comp_BA	−2.30	0.42	1.89	0.57	9.92E-04	2.43E-07

Abbreviations: Comp_BA, dichotomous variable distinguishing those without a bachelor’s degree from those with a bachelor’s degree; Comp_HS+, dichotomous variable distinguishing those with no more than high school education from those with at least some college; DBP, diastolic blood pressure; df, degrees of freedom; Int, interaction effect; Imp, imputed; MAF, minor allele frequency; Main, main effect; PP, pulse pressure; SNP, single nucleotide polymorphism.

The most significant interaction (2 df; P = 2.08×10^–8) was between rs7792298 and the Comp_BA covariable on SBP (Figure 2a). This is one of several SNPs in the chromosome 7 locus with similar effects, minor allele frequencies, and P values (see Table 3). Subjects with 1 minor allele and no bachelor’s degree showed a large increase in average adjusted SBP (5.5mm Hg) relative to those without a minor allele, whereas those with a bachelor’s degree and a minor allele showed a decrease (1.7mm Hg) (Figure 3a). All of the significant SNPs at this locus are located in intron 4 of pleiotrophin (PTN).

Figure 3.

Mean blood pressure by minor allele count (MAC) for selected combinations of blood pressure phenotype and SNP. (a) Mean systolic blood pressure (SBP) by PTN genotype and a dichotomous variable distinguishing those without a bachelor’s degree from those with a bachelor’s degree (Comp_BA). (b) Mean diastolic blood pressure (DBP) by TOX2 genotype and a dichotomous variable distinguishing those with no more than high school education from those with at least some college (Comp_HS+).

A single SNP in an intron of TOX high mobility group box family member 2 (TOX2) on chromosome 20 also reached significance for interaction with Comp_HS+ on DBP (1 df; P = 3.74×10^–8) (Figure 2b). Subjects with no education beyond high school showed an increase in DBP of 4.1mm Hg for each minor allele of this genotyped SNP (rs11086907), whereas those with some post–high school education and 1 or 2 minor alleles showed a modest decrease in DBP (0.7mm Hg per allele) (Figure 3b).

Three other loci showed suggestive association with a BP trait (Table 4). An SNP in an intron of myosin XVI (MYO16) on chromosome 13 had an interaction with Comp_HS+ associated with PP. Pulse pressure was higher by 0.2mm Hg per minor allele among those with no more than high school education but lower by 0.1mm Hg per minor allele among those with more than high school education (1 df; P = 7.61×10^–8) (Figure 2c). Several SNPs on chromosome 8 near hyaluronan synthase 2 (HAS2) had interactions with Comp_HS+ suggestively associated with DBP. The strongest association was with rs1994404, which showed an increase of 2.6mm Hg DBP per minor allele for those with no more than high school education and a decrease of 0.4mm Hg DBP per minor allele for those with at least some college education (1 df; P = 1.41×10^–7) (Figure 2b). Finally, several SNPs on chromosome 13 near deleted in lymphocytic leukemia 2 (DLEU2) had interactions with Comp_BA suggestively associated with DBP. The strongest association was with rs12100048, which showed an increase of 2.4mm Hg per minor allele for those without a bachelor’s degree and a decrease of 0.6mm Hg per minor allele for those with a bachelor’s degree (2 df; P = 1.93×10^–7) (Figure 2d).

One additional SNP, rs6936075, yielded a significant interaction with Comp_BA associated with SBP (2 df; P = 2.8×10^–8) (Figure 2a). This SNP had a minor allele frequency of 0.16%, barely surviving the cutoff based on minor allele count in all 3 cohorts in the most recent visit (see Methods). In the analysis sample (Third Generation cohort only), the total number of minor alleles present in the sample dropped to <30. Because no other SNP supported its association with P < 5×10^–6, this locus was not further considered.

DISCUSSION

The effects of interactions between SNPs and educational attainment on SBP, DBP, MAP, and PP were analyzed. Two loci showed significant interactions: intronic SNPs in pleiotrophin (PTN) and Comp_BA showed significant interaction effects on SBP using either the 1 or 2 df tests (Table 3) and an intronic SNP in TOX2 and Comp_HS+ showed a significant interaction effect on DBP. Loci in 3 genes also suggested interaction effects (Table 4): MYO16 and Comp_HS+ affecting PP, HAS2 and Comp_HS+ affecting DBP, and DLEU2 and Comp_BA affecting DBP.

Previous analyses of the Framingham Heart Study based on SNP main effects did not show significant associations at the PTN locus associated with SBP in this study.¹⁹ Modest associations (2×10^–5 < P < 6×10^–5) with other SNPs in PTN have been reported in dbGaP for several related phenotypes, including high-density lipoprotein cholesterol, BMI, and coronary artery disease,²⁰ but they were all based on main effects only. This growth factor has numerous biological functions, including roles in angiogenesis and kidney development, and it critically regulates both the renin-angiotensin system and catecholamine synthesis in mice.²¹ PTN is expressed in vascularized atherosclerotic plaques and mediates migration of endothelial cells following upregulation by the vasodilator nitric oxide.²² It is also involved in inflammation and has been shown to induce expression of inflammatory cytokines including tumor necrosis factor alpha, interleukin 1 beta, and interleukin 6 in peripheral blood mononuclear cells.²³ Collectively, these studies point to a multifaceted role for PTN in cardiovascular and metabolic physiology with effects on inflammation and BMI, as well as cholesterol. Our results extend this role to BP in particular. In addition, the interaction with educational attainment may reflect PTN’s role in learning and memory, as it has been shown to inhibit long-term potentiation (a crucial neural mechanism of memory) in the hippocampus.²⁴

The second genome-wide significant locus contains a single SNP (Table 3) located in an intron of TOX2 coding a DNA-binding protein about which relatively little is known. As with the PTN locus, the SNP–Comp_HS+ interaction reached significance in the absence of significant associations of SNP main effects at this locus in previous analyses of the Framingham Heart Study.¹⁹ The expression of several hundred genes was affected by small interfering RNA–mediated knockdown of TOX2 in 2 cell lines normally expressing TOX2.²⁵ These genes represented multiple pathways, including vasodilation, blood clotting, angiogenesis, and inflammatory response.²⁵ A meta-analysis of the Family Blood Pressure Program reported linkage evidence with BP for which the confidence interval (logarithm of the odds ratio ≥ 2; approximately 10Mb wide) included TOX2.²⁶ In addition, an association between TOX2 SNPs and conduct disorder²⁷ and a suggestive association with alcoholism²⁸ have been reported, identifying possible mechanisms linking BP regulation with low educational attainment.

The suggestive interaction effect involving MYO16 on PP (Table 4) is supported by previous association results. Zhang et al.²⁹ reported an association with PP in Chinese twins (P = 7.21×10^–7). Associations with other cardiovascular and metabolic traits, including high-density lipoprotein cholesterol, diabetic nephropathies, and coronary artery disease, have also been reported in dbGaP.³⁰ In addition, a suggestive association (P = 3.3×10^–6) has been reported with urinary free dopamine in a Hispanic population.³¹ Expression analysis contrasting 3 strains of spontaneously hypertensive rats with Wistar Kyoto rats also implicated MYO16 in hypertension.³² MYO16 is expressed predominantly in the central nervous system, as well as in the kidney, and collectively these results suggest a role for MYO16 in BP, possibly through involvement in sympathetic nervous activity, which regulates both vasodilation and renal sodium transport. In addition to its role in cardiovascular and metabolic traits, MYO16 has also been implicated in brain development, neurological morphogenesis, and autism,³³ traits which may be associated with educational attainment.

The chromosome 8 locus suggestive of an interaction effect on DBP (Table 4) is located in an intron of HAS2, an important member of the extracellular matrix in atherosclerotic vascular lesions,³⁴ which provides a framework for blood vessel growth. The Framingham Heart Study linked HAS2 with microalbuminuria,³⁵ a risk factor for cardiac and vascular abnormalities, and a SNP near HAS2 has been associated with survival among female black diabetic dialysis patients.³⁶ Rat studies have shown a role for hyaluronan in renal water handling, and response to water deprivation and an F2 intercross (between normotensive and spontaneously hypertensive rats) showed a linkage peak for both MAP and left ventricular mass that contained HAS2.³⁷

The final suggestive locus contains both DLEU1 and DLEU2 (Table 4), 2 noncoding RNAs that harbor 2 microRNAs (MIR15A and MIR16-1). Little is known about these genes, and most studies targeting them are focused on cancers (the name refers to their being frequently downregulated in lymphocytic leukemia). However, epigenetic regulation of DLEU2 has been linked with downregulation of a group of genes regulating the nuclear factor kappa B (NF-κB) complex, a core mediator of inflammation signaling.³⁸ Recent unpublished work (of J. J. Basson) suggests a possible association between BP and SNP–SNP interactions between NF-κB genes and members of the NF-κB–inhibiting IκK complex. Another major inflammatory mediator, transforming growth factor β, is regulated by miRNA-16, which also regulates the renal sodium channel SCNN1B and a serotonin transporter.³⁹ Serotonin is widely expressed in the brain, heart, kidney, and vasculature, organs which play a key role in BP regulation. SNPs in this locus have also been found to have suggestive associations with bipolar disorder;⁴⁰ expression levels of miRNA-16 have been found to be dysregulated in brains from schizophrenic individuals.⁴¹ Once again, a common theme has appeared whereby genes associated with BP regulation are also found to be associated with neuropsychiatric disorders.

SNP interactions with educational attainment have thus revealed 2 significant and 3 suggestive loci associated with BP, none of which were previously implicated in BP by SNP main effects in the Framingham Heart Study or any other study. Although the identified loci are in or near genes not previously associated with BP, biological plausibility, in some cases substantial, provides physiological connection to BP regulation. In this study, as in others, educational attainment likely serves as a surrogate for other unmeasured innate and environmental exposures and behaviors, including neuropsychiatric disorders, smoking, alcohol consumption, physical activity, and stress;⁴² thus, the specific mechanisms of the identified SNP–education interactions remain uncertain. Although further work is necessary to validate these findings and elucidate their mechanisms, these data provide evidence that educational attainment (or other factors for which educational attainment serves as a surrogate) modifies the effect of SNPs on BP and that consideration of SNP–education interactions can improve gene discovery.

DISCLOSURE

The authors declared no conflict of interest.