Influence of smoking status and intensity on discovery of blood pressure loci through gene-smoking interactions

Abstract

Background

Genetic variation accounts for approximately 30% of blood pressure (BP) variability but most of that variability hasn't been attributed to specific variants. Interactions between genes and BP-associated factors may explain some ‘missing heritability.’ Cigarette smoking increases BP after short-term exposure and decreases BP with longer exposure. Gene-smoking interactions have discovered novel BP loci, but the contribution of smoking status and intensity to gene discovery is unknown.

Methods

We analyzed gene-smoking intensity interactions for association with systolic BP (SBP) in three subgroups from the Framingham Heart Study: current smokers only (N = 1,057), current and former smokers (‘ever smokers’, N = 3,374), and all subjects (N = 6,710). We used three smoking intensity variables defined at cutoffs of 10, 15, and 20 cigarettes per day (CPD). We evaluated the 1 degree-of-freedom (df) interaction and 2df joint test using generalized estimating equations.

Results

Analysis of current smokers using a CPD cutoff of 10 produced two loci associated with SBP. The rs9399633 minor allele was associated with increased SBP (5 mmHg) in heavy smokers (CPD>10) but decreased SBP (7 mmHg) in light smokers (CPD≤10). The rs11717948 minor allele was associated with decreased SBP (8 mmHg) in light smokers but decreased SBP (2 mmHg) in heavy smokers. Across all nine analyses, 19 additional loci reached p < 1×10⁻⁶.

Discussion

Analysis of current smokers may have the highest power to detect gene-smoking interactions, despite the reduced sample size. Associations of loci near SASH1 and KLHL6/KLHL24 with SBP may be modulated by tobacco smoking.

INTRODUCTION

High blood pressure (BP) is a major risk factor for cardiovascular disease and was the single biggest health burden globally in 2010 (Lim, SS et al. 2012). Although the heritability of BP is estimated to be approximately 30%, the specific genes and variants responsible have proven elusive. Genome-wide association studies have identified dozens of variants associated with BP, but they collectively explain less than 3% of BP variability (Ehret, GB 2010). Among the factors hypothesized to contribute to this ‘missing heritability’ are interactions between genes and other clinical factors (Manolio, TA et al. 2009) known to influence BP, including age (Shi, G et al. 2009), sex (Ramirez-Lorca, R et al. 2007), BMI (Ramirez-Lorca, R et al. 2007), alcohol consumption (Simino, J et al. 2013), and diet (Zhang, RR et al. 2013).

Cigarette smoking is an environmental exposure that is an important risk factor for cardiovascular disease as well as a variety of other diseases including cancer and lung disease. Smoking has long been known to influence BP directly (Benowitz, NL et al. 1984) and is recently receiving attention for its role in modifying the influence of genetic variants on BP (Sung, YJ et al. 2014). The effects of smoking on BP are complex; epidemiologic studies have shown acute increases in BP but decreased BP with longer tobacco exposure (Green, MS et al. 1986).

Given the influence of tobacco smoking on BP, gene-smoking interactions may enable the detection of novel BP-associated variants. However, the complexity of the relationship between smoking and BP raises questions regarding the methods for investigating gene-smoking interactions that are most likely to contribute to novel gene discovery. In this study, we examined gene-smoking interactions by models that categorized subjects by two different exposure metrics: smoking status, with subjects self-identifying as current smokers, former smokers, or never smokers; and smoking intensity, characterized as the reported number of cigarettes smoked per day (CPD). To investigate the influence these variables have on gene discovery, we performed a genome-wide analysis of gene-smoking interactions in three population subgroups (current smokers only, current and former smokers, and all subjects) defining dichotomous smoking intensity variables based on three different CPD cutoffs (> 10, > 15, and > 20 CPD).

METHODS

Subjects

The analysis used data 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.v12.p9). The Framingham Heart Study is a longitudinal study comprising three Caucasian cohorts: the Original cohort, recruited in 1948; the Offspring cohort, recruited in 1971 comprised of the offspring of the Original cohort and their spouses and children; and the Third Generation cohort, recruited in 2002 and consisting of the biological and adopted offspring of the Offspring cohort. The current study analyzed data from the lone clinic visit of the Third Generation cohort as well as the Original cohort visit and Offspring cohort visit most closely matching the date of that visit.

All subjects analyzed had non-missing genotype and imputed genotype data and complete data for SBP, cigarettes per day (CPD), age, sex, and anti-hypertension medication use. Three sets of subjects were analyzed. The All Subjects sample (N = 6,710) consisted of all subjects with complete data regardless of smoking status. The Ever Smokers sample (N = 3,374) consisted of subjects who were current or former smokers but excluded those who had never smoked. Finally, the Current Smokers sample (N = 1,057) consisted only of current smokers. Descriptive statistics for these samples are shown in Table I.

Table I.

Descriptive Statistics of analysis cohorts.

	All Subjects	Ever Smokers	Current Smokers
Sample Size	6,710	3,374	1,057
CPD10, % (n)	32.7 (2,192)	65.0 (2,192)	65.1 (688)
CPD15, % (n)	27.8 (1,868)	55.4 (1,868)	51.4 (543)
CPD20, % (n)	10.6 (710)	21.0 (710)	15.6 (165)
Never Smokers, % (n)	49.7 (3,336)	NA	NA
Age, years	49.1 ± 13.5	51.6 ± 12.9	46.0 ± 12.2
Male, %	46.5 (3,123)	47.0 (1,586)	47.9 (506)
Taking Anti-hypertensive meds, %	19.1 (1,283)	22.4 (757)	13.8 (146)
SBP, mm Hg	123 ± 19	125 ± 20	121 ± 18

Data presented as % (n) or means ± standard deviation

Phenotypes and Genotypes

The analysis phenotype was the average of three systolic blood pressure (SBP) readings, (one taken by a nurse/technician and two taken by a physician). In subjects taking anti-hypertensive medication, the SBP value was adjusted by adding 15 mmHg.

Three dichotomous smoking covariates were derived from the continuous CPD data using different CPD cutoffs: CPD10, CPD15, and CPD20. Subjects have a dichotomous covariate value of 1 if their CPD value is greater than the cutoff. For example, CPD10 has a value of 1 when CPD > 10. For both the All Subjects and Ever Smokers samples, dichotomous CPD covariates were derived in former smokers using the reported average CPD value during the time they were smoking.

Genotyping was performed on the GeneChip Human Mapping 500K Array Set (Affymetrix, Santa Clara, CA) which assays 487,998 SNPs. A total of 2.5 million SNPs were imputed based on HapMap data using MACH (Li, Y et al. 2010). For all analyses, SNPs were excluded for Mendelian errors, Hardy-Weinberg disequilibrium (p ≤ 10⁻⁶), or poor imputation quality (r² < 0.3). In addition, only SNPs with minor allele frequency (MAF) ≥ 5% (calculated separately for each sample) were analyzed. The resulting SNP set included 2,144,020 SNPs in the All Subjects sample, 2,143,542 SNPs in the Ever Smokers sample, and 2,142,057 SNPs in the Current Smokers sample.

Analysis

The analyses were performed in a generalized estimating equations (GEE) framework using the R package geepack (available through CRAN). This approach uses robust standard errors which help reduce the inflation (and type I error) that can burden analysis of interactions (Voorman, A et al. 2011). The model we analyzed (using CPD10 in this example) is:

$$y={\beta }_0+{{\beta }_1}^{\ast }Age+{{\beta }_2}^{\ast }Sex+{{\beta }_3}^{\ast }CPD10+{{\beta }_4}^{\ast }SNP+{{\beta }_5}^{\ast }SN{P}^{\ast }CPD10+e$$

with each SNP coded using an additive model (0,1,2) and subjects clustered in families. Two tests were performed based on this model: a 1 degree-of-freedom (df) test of the SNP*CPD interactions (β₅) and the 2 df joint test of the SNP main effect and SNP*CPD interaction (β ₄, β₅), and a genome-wide significance threshold of α=5×10⁻⁸ was used.

RESULTS

A total of nine analyses were performed, one for each combination of sample (All Subjects, Ever Smokers, and Current Smokers) and CPD cutoff (CPD10, CPD15, CPD20). One of the goals of this investigation was to determine which combination of the sample and CPD cutoff are promising for discovery in large consortia. For each analysis, the genomic inflation factor (λ, the ratio of the observed median χ² value to the expected median χ² value) for the 1df and 2df tests is shown in Table II (main effect λ values are shown in Supplementary Table I). Nine QQ plots showing both the 1df and the 2df tests are shown in Figure 1. The observed λ values are largely in line with the expectation for highly polygenic traits (Yang, J et al. 2011), although the 2df test shows evidence of moderate inflation, particularly in the All Subjects sample. This inflation is driven by the main effects (Supplementary Table I), while the corresponding 1df tests are well controlled (Table II).

Table II.

Genomic Inflation λ of the SNP*CPD interaction, 1df and 2df.

Sample	1df λ			2df λ
	CPD10	CPD15	CPD20	CPD10	CPD15	CPD20
All Subjects	1.04	1.03	1.01	1.18	1.17	1.16
Ever Smokers	1.03	1.02	1.01	1.09	1.09	1.08
Current Smokers	1.05	1.06	1.04	1.06	1.06	1.05

Table II shows the genomic inflation λ value for the 1df test for each combination of sample and CPD cutoff.

Figure 1. QQ plots by sample and CPD cutoff.

Figure 1 shows the QQ plots for each of the 9 analyses with the indicated combination of sample composition and CPD covariate. The 1df test is shown in red and 2df test is shown in blue. The genomic control lambda values for both tests are inset. The majority of analyses are well controlled, showing modest inflation. The primary signal of interest (−log(P) > 7) is located in the analysis of the Current Smokers group using CPD10 and comes from both 1df and 2df tests.

Across all nine analyses, two loci contained SNPs that reached genome-wide significance using p-values adjusted for genomic inflation (α = 5×10⁻⁸; Table III). Both loci were detected in the Current Smokers CPD10 analysis. For this analysis, both the 1df and 2df tests gave rise to a strong association signal (inflation-adjusted 1df: p = 1.0×10⁻⁸; 2df: p = 1.8×10⁻⁸) on chromosome 6 with multiple SNPs reaching genome-wide significance in a locus located 150 – 200 kb upstream of the nearest gene, SASH1. Among subjects who smoked ≤ 10 cigarettes per day, each C allele at the most significant SNP in this locus (rs9399633, MAF = 14.6%, Table III) was associated with an average SBP decrease of about 5 mmHg, while in subjects who smoked > 10 cigarettes per day, each C allele was associated with an average SBP increase of 5 mmHg (Figure 2).

Table III.

Significant Associations

Sample	CPD cutoff	SNP	Major/minor allele	CHR	BP	MAF	BETA SNP	BETA INT	P SNPGC	P INTGC	P 2dfGC	Gene	Distance (kb)
Current	10	rs9399633	C/T	6	148062948	0.146	5.07	−9.6	1.42E-06	1.00E-08	1.83E-08	SASH1	149
Current	10	rs1534774	C/G	6	148034416	0.121	5.53	−9.8	4.06E-07	7.64E-08	2.94E-08	SASH1	178
Current	10	rs9390526	T/C	6	148030680	0.121	−5.5	9.78	4.07E-07	7.72E-08	2.96E-08	SASH1	181
Current	10	rs1534775	A/G	6	148035906	0.121	5.53	−9.8	4.10E-07	8.07E-08	3.06E-08	SASH1	176
Current	10	rs1358689	T/C	6	148012824	0.121	−5.5	9.74	4.07E-07	8.54E-08	3.16E-08	SASH1	199
Current	10	rs9386204	A/G	6	148013180	0.121	5.54	−9.7	4.06E-07	8.65E-08	3.19E-08	SASH1	199
Current	10	rs1404488	G/A	6	148038927	0.121	−5.5	9.73	4.17E-07	8.90E-08	3.30E-08	SASH1	173
Current	10	rs1881827	A/G	6	148020538	0.121	5.53	−9.7	4.06E-07	1.07E-07	3.66E-08	SASH1	192
Current	10	rs9377097	A/G	6	148019979	0.121	5.53	−9.7	4.06E-07	1.07E-07	3.66E-08	SASH1	192
Current	10	rs1881826	G/A	6	148020622	0.121	−5.5	9.66	4.07E-07	1.07E-07	3.66E-08	SASH1	191
Current	10	rs11717948	G/A	3	183591640	0.0547	−9	7.04	1.47E-08	8.34E-03	5.03E-08*	KLHL6 KLHL24	36
Current	10	rs9872196	C/A	3	183592150	0.0546	−9.3	7.42	1.37E-08	7.08E-03	5.07E-08*	KLHL6 KLHL24	36
Current	10	rs11706582	T/C	3	183591975	0.0552	−9.2	7.24	1.46E-08	8.34E-03	5.09E-08*	KLHL6 KLHL24	36

Figure 2. Mean SBP by SASH1 genotype and CPD10 status in Current Smokers.

Figure 2 shows the average SBP for groups of subjects defined by their genotype at rs9399633 and their CPD10 status. Subjects with CPD ≤ 10 are shown in blue, and subjects with CPD > 10 are shown in red. Error bars indicate standard error for each mean SBP value.

A second genome-wide significant locus was detected with the 2df test (inflation-adjusted p = 5.0×10⁻⁸) on chromosome 3, with KLHL6 and KLHL24 located about 40 kb on either side of the associated SNPs. Among subjects who smoked ≤ 10 cigarettes per day, each G allele at the most significant SNP in the locus (rs11717948 , MAF = 5.4%, Table III) was associated with an average SBP decrease of about 9 mmHg, while among subjects who smoked > 10 cigarettes per day, G alleles were associated with an average SBP decrease of only about 2 mmHg (Figure 3). These genome-wide significant association results are shown in Table III.

Figure 3. Mean SBP by KLHL6/KLHL24 genotype and CPD10 status.

Figure 2 shows the average SBP for groups of subjects defined by their genotype at rs11717948 and their CPD10 status. Subjects with CPD ≤ 10 are shown in blue, and subjects with CPD > 10 are shown in red. Error bars indicate standard error for each mean SBP value.

A total of 19 other loci across all analyses gave rise to suggestive associations (inflation-adjusted p < 10⁻⁶), with the majority (13) also identified from the analysis of the Current Smokers sample: one suggestive locus was detected using a CPD cutoff of 10, one locus was detected using a cutoff of 15, and 11 suggestive loci were detected using a cutoff of 20 (Table IV). Of the six suggestively associated loci detected in other samples, four were detected in the All Subjects sample (two from CPD15 and two from CPD20 analyses) and two were detected in the Ever Smokers sample (one from both CPD10 and CPD15 and one from CPD20 analyses, Table IV).

Table IV.

Suggestive Associations

Sample	CPD cutoff	SNP	Non-coded/coded allele	CHR	BP	MAF	Beta Main	Beta 1df Intxn	P SNP Main (Genomic Control)	P 1df Intxn (Genomic Control)	P 2df Intxn (Genomic Control)	Gene	Distance(kb)
All	15	rs6916459	A/C	6	54734919	0.104	−1.07	5.75	7.91E-02	9.58E-07	2.38E-05	FAM83B	112
All	15	rs12662500	G/C	6	54730930	0.104	1.07	−5.74	7.92E-02	9.64E-07	2.39E-05	FAM83B	116
All	15	rs12469933	T/G	2	139266368	0.102	−4.46	4.28	2.58E-07	4.59E-03	3.82E-07	NXPH2/LRP1B	486
All	20	rs4761750	G/C	12	93928698	0.0742	−2.11	12.7	2.42E-02	2.41E-07	6.60E-06	CRADD	33
All	20	rs12372697	A/G	12	93935745	0.0742	2.1	−12.7	2.43E-02	2.41E-07	6.61E-06	CRADD	40
All	20	rs11107292	A/G	12	93946957	0.0742	2.1	−12.8	2.48E-02	3.22E-07	8.14E-06	CRADD	51
All	20	rs1229982	T/G	4	99322775	0.189	1.59	3.38	4.06E-04	6.79E-03	4.00E-07	ADH1B	intron
Ever	10	rs2355654	C/T	14	49988369	0.36	−3.95	1.54	9.59E-06	1.73E-01	2.56E-07	C14orf182	intron
Ever	15	rs2355654	C/T	14	49988369	0.36	−3.73	1.42	6.68E-06	2.06E-01	5.16E-07	C14orf182	intron
Ever	20	rs12699566	T/C	7	13934836	0.269	0.421	6.75	5.60E-01	5.56E-06	9.31E-07	ETV1	intron
Current	10	rs6934331	T/A	6	148044773	0.132	4.96	−9.12	5.50E-06	1.77E-07	2.60E-07	SASH1	167
Current	10	rs2328843	C/T	6	148047081	0.132	−4.96	9.1	5.54E-06	1.85E-07	2.69E-07	SASH1	165
Current	10	rs9322120	G/A	6	148047385	0.132	4.96	−9.1	5.54E-06	1.87E-07	2.71E-07	SASH1	165
Current	10	rs1973858	T/C	6	148049408	0.132	4.95	−9.08	5.63E-06	2.03E-07	2.88E-07	SASH1	163
Current	10	rs7746768	C/G	6	148056613	0.132	−4.95	9.06	5.71E-06	2.23E-07	3.09E-07	SASH1	156
Current	10	rs351326	A/C	6	93402467	0.0806	−4.31	12.8	5.95E-03	9.16E-07	6.78E-06	EPHA7	intron
Current	15	rs6454802	T/C	6	90104480	0.392	1.83	−7.71	9.57E-02	5.58E-07	1.25E-07	BACH2	intron
Current	15	rs12202119	T/A	6	90105471	0.392	1.82	−7.68	9.65E-02	5.64E-07	1.25E-07	BACH2	intron
Current	15	rs11753332	A/G	6	90109434	0.392	−1.8	7.64	9.81E-02	5.71E-07	1.25E-07	BACH2	intron
Current	15	rs12194007	T/G	6	90113440	0.392	1.79	−7.6	9.93E-02	5.79E-07	1.26E-07	BACH2	intron
Current	15	rs12196749	A/G	6	90110096	0.442	−1.64	7.32	1.35E-01	2.10E-06	5.25E-07	BACH2	intron
Current	20	rs9471517	C/T	6	12341521	0.163	3.62	−13.3	2.33E-04	1.04E-06	8.32E-07	ARG1/MED23	12342
Current	20	rs9493000	C/T	6	131487451	0.0603	−0.115	−13.5	9.40E-01	3.83E-06	3.45E-07	ARG1/MED23	131487
Current	20	rs7740929	G/A	6	131487482	0.0514	0.524	11.4	7.40E-01	3.01E-05	6.69E-07	ARG1/MED23	131487
Current	20	rs9492992	T/G	6	131481287	0.0514	0.543	11.4	7.30E-01	3.22E-05	7.34E-07	ARG1/MED23	131481
Current	20	rs12481734	C/G	21	23699432	0.0967	1.48	−20.4	5.39E-01	2.02E-06	7.11E-07	D21S2088E	23699
Current	20	rs17154431	A/G	5	126342485	0.0668	4.83	−15.9	9.17E-03	7.05E-07	4.78E-06	GRAMD3	126342
Current	20	rs901508	T/C	5	2006893	0.0943	−3.64	16.1	2.95E-02	7.84E-08	3.84E-07	IRX4	2007
Current	20	rs4594252	C/A	16	55286934	0.102	−1.11	15	3.02E-01	1.66E-07	7.71E-07	IRX6	55287
Current	20	rs7973791	C/A	12	91050948	0.0568	−5.53	17.3	2.84E-03	3.64E-07	2.56E-06	KERA	3′ UTR
Current	20	rs7959834	C/T	12	91050279	0.0568	5.53	−17.3	2.84E-03	3.71E-07	2.61E-06	KERA	91050
Current	20	rs12228056	T/C	12	91036196	0.0568	−5.53	17.2	2.85E-03	4.42E-07	3.07E-06	KERA	91036
Current	20	rs3798221	T/G	6	160577116	0.188	−2.32	11.5	2.37E-02	6.92E-07	4.93E-06	LPA	intron
Current	20	rs12033204	A/C	1	212274784	0.0555	−2.06	19.1	3.35E-01	1.01E-06	8.87E-07	PPP2R5A	212275
Current	20	rs16991403	T/C	20	1092170	0.0596	1.67	14.6	3.74E-01	6.47E-05	1.00E-06	PSMF1	1092
Current	20	rs9899604	C/T	17	71414692	0.0549	4.45	−17.3	7.31E-03	2.71E-07	2.00E-06	SOX9	71415

Manhattan plots (using inflation-adjusted p-values) for the analysis of the Current Smokers sample with CPD10 are shown in Figure 4 (1df test) and Figure 5 (2df test). Manhattan plots for the remaining eight analyses are in the supplementary materials.

Figure 4. Manhattan plot for Current Smokers sample, CPD10, 1df test.

Figure 4 shows the manhattan plot for the analysis of the Current Smokers sample using the CPD10 interacting covariate. The –log(P) value adjusted for genomic inflation for the 1df test is shown. A significant locus is located on chromosome 6 (−log(P) = 8.0) 149 kb from SASH1. A suggestively associated locus is also located on chromosome 6 (−log(P) = 6.0) in an intron of EPHA7.

Figure 5. Manhattan plot for Current Smokers sample, CPD10, 2df test.

Figure 4 shows the manhattan plot for the analysis of the Current Smokers sample using the CPD10 interacting covariate. The –log(P) value adjusted for genomic inflation for the 2df test is shown. A significantly associated locus is located on chromosome 6 (−log(P) = 7.7) 149 kb from SASH1. A second significantly associated locus is located on chromosome 7 (−log(P) = 7.3) about 36 kb from the genes KLHL6 and KLHL24.

DISCUSSION

We analyzed the interaction effects of smoking intensity (i.e, cigarettes per day [CPD]) and genetic variants on SBP in three different smoking exposure cohorts (current smokers; current and former smokers; and current, former, and never smokers) using three different CPD interacting covariates (defined at CPD cutoffs of 10, 15, and 20). These nine analyses yielded two loci with genome-wide significance and 19 suggestive (inflation-adjusted p < 10⁻⁶) associations. Both of the significant loci were based on the analysis of the Current Smokers sample and a CPD cutoff of 10. This finding has important implications for the design of future studies analyzing the effects of smoking-SNP interactions on blood pressure. The other two samples, All Subjects and Ever Smokers, are approximately 6.3 and 3.2 times larger than the Current Smokers sample, respectively (Table I). The observation that they produced no genome-wide significant findings compared to the two significant findings from the Current Smokers sample (as well as far fewer suggestive associations) in spite of these much larger sample sizes suggests that the genetic modulation of smoking effects on blood pressure diminishes after a subject stops smoking. As a consequence, analyses of current smokers may be prioritized for discovery of gene-smoking interactions on BP, even if that means a substantial reduction in sample size relative to a sample containing subjects who have never smoked or who have quit smoking.

These data are less conclusive with respect to the CPD cutoff that optimizes gene discovery. While the most significant results were obtained using a cutoff of 10 cigarettes per day, the majority of the results with genomic-controlled p < 10⁻⁶ were obtained using a cutoff of 20. The majority of these suggestive associations were detected in the Current Smokers sample, however, which contained only 165 subjects who smoked more than 20 cigarettes per day. Although the minor allele frequency cutoff of 5% helped ensure that at least 16 copies of the minor allele were present in the heavier smoking group, this value still allows for substantial instability in the regression estimates. As the majority of these suggestive associations were found for SNPs having MAF between 5% and 10%, these results should be treated with caution.

Based on these data, we conclude that current smokers revealed the strongest signals for gene-smoking interaction on BP, including two genome-wide significant associations. The strongest result was for rs9399633 (inflation-adjusted 1df p = 1.00 × 10⁻⁸, 2df p = 1.83 × 10⁻⁸, Table III), located about 150 kb upstream of SASH1 (SAM and SH3 domain containing 1). SASH1 is best known as a tumor suppressor gene for several types of cancer including colon (Nitsche, U et al. 2012), breast (Zeller, C et al. 2003), and melanoma (Lin, S et al. 2012). It is also located in a susceptibility locus for lung cancer (Bailey-Wilson, JE et al. 2004), and SASH1 over-expression reduces the viability, proliferation, and migration of lung cancer cells whereas targeted silencing of SASH1 by RNA interference has opposing effects (Chen, EG et al. 2012). At least three studies investigated the influence of tobacco exposure on gene expression in circulating blood cells found that SASH1 expression was significantly correlated with tobacco exposure (although the direction of the change varied across studies) (Beineke, P et al. 2012; Charles, PC et al. 2008; Verdugo, RA et al. 2013). This smoking-induced change in expression suggests a possible mechanism explaining the interaction between smoking and rs9399633. The effect of SASH1 on blood pressure may be mediated through its expression in endothelial cells of microvascular beds, where it contributes to activation of the NF-κB complex and proinflammatory cytokines (Dauphinee, SM et al. 2013). Motivated by the co-morbidity of inflammatory processes and hypertension in metabolic syndrome, a recent analysis of SNP-SNP interactions among inflammation genes for association with BP suggested a BP-regulatory role for interactions among members and regulators of the NF-κB complex (Basson, JJ et al. 2014). Furthermore, a suggestive association between SASH1 and diabetic nephropathy, a risk factor for hypertension, has been identified in a genome-wide association study of African Americans (Mcdonough, CW et al. 2011). Thus, several lines of evidence support the biologic plausibility that SASH1 plays a role in the modulating effect of tobacco on blood pressure.

The other genome-wide significant association with SBP was located on chromosome 3, with the strongest result (rs11717948, inflation-adjusted 2df p = 5.03 × 10⁻⁸, Table III) located approximately 36 kb downstream of KLHL6 and 44 kb upstream of KLHL24. Although relatively little is known about these genes, they belong to the Kelch-like gene family which contains several evolutionarily conserved domains and several of whose members are associated with Mendelian diseases (Dhanoa, BS et al. 2013); notably, rare mutations in KLHL3 alter renal ion transport and cause familial hyperkalemic hypertension (Louis-Dit-Picard, H et al. 2012). Expression of KLHL6 in CD133(+) cells has also been associated with coronary artery disease (Liu, Det al. 2011), and KLHL24 has been shown to regulate kainite receptors (Laezza, F et al. 2007), which are known to play a role in nicotine dependence (Kenny, PJ et al. 2003; Ma, JZ et al. 2010; Vink, JM et al. 2009). The 2df result at this locus is driven by the SNP main effect (inflation-adjusted main effect p = 1.47 × 10⁻⁸, 1df p = 8.34 × 10⁻³, Table III).

These results indicate that gene-smoking interactions can reveal novel BP loci and that a sample consisting exclusively of current smokers likely provides the greatest power to detect such interactions even when the sample size is considerably smaller than an alternative sample containing former smokers and those who have never smoked. In addition, these analyses implicate SASH1 as modulating the effect of smoking on BP.

Limitations

This study has several limitations. It only analyzed data from one ethnic group, Caucasians, thus limiting generalizability to other racial and ethnic groups. Smoking status and CPD were determined at the time of the visit without reference to the duration of smoking or smoking abstinence; therefore, subjects who recently quit were considered equal to non-smokers and new smokers were considered equal to life-long smokers. The CPD values analyzed were based on self-reports that averaged changing CPD values over time and may not be precise. Finally, although these hypotheses will be investigated in a large CHARGE Consortium (devoted to gene-lifestyle interactions in cardiovascular traits), our findings have not been replicated in external samples.