Novel DNA methylation sites associated with cigarette smoking among African Americans

ABSTRACT

Introduction: Cigarette smoking has been associated with adverse health outcomes for mothers and children and is a major contributor to heart disease. Although cigarette smoking is known to affect the epigenome, few studies have been done in African American populations. In this study, we investigated the association between cigarette smoking and DNA methylation (DNAm) among African Americans from the Intergenerational Impact of Genetic and Psychological Factors on Blood Pressure Study (InterGEN), and the Genetic Epidemiology Network of Arteriopathy (GENOA).

Methods: The InterGEN study aims to examine the effects of genetic and psychological factors on blood pressure among African American women and their children. Current cigarette smoking was assessed at baseline. DNAm of saliva was assessed using the 850K EPIC Illumina BeadChip for Epigenome-Wide Association analyses. A replication study was conducted among 1100 participants in the GENOA study using the same BeadChip.

Results: After controlling for age, body mass index, population structure and cell composition, 26 epigenome-wide significant sites (FDR q < 0.05) were identified, including the AHRR and PHF14 genes associated with atherosclerosis and lung disease, respectively. Six novel CpG sites were discovered in the InterGEN sample and replicated in the GENOA sample. Genes mapped include RARA, FSIP1, ALPP, PIK3R5, KIAA0087, and MGAT3, which were largely associated with cancer development.

Conclusion: We observed significant epigenetic associations between smoking and disease-associated genes (e.g., cardiovascular disease, lung cancer). Six novel CpG sites were identified and replicated across saliva and blood samples.

Cigarette smoking is associated with cancer, cardiovascular and respiratory diseases, and is the leading preventable cause of death worldwide [1]. Nearly 30% of African American (AA) adults smoke [2]. Despite initiating smoking at later ages and smoking fewer cigarettes on average than their Caucasian counterparts, AAs are more likely to die of smoking-related diseases [3,4]. AA adults have higher rates of high blood pressure [5], adverse reproductive outcomes [6], and chronic kidney disease [7] than those in other racial/ethnic groups.

There is accumulating evidence that both genetic and epigenetic mechanisms are at play in the relationship between cigarette smoking and disease [8]. DNA methylation (DNAm) is the most frequently studied epigenetic mechanism and refers to the addition or absence of methyl groups to genes, often but not always, resulting in suppression of gene expression. Epigenome-wide association studies (EWAS) have demonstrated a link between methylation of specific genes and cardiovascular disease outcomes. In the MESA study, for example, the association between smoking, AHRR methylation, and carotid plaque scores were investigated in 1256 participants (51% women, 21% AA adults) [9]. The authors concluded that AHRR methylation may be a potential biomarker for subclinical atherosclerosis in smokers [9]. Others have also shown that smoking is associated with methylation of other genes (such as F2RL3) associated with poor cardiovascular disease prognosis [10]. Although the epigenetic consequences of smoking on DNAm are well-documented, few studies have examined these questions in an all AA cohort using non-blood tissues. This is important as AA women especially represent a demographic group who bear a significant burden of smoking-related morbidity and mortality [3,4]. Therefore, the study of the epigenetic effects of smoking has significant potential for future translational research affecting myriad public health problems including cardiovascular disease, asthma, adverse birth outcomes, and cancers among AAs. In this study, we investigated the association between cigarette smoking and DNA methylation (DNAm) among AAs from the Intergenerational Impact of Genetic and Psychological Factors on Blood Pressure Study (InterGEN), and the Genetic Epidemiology Network of Arteriopathy (GENOA) Study.

Results

Of the 159 women in the discovery sample, three had missing data for smoking, leaving N = 156 for the EWAS analysis. Approximately 20% of women were current smokers, and the mean age of participants was 31 years old (Table 1). In the EWAS analysis, DNAm at 26 autosomal CpG sites were significantly associated with current smoking status after adjustment for age and BMI (FDR corrected p < 0.05) (Figures 1 and 2). No X-chromosome site was significantly associated with current smoking from this analysis. The smoking-associated CpG sites were mapped to known genes, including AHRR, GFI1, EDC3, RARA, SNED1, GNG12, GNG12-AS1, FSIP1, ALPP, XXYLT1, MAFG, PIK3R5, PIK3R, KIAA0087, LINC00299, TACC1, SH2D4A, TKT, and MGAT3.

Table 1.

Participant characteristics, InterGEN study (n = 156).

	Smokers	Non-smokers	Total Sample
N (%)	32 (20.5)	124 (79.5)	156
Maternal Age in years (mean)	30.4	31.9	31.6
Education
High school or less	17 (53.1)	47 (37.9)	64 (41.0)
Some college or Associates degree	14 (43.8)	53 (42.7)	67 (42.9)
Bachelor’s degree or higher	1 (3.1)	24 (19.4)	25 (16.0)
Income
Less than $15,000/year	24 (75)	47 (40.2)	71 (47.7)
$15,000/per year or greater	8 (25)	70 (59.8)	78 (52.3)
BMI Category
Underweight	2 (6.3)	7 (5.6)	9 (5.8)
Normal	6 (18.75)	33 (26.6)	39 (25)
Overweight	9 (28.1)	35 (28.2)	44 (28.2)
Obese	15 (46.9)	49 (39.5)	64 (41.0)

Figure 1.

QQ plot of EPIC 850K current smoking and DNA methylation in women enrolled in the InterGEN study (Inflation Factor = 1.11).

Figure 2.

Manhattan Plot (Red line: FDR corrected p < 0.05).

Most (18 out of 26) CpG sites were found to be hypomethylated among current smokers (Table 2). We compared these findings with previously reported CpG sites that were associated with cigarette smoking from two most recent blood-based EWAS on smoking [11,12] as well as a systematic review that summarized published evidence across 14 earlier EWAS of smoking exposure [13]. Twenty of 26 CpG sites identified in saliva were consistent with the blood-based findings from either previous reports or the replication study (GENOA). Compared to previously reported smoking-associated CpG sites, 12 of 26 CpG sites identified in our analysis were non-overlapping and of these, 8 were available only on the EPIC (850K) BeadChip. A gene-based look-up revealed one gene, PHF14, that was not reported in any previous epigenome-wide smoking studies in blood-based samples in humans. Six CpG sites were novel and replicated in both the discovery InterGEN sample, and in the replication GENOA sample. These sites were mapped to genes including RARA, FSIP1, ALPP, PIK3R5, KIAA0087, and MGAT3.

Table 2.

Discovery and replication of epigenome-wide significant associations (FDR-q < 0.05) with current smoking.

						Discovery (N = 156)				Replication (N = 1100)
CpG	CHR	POS	850K Only	Location	Gene	Beta	SE	P	FDR-q	Beta	SE	P	Genomic Distance	Category
cg17739917	17	38477572	Yes	INT	RARA	−0.063	0.010	1.29E-08	0.001	−0.122	0.006	5.81E-68	1548	Novel and replicated
cg14051805	15	39917220	Yes	INT	FSIP1	−0.053	0.009	1.04E-07	0.006	−0.013	0.002	1.98E-10	26521
cg12956751	2	233246922	Yes	EXO	ALPP	−0.017	0.003	3.15E-07	0.018	−0.029	0.003	7.36E-24	328
cg22996023	17	8803989	Yes	INT	PIK3R5	−0.042	0.008	7.73E-07	0.035	−0.022	0.005	3.48E-06	290
cg07741821	7	26577897	Yes	INT	KIAA0087	−0.037	0.007	8.27E-07	0.035	−0.054	0.004	8.09E-42	201
cg05086879	22	39861490	Yes	INT	MGAT3	−0.022	0.004	1.35E-06	0.044	−0.029	0.002	3.75E-48	8628
cg05644151	7	11035483	No	INT	PHF14	0.011	0.002	3.31E-08	0.003	1.97E-04	4.44E-04	0.658	257189	Novel but not replicated
cg27174698	2	129244189	No			0.011	0.002	6.21E-08	0.004	−2.52E-04	4.73E-04	0.595	7114
cg07439098	3	194878552	Yes	INT	XXYLT1	−0.023	0.004	3.42E-07	0.018	−0.001	0.002	0.681	42250
cg07824483	17	79882042	No	INT	MAFG	−0.039	0.007	7.05E-07	0.033	−0.006	0.003	0.052	499
cg22063959	8	19188603	Yes	INT	SH2D4A	−0.012	0.002	1.19E-06	0.044	−1.81E-05	2.13E-04	0.932	16477
cg00748718	3	53283691	No	INT	TKT	−0.016	0.003	1.30E-06	0.044	−3.23E-04	0.001	0.827	20631
cg05575921	5	373378	No	INT	AHRR	−0.152	0.020	2.61E-12	2.23E-06	−0.154	0.004	2.46E-141	0	Previously reported
cg01940273	2	233284934	No			−0.059	0.008	6.48E-12	2.76E-06	−0.083	0.004	2.52E-67	0
cg21566642	2	233284661	No			−0.090	0.013	3.59E-11	1.02E-05	−0.107	0.005	2.81E-72	0
cg09935388	1	92947588	No	INT	GFI1	−0.080	0.012	4.53E-10	9.65E-05	−0.065	0.003	3.44E-61	0
cg14389122	15	74945851	No	INT	EDC3	−0.027	0.004	1.31E-09	2.23E-04	−0.010	0.002	6.23E-10	0
cg27241845	2	233250370	No			−0.031	0.005	4.50E-09	0.001	−0.029	0.003	3.40E-17	0
cg26703534	5	377358	No	INT	AHRR	−0.039	0.006	7.07E-09	0.001	−0.046	0.002	6.42E-72	0
cg14753356	6	30720108	No			−0.062	0.011	3.52E-08	0.003	−0.040	0.004	2.05E-20	0
cg16937168	2	241936844	No	PRO	SNED1	−0.033	0.006	4.39E-08	0.003	−0.020	0.003	8.40E-11	0
cg25189904	1	68299493	No	INT	GNG12	−0.064	0.011	6.46E-08	0.004	−0.064	0.007	2.42E-19	0
cg00073090	19	1265879	No			−0.034	0.007	4.02E-07	0.020	−0.027	0.003	1.14E-16	0
cg23079012	2	8343710	No	INT	LINC00299	−0.031	0.006	9.73E-07	0.040	−0.006	0.001	5.10E-26	0
cg19695041	8	38615330	No	INT	TACC1	−0.027	0.005	1.19E-06	0.044	−0.002	0.003	0.525	0
cg25648203	5	395444	No	INT	AHRR	−0.027	0.005	1.23E-06	0.044	−0.028	0.002	9.83E-54	0

CHR: chromosome; POS: base-pair position; 850K only: whether the CpG site is only available on EPIC 850K array; Location: Epigenetic properties of the location, i.e., PRO: Promoter; EXO: Exon; INT: Intron; Gene: Nearest gene; Beta: mean change in DNAm; SE: Standard error; P: p-value; FDR-q: False discovery rate q-value; Genomic distance: base-pair distance to the closest smoking-associated CpG site reported previously. Figures 1 & 2 attached in TIFF format.

Discussion

In this preliminary study, we identified significant associations between DNAm of 26 CpG sites and current smoking among African American women enrolled in the InterGEN study. Importantly, in this study, we identified six novel sites, which were replicated across tissues, using the 850k newer chip design. Several well-documented findings were replicated in this study, including DNAm of the AHRR gene on chromosome 5, which has previously been associated with both adult and maternal smoking [14–17]. The aryl-hydrocarbon receptor (AhR) pathway and the methylation of the related AHRR gene has been proposed as a potential biomarker of cardiovascular disease, specifically atherosclerosis, in smokers [9]. We also report the novel finding that smoking is associated with a DNAm change in an intron of PHF14, about 20kb downstream of the transcription start site. This gene has not been associated with human disease previously, however, in mouse models, increased methylation of Phf14 was linked to current smoking [18].

The six sites that were novel and replicated in GENOA were mapped to RARA, FSIP1, ALPP, PIK3R5, KIAA0087, and MGAT3. Five out of the six sites located at introns of the corresponding genes, and one site (cg12956751) was mapped to the 3ʹ UTR of an exon of ALPP. None of the sites are located near known enhancers (Table 2), though enhancer activation by tobacco smoke has been associated with hypomethylation and disease outcomes [19]. RARA, Retinoic Acid Receptor Alpha, is a protein-coding gene that has been implicated in the development of hematological and solid tumor malignancies, including lung cancer [20]. Fibrous Sheath Interacting Protein 1 (FSIP1) is also a protein coding gene, and increased mRNA expression of this gene has been associated with breast tumors [21]. FSIP1 has also been associated with poorer prognosis in lung cancers [22]. ALPP (Alkaline Phosphatase, Placental) is primarily expressed in endometrial and placental tissue and has also been associated with ovarian cancers [23]. PIK3R5 is also a protein-coding gene, known as Phosphoinositide-3-Kinase Regulatory Subunit 5. It is involved in cell growth, proliferation, and differentiation, and some classes of its protein are involved in cancer development [24]. KIAA0087 is a non-coding RNA gene associated with endometrial and pancreatic cancers [25]. Also, a non-coding RNA gene, MGAT3 (Mannosyl (Beta-1,4-)-Glycoprotein Beta-1,4-N-Acetylglucosaminyltransferase) is involved in the function and biosynthesis of glycoprotein oligosaccharides. No known disorders are associated with this gene [26]. Four CpG sites which were available both on the 450K array and 850K array were identified in our discovery cohort using saliva samples but could not be replicated in blood tissue from either our replication cohort, or any of the previous blood-based smoking EWAS. However, one gene – PHF14, has never been reported in earlier studies, which indicates the possibility of a unique profile of smoking-DNAm association in saliva comparing to blood tissue. Examination of the functional consequences of the six replicated genes is necessary and recommended for future studies.

Few studies have examined epigenome-wide effects of smoking in African American samples. Sun and colleagues [27] conducted an analysis of 972 African Americans in GENOA, with replication in 239 African Americans in the Grady Trauma Project (GTP). In this methylome-wide analysis of peripheral blood samples, five DNAm sites were identified and replicated. Dogan and colleagues [28] examined smoking and genome-wide methylation in 111 African American women (mean age = 48) enrolled in the FACHS study from rural Iowa and reported that long-term smokers had altered methylation of genes related to inflammation, immune function, and coagulation. Peripheral blood was collected in FACHS to isolate mononuclear cells for methylation analysis as well as inflammatory markers such as C-reactive protein and interleukin-6 receptor levels, and the genes that were significantly methylated differed from those identified in the current study [28]. Philibert and colleagues [29] studied smoking and methylation in 107 African American men (mean age = 19) and reported methylation of AHRR. They also reported methylation of AHRR in a second study of 399 young African Americans (mean age = 22) with relatively short histories of smoking [30].

The findings in the present study may differ from the previous work described above for several reasons. First, those studies used a different tissue type (peripheral blood) than that used in InterGEN (saliva). However, a substantial portion of cells in saliva are leukocytes, which are the major cell types in the blood. As a result, 50% of smoking-associated DNAm sites discovered in saliva (InterGEN) were replicated in blood (GENOA). Second, most studies used the 450K array [28–30], and one used the 27K and the 450K for replication [27] to assess methylation, while InterGEN used the 850K array. Third, the populations under study differed in age, which may have affected the outcome under study. Finally, smoking was assessed in most studies via self-report [27–29], as in our study, however, one study also verified smoking by measuring serum cotinine [30]. The epigenetic associations with smoking status that we identified in the African American cohort need to be further replicated in independent samples and validated by other technologies (e.g., pyrosequencing) in future studies.

These findings are significant for several reasons. First, the InterGEN sample used the EPIC 850K chip which provides the deepest coverage of DNAm profile comparing to previous studies. Second, we identified novel smoking-associated DNAm sites in saliva and conducted cross-tissue replication in blood samples using the same array. The high level of replication confirmed that our findings were not merely false positives. Third, the novel but unreplicated sites are still important findings, as not all will be captured in blood or epithelial cells. Future saliva-based smoking EWAS can be conducted to confirm these findings. Those that did replicate provide confidence that our analysis demonstrates shared saliva and blood significant sites. Fourth, two of the studies [29,30] did not control for cell type. Finally, the direction of the findings is also significant, as all 6 novel and replicated genes were hypomethylated, which is a characteristic of cancer and may result in overexpression of genes. This finding merits further study as it may be clinically relevant to explain health disparities related to smoking and increased risk of disease development in African Americans.

The example of APOL1 variants [31] and other gene-environment interactions [32] highlight relevant genetic differences between people of different ancestry, and differential epigenetic responses to environmental insults which may explain some of these health disparities. This underscores the need for future studies examining the effects of smoking within racial/ethnic groups.

The major strength of this study is the 850K EWAS analysis of smoking in an understudied population. Despite the limitation of a small sample size, significant associations were identified. Our findings were strengthened by a replication analysis in the GENOA cohort. Though these data were replicated, they are preliminary, and further validation is necessary in future studies to explore the effects on transcription. Future studies may benefit from newer approaches such as multi-omic designs which examine both epigenetic associations and metabolic pathways, as genes and potential pathways related to cigarette smoking have been identified [33] which can improve understanding of biochemical mechanisms.

Methods

Discovery cohort

The InterGEN study is an ongoing, longitudinal cohort study in Connecticut that examines the effects of genetic, epigenetic, and psychological factors on blood pressure in mother/child dyads. Recruitment began in April 2015 and is ongoing through 2018. Eligibility criteria include: mothers (≥21 years old) who self-identify as AA or Black, speak English, and have no mental illness that could interfere with psychological measurements. Women must enroll with a biological child (3–5 years old) who lives with her most of the time and can spit to provide saliva sample. More information on the cohort and psychological measures can be found in previous reports [34]. During the baseline study visit, study personnel take clinical measurements of blood pressure, height, and weight and collect saliva for DNA analysis from both mother and child. Demographic information, health history and psychological measures (including parenting, experiences of perceived racism and discrimination, and depression) are collected through mother’s report using Audio Computer-Assisted Self-Interview (ACASI) software. Self-reported smoking status was assessed at baseline, and participants were categorized as either current or non-current (former or never) smokers. Yale University and New York University’s Institutional Review Board reviewed and approved the study procedures. Data are available via written request to the InterGEN study investigators.

DNA methylation

Researchers collected saliva samples for DNA using the Oragene (OG)-500 format tubes [35], which requires participants to spit into the tube until the contents reach the fill line (2 mL). Detailed DNA collection and analysis procedures have been described elsewhere [36]. Samples were transported from the field to the research laboratory where they were refrigerated at 4ºC until DNA extraction and analysis were completed. Standard protocol for DNA extraction and purification was conducted as indicated in the standard operating procedures guidelines using ReliaPrep kits, and the Illumina Infinium Methylation EPIC (850K) BeadChip was used for epigenome-wide DNAm measurement. Quantile-normalization of beta values for autosomal CpG sites was performed. All individual samples passed laboratory-based quality-control procedures (missing rate < 10% and no sex mismatch). CpG sites were excluded if they had detection p-value greater than 0.01 (n = 71), had a missing rate greater than 10% (n = 514), overlapped with SNPs (n = 15,341), or were listed in the recent Illumina product quality notice. Quality control procedures and all analyses were uniformly performed among autosomal and X-chromosome sites due to a unisex (all female) sample. A total of 831,219 autosomal and 18,895 X-chromosomal CpG sites were included in the association analyses as previously described [37].

Statistical analysis

We conducted an EWAS of smoking (current vs. non-current) among women enrolled in the InterGEN study. DNA methylation beta-value was modeled as the dependent variable using multiple linear regression. Age was controlled for as a potential confounder. We also adjusted for batch effects and potential heterogeneity in cell proportions from saliva using the reference-free EWAS method [38]. False discovery rate (FDR) was used to correct for multiple comparisons. We applied the Benjamini and Hochberg method to calculate the FDR-corrected p-value [39]. CpG sites with FDR-corrected p-value < 0.05 were considered as statistically significant. All statistical analyses were performed in the R statistical environment version 3.4.1 (http://www.r-project.org/).

Replication study

Replication of these findings was conducting using a sample of 1100 participants in the GENOA study using the Illumina Infinium Methylation EPIC (850K) BeadChip.

GENOA sample: The Genetic Epidemiology Network of Arteriopathy (GENOA) is a multi-phase, community-based, prospective study of sibships with two or more siblings diagnosed with primary hypertension before the age of 60. Participants self-identified as Black, additional siblings were invited to participate regardless of hypertension status, and recruitment took place in Jackson, Mississippi area. A total of N = 1,854 AA participants were recruited in Phase I (1995–2000) from 683 sibships, and in Phase II, N = 1,482 returned (2000–2005). Most participants (71%) in GENOA were women. In each phase, demographics, medical history, clinical characteristics, lifestyle factors, and fasting blood samples were collected [40]. In Phase I, DNA methylation was measured from peripheral blood samples, and smoking status was also assessed. Current smoker was defined as anyone who had smoked cigarettes within the past year. Institutional Review Board approval was received for the GENOA study protocol through the University of Mississippi Medical Center and the University of Michigan IRBs.

Genomic DNA was extracted from stored peripheral blood leukocytes that was collected at Phase 1 (N = 1106) or Phase 2 (N = 304) using AutoGen FlexStar (AutoGen, Holliston, MA). Bisulfite conversion was performed with the EZ DNA Methylation Kit (Zymo Research, Irvine, CA), and methylation was assessed using the Illumina 850K BeadChip. After obtaining the raw intensity data, the shinyMethyl R package [41] was used to generate the density plot to identify sex mismatch or sample outliers. Sample identity was further checked using 59 SNP probes implemented in the EPIC chip and mismatched samples were removed. Samples with incomplete bisulfite conversion identified using the QCinfo() function in the ENmix R package [42] were also removed. The Minfi R package was used to perform background correction and normalization [43]. The regression on correlated probes (RCP) method [44] was used to adjust probe-type bias, and ComBat [45] was used to remove confounding effects due to batch or other technical covariates. Principle variance component analysis (PVCA) [46] was performed to quantify the variance explained by the known batch variables before and after batch adjustment to make sure no single batch factors explain more than 3% variance of the data. Detection p-value for each sample at each probe was obtained, and individual probes with detection p-value <10⁻¹⁶ were considered successfully detected. Samples and probes with detection rate<10% were removed. Finally, 857121 CpG sites for 1100 samples at Phase 1 and 294 samples at Phase 2 were available for analysis. The methylation data from 1100 Phase 1 samples were used for this analysis. Cell counts were estimated using Houseman method [38].

DNA methylation data are available with a Data Use Agreement upon reasonable request to the InterGEN study investigators, J.Y. Taylor (jackie.taylor@nyu.edu) and Y.V. Sun (yan.v.sun@emory.edu). Data availability for the GENOA study is as follows: GENOA genotype and phenotype data are available through the Database of Genotypes and Phenotypes (dbGaP accession number phs001238.v1.p1). DNA methylation data are available with a Data Use Agreement upon reasonable request to the study investigators, S. Kardia (skardia@umich.edu) and J. Smith (smjenn@umich.edu).

Funding Statement

This work was supported by the National Heart, Lung, and Blood Institute [HL100185];National Heart, Lung, and Blood Institute [HL133221];National Heart, Lung, and Blood Institute [HL054457];National Heart, Lung, and Blood Institute [T32HL130025];National Heart, Lung, and Blood Institute [HL087660];National Heart, Lung, and Blood Institute [HL119443];National Institute of Nursing Research [K01NR017010];National Institute of Nursing Research [R01NR013520].

Acknowledgments

This work was supported by the National Institute for Nursing Research under grant R01NR013520 (The Intergenerational Impact of Genetic and Psychological Factors on Blood Pressure Study (InterGEN) study); the National Heart, Lung and Blood Institute under grants HL054457, HL100185, HL087660, HL119443, and HL133221 (The Genetic Epidemiology Network of Arteriopathy (GENOA) study); the National Institute for Nursing Research under grant K01NR017010; and the National Heart, Lung, and Blood Institute under grant T32HL130025.

Disclosure statement

No potential conflict of interest was reported by the authors.