Cigarette Smoking Significantly Alters Sperm DNA Methylation Patterns

Abstract

Numerous health consequences of tobacco smoke exposure have been characterized, and smoking’s effects on traditional measures of male fertility are well described. However, a growing body of data indicates that pre-conception paternal smoking also confers increased risk for a number of morbidities on offspring. The mechanism for this increased risk has not been elucidated, but it is likely mediated, at least in part, through epigenetic modifications transmitted through sperm. In this study, we investigated the impact of cigarette smoke exposure on sperm DNA methylation patterns in 78 men who smoke and 78 never-smokers using the Infinium HumanMethylation450 beadchip. We investigated two models of DNA methylation alterations: (1) consistently altered methylation at specific CpGs or within specific genomic regions and (2) stochastic DNA methylation alterations manifest as increased variability in genome-wide methylation patterns in men who smoke. We identified 141 significantly differentially methylated CpGs associated with smoking. In addition, we identified a trend toward increased variance in methylation patterns genome-wide in sperm DNA from men who smoke compared with never-smokers. These findings of widespread DNA methylation alterations are consistent with the broad range of offspring heath disparities associated with pre-conception paternal smoke exposure and warrant further investigation to identify the specific mechanism by which sperm DNA methylation perturbation confers risk to offspring health and whether these changes can be transmitted to offspring and transgenerationally.

INTRODUCTION

In a recent study it was estimated that more than one third of the world’s population is regularly exposed to environmental tobacco smoke (ETS) (Oberg et al., 2011). Further, in 2004 and estimated 603,000 premature deaths and a loss of 10.9 million disability adjusted life years (DALYs) occurred due to involuntary exposure to tobacco smoke (Oberg et al., 2011), and an estimated six million annual deaths are attributable to tobacco smoke exposure. Tobacco smoke contains more than 4000 chemicals including a myriad of known carcinogens. The health consequences of smoke exposure are significant and include numerous diseases and dysfunctions of the respiratory tract, increased risk of multiple types of cancer and increased incidence of cardiovascular disease (DiFranza et al., 2004; Moritsugu, 2007). Clearly tobacco smoke exposure is a global problem, the implications of which are becoming increasingly apparent.

The negative impacts of tobacco smoke on semen parameters are well established. Smoking is associated with an accumulation of cadmium and lead in seminal plasma, reduced sperm count and motility, and fewer morphologically normal sperm (Kiziler et al., 2007; Kulikauskas et al., 1985). In addition, increased sperm chromatin structural abnormalities and sperm DNA damage, and reduced reproductive potential have been reported in tobacco smoke-exposed mice (Polyzos et al., 2009) and humans (Fuentes et al., 2010). Adult male mice exposed to sidestream tobacco smoke display significant increases in sperm DNA mutations at expanded simple tandem repeats (ESTRs) (Marchetti et al., 2011) and aberrations in sperm chromatin structure (Polyzos et al., 2009). In contrast, these male mice exhibit no measurable increase in somatic cell chromosome damage, indicating that germ cells may be more prone to genetic and/or epigenetic insults resulting from smoke exposure than somatic cells. Destabilization of ESTRs in the adult male germline is well documented to be associated with transgenerational effects in the mouse genome (R. Barber et al., 2002; R. C. Barber et al., 2006; Dubrova, 2003; Dubrova et al., 2008; Glen et al., 2012). Thus, these data suggest that smoke exposure might induce lasting genetic and epigenetic changes that could impact subsequent unexposed generations.

The International Association for Research on Cancer (IARC) has recently declared that paternal smoking prior to pregnancy is associated with a significantly elevated risk of leukemia in the offspring (Ji et al., 1997; Secretan et al., 2009) (Chang et al., 2006; K. M. Lee et al., 2009; Pang et al., 2003; Vine, 1996) suggesting smoke-induced genetic or epigenetic changes occur in sperm that are transmitted to offspring. Additionally, children of men who smoke are at increased risk for childhood cancers, asthma (Svanes et al., 2016) and birth defects including cleft palate, urethral stenosis, hydrocephalus (Savitz et al., 1991), congenital heart disease, cardiovascular anomalies (Cresci et al., 2011) anorectal malformations (Zwink et al., 2011), spina bifida (Zhang et al., 1992), and reduced kidney volume (Kooijman et al., 2015).

Smoking has clearly been shown to modify DNA methylation patterns and gene expression in somatic tissues in individuals exposed to first- or second-hand tobacco smoke (Bosse et al., 2012a; Kohli et al., 2012; Word et al., 2012) as well as in newborns of smoking mothers (Breton et al., 2009; Joubert et al., 2012; Perera et al., 2011). A growing body of evidence suggests that exposure could have negative health consequences not only for the exposed individual, but also for the descendants of those who are exposed. A complete understanding of the transgenerational effects of tobacco smoke exposure is of critical relevance to public health.

Given the huge proportion of the population that is exposed to tobacco smoke, either directly or indirectly, complete characterization of the health consequences of this exposure, not only on the exposed individuals, but also potentially on unexposed offspring and progeny is critical in informing health policy moving forward. The aim of the current study was to evaluate the effect of pre-conception tobacco smoke exposure on sperm DNA methylation as a potential mechanism for transmission of health risks to offspring.

MATERIALS AND METHODS

Semen samples were collected from patients being evaluated for couple infertility as well as from men from the general population recruited for research studies through community outreach. All participants gave informed consent for participation in this University of Utah Institutional Review Board-approved study. Semen samples were collected based on WHO criteria after a recommended 2–7 days of sexual abstinence. Individuals with a sperm concentration of less than 2 × 10⁶/ml were excluded from the study due to insufficient sperm DNA to perform the assay. To be included in the non-smoking group, and individual was required to have never smoked cigarettes. Minimum cigarette consumption for the smoking group was one pack year (defined as the number of cigarettes per day × # of years smoking/20). There were no other inclusion or exclusion criteria for the current study. Smoking status was assessed based on a questionnaire filled out by patients at the time of informed consent. Specifically, individuals were asked if they smoked cigarettes, and if so, how many daily and how long they had smoked. The smoking group and control group each consisted of 21 men from the general population and 57 men who presented to the Andrology Laboratory for evaluation for couple infertility. Semen analyses were performed according to WHO IV criteria. Following the diagnostic test, the remaining sample was mixed in a 1:1 ratio with test yolk buffer (TYB; Irvine Scientific, Irvine, CA) and frozen in 1 ml cryovials in liquid nitrogen vapors. Frozen vials were submerged and stored in liquid nitrogen until thawing for methylation analysis.

Descriptive Statistics

Descriptive statistics including mean, standard deviation and 95% confidence intervals were calculated for semen parameters and patient age. Normally distributed variables (age, semen volume, % progressive motility, % normal head morphology, % normal tail morphology, and % viable) were compared between groups using Student’s t-test, while non-parametric variables (sperm concentration, total sperm count, total progressively motile sperm count, somatic cell concentration, and comet score) were analyzed using the Kruskal-Wallis test.

Alkaline Comet assay

Sperm DNA damage was assessed using an alkaline single-cell gel electrophoresis (Comet) assay as modified previously (Donnelly et al., 1999; Hughes et al., 1997). Spermatozoa were considered damaged or normal based on the presence or absence of a visible Comet tail, respectively. A total of 50 to 100 Comets were scored per sample.

Methylation analysis

DNA Isolation

All frozen samples were thawed and processed simultaneously to avoid batch effects. Once thawed, sperm were subjected to a stringent somatic cell lysis protocol to ensure the absence of potential contamination resulting from the presence of leukocytes or other somatic cells. Briefly, samples were run through a 40um cell strainer to remove any large debris or cell clumps. The strained sample was washed in 14 ml of PBS followed by two washes in 14 ml of distilled water. The samples were then centrifuged, and the resulting pellet was resuspended and incubated for a minimum of 60 minutes a 4° C in 14 ml of a somatic cell lysis buffer (0.1% SDS, 0.5% Triton X-100 in DEPC H₂O). Following this lysis step, a visual check was performed under 400× magnification to confirm the absence of somatic cells, then sperm DNA was isolated using a sperm specific modification commonly used in our lab to a column-based extraction protocol using the DNeasy DNA isolation kit (Qiagen, Valencia CA)(Jenkins et al., 2013; Jenkins et al., 2014a). In addition to the visual check, a post hoc analysis of methylation data at the DLK1 locus, a region that is unmethylated in sperm and fully methylated in leukocytes, was evaluated to further confirm a pure sperm population as detailed previously (Jenkins et al., 2016).

Bisulfite Conversion, Array Processing, and Quality Control

Extracted sperm DNA was bisulfite converted with EZ-96 DNA Methylation-Gold kit (Zymo Research, Irvine CA) according to the manufacturer’s recommendations specifically for use with array platforms. Converted DNA was then delivered to the University of Utah Genomics Core Facility hybridized to Infinium HumanMethylation450 BeadChip microarrays (Illumina) and analyzed according to Illumina protocols. Array data for all samples were evaluated for standard data quality indicators, and two samples (both general population samples, one smoker and one non-smoker) were excluded from the study due to failure to meet established standards.

Methylation Data Processing

The Chip Analysis Methylation Pipeline (ChAMP) was used to process array data and generate β-values (fraction methylation values between 0 and 1). This process additionally includes the filtering of poorly performing probes from downstream analysis (probes with a QC p<0.05 were excluded) and SWAN normalization. Normalized β-values were then logit transformed to generate m-values for all further downstream analyses.

Global Methylation Analysis

Global methylation analysis was conducted by averaging β-values across all probes on the array for smoking and control groups. β-values were also averaged across both groups considering CpG island context and gene association.

Differential Methylation Analysis

The Chip Analysis Methylation Pipeline (ChAMP) was used to identify differentially methylated CpGs between smokers and controls. ChAMP is a bioconductor package specifically designed for the analysis of Illumina HumanMethylation arrays that includes functionality for quality control, data normalization, and statistical tests for differential methylation (Morris et al., 2014). Benjamini Hochberg corrected p-values of <0.05 were considered significant. We additionally performed regional differential methylation analysis to identify contiguous differentially methylated CpGs, which are more likely to be biologically relevant. We performed regional analyses in two ways. First, differentially methylated regions were assessed in ChAMP via bump hunter. Second, we utilized the USeq platform to perform a 1000 base pair sliding window analysis commonly used in our lab (Jenkins et al., 2014b). A Benjamini Hochberg corrected Wilcoxon Signed Rank Test FDR of ≤ 0.0001 (≥ transformed FDR of 40) and an absolute log2 ratio ≥ 0.2 were used for our threshold of significance. Any significant findings were subjected to GO Term and Pathway analysis utilizing Genomic Regions Enrichment of Annotations Tool (GREAT) (McLean et al., 2010).

Methylation Variability Analysis

Analysis of variability was performed using the R software package. The overarching goal of this approach was to identify intra-group (smokers and non-smokers) variability for every CpG assessed on the array and to compare these findings between smokers and non-smokers. First, raw, Swan-normalized, β-values were subjected to logit transformation to generate m-values (Fig. 1A) to ensure the absence of heteroscedasticity in our analysis (the utility of logit transformation can be visualized in Fig. 1B). M-values for all samples analyzed in our study undergo “CpG-wise” mean centralization in R resulting in center-scaled (CS) values descriptive of the distance from average for each CpG in every individual. Figure 1C demonstrates the nature of CS values for a representative individual following mean centralization. CS values were then utilized to describe differences in variation between smokers and controls with increased distance from the mean equating to increased variability. With these data we compared average variability across all probes and also assessed regions of the genome that appear to be more variable in smokers compared to controls.

Figure 1. Data transformation steps used for the variability analysis.

A) Density plot displaying the distribution of beta values from a representative individual. B) The values from the same individual following logit transformation of beta values to generate m-values. C) Center scaled (CS) values from a single individual for all ~485k CpGs tiled on the array. CS values represent both the direction and distance from average for each CpG.

RESULTS

Semen parameters and DNA Damage

We observed a modest, but significant, decrease in several semen parameters including semen volume, total sperm count, and total progressively motile sperm count in men who smoke compared with never smokers. In addition, we observed a marginally significant increase in sperm DNA damage (p = 0.05) in men who smoke (Table 1). Box and whisker plots were generated to display the distribution of the various parameters for smokers and nonsmokers (Supplemental Fig 1). While semen parameters were similar in men from the general population and infertility patients, the effects of smoking on semen parameters and DNA damage were generally more significant in men from the general population than in patients. Interestingly, the level of DNA damage did not seem to be influenced by increased smoking duration or amount.

Table 1.

Descriptive statistics of semen parameters and other metrics in smokers and non-smokers.

	Mean ± SEM in smokers (n = 78)	Mean ± SEM in non-smokers (n = 78)	P value
Years smoking	10.6 ± 0.7	0.0	NA
# Cigarettes/day	13.3 ± 0.7	0.0	NA
Pack years	7.4 ± 0.7	0.0	NA
Age (years)	32.4 ± 0.9	31.2 ± 0.6	0.29
Semen volume (ml)	3.2 ± 0.2	4.3 ± 0.2	0.0003
Sperm concentration (M/ml)	81.0 ± 9.4	76.7 ± 6.2	0.82
Total sperm (M)	245.6 ± 30.8	316.0 ± 26.4	0.03
% Progressively motile	48.4 ± 2.6	51.8 ± 2.3	0.33
Total motile count (M)	136.5 ± 19.6	177.8 ± 17.3	0.04
% Normal heads	29.2 ± 1.5	25.6 ± 1.3	0.07
% Normal tails	71.4 ± 1.5	72.8 ± 1.2	0.48
% Viable	56.7 ± 2.1	57.9 ± 1.7	0.67
Concentration amorphous cells (M/ml)	0.7 ± 0.1	0.6 ± 0.1	0.13
Comet positive cells (%)	54.0 ± 3.4	44.1 ± 2.4	0.05

Sperm DNA methylation

Three primary analyses were performed to assess the effect of smoking on sperm DNA methylation: 1) Global methylation differences, 2) locus- or region-specific methylation differences, and 3) differences in methylation variance genome-wide.

We first calculated average beta values across all CpGs in samples from smokers compared to non-smokers, and additionally evaluated average values across different genomic features. We did not identify any differences in methylation globally or at specific genomic features (Fig 2).

Figure 2. Comparison of average beta values across all control versus smoker samples based on genomic context.

No differences in average methylation were observed in smokers versus non-smokers across the whole genome or at CpG islands, gene bodies, or non-CpG loci interrogated by the array.

We next evaluated methylation alterations at specific loci or in specific regions. Site-specific DNA methylation analysis revealed 141 significantly differentially methylated loci associated with smoking following Benjamini-Hochberg correction for multiple comparisons (Table 2, Fig 3). Interestingly, of the 141 significantly altered CpGs, 104 (74%) were hypomethylated, and only 37 (26%) were hypermethylated in samples from men who smoke (p < 0.0001). We previously reported a similar trend in preference toward loss of methylation at significantly altered CpGs in sperm DNA associated with male age (Jenkins et al., 2014b). While the directionality of DNA methylation changes were similar in both studies, we did not identify any DMRs that were shared by both studies. In addition, we found that differentially methylated CpGs were significantly under-represented at CpG islands and significantly over-represented at shore regions (p < 0.0001; Fig 4a), potentially indicating that CpG islands are somewhat protected from environmentally-induced methylation changes. We also found that regions previously reported to escape protamine replacement during spermatogenesis and enriched for H3K4me3 and H3K27me3 (Hammoud et al., 2009) were significantly over-represented in the dataset of differentially methylated CpGs (Fig 4b), indicating that histone-bound regions of the sperm genome may be more prone to environmentally-induced methylation changes. The CpGs located within these histone-bound regions are indicated in Table 2.

Table 2.

Nearest Gene	Probe ID	Chromosome	Map info	Strand	Gene	Feature	CGI	H3K4me3 or H3K27me3- bound region	Residual methylation	Non-smoker DNA methylation (%)	Smoker DNA methylation (%)	Delta DNA methylation (%)	Raw p- value	Benjamini- Hochberg adjusted p- value
FAM92A1	cg08953048	8	94658042	F		IGR	opensea	x		58.3%	65.5%	7.1%	1.5E-13	6.7E-08
FAM92A1	cg01040499	8	94657423	F		IGR	opensea			11.1%	15.5%	4.5%	1.8E-11	4.0E-06
OSMR	cg15599832	5	38845129	R	OSMR	TSS1500	shore	x	x	20.6%	14.7%	−5.9%	2.9E-11	4.3E-06
RP11-474D1.3	cg26959380	12	130498820	F		IGR	opensea	x		42.9%	39.8%	−3.1%	1.2E-10	1.4E-05
ULBP2	cg06471296	6	150259601	F		IGR	island	x		11.7%	8.0%	−3.7%	2.4E-10	2.2E-05
RP11-278H7.1	ch.1.242086458R	1	244019835	F		IGR	opensea			55.0%	50.2%	−4.8%	2.9E-10	2.2E-05
PRSS16	cg01759136	6	27242945	F		IGR	opensea			68.5%	64.6%	−4.0%	1.1E-09	5.9E-05
BCR	cg10480239	22	23522307	R	BCR	TSS1500	shore	x		37.2%	33.8%	−3.4%	1.1E-09	5.9E-05
ZNF784	cg22169206	19	56136215	F	ZNF784	TSS1500	shore			44.6%	42.2%	−2.4%	2.2E-09	1.1E-04
MT1F	cg02527372	16	56692077	F	MT1F	Body	shore	x		18.2%	14.4%	−3.8%	3.2E-09	1.4E-04
LINC01020	cg20662737	5	4941328	R		IGR	opensea	x		60.7%	56.9%	−3.8%	4.4E-09	1.8E-04
NDUFS6	cg07875360	5	1801344	F	NDUFS6	TSS200	island			58.5%	52.2%	−6.4%	5.8E-09	2.1E-04
ZNF253	cg09414724	19	19976638	F	ZNF253	TSS200	opensea	x		30.9%	24.9%	−6.0%	7.2E-09	2.3E-04
NDN	cg18406232	15	23927915	F		IGR	shelf			55.9%	52.1%	−3.8%	7.0E-09	2.3E-04
SHF	cg17911021	15	45493416	F	SHF	TSS200	shelf	x		53.8%	49.6%	−4.2%	7.9E-09	2.4E-04
ZNF280C	cg17758324	X	129403027	R	ZNF280C	TSS200	island			66.5%	61.3%	−5.2%	1.8E-08	5.0E-04
RNF17	cg24609094	13	25319768	F		IGR	island	x		40.6%	36.4%	−4.2%	3.4E-08	8.9E-04
GUSBP2	cg22243996	6	26757644	F		IGR	opensea	x		14.0%	10.2%	−3.8%	3.9E-08	9.7E-04
FAM19A5	cg04377908	1	220250989	R	BPNT1	Body	opensea			75.0%	66.7%	−8.3%	4.9E-08	1.2E-03
KIAA1712	cg20535044	4	175204418	R	KIAA1712	TSS1500	shore	x		7.2%	5.5%	−1.6%	8.4E-08	1.8E-03
ZNF253	cg26933068	19	19976613	R	ZNF253	TSS200	opensea	x		29.2%	22.7%	−6.5%	8.2E-08	1.8E-03
ZIC1	cg01458605	3	147128679	F	ZIC1	1stExon	island			8.1%	6.2%	−1.9%	9.6E-08	1.9E-03
GUSBP2	cg16387141	6	26757842	R		IGR	opensea	x		19.8%	17.4%	−2.4%	1.1E-07	2.2E-03
MIR3973	cg09817641	11	35845023	F		IGR	opensea	x		41.4%	45.3%	4.0%	1.4E-07	2.6E-03
KRTAP5-11	cg10924669	11	71320766	R		IGR	shore			46.8%	43.5%	−3.4%	1.8E-07	3.1E-03
PARP8	cg07757885	5	50219072	R		IGR	opensea	x		35.6%	38.8%	3.2%	2.0E-07	3.2E-03
DCDC2	cg16427109	6	24358683	F	DCDC2	TSS1500	shore			9.0%	12.4%	3.4%	1.9E-07	3.2E-03
ICAM3	cg06855803	19	10450364	F	ICAM3	TSS200	opensea			86.4%	84.5%	−1.9%	2.0E-07	3.2E-03
SLC25A12	ch.2.3493243F	2	172737063	F	SLC25A12	Body	opensea			30.7%	26.1%	−4.6%	2.7E-07	4.2E-03
TMPRSS9	cg25966599	19	2422172	R	TMPRSS9	Body	shore			85.0%	83.2%	−1.8%	3.1E-07	4.6E-03
FAM127C	cg12592455	X	134166801	R	FAM127A	1stExon	island			13.9%	9.7%	−4.2%	3.9E-07	5.2E-03
TMSB4Y	cg17560699	Y	15593045	F	UTY	TSS1500	shore			56.2%	49.2%	−7.0%	3.7E-07	5.2E-03
GUCY2D	cg25581932	17	7892150	R		IGR	shore			18.2%	14.6%	−3.6%	3.8E-07	5.2E-03
MIR5583-2	cg15612221	18	37379468	R		IGR	opensea	x		31.2%	26.0%	−5.1%	4.3E-07	5.6E-03
KRT86	cg16911220	12	52695728	R	KRT86	1stExon	island			16.7%	13.3%	−3.4%	4.5E-07	5.8E-03
ISG15	cg23691733	1	944093	R		IGR	opensea			86.2%	88.3%	2.1%	5.4E-07	6.7E-03
LINC00607	cg09166091	2	216484055	R		IGR	opensea			19.8%	12.6%	−7.1%	5.8E-07	7.0E-03
PDK2	cg20133730	17	48172203	R	PDK2	TSS1500	shore			65.7%	61.8%	−3.8%	6.4E-07	7.5E-03
LMCD1	cg08935301	3	8543732	R	LMCD1	Body	shore			4.4%	5.2%	0.8%	6.9E-07	7.8E-03
ANG	cg16154578	14	21161297	F	ANG	5′UTR	opensea	x		27.8%	23.3%	−4.5%	7.0E-07	7.8E-03
NKAP	cg09092713	X	119133965	R		IGR	island			16.2%	11.7%	−4.5%	8.0E-07	8.6E-03
JAK2	cg03071195	7	16459464	F	ISPD	Body	shore		x	69.2%	72.6%	3.4%	1.0E-06	1.0E-02
PLEKHG5	cg04818845	2	86331859	F	POLR1A	Body	shore		x	29.9%	27.0%	−2.8%	9.9E-07	1.0E-02
NEURL4	cg02831821	8	96040122	R	C8orf38	Body	shelf			89.5%	87.6%	−1.8%	1.0E-06	1.1E-02
SECTM1	cg26214645	17	80292081	R	SECTM1	TSS200	island			34.0%	30.5%	−3.4%	1.1E-06	1.1E-02
THRAP3	cg10212685	1	36748231	R	THRAP3	Body	opensea			46.8%	49.7%	2.8%	1.3E-06	1.3E-02
MIR124-3	cg01747792	20	61806628	R		IGR	island			10.2%	12.9%	2.6%	1.3E-06	1.3E-02
BEST2	cg23867562	19	12861921	F	BEST2	TSS1500	shelf			79.6%	77.7%	−1.9%	1.4E-06	1.3E-02
MIR5087	cg11659025	1	147767989	F		IGR	opensea	x		27.8%	23.0%	−4.7%	1.4E-06	1.3E-02
LOC360030	cg01786216	12	7917888	R	LOC360030	1stExon	opensea			25.3%	21.2%	−4.2%	1.7E-06	1.5E-02
PARK2	cg22291694	6	162294486	F	PARK2	Body	opensea		x	26.2%	31.4%	5.2%	1.7E-06	1.5E-02
TSC22D1-AS1	cg04671914	22	48977382	F	FAM19A5	Body	island			83.0%	80.8%	−2.2%	1.9E-06	1.6E-02
HERC6	cg14212360	4	89302999	F	HERC6	Body	shelf			86.9%	84.6%	−2.3%	2.1E-06	1.8E-02
PCDH20	cg26480584	13	62002615	F		IGR	opensea	x		41.3%	47.1%	5.7%	2.1E-06	1.8E-02
ISPD	cg03037561	19	6711092	F	C3	Body	shore			84.6%	83.1%	−1.6%	2.2E-06	1.8E-02
CEP97	cg23962380	3	101443264	R	CEP97	TSS1500	shore	x		5.8%	7.5%	1.6%	2.2E-06	1.8E-02
SIRT4	cg05178654	1	6580167	R	PLEKHG5	TSS200	opensea			78.8%	75.5%	−3.3%	2.5E-06	1.9E-02
ATP6V0E1	cg07419740	5	172410441	F	ATP6V0E1	TSS1500	shore	x		9.5%	7.5%	−2.0%	2.5E-06	1.9E-02
SHISA7	cg13752254	19	55940813	F	SHISA7	3′UTR	shelf			79.3%	76.4%	−2.8%	2.5E-06	1.9E-02
KCNIP3	cg03891849	12	112847769	F	RPL6	TSS1500	shore			50.1%	46.5%	−3.6%	2.7E-06	2.0E-02
SLC6A19	cg05972316	5	1204924	R	SLC6A19	Body	shelf			87.6%	86.3%	−1.3%	2.7E-06	2.0E-02
SNORD109B	cg23225193	15	25285663	F	SNORD109B	TSS1500	opensea			80.8%	77.0%	−3.7%	2.8E-06	2.0E-02
EPRS	cg21573359	1	220132728	R		IGR	opensea	x		11.0%	13.5%	2.5%	2.9E-06	2.1E-02
RPL6	cg03788567	9	21696666	F		IGR	island	x		52.6%	55.5%	3.0%	3.0E-06	2.1E-02
RHOV	cg09166898	15	41167165	R	RHOV	TSS1500	shore	x		51.2%	46.5%	−4.6%	3.1E-06	2.1E-02
ZNF784	cg11306428	19	56136210	F	ZNF784	TSS1500	shore			35.9%	32.9%	−3.0%	3.2E-06	2.2E-02
SNHG3-RCC1	cg23682641	1	28857538	R	SNHG3-RCC1	Body	shore			25.5%	21.8%	−3.8%	3.3E-06	2.2E-02
ZNF675	cg24101933	19	23870686	F	ZNF675	TSS1500	opensea			66.0%	61.1%	−5.0%	3.3E-06	2.2E-02
MRGPRX4	cg04272710	12	118499294	R	WSB2	TSS1500	island	x		23.5%	28.4%	4.9%	3.6E-06	2.2E-02
CANT1	cg08088171	17	77003113	R	CANT1	5′UTR	shelf			87.3%	85.8%	−1.4%	3.5E-06	2.2E-02
CHEK1	cg15190354	11	125494874	F	CHEK1	TSS1500	shore			14.8%	11.9%	−2.9%	3.7E-06	2.2E-02
GUSBP2	cg16675381	6	26757706	R		IGR	opensea	x		13.9%	11.7%	−2.2%	3.6E-06	2.2E-02
INPP5J	cg18053607	22	31518963	R	INPP5J	5′UTR	opensea			47.2%	42.3%	−4.9%	3.5E-06	2.2E-02
DIP2C	cg18072147	10	712592	F	DIP2C	Body	shore	x	x	35.6%	24.2%	−11.4%	3.7E-06	2.2E-02
CEP97	cg23048036	3	101442954	R	CEP97	TSS1500	shore			9.0%	12.2%	3.3%	3.6E-06	2.2E-02
PISD	cg27454842	22	32027588	R	PISD	TSS1500	shore	x		57.1%	51.8%	−5.3%	3.8E-06	2.2E-02
RP11-465B22.5	cg16989340	1	1084147	F		IGR	opensea			75.6%	73.6%	−2.0%	4.1E-06	2.3E-02
INPP5J	cg26373518	22	31518942	R	INPP5J	TSS200	opensea			45.1%	39.6%	−5.6%	4.1E-06	2.3E-02
FAM91A1	ch.8.2465681F	8	89645708	F		IGR	opensea			5.1%	6.3%	1.2%	4.1E-06	2.3E-02
HGS	cg22519420	17	79668370	F	HGS	Body	shore			79.7%	76.8%	−2.9%	4.2E-06	2.3E-02
C3	cg03015114	1	6638504	R	TAS1R1	Body	shore			62.7%	58.6%	−4.0%	4.8E-06	2.4E-02
CTNNA3	cg05626013	10	68685252	R	CTNNA3	Body	opensea		x	46.8%	51.6%	4.8%	4.7E-06	2.4E-02
HIST1H2AI	cg08039385	6	27776329	F	HIST1H2AI	1stExon	shore	x		22.7%	20.2%	−2.5%	4.5E-06	2.4E-02
SLC25A31	cg11060673	4	128651206	F	SLC25A31	TSS1500	shore	x		52.2%	56.6%	4.4%	4.7E-06	2.4E-02
AZI1	cg13535852	17	79173823	R	AZI1	Body	shelf			89.8%	88.8%	−1.0%	4.8E-06	2.4E-02
C19orf35	cg17437218	19	2278847	F	C19orf35	Body	island			87.1%	85.9%	−1.2%	4.8E-06	2.4E-02
C3orf45	cg18343556	3	50316384	R	C3orf45	TSS200	shore	x		80.5%	77.6%	−2.9%	4.9E-06	2.4E-02
HIST1H2AJ	cg20478264	6	27782176	R	HIST1H2AJ	1stExon	shore			31.3%	27.0%	−4.3%	4.8E-06	2.4E-02
MID1IP1	cg26129027	X	38662279	F	MID1IP1	5′UTR	shore			40.4%	35.9%	−4.5%	4.8E-06	2.4E-02
CGB2	cg11177404	19	49535079	R	CGB2	TSS200	opensea			38.9%	36.0%	−2.9%	5.1E-06	2.5E-02
WSB2	cg04106489	2	95963068	F	KCNIP3	TSS200	opensea			85.7%	84.2%	−1.5%	5.4E-06	2.6E-02
DHX15	cg12857541	4	24583375	R	DHX15	Body	shelf			55.3%	59.9%	4.6%	5.6E-06	2.7E-02
API5	cg22764193	11	43333145	F	API5	TSS1500	shore			9.8%	12.3%	2.5%	5.6E-06	2.7E-02
MVP	cg13906968	16	29831676	R	MVP	TSS200	shelf			73.4%	69.0%	−4.4%	5.9E-06	2.8E-02
RPS21	cg26145511	20	60953415	R		IGR	island			88.0%	86.6%	−1.4%	5.9E-06	2.8E-02
TBXA2R	cg05343404	19	3608327	F	TBXA2R	TSS1500	shore	x		87.1%	85.4%	−1.7%	6.2E-06	2.9E-02
TMEM145	cg13542542	19	42821106	R	TMEM145	Body	shelf			78.3%	76.5%	−1.8%	6.3E-06	2.9E-02
BPNT1	cg04282258	11	18211034	F		IGR	island	x		36.2%	37.9%	1.6%	7.3E-06	3.3E-02
KRT81	cg12231340	12	52685221	F	KRT81	1stExon	island			13.7%	11.3%	−2.4%	7.2E-06	3.3E-02
FAM118A	cg12262698	22	45704988	R	FAM118A	TSS1500	shore			8.5%	10.6%	2.1%	7.4E-06	3.3E-02
C7orf36	cg27377749	7	39605536	R	C7orf36	TSS1500	shore			8.5%	11.1%	2.6%	7.6E-06	3.4E-02
FLJ45079	cg05264101	12	120729684	F		IGR	opensea	x		6.3%	5.3%	−1.0%	7.8E-06	3.4E-02
TLK2	cg07466783	17	60691727	F	TLK2	3′UTR	opensea			46.8%	50.3%	3.5%	8.4E-06	3.5E-02
CDC42EP2	cg12562828	11	65076843	R		IGR	opensea			85.1%	83.1%	−2.1%	8.4E-06	3.5E-02
DIP2C	cg12760625	10	712442	R	DIP2C	Body	shore			26.2%	18.4%	−7.7%	8.4E-06	3.5E-02
GNGT2	cg23572751	17	47270556	R		IGR	shore	x		24.7%	21.0%	−3.7%	8.4E-06	3.5E-02
SERPINH1	cg26104986	11	75275303	F	SERPINH1	5′UTR	shore			83.3%	81.0%	−2.3%	8.2E-06	3.5E-02
TAS1R1	cg02954884	19	7910696	F	EVI5L	5′UTR	shore			85.9%	84.1%	−1.8%	8.5E-06	3.5E-02
C13orf45	cg10645091	13	76445055	R		IGR	opensea	x		6.4%	7.6%	1.2%	9.2E-06	3.7E-02
CAP1	cg15568225	1	40537438	F	CAP1	3′UTR	opensea			58.1%	62.0%	3.8%	9.1E-06	3.7E-02
MTAP	cg03724238	22	32598681	F	RFPL2	1stExon	opensea			26.8%	28.3%	1.5%	9.4E-06	3.8E-02
PML	cg01947066	15	74289586	R	PML	Body	shore			70.4%	63.4%	−7.0%	1.0E-05	3.8E-02
SLC16A9	cg13525697	10	61320994	R		IGR	opensea			89.0%	86.9%	−2.1%	9.7E-06	3.8E-02
SLC44A4	cg18949702	6	31832486	F	SLC44A4	Body	shore			85.2%	83.0%	−2.1%	1.0E-05	3.8E-02
SLC29A2	cg19347288	11	66135554	R	SLC29A2	Body	shelf			84.9%	82.9%	−2.0%	1.0E-05	3.8E-02
MTRNR2L1	cg25219874	17	22193751	F		IGR	shore	x		20.5%	16.2%	−4.3%	1.0E-05	3.8E-02
FAM91A1	ch.8.2465681F	8	124800139	F	FAM91A1	Body	opensea			6.4%	7.8%	1.4%	9.9E-06	3.8E-02
ITPKC	cg21869046	19	41225005	R	ITPKC	Body	shore			28.2%	17.9%	−10.3%	1.0E-05	3.8E-02
PDCD1	cg00795812	2	242802009	F	PDCD1	TSS1500	shelf			86.2%	83.7%	−2.5%	1.0E-05	3.8E-02
APC2	cg23291200	19	1473179	R	APC2	3′UTR	island			83.1%	82.0%	−1.1%	1.1E-05	4.0E-02
POLR1A	cg04771206	13	45153489	F		IGR	shore			9.1%	10.8%	1.7%	1.1E-05	4.0E-02
DCAF4L1	cg16000989	4	41983716	F	DCAF4L1	5′UTR	shore	x		4.1%	4.9%	0.8%	1.2E-05	4.2E-02
ABR	cg17018021	17	1029394	F	ABR	Body	shore		x	83.8%	82.2%	−1.5%	1.2E-05	4.3E-02
TRIO	cg00940867	5	14055789	F		IGR	opensea		x	86.0%	83.4%	−2.6%	1.2E-05	4.3E-02
RFPL2	cg03477302	9	5109889	R	JAK2	Body	shore			45.7%	49.5%	3.8%	1.2E-05	4.3E-02
BLM	cg22690576	15	91260127	R	BLM	TSS1500	shore			12.8%	10.4%	−2.3%	1.2E-05	4.4E-02
EVI5L	cg02836017	17	7231188	R	NEURL4	Body	shore			85.6%	84.0%	−1.6%	1.3E-05	4.4E-02
GABBR1	cg17053201	6	29593246	R	GABBR1	Body	shelf			84.3%	81.1%	−3.2%	1.3E-05	4.4E-02
DUSP8	cg02248763	11	1580098	F	DUSP8	Body	shore			92.9%	94.0%	1.1%	1.3E-05	4.4E-02
SNAR-F	cg07516355	19	51073535	F		IGR	shelf			82.6%	79.6%	−3.0%	1.3E-05	4.4E-02
PTGIR	cg07780534	19	47116777	R		IGR	opensea			37.3%	34.4%	−3.0%	1.3E-05	4.4E-02
FCER2	cg11445324	19	7751013	F		IGR	shelf			89.4%	88.2%	−1.2%	1.3E-05	4.4E-02
MARCH10	cg17050097	17	60783168	F	MARCH10	Body	shore	x		74.0%	70.0%	−4.0%	1.3E-05	4.4E-02
PRDX1	cg18288967	1	45987694	F	PRDX1	TSS200	island	x		18.1%	19.9%	1.8%	1.3E-05	4.4E-02
PSORS1C1	cg21084702	6	31107663	R	PSORS1C1	Body	opensea			79.0%	76.4%	−2.6%	1.3E-05	4.4E-02
PRIM2	ch.6.57776846F	6	57668889	F		IGR	opensea			6.9%	9.4%	2.5%	1.4E-05	4.4E-02
MIR5093	cg00453202	16	85482380	F		IGR	opensea			72.4%	69.5%	−3.0%	1.4E-05	4.5E-02
OR1F1	cg27116061	16	3239638	R		IGR	shore	x		17.8%	14.2%	−3.5%	1.4E-05	4.5E-02
TNS1	cg09859492	2	218723349	R	TNS1	Body	opensea			86.9%	85.6%	−1.3%	1.5E-05	4.7E-02
MMP16	ch.8.1820050F	7	141059306	F		IGR	opensea			4.3%	5.2%	0.8%	1.5E-05	4.7E-02
C8orf38	cg02693150	1	963599	F	AGRN	Body	shore			91.8%	90.9%	−1.0%	1.5E-05	4.7E-02

Figure 3. Volcano and Manhattan plots illustrating the region, magnitude and statistical significance of methylation changes in smokers compared with controls.

The volcano plot (left) clearly shown the bias toward reduced methylation in smokers compared with controls. Blue points meet the threshold for statistical significance after Benjamini-Hochberg correction. The Manhattan plot (right) shows general dispersion of significantly differentially methylated sites across the genome. The horizontal line at –log₁₀(p) ~4.8 indicates the Benjamini-Hochberg threshold for significance.

Figure 4. The association of significantly differentially methylated CpGs with GpG island context (A) and sites associated with histone retention and histone tail modification in sperm (B).

These associations were compared to what would be expected based on the frequency of associations for the background data set (all probes tiled on the 450k array). We found that CpGs differentially methylated in smokers were significantly reduced at CpG Islands (p<0.0001) compared to background and significantly enriched at CpG Shores (p<0.0001) based on expected values. Further, we identified significant enrichment in regions with histone retention in sperm in general (p=0.0042) and at sites with H3K4 (p=0.018). *** indicates p<0.001, ** indicates p<0.01, * indicates p<0.05.

While the locations of differentially methylated CpGs were strongly associated with specific genomic features and chromatin architecture, they were not associated with a specific gene class based on GO term analysis.

Several studies have found that while the great majority of the genome undergoes demethylation during early embryonic development and again during primordial germ cell migration to the gonadal ridge, a small proportion of the genome retains methylation through one or both waves of demethylation. Those regions include LINE1, SINE, intracisternal A particles (IAPs), Alu, ERVK, and alpha satellite regions (Guo et al., 2015; Seisenberger et al., 2012) as well as several hundred genes (Gkountela et al., 2015). Repeat elements are not represented on the array; however we evaluated the overlap of those genes that escape demethylation (Gkountela et al., 2015) with the differentially methylated CpGs we identified in the current study. Of the 141 differentially methylated CpGs (132 unique genes) identified in this study, eight (6.1%) were among those that display persistent methylation through early development, which represent approximately 2.7% of genes in the genome. This difference is marginally significant based on Chi-square analysis with Yates correction (p = 0.05). These regions are indicated in the “Residual Methylation” column of Table 2.

In addition to single CpG analysis, we performed regional analysis in two distinct ways to determine whether there were specific regions of the sperm genome that were differentially methylated in men who smoke. This work was performed using the differentially methylated region finder in ChAMP, and a window analysis using the methylation array scanner application in USeq, as has been previously performed in our lab (Jenkins et al., 2014a). We did not identify any significantly differentially methylated regional changes associated with smoking using either approach.

We were also interested in determining whether smoking was associated with stochastic sperm DNA methylation changes that would be apparent by evaluating the average variance from the mean for each CpGs between smokers and controls. We found a shift toward increased variation in DNA methylation in smokers compared with controls that was driven by a subset of samples though this change failed to reach significance (Fig 5). The data indicate that smoking may disrupt sperm DNA methylation fidelity to some degree, resulting in less tightly controlled sperm DNA methylation patterns, though this effect appears to vary between patients.

Figure 5. The distribution of each individual’s average standard deviation from the mean centralized normal value at each CpG as a measure of general variability.

A) Histogram with overlaid density plot of all individuals in the smokers and control groups to illustrate the generally wider distribution in the smoking group. B) Bean plot of the same distribution to illustrate the increased density of smoking samples with average standard deviations greater than 0.5. Average SD in the smoking group was slightly higher than the average in the control group due to the generally wider distribution of variability values, but this difference was not found to be significant.

An inherent complexity in performing these types of studies in humans is the inability to perfectly isolate a single lifestyle factor. In the current study, efforts were made to isolate a single variable, cigarette smoking. Smoking and non-smoking groups did not differ by age or BMI; however alcohol consumption was common in smokers selected for the study, while none of the non-smokers in the study reported alcohol consumption. In order to rule out the contribution of alcohol consumption, we performed the same differential methylation analysis described previously to evaluate methylation profiles based on alcohol consumption. We compared methylation profiles in men who smoked and consumed less than one serving of alcohol per week (n = 29) against those who smoked and consumed one or more servings per week (n = 43). Additionally, we performed the same analysis using three servings of alcohol per week as a cutoff (n = 40 consumed < 3 servings/week and n = 30 consumed ≥ 3 servings per week). In both cases, no significantly differentially methylated CpGs were identified, indicating that alcohol consumption is not influencing the results.

DISCUSSION

Our finding of a modest decline in several semen parameters is in agreement with previous studies. A recent systematic review and meta-analysis that included 20 studies with 5865 participants found that cigarette smoking was associated with significantly reduced sperm count, motility and morphology (Sharma et al., 2016). Analysis of basic semen parameters of the individuals included in the current study indicated a modest but significant reduction in semen volume (in disagreement with the meta-analysis), total sperm count, and total number of progressively motile sperm.

Previous studies have reported that cigarette smoking is associated with increased sperm DNA damage (Dai et al., 2015). As DNA damage has been proposed as a mechanism for DNA methylation changes (Russo et al., 2016), we assessed DNA damage in the samples analyzed in the current study. We used the alkaline comet assay to measure sperm DNA damage and, as reported previously, we found marginally increased DNA damage in sperm from men who smoke compared with controls (p = 0.05). While sperm DNA damage has not been precisely mapped in the context of chromatin structure, the histone-bound regions of the sperm genome are thought to be more susceptible to DNA damage than protamine-bound DNA. The current data offer evidence for an association between DNA damage in sperm and methylation changes; however, additional studies are required to characterize the nature of the association and to establish causation.

Pre-conception paternal smoking has been associated with increased incidence of numerous disorders in offspring including several types of cancer and birth defects. While cigarette smoke contains a variety of mutagenic agents, it has also been reported to impact DNA methylation patterns in peripheral blood, buccal cells and lung tissue (Ambatipudi et al., 2016; Bosse et al., 2012b; Breitling et al., 2011; Joehanes et al., 2016; K. W. Lee et al., 2013; Teschendorff et al., 2015). We hypothesized that smoking would impact DNA methylation profiles in sperm, and altered sperm DNA methylation could explain some of the increased risk to offspring health associated with paternal smoking. The primary objective of the current study was to evaluate sperm DNA methylation patterns in men who smoke compared with nonsmokers using the Illumina HumanMethylation 450k array. We identified a number of interesting and potentially important methylation alterations associated with smoking.

Global methylation levels were not different in men who did and did not smoke. However, we did find evidence for reduced sperm DNA methylation fidelity, or increased variance in methylation patterns in a subset of smokers. Interestingly, this increased variance did not seem to be related with smoking quantity or duration, suggesting that other factors might work in concert with smoke exposure to alter methylation patterns in sperm.

The most striking finding or the current study was the relatively large number of significantly differentially methylated CpGs in men who smoke as well as the genomic context of differentially methylated CpGs. We identified 141 significantly differentially methylated CpGs in sperm from smokers compared with non-smokers. In spite of the relatively large number of altered methylation sites identified, similar studies investigating the effect of cigarette smoking on DNA methylation in peripheral blood, buccal cells and lung tissue report many more differentially methylated loci (Ambatipudi et al., 2016; Bosse et al., 2012b; Breitling et al., 2011; Teschendorff et al., 2015), indicating that sperm may be less susceptible to environmentally-induced methylation alterations compared with other tissues. This is not surprising considering the importance of genetic and epigenetic integrity of male gametes for offspring health.

While the differentially methylated CpGs were not associated with a specific biological pathway or GO term, we did observe significant bias in the genomic context of altered CpGs. Differentially methylated CpGs were found at CpG islands, shores, shelves, and open seas; however, there was significant depletion of differentially methylated CpGs at islands and significant enrichment at shores. The mechanism for this bias in unclear, but it does indicate that genomic context is a driver for susceptibility to differential methylation as a result of tobacco smoke exposure. Additionally, we found that differential methylation occurred more frequently at regions previously reported to display H3K4 and H3K27 methylation in mature sperm, again suggesting that histone-bound regions may be more susceptible to methylation perturbations compared with protamine-bound regions. The apparent protection of methylation patterns conferred by protamination may explain, at least in part, the reduced effects of smoking on DNA methylation perturbations in sperm compared with other tissues. While protamination may confer a protective effect on DNA methylation perturbations in sperm, it is important to note that altered sperm DNA methylation at histone-bound loci may be more impactful on gene expression, given the importance of these regions in early embryonic development.

CONCLUSIONS

While the current study offers compelling evidence that smoking can impact sperm DNA methylation patterns in a consistent manner, replication studies in larger cohorts as well as functional studies are required to understand the potential for sperm DNA methylation alterations to be transmitted to offspring or to otherwise affect offspring phenotype. In order for sperm DNA methylation patterns to impact offspring phenotype, alterations must escape the epigenetic reprogramming that occurs during early embryogenesis, when most, but not all DNA methylation marks are erased and subsequently re-established in a tissue-specific manner. Alternatively, even if altered DNA methylation is not directly passed on to offspring, the alterations might impact early embryonic gene expression or modify reprogramming or early development in other ways. These are important areas for future research.

Altered sperm DNA methylation resulting from paternal exposures or lifestyle factors is a plausible mechanism for phenotype modification in offspring, however, additional studies in humans and animal models are needed to characterize the types of exposures that can impact sperm epigenetics. Additional studies similar to the one described here targeting other exposures will offer insight into the nature of those genomic regions that are particularly susceptible to methylation change. Finally, controlled, multigenerational animal studies are required to assess the transmission of altered sperm DNA methylation patterns to offspring and the potential for those alterations to be perpetuated across generations.