Cigarette smoking significantly alters sperm DNA methylation patterns

Abstract

Numerous health consequences of tobacco smoke exposure have been characterized, and the effects of smoking 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 spermatozoa. 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 Human Methylation 450 beadchip. We investigated two models of DNA methylation alterations: (i) consistently altered methylation at specific CpGs or within specific genomic regions and (ii) 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.

Keywords: epigenetics; genome-wide; smoking; sperm DNA methylation; transgenerational inheritance.

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.

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.

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.

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.