Table 1.
Donor demographics.
Figure 1.
Pyrosequencing results for the LINE-1 global methylation assay.
(A) The box plot depicts significantly increased average global methylation with age based on a non-paired t-test of all samples ≤45 (n = 17) years of age vs. all samples >45 (p = 0.001; n = 17). Global methylation was also stratified based only on age at the time of collection for each sample from all 17 donors (a total of 34 samples with each donor represented twice). (B) Linear regression analysis confirmed the significant increases in global sperm DNA methylation with age (p = 0.0062).
Figure 2.
(A) The magnitude of alterations in terms of amount of change per year (reported as slope magnitude) for all regional changes that occur at CpG islands, shores and outside of these regions (other).
Average alterations per year were approximately 0.281%. (B) Average β-values for all significant windows (hypomethylation and hypermethylation events) for both aged and young. Average decrease in β-value was approximately 3.9% for intra-individual hypomethylation events and 3.2% for hypermethylation events. (C) the percent of regions of hypermethylation and hypomethylation at CpG islands, shores and outside of these regions (Other). Hypermethylation events were significantly more enriched at islands than were hypomethylation events based on a fisher exact test (p = 0.0056). Hypomethylation events were significantly more enriched at shores in comparison to hypermethylation events (p = 0.0015). Hypermethylation and hypomethylation events were similarly enriched in regions outside of islands and shores. (D) We also investigated the co-localization of nucleosomes (every region of known histone retention) as well as histone modifications (H3K4 methylation, and H3K27 methylation) with our windows of interest. Hypermethylation events were less frequently associated with all retained histones (nucleosomes) or loci with H3K27 methylation when compared to hypomethylation events based on Fisher's Exact Test (p = 0.002; p = 0.0107). Co-localization of hypermethylation or hypomethylation events with H3K4 methylation was statistically similar.
Figure 3.
(A) Chromosomal loci of each altered region are depicted where blue marks represent hypomethylation events and red marks hypermethylation events.
(B) The Correlation Maps app on the USeq platform was used to locate any specific chromosomal enrichment of altered methylation windows. Specifically, the application called any 100 kb region where at least two significantly altered methylation marks were found. All called chromosomal enrichment regions are displayed though none were found to be significantly enriched over the background.
Figure 4.
(A) Comparison of MiSeq results to our array results at 21 representative regions.
Because beta-values and fraction methylation are generated in a different manner (array vs. sequencing respectively) they are not directly comparable. For this reason we compared the fractional difference for each loci and each technology. This is accomplished by the following equation: fractional difference = (aged value/young value)−1. (B) the fractional difference between young and aged samples at 15 selected loci as measured by array in the 17 donor samples as well as in the independent cohort (19 samples from individuals > = 45 years of age and 47 samples from individuals <25 years of age taken from the general population). On average the fractional difference identified in the independent cohort (as measured by sequencing) was approximately 2.2 times greater in magnitude than was identified in the 17 donors.
Figure 5.
Single molecule analysis reveled 3 distinct alterations that occur with age.
(A) DRD4 has only slight alterations associated with age because the young cohort (<45) is strongly hypomethylated initially, and aging simply amplifies this. RDMR_2 is representative of many alterations observed in this analysis which had a strong population shift from moderately hypomethylated to hypomethylated. TBKBP1 is representative of sites that had a bimodal distribution methylation patterns in the young group that becomes stabilized with age. (B) in every case (DRD4, RDMR_2, TBKBP1) each region has significant demethylation with age though the magnitude of change varies.
Figure 6.
The frequency of disease associations within our gene set was analyzed and compared to the frequency of disease associations for all genes known to be associated with at least a single disease based on GAD annotation.
Schizophrenia, bipolar disorder, diabetes mellitus and hypertension were selected, as there were at least 3 genes in our small set of identified genes that are associated with these diseases. Only bipolar disorder was more frequently associated with our identified genes than the background set of genes, based on chi-squared analysis with multiple comparison correction (Bonferonni) of the 117 age associated genes analyzed (p = 0.012), and schizophrenia also trended toward increased frequency (p = 0.07). However, these are not considered significant enrichments if considering all genes in the genome (omitting the filter for a disease connection). The frequency of genes associated with hypertension and diabetes mellitus in the two groups was statistically similar.
Figure 7.
Various descriptive statistics are presented for both TNXB and DRD4; 2 regions of representative methylation alterations.
(A,B) The alignment track for each gene is displayed in Integrated Genome Browser (IGB) with the associated false discovery rate (FDR) denoting the significance of the change and the absolute log 2 ratio reflecting the magnitude of the alteration. (C,D) Scatter plots for each sample from all 17 donors (a total of 34 samples with each donor represented twice) with linear regression lines and associated r2 values were generated. Regression analysis revealed a significant decrease in methylation with age at both DRD4 and TNXB (p = 0.0005 and p = 0.003 respectively). (E,F) The average methylation within each window (DRD4 and TNXB) was plotted for each paired sample set and is displayed for each donor.