Intra- and interindividual epigenetic variation in human germ cells

Abstract

Epigenetics represents a secondary inheritance system that has been poorly investigated in human biology. The objective of this study was to perform a comprehensive analysis of DNA methylation variation between and within the germlines of normal males. First, methylated cytosines were mapped using bisulphite modification-based sequencing in the promoter regions of the following disease genes: presenilins (PSEN1 and PSEN2), breast cancer (BRCA1 and BRCA2), myotonic dystrophy (DM1), and Huntington disease (HD). Major epigenetic variation was detected within samples, since the majority of sperm cells of the same individual exhibited unique DNA methylation profiles. In the interindividual analysis, 41 of 61 pairwise comparisons revealed distinct DNA methylation profiles (P=.036 to 6.8 x 10(-14)). Second, a microarray-based epigenetic profiling of the same sperm samples was performed using a 12,198-feature CpG island microarray. The microarray analysis has identified numerous DNA methylation-variable positions in the germ cell genome. The largest degree of variation was detected within the promoter CpG islands and pericentromeric satellites among the single-copy DNA fragments and repetitive elements, respectively. A number of genes, such as EED, CTNNA2, CALM1, CDH13, and STMN2, exhibited age-related DNA methylation changes. Finally, allele-specific methylation patterns in CDH13 were detected. This study provides evidence for significant epigenetic variability in human germ cells, which warrants further research to determine whether such epigenetic patterns can be efficiently transmitted across generations and what impact inherited epigenetic individuality may have on phenotypic outcomes in health and disease.

Figure 1.

Intraindividual variability of DNA methylation. A, DNA methylation profiles of the promoter CpG islands of BRCA1 and PSEN2, determined on the basis of sequencing 60 clones of bisulphite-modified sperm DNA. The BRCA1 locus covered 32 CpGs, and the PSEN2 region included 45 CpGs. Nine monomorphic (unmethylated) CpGs (BRCA1 or PSEN2) were excluded from the figure. Each individual is represented, with individual CpG dinucleotides from left to right (black = methylated cytosines; white = unmethylated cytosines) and individual clones from top to bottom. Like the presented BRCA1 and PSEN2 cases, a substantial proportion of clones in other loci (HD, DM1, BRCA2, and PSEN1) revealed unique DNA methylation profiles. B, Estimates of the proportion of unique methylation profiles in the promoter regions of the six analyzed genes. The Y-axis shows the proportion of clones carrying unique methylation profiles over the total number of sequenced clones; the X-axis shows the proportion of unique profiles that contain at least 1, 2, and 3 differences (left, middle, and right bars, respectively), compared with the other profiles at the same locus in the same individual.

Figure 2.

Interindividual variability of DNA methylation in six human disease genes. Bisulphite modification–based mapping of methylated cytosines in BRCA1, BRCA2, HD, DM1, PSEN1, and PSEN2. Thirty individual clones were sequenced from three to seven individuals. Analysis for each gene is represented in two panels. Left panels, graphical profile of the percentage of methylation (Y-axis, ranging from 0% to 40%) for every CpG dinucleotide (X-axis, ranging from 32 to 108 CpG dinucleotides), out of the total number of clones for each individual. Right panels, Euclidean distances (Y-axis) of pairwise comparisons between individual methylation profiles (X-axis). The blue line is the mean distance, and red lines are ±2 SD from the mean, both obtained for each gene from the permutation study (see the “Material and Methods” section). Pairwise comparisons are annotated—for example, as “16”—for the comparison of the Euclidean distance of individual 1 with that of individual 6. Primed individual numbers (e.g., 4′) represent a second set of 30 clones from those individuals. The error bars on some data points represent SDs from 100,000 permutations of 30 clone groups from the individuals from whom 60 clones were sequenced.

Figure 3.

Chromosomal view of methylation variability by CpG island microarray analysis. A, Unmethylated fraction of genomic DNA extracted from sperm samples (n=21) hybridized individually (Cy5), in contrast to the pooled reference control (Cy3). The CV of the Cy5/Cy3 ratio was calculated for each spot across the 21 individuals and was mapped to the corresponding genomic location. Each chromosome ideogram is overlaid with red bars that represent the position of each clone on the CpG island microarray. The bars highlighted in green are the loci that showed variance in the 90th percentile (the top 10% of loci exhibiting the largest degree of DNA methylation variation). B, Screenshot of the custom annotation track on the UCSC Genome Browser (available from the Center for Addiction and Mental Health). Shown is chromosome 6, which includes the major histocompatibility complex locus that was screened for epigenetic variability by the Human Epigenome Project pilot study.

Figure 4.

Genomewide view of brain DNA–HpaII (A) and sperm DNA–HHA (B) data sets. The unmethylated fraction of genomic DNA was enriched from brain DNA (n=22) or sperm samples (n=25), and each was hybridized individually (Cy5), in contrast to the pooled samples (Cy3). The CV of ratio Cy5/Cy3 was calculated for each spot across the 22 or 25 individuals and was mapped to the corresponding genomic location. Each chromosome ideogram is overlaid with red bars that represent the position of each clone on the array. The bars highlighted in green are the loci that showed statistically significant variance (90th percentile).

Figure 5.

Age-related DNA methylation changes in the sperm. Individuals were ordered by increasing age (top left panel), and gene-specific DNA methylation dynamics were investigated using the individual ages (sperm DNA–HpaII age range 24–56 years) as a covariate. Pearson correlation was calculated for each locus, and the one-tailed P value of the coefficient was obtained. In the sperm DNA–HpaII data set, 105 loci were identified as significantly (P<.05) correlated (r>0.5) or inversely correlated (r<-0.5) with age. Since the unmethylated fraction of DNA was interrogated, positive correlation indicates decreasing DNA methylation with age, whereas inverse correlation reflects increasing methylation with age. The genes CTNNA2, EED, CALM1, CDH13, and STMN2 are shown as examples. Other genes for the sperm DNA–HpaII, sperm DNA–HHA, and brain DNA–HpaII data sets are available in tables 3 and 4.

Figure 6.

Repetitive-element analysis in sperm DNA–HpaII and brain DNA–HpaII data sets. The microarray loci that contain a single repetitive element were separated into each repeat class, and the mean CV (±SD) was calculated (A). The repeat classes include DNA transposons (n=209), long interspersed transposable elements (LINEs [n=771]), low-complexity repeats (n=461), long terminal repeats (LTRs [n=360]), satellites (n=208), simple repeats (n=346), SINEs (n=1,058), small nuclear RNA (snRNA [n=30]), and tRNA (n=40), and the nonrepetitive loci (n=6,976) are presented for comparison. The satellite repeats were the only class to show significantly increased variability in the sperm DNA–HpaII (P=6.12×10^-17) and less-significantly increased variability in the brain DNA–HpaII (P=.0027) data sets. B, Breakdown of the satellite repeats into specific satellite-repeat classes, which reveals a number of repeat classes with increased variability—predominantly, the centromeric satellite repeats, including (GAATTC)_n (P=8.44×10^-17; n=55), ALR/α (P=4.08×10^-25; n=119), CER (P=.0026; n=6), and HSATII repeats (P=3.91×10^-5; n=19) but not BSR/β repeats (P>.05; n=7).

Figure 7.

MS-SNuPE analysis of densities of methylated cytosines in CpG dinucleotides of selected genes. Genomic DNA from 11 individuals was treated with sodium bisulphite and then was PCR amplified for each gene. The genes NELL2, SCAM1, NEIL2, MKL2, CDH13, and OLR1 are represented. The methylation status of CpG dinucleotides within each of the restriction-enzyme sites was interrogated using the primer-extension reactions. Methylation of each of the CpG dinucleotides is represented as a percentage of methylated PCR products: completely unmethylated (white circles), partially methylated (partially black circles), or completely methylated (black circles).

Figure 8.

Methylation profiles of CDH13. Methylation status of 16 CpG sites surrounding the CDH13 C/G SNP across 30 clones sequenced in each of five tested individuals. Seventy-seven percent (67/87) of the G alleles are methylated (four or more methylated CpGs), whereas 78% (49/63) of the C (bisulphite-converted to T) alleles are unmethylated. The first seven CpG dinucleotides interrogated by MS-SNuPE in figure 7 are represented in this figure as CpGs 5, 6, 9, 10, 13, 15, and 16. CpG 9 is the third MS-SNuPE primer that was predominantly unmethylated in all individuals. Each individual is represented, with single CpG dinucleotides from left to right (black = methylated; white = unmethylated) and with individual clones from top to bottom.