A summary of the methylation of all amplicons in different individuals is shown in Figure 1a. The methylation levels of highly methylated or completely unmethylated amplicons were usually similar among individuals, while the intermediately methylated amplicons showed higher variability (Figure 1b). Four amplicons showed a maximal inter-individual difference in their methylation levels of >50% and another five amplicons showed variances of >30%. As an example, the methylation pattern of amplicon 23_1 in 20 individuals is shown in Figure 1c. When comparing the methylation difference among individuals, we did not observe any significant methylation difference between old and young individuals or males and females. This is consistent with former results where no significant correlation of methylation changes with gender or age was detected 14.

We observed that amplicons with a high methylation variance among individuals often showed a biphasic distribution of methylation comprising clones that were mainly unmethylated or highly methylated. This led to the question of whether the DNA methylation varied from one cell to another or between the two alleles in each cell. In some of these cases, the alleles could be differentiated by the presence of a single nucleotide polymorphism (SNP) in the sequenced region such that the sequencing reads could be sorted according to the allelic origin. In many of these cases we observed a clear correlation of genotype and methylation levels, indicating the presence of ASM (Table 1). To exclude that clonal PCR after bisulfite conversion might have caused these results, we performed individual bisulfite genomic sequencing analysis twice for cases where ASM was observed and always obtained similar results (Additional file 3). In addition, we used digestion with methylation-sensitive restriction enzymes to confirm the ASM (Additional file 3). For amplicons with ASM, we included more individuals in the analysis.

We observed strong ASM on the first exon of the C21orf81 gene where we studied the methylation pattern of amplicon 23_2 and 23_1, which are located next to each other and cover almost the entire CpG island (Figure 2a). The analysis for amplicon 23_1 was performed in 38 individuals, 17 of which had an A/C SNP in the sequenced region. Out of them, we observed ASM with a methylation difference of >30% in eight cases, always having the A allele hypomethylated (Table 1; Figure 2; Additional file 4). Some individuals showed a weaker DNA methylation difference between alleles, and in other cases no DNA methylation difference was detectable. Within the 21 homozygous individuals, we observed a biphasic distribution of DNA methylation levels in 9 cases, suggesting the potential presence of ASM, although this could not be identified because of the lack of a characteristic SNP.

The methylation pattern of amplicon 23_2 was studied in 19 individuals, including one (called 'N' in Figure 2 and the following) previously analyzed 12. Altogether, eight individuals had an A/C SNP in this region and seven of them showed strong ASM (Table 1; Figure 2d; Additional file 4). In all ASM cases, the A allele was less methylated than the C allele (Figure 2e). For the 11 homozygous individuals, biphasic DNA methylation was observed in 7 cases.

We identified the correlation between the SNPs in amplicon 23_1 and 23_2 in the genomic DNA sequence of all individuals with ASM. When both SNPs were present, we always observed A-A and C-C haplotypes. In three such cases (individuals 9, 11 and 17), the A alleles in both amplicons were hypomethylated, indicating that pronounced ASM can span a whole CpG island (Figure 2a; Additional file 4). In individuals 3 and N, ASM was observed in one amplicon (either 23_1 or 23_2) and the other amplicon showed biphasic methylation, but did not contain the SNP, suggesting that the whole CpG island has ASM as well in these cases. However, there were also cases in which the methylation states of alleles in these two amplicons were not coupled. In individuals 4, 12 and 20, ASM was observed in 23_2, but there was no obvious methylation difference between alleles in amplicon 23_1. No methylation difference between alleles in both amplicons was observed in one individual (individual 1).

To determine if ASM leads to monoallelic expression of the C21orf81 gene, we prepared RNA from leukocytes of individuals 9 and 11, which showed strong ASM of both amplicons, and converted it into cDNA. Sequencing of the cDNA revealed that the gene was exclusively expressed from the unmethylated A allele, indicating the ASM correlates with allele-specific gene expression (Figure 2a).

We were able to check the methylation pattern of amplicon 23_1 in the parents of individual 17 (male). Like individual 17, the mother (17_M) had an A/C SNP with the C allele methylated. The father (17_P) had an A/A genotype and showed slight biphasic methylation but did not show clones with high methylation characteristic for C alleles (Additional file 4).

We also identified ASM on amplicon 262 containing 42 CpG sites, which is located on a CpG island in the HSF2BP gene (Figure 3a). The methylation state of this region was studied in 33 individuals. In many cases, a clear boundary between methylated CpG sites and unmethylated CpG sites was observed in the same amplicon (Figure 3a, b). Out of 20 individuals with an A/G SNP, ASM was observed in 19 cases (Table 1; Figure 3; Additional file 5), which showed two different patterns of ASM (Figure 3b). In 18 individuals, the first 28 CpG sites were methylated in both alleles, whereas in the remaining part of the amplicon, the A allele was hypomethylated. In one case (individual 2), the G allele was hypomethylated throughout the entire amplicon, whereas the A allele was methylated in the first 28 CpG sites but low methylated on the remaining 14 CpG sites similar to what was observed for the A allele in the other individuals. Absence of ASM was only observed in one heterozygous individual (individual 28). In this case, both alleles were highly methylated throughout the whole amplicon (Figure 3b). The methylation difference between alleles of all heterozygous individuals in amplicon 262 is summarized in Figure 3c. We analyzed the expression state of the alleles in leukocytes from individuals 10 and 11 and the results show that both alleles were expressed (Figure 3a).

We investigated the methylation levels of amplicon 262 in homozygous individuals and observed a strong correlation between DNA methylation and genotype - individuals with A/A genotype were always low methylated while individuals with G/G were highly methylated (Figure 3d). The methylation levels of the A and G alleles from heterozygous individuals were similar to those of individuals with A/A or G/G genotypes, respectively (Figure 3d), suggesting that the methylation difference is determined mainly by the polymorphisms between the alleles. A similar observation was made for amplicons 23_1 and 23_2 (Additional file 6).

Amplicon 232 is located on a CpG island in an intron of the gene RIPK4 near the transcriptional start site (Figure 4a). We studied the DNA methylation pattern of this region in 36 individuals. Out of the 11 individuals heterozygous at an A/G SNP in the studied region, ASM was observed in 3 cases where the A allele was always hypomethylated. In the remaining heterozygous individuals, both alleles were low methylated (Table 1; Figure 4; Additional file 7). The gene expression analysis in leukocytes from individual 10 revealed that both alleles were expressed (Figure 4a).

We previously identified ASM of two amplicons (176_1 and 176_2) of the CBR1 gene in blood derived from one individual 12. The allele with a deletion in 176_1 (dbsnp rs41563015) and with C at a C/G SNP site in 176_2 (dbsnp rs25678) was hypermethylated when compared with the allele without deletion and with G at the SNP site. Here, we studied the methylation pattern of these two regions in 20 additional individuals (Additional file 8). For amplicon 176_1, none of these individuals had the deletion or any other SNP in the sequence and all showed low methylation. For amplicon 176_2, 7 out of 20 individuals were heterozygous, but both alleles were always unmethylated. This result suggests that the deletion was more important for ASM to occur in the CBR1 gene promoter than the SNP in amplicon 176_2.

Blood DNA was collected from 37 apparently healthy individuals, including 36 of caucasian origin, and 1 from Asia (Additional file 2). All individuals were informed about the study and declared consent in written form. Genomic DNA from blood was extracted using a QIAamp DNA Blood Mini Kit (Qiagen, Hilton, Germany).

DNA methylation analysis was carried out as described 22. Briefly, 200 to 300 ng genomic DNA was converted by bisulfite and used as template for PCR with primer designed in the region of interest. Information on amplicons selected and primer sequences are listed in Additional file 1. Around 40 clones for each amplicon were picked and sequenced using ABI BigDye Terminator chemistry (BigDye Terminator v3.1 K (Applied Biosystems Inc, Foster City, CA, USA) and 3730 × l ABI 96-capillary sequencer systems equipped with capillaries of 50-cm separation length. In all ASM cases, the presence of the SNP was confirmed by sequencing of unconverted DNA using primers listed in Additional file 1. In cases where ASM was observed, the bisulfite conversion, subcloning and sequencing analysis was performed twice in independent experiments for the same amplicon to exclude clonal PCR amplification. The final results presented here are a combination of the results from both experiments. The separate results are shown in Additional file 3.

We confirmed the methylation difference between alleles by using the methylation-sensitive restriction enzymes HpaII (for amplicons 23_1, 23_2, and 232) and BstUI (for amplicon 262) (New England Biolabs, Ipswich, MA, USA). Genomic DNA (250 ng) was digested at the appropriate temperature (37°C for HpaII, 60°C for BstUI) for 12 h in buffer recommended by the supplier. Afterwards, the PCR products were amplified from the digested genomic DNA using the following thermal cycling parameters: 15 minutes at 95°C, 35 cycles of 45 s at 95°C, 45 s at 50°C and 1 minute at 72°C, and finally 72°C for 5 minutes. The sequences of the PCR products were analyzed by direct sequencing or sequencing of subclones. The PCR products amplified from undigested genomic DNA were used as control. The primer sequences are listed in Additional file 1.

Total RNA was extracted from leukocytes (QIAamp RNA Blood Mini Kit, Qiagen) and was reverse-transcribed. The cDNA was used as templates for PCR by using the following thermal cycling parameters: 15 minutes at 95°C, 35 cycles of 45 s at 95°C, 35 s at 55°C and 50 s at 72°C and finally 72°C for 5 minutes. The primers were designed to span different exons of genes C21orf81, HSF2BP and RIPK4 and include exonic SNPs (listed in Additional file 1). The RT-PCR product was analyzed by direct sequencing.

BiQ analyzer software was used to perform quality control and to derive DNA methylation patterns from the sequencing results 23. The BDPC (Bisulfite sequencing Data Presentation and Compilation) program was used to present the methylation pattern of single amplicons, single alleles and prepare the methylation map of all amplicons studied 24. All data can be accessed at 25.