Six siblings who presented to the cancer genetics clinic at the Institute of Oncology, Istanbul University in 2012 for BRCA1 and BRCA2 mutation testing were included in the study. Each individual signed an informed consent form. The study was approved by the Ethics Board of Istanbul University (approval no. 1552, dated May 18, 2015) in accordance with The Declaration of Helsinki (20).

DNA samples isolated from the peripheral blood lymphocytes of the six siblings were used in the study. Genomic DNA was isolated with a QIAamp DNA Mini QIAcube kit (cat. no./ID: 51326) using the QIAcube automated nucleic acid extraction system (both Qiagen N.V.). The integrity of the isolated DNA was measured with the Invitrogen Qubit 4 Fluorometer (Thermo Fisher Scientific, Inc). The BRCA MASTR Plus Dx (cat. no. MR-2015.024; Multiplicom N.V., Agilent Technologies GmbH) kit was used for sequencing on the Illumina MiSeq next generation sequencing (NGS) platform (Illumina, Inc.) with paired end libraries. Each reaction was conducted using 10 ng DNA. All BRCA1 and BRCA2 coding regions, including 50-bp intron-exon junctions, were covered with the BRCA MASTR Plus Dx kit. Sequencing was performed with 200× coverage. During the run, the presence of small indel mutations and large deletions and duplications was also investigated and evaluated. The sequencing run was performed using MiSeq Reagent kit v2 (cat. no. MS-102-2003; Illumina, Inc.). Sequencing data were analyzed with the SOPHiA™ DDM clinical NGS data analysis platform (v4; Sophia Genetics).

Methylation differences were evaluated among the six siblings, who comprised MZ twins both with BRCA1 pathogenic mutations but discordant for ovarian cancer, three sisters with BRCA1 mutations, and one healthy brother with non-mutated BRCA1. The data and codes of the individuals are presented in Table I, and their family tree is shown in Fig. 1.

Four different comparison groups were generated, each containing a case and control group, in order to evaluate the differences in methylation levels according to diagnosis and BRCA mutation conditions. These groups were as follows: Group 1, MZ twin siblings with and without ovarian cancer; Group 2, the MZ twin with ovarian cancer and healthy non-twin siblings; Group 3, all siblings with the BRCA1 mutation and the sibling with no BRCA1 mutation; and Group 4, the healthy MZ twin, and all other healthy siblings. The groups and the codes of individuals in the groups are presented in Table II.

The CpG islands at which differences in methylation level were detected between the case and control groups according to ovarian cancer etiology and BRCA1 mutation-carrying status were compared and evaluated. The methylation differences between the groups were evaluated as 10-fold and in some groups as 25-fold or more to obtain more specific regions.

Following isolation of the DNA, bisulfite modification was performed for a 500-ng DNA sample in each case. The bisulfite conversions of DNA samples were conducted using the EZ DNA Methylation Kit (cat. no. #D5001; Zymo Research Corp.).

The genome-level methylation profiles of the modified DNA samples were investigated using the Infinium MethylationEPIC BeadChip Array on an iScan device (Illumina, Inc). The Infinium MethylationEPIC Array is a genome-wide DNA methylation analysis system based on bisulfite conversion and Infinium HD sequencing technology that queries differentiated loci using region-specific probes designed for methylated and non-methylated regions. The total methylation level for a queried locus is determined by calculating the ratio of fluorescent signals from the methylated and unmethylated regions (21). Data analyses of the experimental results were conducted using the Lumi libraries within the Illumina GenomeStudio v2011.1 Methylation Module v1.9.0 (https://www.illumina.com/techniques/microarrays/array-data-analysis-experimental-design/genomestudio.html) and R 3.0.2 (http://www.r-project.org). The differences in methylation for >850,000 regions on a point basis were investigated using this chip system. All six cases in the study group were evaluated at >850,000 different CpG points.

Background corrections and dye bias equalization filtering, transformation and normalization of the data were conducted using library(lumi) in R 3.0.2 and were performed to minimize the rate of possible systematic statistical error. After filtering all samples by P-value, a mean of 866,309.3 CpG regions were identified by P<0.01 and 866,518.5 CpG regions were identified by P<0.05. Probes were regarded as erroneous and excluded from the CpG analysis when no detection could be taken from the same probe in >25% of all samples (P≥0.05).

Probes readable in all samples or having only 1 sample with no readable value were included in the analysis, and the other probes were filtered out (Fig. 2). Accordingly, a total of 563 CpG regions that were found to have insignificant results according to their detection P-value were excluded from the analysis. Thus, 866,332 CpG regions were analyzed in accordance with the Beta Mixture Quantile (BMIQ) normalization procedure using the BMIQ function in R 3.0.2 (22).

Boxplot and density diagrams were drawn for comparison of the distributions before and after BMIQ normalization and data conversion to avoid false results and reduce systematic bias.

A rating diagram was prepared to observe the degree of repetitiveness between the samples using the M-value with Pearson's correlation. The M-value is calculated as the log2 ratio of the densities of the methylated probe and the unmethylated probe (23). The interval of this rating diagram was established to provide a correlation coefficient (r) of −1≤ r ≤1. The samples were identified to have a strong positive correlation if r was close to +1. Our samples were identified as r=0.99 with strong positive correlation.

A dendrogram was drawn using M-values for the samples grouped using hierarchical clustering with Euclidean distance and complete linkage methods (Fig. 3). The diseased and healthy MZ twins were classified together into one group in accordance with the Euclidean distance clustering approach, and the other 4 healthy siblings were classified into a separate group. Investigation of the healthy siblings showed that the BRCA negative brother was classified into a different group from the BRCA mutation-carrying sisters. After clustering with the Euclidean distance method, methylation expression levels were analyzed in the aforementioned case and control groups. The differences between these groups were investigated with regard to two different aspects, namely association with disease and the presence of BRCA1 mutation.

Protein-protein interaction (PPI) analysis provides new data on protein functions and the general organizational principles of functional cellular networks (24). The STRING scores are indicators of confidence and rank from 0 to 1, with 1 being the highest possible confidence (25).

The biological functions of the genes were evaluated using the Protein Analysis Through Evolutionary Relationships (PANTHER) classification system, which is designed to classify proteins and genes according to their functions (26).

The sites with variations in methylation levels were identified in four different comparison groups for analysis according to diagnosis and BRCA1 mutation conditions.

The regions that were identified to have >10-fold differences in methylation levels by comparison of the MZ twins that were discordant for the presence of ovarian cancer in Group 1 are presented in Table III. In the MZ twin with ovarian cancer, hypermethylation was detected in the promoter region of the PR/SET domain 6 (PRDM6) (NM_001136239), RB-binding protein 7, chromatin repair factor (RBBP7) (NM_002893), ankyrin repeat domain 23 (ANKRD23) (NM_144994), RIB43A domain with coiled-coils 1 (RIBC1) (NM_001031745), C6orf227 (NR_027908) and clustered mitochondria homolog (CLUH) (NM_015229) genes, and hypomethylation was detected in the promoter region of the cytochrome B5 reductase 4 (CYB5R4) (NM_016230) gene. The sites of hypermethylation were located in CpG islets of the RBBP7, ANKRD23, RIBC1 and C6orf227 genes, and the northern (N) shore region of the CLUH gene, while the CYB5R4 gene was hypomethylated in the southern (S) shore region.

The regions with >10-fold difference in methylation levels when the MZ twin with ovarian cancer was compared with the other healthy siblings in Group 2 are presented in Table IV. The differentially methylated sites of the genes were as follows: RPL9, LIAS and CYP2U1 in CpG islets; ACTN3 and ZFA in S shelf regions; MYCBP, GJA9 and SEMA4D in S shore regions; and ZNF714, OR4D and CLUH in N shore regions.

Regions with >10-fold hypomethylation and with >25-fold hypermethylation (there were too many regions over 10-fold in this group, therefore 25-fold was used, for which more specific regions should be given) when the BRCA1 positive cases were compared with the BRCA1 negative case in Group 3 are presented in Table V. The hypermethylation sites of genes PQBP1, TIMM17B, FMR1 and AIFM1 were located in CpG islets, while those of the ARHGEF9, AR and RPL36A genes were in N shore regions. The TSC22D3 and DOCK11 genes were found to be hypermethylated at S shore sites.

The regions with >10-fold differences in methylation levels when the healthy MZ twin was compared with the other healthy siblings in Group 4 are presented in Table VI. With regard to hypomethylated sites, those of RPL9, LIAS and PRDM6 were located in CpG islets, ACTN3 and ZFAT were in S shelf regions, MYCBP, GJA9 and KLHL36 were in S shore regions, and OR4D1 was in the N shore region. The CYP2U1 gene was hypermethylated on CpG islets, the LOC25 3724 gene was hypermethylated in the S shelf region, and the CYB5R4 gene was hypermethylated in the S shore region.

Comparison of the MZ twin with ovarian cancer and the healthy MZ twin (Group 1) showed that the PRDM6 gene was hypermethylated in the MZ twin with ovarian cancer. However, PRDM6 was found to be hypomethylated in the healthy MZ twin compared with the other healthy siblings (Group 4). The CYB5R4 gene was demonstrated to be hypomethylated in the MZ twin with ovarian cancer compared with the healthy MZ twin (Group 1), but hypermethylated in the healthy MZ twin compared with the other healthy siblings (Group 4). Furthermore, the CYB5R4 gene was found to be hypermethylated in all healthy individuals, regardless of whether they were negative or positive for the BRCA1 mutation (Group 3). The genes RPL9, LIAS, TGFBI, ACTN3, SLC2A1-AS1, MYCBP, GJA9, KLHL36, LUZP1, HDAC4, OR4D1, UPF1, SHANK2, TG, FBXW12, FAM114A2, DYRK4, SLC25A13 and ZFAT were found to be hypomethylated in the MZ twin with ovarian cancer compared with the non-twin healthy siblings (Group 2) and the healthy MZ twin compared with the other healthy siblings (Group 4), while the genes TRIO, DNTTIP2, HIVEP2, CTNND2, CASQ2, C9orf171, FLNB, C7orf45, ACOT11, CYP2U, and ITIH3 were found to be hypermethylated in these comparison groups.

The genes RBBP7, ANKRD23, RIBC1, C6orf227 and CLUH were hypermethylated in the MZ twin with ovarian cancer compared with the healthy MZ twin (Group 1), while the CYB5R4 gene was hypomethylated. The ZNF714, OR52M1 and SEMA4D genes were observed to be hypomethylated, while the CHD1L, CAPZB and CLUH genes were hypermethylated in the MZ twin with ovarian cancer compared with non-twin healthy siblings (Group 2). Evaluation of the results obtained from all the comparison groups in the present study suggests that the methylation conditions of the PRDM6, CYB5R4, ZNF714, OR52M1, SEMA4D, CHD1L, CAPZB, CLUH, RBBP7, ANKRD23, RIBC and C6orf227 genes may be effective for the differentiation of ovarian cancer.

The comparison of BRCA1 positive and BRCA1 negative cases (Group 3) showed that the genes NADK2, KRT38, KIAA0513, ASAM, FNDC1, GSDMA, SFT2D1, C5orf33, CD24, TTTY14, TXNDC16, XG and TRAPPC12 were hypomethylated, and the genes CXorf26, FAM122C, ARHGEF9, PQBP1, TIMM17B, FMR1, AR, AIFM1, TSC22D3, RPL36A and DOCK11 were hypermethylated between the comparison groups. These methylation changes may be associated with the presence or absence of the BRCA1 gene mutation.

A PPI network was established in the present study to analyze the functions of the proteins encoded by the differentially methylated genes. The experimental data and the STRING functional protein association network v.10.5, which provides data on estimated interactions, was used in the analysis of the interactions. The protein network obtained was shown to have significantly higher protein interaction than expected (P=0.000331; Fig. 4). Two more STRING analyses were performed, which reported connections between CAPZB and FLNB, and BRCA1 and AR (data not shown).

The results of the PANTHER analysis revealed that the genes identified in the study had roles in biological processes including ‘biologic adhesion’, ‘regulation’, ‘cellular components organization’ and ‘biogenesis’, ‘immune system functioning’, ‘metabolic functioning’ and ‘localization’. Also, these genes were found to have roles in molecular functions including ‘attachment’, ‘catalytic activity’, ‘receptor activity’, ‘signal transmission’ and ‘translational regulation’.