Objective
The purpose of this study was to analyze DNA methylation profiles among different types of ovarian serous neoplasm, which is a task that has not been performed.
Study Design
The Illumina beads array (Illumina Inc, San Diego, CA) was used to profile DNA methylation in enriched tumor cells that had been isolated from 75 benign and malignant serous tumor tissues and 6 tumor-associated stromal cell cultures.
Results
We found significantly fewer hypermethylated genes in high-grade serous carcinomas than in low-grade serous carcinoma and borderline tumors, which in turn had fewer hypermethylated genes than serous cystadenoma. Unsupervised analysis identified that serous cystadenoma, serous borderline tumor, and low-grade serous carcinomas tightly clustered together and were clearly different from high-grade serous carcinomas. We also performed supervised analysis to identify differentially methylated genes that may contribute to group separation.
Conclusion
The findings support the view that low-grade and high-grade serous carcinomas are distinctly different with low-grade, but not high-grade, serous carcinomas that are related to serous borderline tumor and cystadenoma.
DNA methylation is an important mechanism in the regulation of gene function and serves as an epigenetic marker for tumor detection, classification, and prognostication. DNA hypermethylation refers to the addition of a methyl group to the cytosine ring of those cytosines that precede a guanosine (referred to as CpG dinucleotides) to form methyl cytosine (5-methylcytosine). CpG dinucleotides are found at increased frequency in the promoter region of many genes. These regions have been termed CpG islands ; and, with the exception of genes on the inactive X chromosome and imprinted genes, CpG islands are protected from methylation in normal cells. This protection is critical, because methylation of CpG islands usually is associated with loss of expression of that particular gene. It has been demonstrated that the decreased expression of tumor suppressor genes and DNA repair genes are associated with promoter hypermethylation, which is a common feature in human cancer and serves as an alternative mechanism for loss of their function. For example, hypermethylation in the promoters of CDKN2 (p16) , VHL, WT1 , and MLH1 occurs in many solid tumors and is associated with loss of its expression, which indicates that aberrant DNA methylation is 1 of the main mechanisms in the pathogenesis of human cancer. To this end, 5-azacitidine has been shown to demethylate DR4, which results in its reexpression and leads in turn to enhanced sensitivity of platinum-resistant ovarian cancer cells to carboplatin through the induction of apoptosis.
Aberrant methylation of multiple CpG islands is frequently observed in ovarian carcinoma, compared with normal ovarian surface epithelium or benign ovarian neoplasms. For example, frequent CpG island hypermethylation in BRCA1 , RASSF1A , and OPCML is detected in ovarian carcinoma and plays a role in tumor development. Different histologic subtypes of ovarian cancer that include serous, endometrioid, and clear cell carcinoma demonstrate distinct DNA methylation profiles. Moreover, DNA methylation in certain gene promoters has been found to be a reliable epigenetic marker to predict treatment outcome in several types of human cancer, which include ovarian carcinoma. Thus, it has been suggested that DNA methylation changes have implications for ovarian cancer diagnosis, prognostication, and treatment. Although promoter methylation has been studied in ovarian carcinoma, a comprehensive analysis of methylation profiles has not yet been performed in different types of benign and malignant ovarian serous neoplasms that include serous cystadenoma, serous borderline tumor (SBT), and low-grade and high-grade serous carcinoma.
Recently, the Cancer Genome Atlas has performed the methylome profiles in a relatively large number of ovarian carcinoma, but the consortium only focuses on primary high-grade serous carcinomas. Accordingly, recurrent high-grade serous carcinoma, low-grade serous carcinoma, SBT, and benign serous cystadenomas were not analyzed. In this study, we applied the Illumina GoldenGate array (Illumina Inc, San Diego, CA) to profile DNA methylation and compare their patterns in high-grade serous carcinoma, low-grade serous carcinoma, SBT, and benign cystadenoma together with normal stromal cells. Our results provide new evidence that demonstrates that low-grade serous carcinoma is epigenetically distinct from high-grade serous carcinoma and is more closely associated to SBT and serous cystadenoma, which provides further support to the dualistic model of ovarian serous carcinogenesis. Therefore, the current study complements the Cancer Genome Atlas database and offers a unique opportunity to investigate the biologic significance of gene methylation in the pathogenesis of different types of ovarian serous neoplasms.
Materials and Methods
Tissue samples
Eighty-one samples that included 6 normal tissues (tumor adjacent stromal cells), 12 serous cystadenomas, 9 SBTs, 8 low-grade serous carcinomas, and 46 high-grade serous carcinomas (26 primary tumors, 13 recurrent tumors, and 7 specimens collected from primary ascites) were obtained from the Johns Hopkins Hospital except the tumor ascites were obtained from the Norwegian Radium Hospital. The high-grade serous carcinomas were all advanced stage (IIIC and IV). Samples were collected after appropriate review board approval according to the institutional guidelines. For low-grade and high-grade serous tumors, the tumor cells were enriched by Epi-CAM-conjugated magnetic beads according to the previously described method. The epithelial cells from SBT and cystadenoma were isolated by carefully scrapping the tumor surface on fresh tissues.
DNA preparation and methylation beads array
Genomic DNA was extracted with a Qiagen DNA extraction kit (Qiagen, Valencia, CA). Fully methylated DNA was prepared by treating genomic DNA with the SssI methylase (New England Biolabs, Beverly, MA) as the positive control. DNA that was prepared by whole genomic amplification was used as the unmethylated DNA control. Bisulfite conversion was performed with EZ DNA Methylation Kit (Zymo Research, Orange, CA); the Illumina GoldenGate beads arrays, which contained probes for 1505 CpG sites, were used to detect methylation profiles in different types of ovarian serous neoplasms. The methylation level was determined by the fluorescence intensity of methylated vs unmethylated alleles.
Clustering analysis
For unsupervised sample clustering, the methylation data were normalized or standardized for each CpG locus to mean and 1 standard deviation across all samples. The hierarchical sample clustering method was carried out based on all CpG loci with the Hierarchical Clustering Explorer tool, where the Pearson correlation coefficient was chosen as the distance metric and the clusters were merged with the average linkage criterion.
Comparative analysis
Significant Analysis of Microarrays (SAM) analysis was implemented on 2 groups of methylation data for the identification of the statistically significant hypermethylated or hypomethylated loci. In this study, a threshold cutoff of q value, 0.05, was used to define the number of significant loci. To assess the alterations in the overall methylation pattern among different clusters of tumor groups, we calculated the number of genes that showed hypomethylation and hypermethylation in each sample. This gene count was based on the probability value of < .05 in an individual gene, compared with the group of normal cells. To identify the differentially methylated genes between recurrent high-grade and primary high-grade serous carcinoma and between low-grade serous carcinoma and SBT, we used a supervised method to group samples under specific diagnosis, because unsupervised clustering failed to distinguish them. For each comparison, we carried out a significance test and selected the genes with q value of <0.05 using SAM. For each comparison, based on the genes previously selected, we tested the methylation level and calculated the mean and standard deviation of the normal samples for each selected gene. Then we used a threshold of ±1.65 standard deviation, which was equivalent to the probability value if .05 in a 1-tailed test. For each sample, we define those genes with a z -score larger than ±1.65 as hypermethylated and those with a z -score smaller than –1.65 as hypomethylated. Finally, the genes, which were significantly hypomethylated or hypermethylated, were counted for each sample.
Results
Unsupervised hierarchic clustering was performed to determine the similarity in methylation patterns among all the samples ( Figure 1 ). The results demonstrated that most serous cystadenomas, SBTs, and low-grade serous carcinomas clustered under a major branch; the normal tissues clustered in another distinct group. Most high-grade serous carcinomas, which included primary and recurrent tumors and tumor ascites, clustered into 2 separate branches and 1 additional minor branch that were embedded within the normal tissue branch. The morphologic features were similar in high-grade serous carcinomas in these different groups. Only 2 of 8 low-grade serous carcinomas and 3 of 9 SBTs did not fall within the cystadenoma/SBT/low-grade carcinoma group, whereas the remaining 24 (83%) of 29 cystadenoma/SBT/low-grade carcinoma samples clustered into 1 group. For high-grade serous carcinoma, only 1 specimen did not fall into 1 of the 3 main high-grade carcinoma clusters, whereas the remaining 45 (98%) of 46 high-grade serous carcinomas grouped into 3 clusters ( Figure 1 ). Interestingly, the high-grade serous carcinoma that was “misclassified” into the cystadenoma/SBT/low-grade carcinoma group that had been developed from a preexisting low-grade serous carcinoma.
