Aberrant methylation of promoter CpG islands is a well-documented tumor suppressor silencing mechanism that occurs in malignancy progression. Additionally, the onset of resistance to specific chemotherapeutics, such as the DNA-crosslinking agent cisplatin, has been linked with increased DNA methylation of specific genes, including hMLH1, which encodes a mismatch repair enzyme associated with apoptosis. To assess alterations in DNA methylation correlated with cisplatin resistance, including possible mechanisms, we utilized a commercially available methylation microarray of 82 gene promoters (Panomics, Redwood City, CA) to determine “methylation signatures” for cisplatin-sensitive A2780, cisplatin-resistant CP70, and normal ovarian surface epithelial (NOSE) cells. The methylation signatures of NOSE and A2780 cells were nearly identical, with 13 methylated and 68 unmethylated genes in common. In contrast, CP70 cells possessed 24 genes exclusively methylated, including hMLH1 and another putative tumor suppressor, NES1, suggesting increased methylation associated with the chemoresistant phenotype. To determine possible mechanisms of cisplatin-associated DNA methylation, we developed a cell culture model system. Single cells of cisplatin-sensitive A2780 were isolated and expanded. Chemosensitive clones were treated with variable doses of cisplatin, followed by assessment of cell number (by MTT assay) and determination of the 50% growth inhibitory dose (IC50). Each subsequent cycle entailed selection of resistant cells using the 70% growth inhibitory dose (IC70) of cisplatin, followed by redetermination of the IC50 values of the surviving cells. Following three cisplatin treatment cycles, IC50 doses increased from an average of 5 μM to 22 μM. To examine changes in DNA methylation, genomic DNA was isolated following each selection cycle and examined by the Panomics array and/or methylation-specific PCR (MSP). Following one treatment cycle, increased methylation of NES1 and hMLH1 was observed by MSP, validating our previous microarray result implicating those genes in cisplatin-associated methylation. To classify methylated and unmethylated sequences associated with cisplatin resistance, we built a computational model using eight sequence motifs found by MEME, a pattern discovery method. Using 10-fold cross validation, this model achieved a prediction accuracy of 79%. Such a model could be used to determine whether an unknown sequence could be methylated in association with cisplatin resistance. In summary, we demonstrate that DNA methylation plays a significant role in the development of platinum resistance in ovarian cancer cells; further studies are now underway to determine possible mechanisms and strategies for chemosensitization.

[Proc Amer Assoc Cancer Res, Volume 47, 2006]