To improve the poor prognosis of ovarian cancer, we attempted to develop a prediction system for individual response to chemotherapy. First we tried to identify useful prediction marker genes for paclitaxel and cisplatin responses from 22 candidate genes, including 12 marker genes suggested by our previous cDNA microarray analyses mainly using gastrointestinal cancer cell lines (Int. J. Cancer 111:617-26, 2004). Although some of these genes showed different expression patterns in paclitaxel-, cisplatin-, and the 2-drug resistant cell lines, resistance to paclitaxel and cisplatin in ovarian cancer cells could not be consistently predicted. Furthermore, quantitative prediction of resistance by using their expression data (95th AACR Annual Meeting) was inconclusive. To obtain genes that would account for differing sensitivity to the drugs in ovarian cancer cells, we thus extended the screening field to 37 genes generally known as sensitivity determinants for the 2 drugs. We performed MTT assay and real-time RT-PCR in 9 ovarian cancer cell lines and sorted out the genes correlated with resistance to the 2 drugs in expression levels. Among the 37 genes investigated, expression levels of ABCB1, CYP2C8, MGMT and NQO1 for paclitaxel and those of IL6, BCL2, VEGF, ERCC2 and NQO1 for cisplatin were found to correlate with resistance to each drug (P< 0.05). ABCB1 encodes an efflux pump of paclitaxel, and the product of CYP2C8 plays an important role in paclitaxel metabolism. MGMT and ERCC2 are DNA-repair enzyme genes and NQO1 is a bioreductive enzyme gene. IL6 encodes a cytokine promoting tumor growth, the product of BCL2 inhibits apoptotic cell death, and VEGF encodes a vascular endothelial growth factor. Nevertheless, none of the 9 genes alone could accurately predict resistance in KFr13Tx cells, a paclitaxel and cisplatin-resistant cell variant of KF28 ovarian cancer cell line, used as a test sample and not included in the 9 cell lines used in obtaining the formulae. By using multiple regression analysis, we fixed prediction formulae which embraced the variable expressions of all 4 genes for paclitaxel and all 5 genes for cisplatin selected above, and arranged them to predict resistance to each drug by referring to the value of Akaike’s information criterion for each sample. The obtained formulae quite accurately predicted in vitro resistance to each drug of KFr13Tx, so these genes are proposed as powerful prediction marker genes. Extension of the selection field may provide novel prediction marker genes and allow us to develop a more potent prediction formula. To identify more powerful ovarian cancer-specific drug sensitivity maker genes, we are now performing oligonucleotide microarray analysis to search for them among tens of thousands of genes.

[Proc Amer Assoc Cancer Res, Volume 46, 2005]