DNA methylation signature for the assessment of metastatic risk in primary renal cell cancer

516 Background: Transition from localized renal cell cancer (RCC) to metastatic disease is associated with an immense mortality. Beside clinicopathological risk estimation, a molecular prediction of metastatic risk from primary cancers with a sufficient diagnostic accuracy is not available. We biometrically identified a candidate metastasis associated methylation signature (MAMS) in a genome-wide in silico DNA methylation analysis of TCGA data and carried out a double evaluation study using tissues from localized RCC, primary metastatic RCC, and distant metastases. Methods: Candidate MAMS was identified by genome-wide random forest analyses of TCGA methylation level 3 data aiming for classification of 230 primary RCC without distant metastases and 52 metastasized tumors. Forty-nine pyrosequencing and/or quantitative methylation specific PCR analyses were established for a total of 20 candidate methylated loci. For evaluation of MAMS, DNA was isolated and bisulfite converted from the primary RCC tissue cohort (n=187) as well as 99 distant metastases. Localized RCCs consisted of 92 pT1a/b tumors (clear cell, 73; papillary, 16; chromophobe, 2; not classified, 1) without lymph node or distant metastases. RCC samples of primary metastatic disease were obtained from 31 patients. Results: Single candidate loci showed specific hypermethylation when metastatic tissues were compared to primary RCC samples. Random forest analysis showed that MAMS including nine methylation loci achieved a sensitivity of 93% and a specificity of 89% (AUC 0.95) for differentiation of localized RCC and metastases (chi square, p = 3.4*10-29). Positive, negative likelihood, and diagnostic odds ratios were 8.6, 0.08, and 108. Using this random forest model for prediction of primary RCCs associated with distant or lymph node metastases revealed a sensitivity of 58% and specificity of 94% (AUC 0.84, p = 1.0*10-9). Positive, negative likelihood, and diagnostic odds ratios were 9.2, 0.45, and 20.5. Conclusions: A negative MAMS test result retrospectively reduced the probability of metastasis 15-fold in the localized disease cohort, suggesting a prospective evaluation for a possible clinical translation of MAMS in handling RCC patients.