DNA methylation subtypes inferred from transcriptomes guide sunitinib and immunotherapy selection in renal cancer

Abstract

Responses to targeted and immune checkpoint therapies in metastatic clear cell renal cell carcinoma (ccRCC) are heterogeneous, and no predictive biomarker is available. We performed a meta-analysis of DNA methylation profiles from 727 ccRCC tumors, benchmarked four published signatures, and identified a hyper-methylated subtype (15% of tumors) associated with shorter overall survival. Integration of 589 matched transcriptomes showed that hyper-methylated tumors were more proliferative, depleted in endothelial tip cells, and overexpressed DNMT3B while underexpressing TET2. We then developed a gene expression classifier based on nine gene ratios that predicted methylation subtypes in an independent cohort (AUC-ROC = 0.96). In four treated cohorts, predicted hyper-methylated tumors had shorter progression-free survival on sunitinib but not on immune checkpoint-based regimens. Hyper-methylated primary tumors responded better to atezolizumab plus bevacizumab than to sunitinib, whereas hypo-methylated metastases benefited more from sunitinib. Methylation subtypes may thus help guide treatment selection in metastatic ccRCC.