Driver mutations of intrahepatic cholangiocarcinoma shape clinically relevant genomic clusters with distinct molecular features and therapeutic vulnerabilities

Abstract

Purpose: To establish a clinically applicable genomic clustering system, we investigated the interactive landscape of driver mutations in intrahepatic cholangiocarcinoma (ICC). Methods: The genomic data of 1481 ICCs from diverse populations was analyzed to investigate the pair-wise co-occurrences or mutual exclusivities among recurrent driver mutations. Clinicopathological features and outcomes were compared among different clusters. Gene expression and DNA methylation profiling datasets were analyzed to investigate the molecular distinctions among mutational clusters. ICC cell lines with different gene mutation backgrounds were used to evaluate the cluster specific biological behaviors and drug sensitivities. Results: Statistically significant mutation-pairs were identified across 21 combinations of genes. Seven most recurrent driver mutations (TP53, KRAS, SMAD4, IDH1/2, FGFR2-fus and BAP1) showed pair-wise co-occurrences or mutual exclusivities and could aggregate into three genetic clusters: Cluster1: represented by tripartite interaction of KRAS, TP53 and SMAD4 mutations, exhibited large bile duct histological phenotype with high CA19-9 level and dismal prognosis; Cluster2: co-association of IDH/BAP1 or FGFR2-fus/BAP1 mutation, was characterized by small bile duct phenotype, low CA19-9 level and optimal prognosis; Cluster3: mutation-free ICC cases with intermediate clinicopathological features. These clusters showed distinct molecular traits, biological behaviors and responses to therapeutic drugs. Finally, we identified S100P and KRT17 as “cluster-specific”, “lineage-dictating” and “prognosis-related” biomarkers, which in combination with CA19-9 could well stratify Cluster3 ICCs into two biologically and clinically distinct subtypes. Conclusions: This clinically applicable clustering system can be instructive to ICC prognostic stratification, molecular classification, and therapeutic optimization.