The Global Burden of Cholangiocarcinoma Cholangiocarcinoma (CCA) is the second most common hepatobiliary malignancy, accounting for 10-20% of primary liver cancers. Its epidemiology is strikingly geography-dependent: in Southeast Asia (particularly Northeast Thailand, Cambodia, and Laos), CCA is highly prevalent due to endemic infection by liver flukes (Opisthorchis viverrini and Clonorchis sinensis). In Western countries, the main risk factors are primary sclerosing cholangitis and hepatolithiasis. Despite these distinct etiologies, current classification is primarily anatomical or pathological rather than molecular.
Poor Prognosis and Limited Therapy Five-year survival rates after surgery and chemotherapy remain below 20%. Clinical trials of targeted therapies in unselected CCA populations have shown minimal benefit, suggesting that patient selection by molecular subtype is critical for therapeutic success.
A Global Genomics Effort On behalf of the International Cancer Genome Consortium, this study analyzed 489 CCAs from 10 countries using whole-genome sequencing (71 cases), exome/targeted sequencing, copy-number profiling, methylation analysis, and gene expression. Integrative clustering across 94 fully-profiled cases defined four molecular subtypes with distinct genetic, epigenetic, and clinical features linked to different etiological drivers.
Sample Collection 489 CCA samples were collected from 10 countries, including both fluke-associated cases (predominantly from Thailand) and non-fluke cases (predominantly from Western countries). Samples were analyzed using different platforms based on availability: 71 cases by whole-genome sequencing to an average depth of 64x, 200 by exome sequencing, 188 by high-depth targeted sequencing, with subsets analyzed for copy-number, methylation, and expression.
Integrative Clustering iClusterPlus was used to perform integrative clustering on 94 tumors with all four data types (mutations, copy-number alterations, mRNA expression, and methylation). Robustness was confirmed by randomized subsampling and expanded clustering with partial data profiles. Importantly, clusters persisted when samples were stratified by anatomical location, confirming that molecular clusters are not simply reflecting anatomical differences.
Non-Coding Genome Analysis A novel analysis framework incorporating experimentally-derived protein-DNA binding affinities and pathway information was applied to identify functional promoter mutations in the non-coding genome. This enabled detection of non-coding driver mutations that would be missed by exome-focused approaches.
Clusters 1 and 2 - Fluke-Positive CCAs Clusters 1 and 2 were predominantly fluke-associated CCAs, enriched in TP53 mutations and ERBB2 amplifications. Cluster 1 showed CpG island hypermethylation, ARID1A/BRCA1/2 mutations, and high promoter mutations affecting H3K27me3-regulated sites. Cluster 2 showed upregulation of CTNNB1, WNT5B, and AKT1. Both clusters were enriched in extrahepatic (perihilar and distal) anatomical locations and associated with poorer overall survival.
Cluster 3 - High Copy-Number, Immune-Active CCAs Cluster 3 comprised predominantly fluke-negative intrahepatic CCAs and displayed the highest level of somatic copy number alterations, including frequent chromosome 2p and 2q amplifications. Strikingly, Cluster 3 showed specific upregulation of immune checkpoint genes (PD-1, PD-L2, BTLA) and pathways related to antigen cross-presentation and T cell signaling, suggesting potential responsiveness to immune checkpoint therapy.
Cluster 4 - Epigenetic Mutation-Driven CCAs Cluster 4 was also intrahepatic and fluke-negative, characterized by IDH1/2 mutations, BAP1 mutations, and FGFR alterations with upregulated FGFR family and PI3K pathway signatures. Rather than CpG island hypermethylation (as in Cluster 1), Cluster 4 showed hypermethylation at CpG shores, suggesting a distinct mechanism of epigenetic carcinogenesis driven by intrinsic genetic insults rather than extrinsic carcinogens.
Newly Identified Driver Genes Whole-genome analysis uncovered new CCA driver genes not previously identified by exome-focused studies: RASA1 (a RAS GTPase-activating protein), STK11 (encoding LKB1, a tumor suppressor), MAP2K4 (a MAPK pathway kinase), and SF3B1 (a splicing factor with known driver roles in hematologic malignancies). These genes represent new therapeutic targets in CCA.
FGFR2 3-prime UTR Deletion A novel structural variant mechanism was discovered: deletion of the FGFR2 3-prime UTR. This deletion removes regulatory sequences that normally limit FGFR2 mRNA stability and expression, providing a non-coding mechanism for FGFR2 upregulation. This finding explains FGFR2 overexpression in some CCAs lacking FGFR2 fusion genes and expands the repertoire of patients who might benefit from FGFR inhibitors.
Non-Coding Promoter Mutations The novel analysis framework for non-coding genome mutations identified pervasive modulation of H3K27me3-associated promoter sites. These promoter mutations in repressed chromatin regions could reactivate silenced oncogenes, representing a mechanism of epigenetic dysregulation that would be invisible to protein-coding genomic analyses.
Two Types of CCA Hypermethylation A striking finding was that two CCA clusters showed distinct and complementary patterns of DNA hypermethylation targeting different genomic regions. Cluster 1 (fluke-associated) showed hypermethylation enriched at CpG islands - transcription start site-proximal regions where methylation typically silences gene expression. Cluster 4 (IDH-mutant/BAP1 non-fluke) showed hypermethylation preferentially at CpG shores - regions 2kb flanking CpG islands with regulatory functions.
Two Mechanisms of Carcinogenesis These distinct methylation patterns suggest fundamentally different carcinogenic mechanisms. Fluke-associated CCA (Cluster 1) methylation resembles the pattern induced by extrinsic carcinogens - the liver fluke secretes pro-inflammatory and potentially carcinogenic molecules into bile ducts, driving CpG island methylation. In contrast, Cluster 4's IDH1/2 mutations produce 2-hydroxyglutarate which directly inhibits TET demethylases and histone demethylases, causing the shore-predominant methylation pattern.
Subclonality Analysis Mutational signature and subclonality analysis provided evidence that these different methylation patterns reflect early versus late events in tumor evolution. Understanding the temporal sequence of epigenetic changes relative to mutational events in each subtype has implications for understanding cancer initiation and identifying windows for intervention.
FGFR Inhibitors Cluster 4's FGFR alterations (fusion genes and the newly identified 3-prime UTR deletion) identify a population that may benefit from FGFR inhibitors - a class of drugs that has since shown impressive results in clinical trials for FGFR2-altered CCA. Pemigatinib was FDA-approved for FGFR2-fusion CCA in 2020, validating this approach.
Immune Checkpoint Blockade Cluster 3's upregulation of PD-1, PD-L2, and immune activation pathways identifies this subtype as the most likely to respond to checkpoint inhibitors. The combination of high copy-number alterations (which may generate neoantigens) and an active immune microenvironment supports a strategy of PD-1/PD-L1 blockade specifically in Cluster 3 patients.
ERBB2 and IDH Targeting ERBB2 (HER2) amplification in Clusters 1 and 2 identifies patients for HER2-targeted therapy (trastuzumab, tucatinib). IDH1/2 mutations in Cluster 4 are targetable by approved inhibitors (ivosidenib for IDH1), and the IDH-mutant methylation phenotype may predict sensitivity to other epigenetic therapies. Together, these findings provide a framework for matching CCA patients to specific molecular therapies.
Sample Heterogeneity Challenges The study combined samples from 10 countries analyzed on different platforms, introducing potential batch effects despite careful statistical controls. Prospective collection of uniformly processed samples with comprehensive multi-platform profiling would strengthen confidence in the molecular subtype definitions and enable refinement of cluster boundaries.
Functional Validation of New Drivers The newly identified driver genes (RASA1, STK11, MAP2K4) require functional validation in cell and animal models to confirm their roles in CCA initiation and progression. Understanding how these genes interact with the etiology-specific molecular features of each CCA subtype is important for developing rational combination therapies.
Clinical Trial Design The most urgent research priority is designing biomarker-stratified clinical trials that test subtype-specific therapies in matched patient populations. Given the rarity of CCA, international multi-center collaboration (as demonstrated by this study) is essential for accruing sufficient patients to each molecular subtype to achieve statistical power in randomized trials.