Whole-Genome and Epigenomic Landscapes of Etiologically Distinct Subtypes of Cholangiocarcinoma

Abstract

Cholangiocarcinoma (CCA) is a hepatobiliary malignancy exhibiting high incidence in countries with endemic liver-fluke infection. We analyzed 489 CCAs from 10 countries, combining whole-genome (71 cases), targeted/exome, copy-number, gene expression, and DNA methylation information. Integrative clustering defined 4 CCA clusters-fluke-positive CCAs (clusters 1/2) are enriched in ERBB2 amplifications and TP53 mutations; conversely, fluke-negative CCAs (clusters 3/4) exhibit high copy-number alterations and PD-1/PD-L2 expression, or epigenetic mutations (IDH1/2, BAP1) and FGFR/PRKA-related gene rearrangements. Whole-genome analysis highlighted FGFR2 3' untranslated region deletion as a mechanism of FGFR2 upregulation. Integration of noncoding promoter mutations with protein-DNA binding profiles demonstrates pervasive modulation of H3K27me3-associated sites in CCA. Clusters 1 and 4 exhibit distinct DNA hypermethylation patterns targeting either CpG islands or shores-mutation signature and subclonality analysis suggests that these reflect different mutational pathways. Our results exemplify how genetics, epigenetics, and environmental carcinogens can interplay across different geographies to generate distinct molecular subtypes of cancer.Significance: Integrated whole-genome and epigenomic analysis of CCA on an international scale identifies new CCA driver genes, noncoding promoter mutations, and structural variants. CCA molecular landscapes differ radically by etiology, underscoring how distinct cancer subtypes in the same organ may arise through different extrinsic and intrinsic carcinogenic processes. Cancer Discov; 7(10); 1116-35. ©2017 AACR.This article is highlighted in the In This Issue feature, p. 1047.

Figure 1. Integrative Clustering Defines Four Molecular Subtypes of CCA

(A) Heatmap showing four clusters identified by iClusterPlus based on clustering of mutation, copy-number, gene expression and methylation data. Top rows indicate clinical characteristics, risk factors, geographical region, and sequencing platform. Microsatellite instability (MSI) status was defined by indel counts (≥ 6 indels) in simple repeat sequences. Bottom rows indicate selected genetic alterations. (B) High expression of PD-1, PD-L2 and BTLA in Cluster 3 relative to other clusters. Brown dots indicate MSI cases. Pink dots indicate cases with DNA polymerase epsilon (POLE) proofreading deficiency. (C) Survival analysis showing improved survival in Cluster 3 and 4 CCAs compared to other clusters. Multivariate analysis confirmed this difference even after accounting for fluke association, anatomical location, and clinical staging. (D) Representative genetic, epigenetic and gene expression features of CCA clusters.

Figure 2. Significantly Mutated Genes in CCAs

(A) Significantly mutated genes in Fluke-Pos and Fluke-Neg CCA. Genes in purple are mutated in both Fluke-Pos and Fluke-Neg CCAs. Novel significantly mutated genes are highlighted in red. (B) Alterations in Kinase-Ras/Raf pathway components across Fluke-Pos and Fluke-Neg CCAs. CCAs with FGFR/PRKACB rearrangements are also highlighted (arrows). (C) Matrix of genes (rows) and tumors (columns) showing occurrence of 32 somatic mutated genes. The bar chart at right shows frequencies of affected cases in Fluke-Pos and Fluke-Neg tumors. Asterisks indicate genes with significant differences between Fluke-Pos and Fluke-Neg CCAs. P-values were computed using the Fisher’s exact test. (D) Distribution of somatic mutations in RASA1. (E) RASA1 expression in tumors without RASA1 alterations (Wild-type) compared to tumors with RASA1 deletions and inactivating mutations (nonsense mutations or frame-shift indels). (F) RASA1 shRNA silencing inhibits CCA migration and invasion in vitro. Expression levels (mRNA and protein) of RASA1 in M213 (left) and HUCCT1 (right) cells transduced with two independent shRNAs (RASA1 shRNA#1 and RASA1 shRNA#2) targeting different regions of RASA1 were assessed by qPCR and Western blotting (first and second panel, respectively). Migration and invasion of RASA1 knockdown cells were assessed by transwell assays. Mean ± SEM of three independent experiments were analyzed. (G–H) Distribution of somatic mutations in STK11 (G) and SF3B1 (H). The red box indicates mutations in previously described SF3B1 hotspots.

Figure 3. FGFR and PRKA Gene Rearrangements in CCA

(A) Identification of FGFR2-STK26, FGFR2-WAC and FGFR2-TBC1D1 and FGFR2-BICC1 rearrangements in CCA. All fusions were validated by RT-PCR and sequence chromatograms are shown. FGFR2-STK26, FGFR2-WAC and FGFR2-TBC1D1 were validated in this study, while FGFR2-BICC1 was validated in (8). (B) Identification of a FGFR3-TACC3 gene fusion. Transcript validation was performed confirming a 7 bp insertion (red dotted lines). (C) Recurrent loss of 3′ UTRs in FGFR2 due to rearrangements with intergenic regions. (D) Relative luciferase activity between empty luciferase vector (LUC) and FGFR2-3′UTR in HEK293T and H69 immortalized cholangiocyte cell lines. Data is presented in Mean ± SD. Three individual experiments were performed. (E) FGFR2 gene expression levels between FGFR2-wildtype CCAs and CCAs exhibiting different categories of FGFR2 alterations, as shown by the color chart. (F) Identification of LINC00261-PRKACB and ATP1B1-PRKACB fusions. Both fusions were validated by RT-PCR and sequence chromatograms.

Figure 4. FIREFLY Analysis of Pathways Systematically Dysregulated by Somatic Promoter Mutations that Alter Transcription Factor Binding

(A) Changes in transcription factor (TF)-DNA binding estimated from PBM (protein binding microarray) data for 486 mammalian TFs. Changes in binding specificity were computed using PBM-derived binding scores for 8-mer sequences overlapping each mutation. To determine whether a given gene set was preferentially enriched for binding-change mutations, we computed the statistic M (the number of genes in the gene set with TF binding-change mutations in the promoter) summed over all tumors. FIREFLY assessed systematic enrichment of binding-change mutations with 2 statistical tests: (i) Fisher’s exact test of whether M is greater than expected by chance given the number genes in the gene set and the total number of genes affected by binding-change mutations, (ii) a comparison of M in actual data to a null distribution of M over 1,000 sets of 70 in-silico mutated tumor sequences, based on patient-specific trinucleotide contexts of mutations for each tumor. FIREFLY then tests for putative transcriptional dysregulation associated with the binding-change mutations by performing a Gene Set Analysis (GSA) to associate gene expression dysregulation with the number of binding-change mutations. (B) Details of the four gene sets meeting FIREFLY’s criteria of q < 0.1 for all three statistical tests: MIKKELSEN_MCV6_HCP_WITH_H3K27ME3, MIKKELSEN_MEF_ICP_WITH_H3K27ME3, MARTORIATI_MDM4_TARGETS_NEUROEPITHELIUM_DN, and WONG_ENDOMETRIUM_CANCER_DN. (C) Details of an example non-significant gene set.

Figure 5. Epigenetic Clusters and Integration of Mutation Signatures in CCA

(A) Heatmap showing three DNA methylation clusters: two hypermethylated clusters (Cluster 1 and Cluster 4), and a low methylation cluster (mixed Clusters 2 and 3). (B) Distinct methylation patterns in Cluster 1 versus Cluster 4. Top: typical organization of a gene promoter with CpG island and shores. Vertical ticks represent CpG sites. Bottom: Levels of promoter hypermethylation in CpG islands and shores in Cluster 1 and Cluster 4. (C) Left: enrichment of mutation Signature 1 (CpG>TpG) in Cluster 1. Right: similar levels of mutation Signature 5 among methylation clusters. Circled tumors represent MSI tumors. (D) Proximity of somatic mutations to hypermethylated CpGs in Cluster 1 and 4. Left: CpG>TpG mutations are located preferentially near hypermethylated CpGs in Cluster 1, but not in Cluster 4. Right: Non C>T mutations are not located preferentially near hypermethylated CpGs in either Cluster 1 or 4. (E) Histograms of corrected variant allele frequencies (VAF) of point mutations in Cluster 1 and 4.

Figure 6. Model for Distinct Pathways of CCA Tumorigenesis

A proposed model for CCA development in Clusters 1 and 4 being driven by distinct mechanisms. Cluster 1 may be initiated by extrinsic carcinogens (fluke-infection) causing genome-wide epigenetic derangement and subsequent spontaneous 5-methylcytosine deamination and CpG>TpG mutations. In contrast, in Cluster 4 CCAs, intrinsic genetic mutations in strong driver genes such as IDH1 reflect a primary initiating event and consequently drive DNA hypermethylation. See Discussion for details.