Mutational spectrum and precision oncology for biliary tract carcinoma

Abstract

Background: The genomic spectrum of biliary tract carcinoma (BTC) has been characterized and is associated with distinct anatomic and etiologic subtypes, yet limited studies have linked genomic alterations with personalized therapies in BTC patients. Methods: This study analyzed 803 patients with BTC:164 with gallbladder cancer, 475 with intrahepatic cholangiocarcinoma (ICC) and 164 with extrahepatic cholangiocarcinoma. We determined genomic alterations, mutational signatures related to etiology and histopathology and prognostic biomarkers. Personalized targeted therapies for patients harboring potentially actionable targets (PATs) were investigated. Results: The median tumor mutation burden (TMB) was 1.23 Mut/Mb, with 4.1% of patients having hypermutated BTCs. Unlike the results obtained from the Western population, the most frequently altered cancer-related genes in our cohort included TP53 (53%), KRAS (26%), ARID1A (18%), LRP1B (14%) and CDKN2A (14%). Germline mutations occurred mostly in DNA damage repair genes. Notably, 35.8% of the ICCs harbored aristolochic acid related signatures and an elevated TMB. TP53 and KRAS mutations and amplified 7q31.2 were demonstrated to negatively affect patient prognosis. Moreover, 19 genes were proposed to be PATs in BTCs, with 25.4% of patients harboring these PATs. Forty-six patients received PAT-matched targeted therapies, achieving a 26.1% objective response rate; the median progression-free survival (PFS) was 5.0 months, with 56.8% of patients obtaining PFS benefits. Conclusions: Extensive genomic diversity and heterogeneity were observed among BTC patients, with contributions according to potential etiology exposures, anatomical subtypes and clinicopathological characteristics. We also demonstrated that patients with refractory BTCs who have PATs can derive considerable benefit from receiving a matched therapy, initiating further prospective clinical trials guided by molecular profiling among this aggressive cancer.

Keywords: biliary tract cancer; genomic alterations; molecular screening; precision medicine; targeted therapy.

Figure 1

Genomic mutation profiles of 803 BTC patients. (A) Numbers and proportions of the three anatomical subtypes of BTC. (B) Landscape of tumor mutation burden (TMB) across the major tumor types; the median level of TMB for each tumor type and our cohort is highlighted. (C) Mutation profiles of driver genes detected by MutSigCV and frequently mutated BTC-related genes. Mutant frequencies in the cohort are shown on the right, and associated clinicopathological characteristics for all 803 patients are shown at the bottom.

Figure 2

Identification of chromosomal somatic copy-number alterations (SCNAs) in putative cancer driver genes. (A) Significant (q < 0.1) focal SCNAs along all chromosomes. The vertical axis indicates the G-scores generated from GISTIC2, which considers the amplitude of the aberration and the frequency of its occurrence across samples. Recurrent SCNAs of putative cancer driver genes are also highlighted. (B) Proportions of patients from the entire cohort (n = 803) for with significantly amplified (red) or deleted (blue) cancer driver genes. (C) Correlations between driver mutations and clinical phenotypes in the entire cohort; significant correlations are highlighted in red. Two-sided Fisher's exact tests were then performed, and an FDR cutoff of 0.05 was used for reported genes. (D) Scatter plots depicting the mutational frequencies (percentage of patients) between patients with hypermutated tumors and non-hypermutated tumors in our cohort. Each dot represents one gene, and dots are color coded according to the P-values (-log10(P) uncorrected) shown in the legend. Statistics shown were derived from two-sided Fisher's exact tests. (E-F) Comparisons of TMB between mutant and wild-type KMT2D (E) or FAT3/4 (F) separated by histological subtypes.

Figure 3

Identification of DDR mutants and germline variants. (A) Percentage of patients with the three different BTC subtypes with somatic mutations in DDR genes classified by TP53 and various functional families of non-TP53 DDR genes. (B) Somatic mutation spectrum of non-TP53 DDR genes in BTC ranked by their prevalence. The color of the bars represents the type of genomic alteration. (C) Patients harboring mutated DDR genes have a significantly higher TMB than those with wild-type DDR; this was consistent across the three individual BTC subtypes. (D) Numbers of patients with pathogenic or pathogenic-likely germline variants; the findings indicate that PRSS1, BRCA2, MUTYH, SPINK1 and BRCA1 mutations are recurrent germline mutations in BTC patients. (E) Annotations and locations of PRSS1 (upper) and SPINK1 (bottom) mutations in our cohort.

Figure 4

Analyses of mutational signatures in ICC and prognostic alterations. (A) Contribution of the three mutational signatures among 148 ICC patients. (B) Signature A identified from ICC samples is linked to COSMIC age-related single-base substitutions (SBSs) of SBS1. (C) Aristolochic acid (AA) signature (signature B) mutations were identified in 53 of 148 ICC tumors. The relative mutation frequencies of all 96 trinucleotide mutation patterns are plotted, with AA-related mutation patterns labeled in gray. (D) Signature C is linked to COSMIC age-related SBS40. (E) Comparisons of TMB in tumors with and without the AA signature. The line and box represent the median and upper and lower quartiles, respectively. (F) Negative impacts of TP53 and KRAS mutations and 7q31.2 amplification on postoperative disease-free survival (DFS) in 140 patients. (G) Negative impacts of TP53 and KRAS mutations and 7q31.2 amplification on postoperative overall survival (OS) in 160 patients.

Figure 5

Potentially actionable targets (PATs) and biomarker-guided targeted therapies in BTC patients. (A) Pie graph of the percentage of patients harboring at least one PAT in the entire cohort (803 patients with BTCs); the populational proportion of PATs in each subtype of BTCs is shown as a bar plot. (B) Pie graphs show the percentage of PATs in mutation-originated targets (left) for all 803 BTCs and fusion-originated targets (right) among 643 patients who underwent CSYS panel analysis. As some patients had multiple PATs, the percentages do not add up to 100%. (C) The stacking diagram represents the number of patients and number of mutational sources for each PAT. (D) The upper scattered heatmap indicates the histotype, PAT count, relative highest OncoKB tier and highest ESCAT tier for each patient with available PATs among 204 patients. The colored bars at the bottom indicate the different levels of each parameter. (E) Efficacy-pathologic-genomic landscape of umbrella-setting BGTs in 46 patients with advanced or metastatic BTCs. The central scatter plot shows the response outcomes (x-axis) and PFS2/PFS1 ratio (y-axis) for each PAT. Targets with a PFS2/PFS1 ratio ≥ 1.3 and a responsive status (CR or PR) are highlighted in the center plot. (F) Survival analysis of progression-free survival (PFS) estimated by Kaplan-Meier curves among 46 patients who received BGTs. (G) The flow diagram in the left panel illustrates the list of druggable targets, the corresponding therapeutic response and the PFS2/PFS1 ratio. The colors of the curved belts indicate whether the druggable targets belonged to the 19 genes proposed as PATs in BTC, and the widths of the belts indicate different frequencies for each target at every level. Abbreviations: MSI-H: microsatellite instability-high; CR: complete response; PR: partial response; SD: stable disease; PD: progressive disease; PFS: progression-free survival; ECOG: Eastern Cooperative Oncology Group; dMMR: deficient mismatch repair.