Abstract
Cholangiocarcinoma (CCA) arises within the peritumoral bile microenvironment, yet microbial translocation from bile to intracholangiocarcinoma (IntraCCA) tissues remains poorly understood. Previous studies on bile microbiota alterations from biliary benign disease (BBD) to CCA have yielded inconsistent results, highlighting the need for cross-study analysis. We presented a comprehensive analysis of five cohorts (N = 266), including our newly established 16S rRNA gene profiling (n = 42), to elucidate these microbiota transitions. The concordance of bacteria between CCA bile and intraCCA tissue, represented by Enterococcus and Staphylococcus, suggested microbiota migration from bile to intratumoral tissues. A computational random forest machine learning model effectively distinguished intraCCA tissue from CCA bile, identifying Rhodococcus and Ralstonia as diagnostically significant. The model also excelled in differentiating CCA bile from BBD bile, achieving an AUC value of 0.931 in external validation. Using unsupervised hierarchical clustering, we established Biletypes based on microbial signatures in our cohort. A combination of 17 genera effectively stratified patients into Biletype A and Biletype B. Biletype B robustly discerned CCA from BBD, with Sub-Biletype B1 correlating with advanced TNM stage and poorer prognosis. Among the 17 genera, bacterial Cluster 1, composed of Sphingomonas, Staphylococcus, Massilia, Paenibacillus, Porphyrobacter, Lawsonella, and Aerococcus, was enriched in Biletype B1 and predicted CCA with an AUC of 0.96. Staphylococcus emerged as a promising single-genus predictor for CCA diagnosis and staging. In conclusion, this study delineates a potential microbiota transition pathway from the gut through CCA bile to intra-CCA tissue, proposing Biletypes and Staphylococcus as biomarkers for CCA prognosis.