CCA patients exhibit significant gut microbiota dysbiosis, characterized by reduced microbial diversity, depletion of beneficial taxa (e.g., Faecalibacterium), and enrichment of pathobionts (e.g., Escherichia-Shigella), which collectively disrupt gut-liver axis homeostasis ( Table 1 ). Of note, α-diversity alterations exhibit marked heterogeneity across CCA subtypes, with conflicting reports in extrahepatic and hilar variants (30). Emerging evidence on iCCA consistently demonstrate microbial dysbiosis accompanied by reduced α-diversity, likely reflecting tumor microenvironment-driven ecological pressures. In a study by Zhang et al., it was revealed that patients with iCCA exhibited significantly reduced abundances of beneficial gut bacteria Bacteroides, Faecalibacterium, and Roseburia, along with increased levels of pathogenic taxa Escherichia-Shigella and Subdoligranulum, compared to healthy controls (16). This study firstly unveiled that gut microbiota modulates glutamine metabolism to downregulate the ALK5/NOX1 axis, thereby inhibiting ferroptosis in CCA cells, offering a promising therapeutic strategy for precision treatment in iCCA. Moreover, Ito et al. reported enrichment of Enterobacteriaceae (harbouring the polyketide synthase carcinogenic island) in biliary tract cancer (BTC) faecal samples, suggesting its direct pro-carcinogenic role in biliary tract carcinogenesis (31). A case-control study and a cohort analysis found no significant difference in gut microbiota α-diversity between CCA patients and healthy controls (21, 32). Notably, within this cohort study, a machine learning-driven approach leveraging a random forest algorithm identified eight discriminative bacterial genera (e.g., Faecalibacterium, Klebsiella) to develop a tripartite microbial signature-based classifier, achieving exceptional diagnostic accuracy (AUC: 0.92–0.99) in effectively differentiating CCA from HCC (21). Furthermore, CCA patients exhibited increased abundance of Gram-negative bacteria (Enterobacteriaceae), correlating with elevated systemic inflammation markers (neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio). In contrast, Jia et al. revealed elevated α-diversity patterns in iCCA patients through a multi-omics approach integrating taxon-specific microbial biomarkers (notably Lactobacillus and Alloscardovia) with functional BA metabolism signatures (33). Despite established links between specific gut microbiota alterations​​ (e.g., decreased Faecalibacterium, increased Escherichia-Shigella/Enterobacteriaceae) ​​and CCA pathogenesis​​ (16, 31, 33), ​​interpretation remains constrained by methodological limitations. Predominantly small cohort sizes​​ compromise statistical power and hinder validation/generalizability of findings, including biomarker classifiers or subtype-specific diversity patterns (e.g., the α-diversity heterogeneity observed) (16, 21, 31–33). ​​Furthermore, reliance on observational designs​​ impedes causal inference due to uncontrolled, dynamic confounders such as recent antibiotic use, dietary fluctuations, and variable underlying liver pathology (e.g., cirrhosis severity). Consequently, ​​attributing dysbiosis directly to CCA versus confounders/disease consequences remains challenging.​​ Future research necessitates ​​large-scale, multi-center prospective cohorts​​ with ​​rigorous longitudinal monitoring of key confounders (antibiotics, diet, liver function)​​ and ​​robust causal inference methods.​.

Mendelian randomization analyses leveraging MiBioGen gut microbiota genome-wide association study (GWAS) data have identified host genetic variants causally associated with microbial compositional changes. These analyses revealed that elevated abundances of Eubacterium hallii group, Candidatus Soleaferrea, and Flavonifractor confer increased BTC risk, while Dorea and Lachnospiraceae ND3007 demonstrated protective effects (34). Complementary Mendelian randomization studies further implicated Veillonellaceae, Alistipes, Enterobacteriales, and Firmicutes as iCCA-promoting taxa (35). Mechanistically, protective genera such as Anaerostipes, Paraprevotella, Parasutterella, and Verrucomicrobia enhance intestinal barrier integrity via butyrate production or exert anti-inflammatory activity, underscoring the dual diagnostic and therapeutic potential of gut microbiota in hepatobiliary malignancies. Notably, iCCA patients demonstrate gut mycobiota dysbiosis characterized by Candida spp. (particularly C. albicans) enrichment and depletion of beneficial Saccharomyces cerevisiae, with advanced TNM stages (III–IV) correlating strongly with pathogenic fungal predominance (36). These findings position gut microbiota dysbiosis as a driver of iCCA progression, warranting exploration of antifungal therapies ( Supplementary Figure 1 ). However, critical gaps persist in addressing bacterial-fungal ecological interactions central to dysbiosis mechanisms and establishing experimental validation of causality between specific fungal taxa (e.g., C. albicans) and oncogenesis through animal models or in vitro systems.

While emerging evidence reveals subtype-specific gut microbial configurations differentiating CCA and healthy controls, current investigations remain constrained by geographically restricted cohorts. Future studies should prioritize mechanistic validation through germ-free animal models ​​integrated with​​ multi-ethnic validation cohorts—​​stratified into three principal groups​​: (1) Southeast Asian cohorts (≥40% allocation, bearing >50% of the global CCA burden), (2) European/North American cohorts (addressing industrial/environmental confounders), and (3) latent-risk diaspora cohorts. Crucially, ancestry-based principal component analysis with structured recruitment quotas must mitigate cryptic population substructure and allelic sampling biases, ​​thereby ensuring robust control​​ of geographic and migrant heterogeneity. ​​This integrated methodology​​ will enable rigorous delineation of causal host-microbiota interactions while controlling population stratification artifacts.

Emerging evidence positions the gut microbiota as a pivotal determinant of CCA progression through multifaceted molecular mechanisms. Prognostically, tumor vascular invasion (VI) shows strong correlations with specific gut microbial alterations (37) ( Supplementary Figure 1 ). VI-positive patients exhibit distinct dysbiotic patterns marked by Oscillospiraceae enrichment and depletion of beneficial taxa (Eubacteriaceae, Allobaculum, Pediococcus), patterns that independently predict reduced survival and elevated recurrence risks (33). Notably, advanced TNM stages (III-IV) in intrahepatic CCA show marked Candida albicans overgrowth in gut mycobiota, suggesting fungal dysbiosis as a potential progression biomarker (36). The microbial-immune interface further reveals VI-associated immunological shifts, with Ruminococcaceae-enriched cases displaying elevated IL-4 and suppressed IL-6 levels (21), highlighting microbiota-driven immune microenvironment remodeling.

Therapeutically, gut microbial signatures demonstrate predictive value for treatment responses across modalities. While Bacteroidetes enrichment (particularly Alistipes sp. Marseille-P5997) associates with improved anti-programmed cell death protein 1(PD-1) immunotherapy outcomes in biliary tract cancers (38), Proteobacteria dominance inversely correlates with Sintilimab-anlotinib combination efficacy in advanced cases (39). Mechanistically, microbial metabolites like cyclic dinucleotide c-di-AMP enhance radiotherapy-induced antitumor immunity through STING pathway activation (40), revealing novel microbiome-mediated therapeutic sensitization strategies. Nevertheless, critical knowledge gaps persist in delineating precise microbiota-immune crosstalk mechanisms and validating microbial biomarkers for clinical translation. Systematic investigation of host-microbiota metabolic interactions and standardized multi-omics approaches remain imperative to harness the full therapeutic potential of gut microbiome in CCA management.

The intestinal barrier serves as a critical interface that maintains microbial homeostasis through selective permeability, effectively restricting the translocation of exogenous and microbiota-derived molecules under physiological conditions (41). While inherently plastic adapts to physiological demands, this barrier exhibits vulnerability to age-related deterioration, environmental stressors, and pathological insults. Gut microbiota-derived metabolites, including those directly synthesized by bacteria or enzymatically transformed from dietary and host-derived molecules, exert oncogenic effects via bidirectional crosstalk with host signaling pathways. Dysbiosis disrupts intestinal barrier integrity, enabling systemic translocation of microbial components, including structural molecules (e.g., lipopolysaccharides, LPS) and metabolic byproducts (e.g., short-chain fatty acids (SCFAs)) (42–44). These translocated products, particularly endotoxins, access the liver via portal circulation, activating hepatic Kupffer cells (KCs) to initiate pro-inflammatory cytokine cascades (e.g., tumor factor alpha [TNF-α], IL-6) that perpetuate chronic inflammation, bile duct injury, and metabolic dysregulation (45, 46).

Elevated circulating LPS, a hallmark of Gram-negative bacterial dysbiosis, is clinically associated with chronic hepatobiliary diseases and malignancy progression. Impaired tight junction integrity (e.g., reduced Claudin-1/Occludin expression) facilitates LPS translocation to the liver (47). where it drives tumorigenesis via TLR4-mediated activation of oncogenic pathways (e.g., NF-κB, PI3K/AKT) (48, 49). Diet-induced metabolic perturbations exacerbate this cycle by promoting barrier dysfunction and endotoxemia, as evidenced in nonalcoholic fatty liver disease models (50, 51). In PSC models, dysbiosis-driven LPS translocation activates hepatic TLR4/MyD88 signaling, exacerbating cholangitis and fibrosis to establish a pro-carcinogenic niche (52). Patients with iCCA exhibit gut microbiota shifts marked by opportunistic pathogen enrichment (e.g., Veillonella, Klebsiella), with pathogen-derived LPS activating TLR4 on cholangiocytes and KCs to induce NF-κB-driven IL-6/TNF-α production, amplifying biliary injury and cholestasis (21). Mechanistically, LPS enhances tumor aggressiveness via METTL3-mediated PI3K/AKT pathway activation, promoting tumor cell migration and invasion (17). Additionally, LPS orchestrates immunosuppression in the tumor microenvironment (TME) by triggering TLR4-dependent hepatocyte C-X-C motif chemokine ligand (CXCL)1 secretion, which recruits polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs) to dampen antitumor immunity (19). These findings collectively position the LPS/TLR4 axis as a therapeutic target, suggesting microbiota modulation and microenvironment reprogramming as viable strategies to combat iCCA progression.

The biliary system harbors gut-associated microbes including Escherichia coli and Enterobacter spp., translocated via the gut-liver axis or establish local colonization (53). Systematic characterization by Carlo et al. revealed that biliary enrichment of E. coli and Klebsiella pneumoniae serves as an independent prognostic marker for reduced survival in pancreaticobiliary malignancies, highlighting their potential as diagnostic biomarkers (54). Notably, Alcaligenes faecalis demonstrates TME-specific colonization in eCCA, exhibiting diagnostic potential for early tumor detection and therapeutic monitoring (55). Contrasting with cholangitis-associated microbiotas, perihilar CCA exhibits a Gram-positive predominance featuring Enterococcus spp., which independently predicts portal vein thrombosis and multidrug resistance development (56). Furthermore, biliary colonization by Enterococcus or Candida spp. correlates strongly with clinical progression in PSC, solidifying microbial dysbiosis as a prognostic determinant (57). Critically, antibiotic-resistant strains such as Enterococcus faecalis and Enterobacter spp. in hepatic ducts directly exacerbate postoperative infection rates and mortality in eCCA, underscoring the clinical imperative for microbiota-guided perioperative management (58). These findings collectively establish biliary microbiota as both a driver of oncogenesis and a therapeutic target in CCA progression.

Comprising approximately 770 phylogenetically diverse species, the oral microbiota emerges as a pivotal regulator of systemic immunity and metabolic homeostasis with broad disease implications (59, 60). The functional crosstalk between oral and gut microbiotas has gained prominence given their collective influence on human pathophysiology (61), particularly through pathobiont translocation along the gut-liver axis during dysbiosis (62). Clinically, Rao and colleagues developed an oral microbiota-based diagnostic algorithm incorporating three bacterial biomarkers (Lautropia, Alloprevotella, and Actinomyces), achieving exceptional accuracy (AUC=0.981) in differentiating iCCA from HCC (63). A complementary Korean cohort study further demonstrated stratified oral microbial signatures across upper gastrointestinal malignancies, revealing significantly elevated α-diversity in oesophageal/gastric cancers compared to BTC and pancreatic cancer cohorts (64). Intriguingly, while distinct diagnostic microbial profiles were identified for oesophageal/gastric malignancies, BTC and pancreatic cancers lacked comparable signatures. This disparity suggests an anatomic microbial gradient along the gastrointestinal tract, where tumor proximity to the oral cavity correlates positively with both microbial dysbiosis severity and diagnostic biomarker detectability. While BTC-specific oral microbial biomarkers remain unidentified, these findings provide foundational evidence for the oncogenic involvement of the oral-gut-liver axis in BTC pathogenesis. Subsequent research must prioritize multi-institutional cohorts with longitudinal sampling to resolve spatiotemporal dynamics of microbial translocation and enable mechanistic dissection of tumor-promoting crosstalk within this axis.

Serum total bile acid levels progressively escalate across the disease continuum from healthy controls to benign biliary disease (BBD) and CCA patients, correlating with biliary obstruction severity (68). Conjugated bile acids exhibit marked accumulation, particularly primary conjugated BAs (glycocholic acid [GCA], taurochenodeoxycholic acid [TCDCA]) showing marked increases in both serum and bile compared to BBD and healthy controls (69, 70) ( Table 2 ). Paradoxically, secondary conjugated species (glycolithocholic acid [GLCA], glycoursodeoxycholic acid [GUDCA]) demonstrate malignant-specific reduction (71). Concurrently, primary unconjugated BAs (cholic acid [CA], chenodeoxycholic acid [CDCA]) accumulate due to cholestasis-induced excretion impairment (69), whereas secondary unconjugated derivatives (deoxycholic acid [DCA], lithocholic acid [LCA]) decline in plasma, reflecting gut microbial 7α-dehydroxylation defects (72). This conjugated/unconjugated BA ratio imbalance strongly correlates with suppressed bacterial deconjugation activity, establishing a pathogenic feedback loop between hepatic BA synthesis and gut microbial metabolic reprogramming that drives carcinogenesis. Thus, between-study heterogeneity in BA levels (particularly CCA vs BBD discrepancies) arises from methodological variables including biospecimen type (serum/bile/tissue), cohort heterogeneity, and gut microbiota-mediated BA metabolic variation. Critically, current BA metabolomic validation in CCA faces three critical gaps: (1) ​​absence of multi-center harmonization​​ in pre-analytical (sample collection/storage) and analytical (LC-MS) workflows; (2) ​​deficient dynamic monitoring​​ failing to track tumor progression/therapy-induced BA flux changes; and (3) ​​inadequate pathological specificity​​ against cirrhosis/cholangitis confounders. These limitations mandate prospective longitudinal cohorts with protocolized multi-omics integration to resolve CCA-specific BA dynamics.

Cholestasis-induced BA overload initiates a self-perpetuating pathogenic cycle central to CCA progression. Excessive BAs activate hepatocyte-cholangiocyte transdifferentiation via dual receptor systems: nuclear receptors (farnesoid X receptor, FXR) and membrane-bound receptors (Takeda G protein-coupled receptor 5, TGR5; sphingosine-1-phosphate receptor 2, S1PR2) (73, 74) ( Figure 2 ). Sustained BA-receptor engagement generates a pro-tumorigenic niche marked by oxidative DNA damage (via reactive oxygen species [ROS] overproduction) and apoptotic resistance (through CD95 inactivation), thereby completing malignant transformation. This BA signalling duality manifests through distinct pathways: (1) Nuclear receptor modulation: Although frequently downregulated in CCA (84), FXR activation by CDCA or synthetic agonists (e.g., GW4064) exerts tumor-suppressive effects (75, 76). Mechanistically, FXR activation induces small heterodimer partner (SHP) expression, which suppresses STAT3 phosphorylation to downregulate Bcl-xL in BTC cells, ultimately triggering apoptosis and inhibiting proliferation (77–79). (2) Membrane receptor cascades: TGR5 promotes cholangiocyte proliferation through extracellular signal-regulated kinase (ERK)1/2 phosphorylation via a cascade comprising ROS generation, Src kinase activation, and epidermal growth factor receptor (EGFR) transactivation. Simultaneously, TGR5 triggers PKA-mediated CD95 serine/threonine phosphorylation that blocks death receptor signalling, conferring apoptotic resistance, with these dual mechanisms synergistically driving cholangiocarcinogenesis (80–82). S1PR2 axis: TCA-mediated S1PR2 activation triggers ERK1/2/Akt signalling through both G protein-dependent and EGFR-mediated transactivation pathways, subsequently inducing NF-κB activation and cyclooxygenase-2 (COX-2) upregulation. This signalling axis coordinately upregulates proinflammatory cytokines (IL-6, TNF-α), thereby establishing a self-sustaining inflammatory niche that promotes CCA progression (83–85).

BA signalling orchestrates TME immunosuppression and metastatic priming through macrophage polarization and T-cell functional suppression. S1PR2 activation by conjugated BAs upregulates COX-2/PGE2 signalling, which directly suppresses CD8+ cytotoxic T lymphocyte activity while promoting regulatory T cell infiltration, thereby establishing an immune-evasion niche conducive to tumor progression (83, 84). The functional duality of TGR5 pivots on ​​disease stage-specific microenvironmental cues​​. In ​​cholestatic conditions​​ (e.g., early-stage PSC), TGR5 activation exerts hepatoprotective effects by mediating M2 macrophage polarization and suppressing NF-κB-mediated inflammation​​ via cAMP-PKA signalling (52). Conversely, within the ​​established TME of advanced CCA, TGR5 signalling shifts toward pro-tumorigenic activity by conferring apoptosis resistance and enhancing proliferation in malignant transformed biliary epithelial cells (81). The precise molecular mechanisms and pathological staging thresholds governing the transition of TGR5 from hepatoprotective to protumorigenic function remain incompletely characterized. Current evidence suggests this transition may correlate with ​​progression of biliary intraepithelial neoplasia, ​​advanced cirrhotic remodelling​​, ​​sustained BA accumulation​​, and ​​immune reprogramming. This receptor dichotomy underscores the delicate equilibrium between BA-mediated immune regulation and malignant transformation in hepatobiliary diseases. Elucidating these context-dependent switches represents a critical research priority for developing stage-adapted TGR5-targeted therapies in CCA.

Clinical evidence highlights the critical association between dysregulated BA metabolism and the diagnosis, prognostic evaluation, and therapeutic strategies for CCA. BA profiling has emerged as a potential source of non-invasive biomarkers for CCA diagnosis ( Table 2 ). Zhang et al. (70), developed a diagnostic model based on CDCA and TCDCA, which effectively distinguished CCA from BBD, gallbladder cancer, and HCC, demonstrating significantly superior specificity compared to the conventional biomarker CA19-9. Multi-omics studies further identified a quadruple diagnostic panel comprising hyodeoxycholic acid, isoLCA, bCDCA, and DCA, achieving 93.3% sensitivity and 86.7% specificity (72). Notably, serum GCA exhibits diagnostic potential but shows limited specificity in differentiating malignant from benign biliary disorders (69).

Metabolomic stratification studies demonstrate that BA metabolic activity serves as an independent prognostic determinant in CCA. Molecular subtyping based on BA metabolism features classifies CCA into metabolically active and inactive subgroups, where the active subtype exhibits significantly shorter OS and impaired immunotherapy response (86). These findings establish BA metabolic reprogramming as a mechanistic basis for precision oncology, providing a molecular classification framework for personalized therapeutic strategies. Intriguingly, while conjugated primary BAs (e.g., TCA, TCDCA) demonstrate strong correlation with HCC mortality, this association remains absent in BTCs, suggesting tumor-specific metabolic rewiring mechanisms (87). Overcoming these challenges necessitates collaborative multicenter studies that integrate machine learning pipelines to discriminate disease-relevant metabolic signatures from inter-individual physiological variability.

Gut microbiota-derived enzymes convert primary BAs to tumorigenic secondary species (DCA, LCA) that activate oncogenic signalling cascades (91, 92). Clostridium scindens and other 7α-dehydroxylase-expressing species generate secondary BAs associated with diet-induced hepatocarcinogenesis models, where dysbiosis elevates circulating DCA. This hydrophobic secondary BA drives tumorigenesis by inducing senescence-associated secretory phenotype activation, ROS overproduction, and NF-κB-mediated inflammation via gut barrier disruption (93). Notably, DCA impairs hepatic CXCL16 secretion in liver sinusoidal endothelial cells, reducing CXCR6+ natural killer cell (NKT) recruitment and compromising antitumor immunity. Pharmacological depletion of Gram-positive 7α-dehydroxylase bacteria (e.g., vancomycin) attenuates secondary BA production and exerts chemopreventive effects in preclinical models (94). Furthermore, microbial bile salt hydrolases dynamically remodel the BA pool, facilitating intestinal absorption or further modification of BA (95). Therapeutic microbiota modulation demonstrates dual mechanisms: (1) Lactobacillus rhamnosus GG (LGG) activates FXR to reduce BA accumulation in cholestatic liver disease (96, 97); (2) Pediococcus pentosaceus Li05 enhances bile salt hydrolase activity while suppressing NLRP3 inflammasome activation in PSC (98, 99). These findings position gut microbiota as master regulators of BA homeostasis, linking microbial ecology to inflammation-driven carcinogenesis.

BAs critically shape gut microbiota composition and homeostasis through their chemical properties and signalling functions (67, 100). Primary BAs (e.g., CA, CDCA) exhibit surfactant-like properties that disrupt bacterial cell membrane integrity and interfere with DNA stability, thereby inhibiting pathogens such as Clostridium difficile, Salmonella spp., and Listeria spp. Notably, Gram-positive bacteria (Firmicutes) demonstrate heightened susceptibility to these antimicrobial effects compared to Gram-negative counterparts, a phenomenon attributable to structural differences in their cell wall architectures (101). Moreover, gut microbiota converts primary BAs into secondary BAs (e.g., DCA) through bile salt hydrolase and 7α-dehydroxylase activities. This biotransformation not only modulates BA toxicity and signalling activity but also reshapes microbial ecology via metabolic crosstalk. DCA selectively inhibits beneficial bacteria (e.g., Lactobacillus) while promoting the expansion of pathobionts such as Enterobacteriaceae, fostering a pro-inflammatory microenvironment (22).

Furthermore, BA signalling orchestrates gut microbial homeostasis through BA receptor-mediated pathways, predominantly FXR. Mechanistically, FXR activation stimulates the expression of antimicrobial peptides such as angiogenin-1, effectively curbing the expansion of pathobionts within the Proteobacteria phylum (102), while concurrently fortifying intestinal barrier function by upregulating tight junction proteins (e.g., occludin, ZO-1) (103). This dual regulatory axis not only maintains microbial equilibrium but also mitigates systemic inflammation triggered by bacterial endotoxin leakage. Future studies should delineate the mechanisms through which BA pool-mediated microbiota modulation governs host-microbial interactions.

Targeted microbial modulation—via antibiotics, probiotics, or faecal microbiota transplantation (FMT)—represents a potential therapeutic strategy for CCA by restoring microbiota homeostasis (104) ( Table 4 ). These interventions mechanistically counteract dysbiosis-driven oncogenesis through pathogen depletion, intestinal barrier restoration, and suppression of protumorigenic signalling. However, translational progress remains limited, with current evidence largely confined to preclinical models.

Metabolic dysregulation, characterized by activation of BA, fatty acid, and xenobiotic pathways, represents a hallmark of CCA (121). Emerging strategies aim to disrupt BA-driven oncogenic networks while preserving metabolic homeostasis. S1PR2 inhibitors (e.g., JTE-013) suppress tumor invasiveness in CCA preclinical models by inhibiting ERK/AKT/NF-κB signalling cascades (85). Similarly, TGR5 activation in CCA cells promotes tumor progression, its downregulation in cholangiocytes and tumor-associated macrophages attenuates inflammation through NF-κB and MAPK/ERK pathway inhibition (52, 122). Intriguingly, TGR5 signalling exhibits ​​a stage-dependent duality​​: it ​​exerts hepatoprotective effects​​ during early cholestasis but ​​becomes pro-tumorigenic in established CCA (123). This functional switch necessitates carefully timed therapeutic interventions.​​ During the therapeutic window of biliary intraepithelial neoplasia, agonists​​ like INT-777 ​​may​​ ​​mitigate​​ malignant transformation ​​by restoring​​ anti-inflammatory and epithelial barrier functions. Conversely, in advanced CCA characterized by CAF-rich microenvironments, antagonists ​​such as​​ SBI-115 ​​can block​​ pro-fibrotic and pro-metastatic signalling cascades, ​​requiring stage-adapted regimens to optimize the therapeutic index.

FXR agonists (e.g., INT-747/OCA) demonstrate therapeutic potential in attenuating sepsis-induced cholestasis and intestinal injury. In cecal ligation and puncture -induced septic mice, INT-747 upregulates intestinal FXR/FGF15 signalling, preserving gut barrier integrity and reducing hepatic cholestasis (124). Clinically, OCA is approved for PSC treatment (125), with recent advances utilizing ROS-responsive nanoparticles for targeted OCA delivery to enhance efficacy (126). FXR expression was found to be markedly downregulated in samples of primary CCA (127, 128). OCA counteracts this deficit by restoring BA homeostasis via SHP/LRH-1-dependent pathways, suppressing CCA proliferation and migration (127). Critically, ​​heterogeneous FXR expression across CCA subtypes directs subtype-specific therapeutic strategies.​​ ​​In chemotherapy-resistant iCCA models​​, the ​​combination of FXR agonists (e.g., GW4064, CDCA)​​ with standard chemotherapy (cisplatin/gemcitabine) ​​potentiates anti-proliferative and pro-apoptotic responses​​ (129). ​​Conversely, pCCA and dCCA subtypes frequently exhibit low-to-moderate FXR levels​​ (H-score <120), ​​necessitating the priming use of epigenetic modulators​​ (e.g., DNMT inhibitors) ​​to restore agonist sensitivity​​ (78, 130). ​​Concerning carcinogenic risk, chronic FXR activation displays a complex tissue-selective duality:​​ while systemic FXR loss promotes hepatocarcinogenesis and tumor-specific FXR silencing characterizes the HCC/CCA microenvironment (131), hyperactivation within specific gastrointestinal tissues may drive β-catenin-dependent oncogenesis (132). Therefore, the tissue-selective functional duality of FXR signalling necessitates rigorous long-term safety evaluation in forthcoming preclinical and clinical investigations.

Certain genetic mutations indirectly disrupt BA metabolism. For instance, frequent IDH1/2 mutations in iCCA drive 2-hydroxyglutarate accumulation, which epigenetically impairs cell differentiation and may suppress BA synthesis genes (e.g., CYP7A1) via DNA hypermethylation (133, 134). Ivosidenib (AG-120), a selective IDH1 inhibitor, demonstrated improved progression-free survival in the phase III ClarIDHy trial, becoming the first targeted therapy approved for IDH1-mutant CCA (135). HNF4A, a master regulator of hepatobiliary differentiation, directly modulates bile acid synthesis enzymes and transporters (e.g., BSEP/ABCB11) (136, 137). While HNF4A activation may suppress tumor progression by restoring BA homeostasis, direct targeting agents remain unavailable. Despite preclinical evidence supporting BA metabolic reprogramming as a therapeutic strategy for CCA, BA-targeted agents remain largely experimental with limited clinical translation. This gap necessitates large-scale trials and mechanistic studies to rigorously validate efficacy and safety, accelerating their clinical integration.

The pronounced heterogeneity of CCA necessitates therapeutic targeting based on robust molecular subtyping. Multi-omics approaches—encompassing genomics, metagenomics, and spatial metabolomics—systematically dissect the core mechanisms underlying gut-liver axis dysregulation, including critical microbiota-bile acid-host interactions. These approaches yield quantifiable and targetable biomarkers (e.g., dysbiosis subtypes, cancer-associated fibroblast BA dependency phenotypes) essential for subsequent therapeutic development phases. However, clinically applicable gut-liver axis biomarkers validated through large-cohort studies remain lacking. Phase I prioritizes the validation of foundational biomarkers through global consortia (ICCA-GUT), standardizing profiling to map BA distribution. Leveraging single-cell spatial transcriptomics will reveal cancer-associated fibroblast-specific BA dependencies. Cross-cancer comparative analyses will identify conserved vulnerabilities like pan-hepatobiliary butyrate deficiency. Machine learning frameworks will correlate IDH1/2status with microbiota-metabolome covariation, culminating in a validated dysbiosis subtype classifier (AUC>0.90). This establishes a closed-loop pipeline for refining precision medicine, from multi-omics discovery to mechanism-guided translation.

Building on validated biomarkers, this phase utilizes AI platforms (CCA-TIME v1.0) to decode the multicellular interactome of tumor microenvironment, identifying therapeutic targets. The platform quantifies multidimensional efficacy-toxicity landscapes, forecasting synergistic interactions (e.g., probiotic-enhanced dendritic cell function augmenting anti-PD-1 responses) (138). Real-time toxicity monitoring flags risks like antibiotic-induced immunotherapy resistance. These insights guide master protocol trials: (1) Arm A targets dysbiosis index ≥5.2 tumors with FMT + anti-PD-1 to correct gut-liver axis dysregulation, and (2) Arm B addresses FXR-methylated subtypes via DNMT inhibitor + OCA combinations that destabilize chemotherapy-resistant microenvironments through cholesterol homeostasis modulation (121). This phase systematically converges microbial engineering, metabolic targeting, and AI-driven design to transition CCA management from empirical to mechanism-guided interception.

Synthetic biology tackles CCA heterogeneity through precision-engineered bacteria with logic-gated (“sense-decide-act”) capabilities, targeting dysregulation via tripartite mechanisms: microbiota modulation, BA metabolic intervention, and antitumor immunity activation (139). Crucially, this phase integrates outputs from prior stage: Phase I-derived single-cell metatranscriptomics enables AI-powered prediction of strain-engineered metabolic perturbations via hepatic flux balance analysis, informing designs like multifunctional consortia (BA regulation), Lactobacillus reuteri (IL-22-mediated mucosal repair), and attenuated Salmonella typhimurium (hypoxia-targeted gemcitabine delivery). Phase II-validated synergy models (e.g., butyrate-enhanced dendritic cell priming) guide strain-immunotherapy combinations. Convergence with nanomedicine (e.g., liver-tropic nanoparticles co-loaded with FXR agonists/probiotics) or immunotherapies (e.g., anti-PD-1/butyrate) amplifies therapeutic indices (126). Short-term priorities include strain optimization (kill switches, hypoxia-triggered gemcitabine) and ecological integration via butyrate-producing consortia for pan-hepatobiliary malignancies. The pivotal Phase III milestone requires demonstrating ≥30% reduction in CAF, validating engineered microbiota against desmoplastic CCA progression.

Therapeutic development for CCA faces significant interspecies translational barriers. Murine models (e.g., KRAS/p53 KO) inadequately recapitulate human cholestatic pathophysiology due to species-specific BA metabolism—notably CYP8B1 deficiency—and the absence of key immune components, including Kupffer cells (140, 141). Further complexity arises from BA-specific receptor (e.g., TGR5) and metabolite (e.g., butyrate) exhibiting context-dependent dual roles (142), necessitating single-cell spatiotemporal resolution to dissect microenvironmental interactions. Biomarker standardization remains elusive, with analytical variability in quantifying >40 carcinogenic BA isoforms and microbial interference with BA gene signatures (e.g., CYP7A1/SLC10A2 ratios) confounding treatment predictions (70). Compounding these challenges, interindividual variability driven by genetic polymorphisms, geographic disparities in microbiota composition (e.g., Bacteroides abundance in Asian versus Western populations) (143, 144), and dietary influences (e.g., high-fat diet-induced TGR5 activation (93)) collectively modulate therapeutic responsiveness, underscoring the need for precision stratification models.

The clinical translation of gut-liver axis modulators faces unresolved safety-efficacy dilemmas. Probiotic-related bacteremia, particularly septicemia caused by Clostridium butyricum, is predominantly concentrated in high-risk post-hepatectomy patients (e.g., with neutropenia or bilioenteric anastomoses), contrasting sharply with therapeutic benefits (145). This risk is mechanistically linked to intestinal barrier disruption (e.g., anastomotic leaks) facilitating bacterial translocation. Broad-spectrum antibiotics deplete oncobiotic taxa but concomitantly disrupt chemotherapy efficacy—for instance, by depleting folate-producing Faecalibacterium, thereby contributing to 5-FU resistance. This paradox highlights the critical need for optimizing this risk-benefit balance. Synthetic microbial therapies present significant biocontainment challenges. Engineered bacteria (e.g., utilizing the Escherichia coliNissle 1917 chassis) demonstrate concerning horizontal gene transfer frequencies (10^-³–10^-5 CFU/recipient via plasmids) in murine models, potentially facilitating dissemination of virulence genes (e.g., ctxfrom Vibrio cholerae) (146, 147). Furthermore, the limited standardization of FMT protocols and unclear pathogenic thresholds impede clinical reproducibility.

Current static multi-omics approaches fail to capture the dynamic heterogeneity of CCA, particularly evolving microbiota-immune-metabolite interactions needed for adaptive therapy. While spatial metabolomics and AI enable microenvironment mapping, clinical translation is limited by inconsistent analytical standardization across workflows. The absence of closed-loop systems integrating longitudinal multi-omics data (e.g., IDH-mutation gut-liver signatures) with adaptive algorithms impedes progress from empirical treatment towards AI-driven precision oncology. Addressing this requires collaborative validation of predictive biomarkers through prospective trials and development of real-time monitoring framework.