Abstract

Background

   Cholangiocarcinoma is a highly heterogeneous tumor with bile acid
   metabolism involving in its development. The aim of this study was to
   characterize bile acid metabolism and identify specific subtypes to
   better stratify cholangiocarcinoma patients for individualized
   treatment and prognostic assessment.

Methods

   A total of 30 bile acids were quantified using the ultra-performance
   liquid chromatography tandem mass spectrometry. Using Consensus
   clustering, the molecular subtypes related to bile acid metabolism were
   identified. The prognosis, clinicopathologic characteristics, immune
   landscape, and therapeutic response were compared between these
   subtypes. The single-cell RNA sequencing (scRNA-seq) analysis and
   preliminary cell experiment were also conducted to verify our findings.

Results

   The altered bile acid profile and genetic variation of bile acid
   metabolism-related genes in cholangiocarcinoma were demonstrated. The
   cholangiocarcinoma was categorized into bile acid metabolism-active and
   -inactive subtypes with different prognoses, clinicopathologic
   characteristics, tumor microenvironments (TME) and therapeutic
   responses. This categorization was reproducible and predictable.
   Specifically, the bile acid metabolism-active subtype showed a poor
   prognosis with an immunosuppressive microenvironment and an inactive
   response to immunotherapy, while the bile acid metabolism-inactive
   subtype showed the opposite characteristics. Moreover, the scRNA-seq
   revealed that immunotherapy altered bile acid metabolism in TME of
   cholangiocarcinoma. Finally, a prognostic signature related to bile
   acid metabolism was developed, which exhibited strong power for
   prognostic assessment of cholangiocarcinoma. Consistently, these
   results were verified by immunohistochemistry, cell proliferation,
   migration, and apoptosis assays.

Conclusion

   In conclusion, a novel cholangiocarcinoma classification based on bile
   acid metabolism was established. This classification was significant
   for the estimation of TME and prognosis.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12885-024-13081-0.

   Keywords: Bile acids, Subtypes, Prognosis, Cholangiocarcinoma, Tumor
   microenvironment

Background

   Cholangiocarcinoma (CCA) is a highly heterogeneous malignant tumor with
   the characteristics of biliary epithelium [[32]1, [33]2]. The genome,
   transcriptome, epigenome, and proteome of CCA all exhibit extensive
   heterogeneity [[34]3], which is essential for the carcinogenesis and
   drug resistance and influences the prognosis of CCA patients [[35]4,
   [36]5]. 5-year survival rate of CCA is about 7–20%, and both its
   morbidity and mortality have been rising globally [[37]6]. Hence, it is
   of the essence to explore the molecular characteristics, distinguish
   different subtypes, precisely predict the therapeutic response and
   prognosis, and thus develop individualized treatment strategies of CCA.

   Bile acid is an important component of bile and is synthesized by
   cholesterol in the liver [[38]7]. Under the joint contribution of the
   host and gut microbes, more than 50 different bile acids (including
   primary/secondary bile acids and free/conjugated bile acids) are
   produced [[39]8, [40]9]. It has been reported that bile acid
   accumulation may contribute to bile duct carcinogenesis by triggering
   cholangiocyte inflammation and proliferation [[41]10, [42]11].
   Additionally, diverse types of bile acids have been shown to have
   different effects on cholangiocarcinoma. Specifically, free bile acids
   inhibit CCA, while conjugated bile acids motivate it [[43]12, [44]13].
   In addition, recent studies have manifested that bile acids serve as
   essential molecules in the regulation of intracellular signaling
   pathways through receptors such as FXR or TGR5 [[45]14–[46]17]. By
   activating these different signaling pathways, bile acids modulate not
   only glucose homeostasis, cholesterol, triglyceride, and energy
   [[47]14], but also immune cell differentiation, recruitment, and
   function [[48]15–[49]17]. Therefore, the importance of bile acid
   metabolism in the tumor microenvironment (TME) has attracted increasing
   attention [[50]18–[51]20]. However, the molecular alteration and
   heterogeneity of bile acid metabolism in the TME of CCA are poorly
   understood. The exploration of the characteristics and subtypes of bile
   acid metabolism and associated TME features may provide insight into
   the crosstalk between metabolism and immune to better predict prognoses
   and therapeutic responses and develop individualized therapies of CCA.

   This study analyzed the altered bile acid profile in CCA. The CCA was
   categorized into bile acid metabolism-active and -inactive subtypes
   with different prognoses, clinicopathologic characteristics, TME and
   therapeutic responses. This categorization was reproducible and
   predictable. In addition, the association between immunotherapy and
   bile acid metabolism was validated using single-cell sequencing
   (scRNA-seq) data from the TISCH database [[52]21]. Finally, a
   prognostic signature based on bile acid metabolism was developed and
   validated experimentally. The results emphasized the bile acid
   metabolic heterogeneity of CCA, and the obtained stratification has
   clinical implications for driving the development of bile acid
   metabolism-targeted diagnosis and treatment.

Methods

Patients and datasets

   In this study, healthy individuals(n = 16) and patients with
   CCA(n = 13) from the First Affiliated Hospital of Zhejiang University
   School of Medicine were enrolled. The data of FU-iCCA cohort (n = 255)
   were obtained from the biosino NODE database (OEP001105,
   [53]https://www.biosino.org/node/project/detail/OEP001105) [[54]22].
   [55]GSE89749 (n = 118), [56]GSE26566 (n = 104), and [57]GSE107943
   (n = 30) from the GEO database were employed for the following
   verification (Fig. [58]1) [[59]23–[60]25]. Among them, the clinical
   data of [61]GSE26566 were not available (Supplementary Table [62]1).
   The single-cell sequencing analysis of [63]GSE125449 was performed in
   the TISCH database [[64]21, [65]26]. A total of 82 bile acid
   metabolism-related genes (BRGs) were retrieved from MSigDB (Fig. [66]1)
   (Supplementary Table [67]2) [[68]27].

Fig. 1.

   [69]Fig. 1
   [70]Open in a new tab

   The workflow of this study BRGs, bile acid-related genes; CCA,
   cholangiocarcinoma; CNVs, copynumber variations. scRNA-seq, single cell
   RNA sequencing

Bile acid quantification

   Bile acids were quantified as previously described by an
   ultra-performance liquid chromatography tandem mass spectrometry
   (UPLC-MS/MS) system (ACQUITY UPLC-Xevo TQ-S, Waters Corp., Milford, MA,
   USA) [[71]28]. MassLynx software(v4.1, Waters, Milford, MA, USA) was
   used to process the raw data generated by UPLC-MS/MS to establish a
   standard curve and quantify for each bile acid.

Consensus clustering

   Using the “ConsensusClusterPlus” R package, Consensus clustering was
   performed. The expression of BRGs was standardized by the “sweep”
   function and then clustered. Principal component analysis was applied
   to verify differences in gene expression between subtypes. To compare
   prognostic differences between subtypes, the “survminer” R program was
   utilized. In addition, the consistency of subtype classification across
   several datasets was examined using an unsupervised Subclass Mapping
   approach (SubMap) [[72]29].

Functional enrichment analysis

   Differentially expressed genes among subtypes were identified using the
   “Limma” or “Deseq2” package, and the threshold was set as adjp < 0.05,
   |log2Foldchange|>1. KEGG, GO and GSEA pathway enrichment analysis was
   conducted through the “clusterProfiler” package. For GSVA analysis, the
   gene set “H.ALE.V2023.1.hS. symbols” was downloaded from MSigDB and
   analyzed using the “GSVA” package.

Evaluation of immune cell infiltration

   CIBERSORTx (stanford.edu) was employed to explore the infiltration of
   22 immune cells in cholangiocarcinoma. To calculate the enrichment
   levels of 28 immunological signatures, single-sample gene-set
   enrichment analysis (ssGSEA) was performed by the “GSVA” package. The
   correlation of immune cells with bile acid-related genes was portrayed
   using the R-package “corrplot”.

Drug sensitivity of bile acid metabolism-related subtypes

   The sensitivity of common drugs was analyzed by gene expression
   profiling using the R package “pRRophetic”. Immunophenoscore (IPS) was
   used to evaluate the effect of immunotherapy against PD-1 and CLAT-4 in
   patients, and the “IOBR” package was used for this purpose [[73]30,
   [74]31].

Single-cell RNA sequencing analysis

   The single-cell RNA sequencing analysis of [75]GSE102988 was performed
   in the Tumor Immune Single Cell Hub (TISCH) as previously reported
   [[76]21]. Datasets in TISCH were processed using a standard analysis
   method according to MAESTRO v1.1.0.

Identification and verification of a bile acid metabolism-related signature
with prognostic significance

   BRGs with prognostic significance were initially identified by
   univariate COX analysis. Then, LASSO regression was performed using the
   “glmnet” package. Finally, multivariate COX regression was applied to
   determine a bile acid-related signature with prognostic significance.
   For receiver operating characteristic (ROC) analysis and Kaplan-Meier
   analysis, the “survivalROC” package and the “survival” package were
   utilized, respectively. This prognostic signature was validated in
   [77]GSE89749 (n = 118) and [78]GSE107943 (n = 30).

Cell culture

   The human cholangiocarcinoma cell lines RBE and HUCC-T1 were obtained
   from the MeisenChinese Tissue Culture Collections (Hangzhou, China).
   Cells were cultured using Dulbecco’s modified Eagle’s medium (DMEM)
   with 10% fetal bovine serum (FBS) (Gibco by Life Technologies,
   Bleiswijk, the Netherlands). All cells were incubated under 37 °C at 5%
   CO[2].

Cell transfection

   Vector (pcDNA3.1) and pcDNA3.1-SLCO1B3, pcDNA3.1-CEACAM1 were purchased
   from GenePharma (Shanghai, China). 2 µg plasmid and jetPRIME^®
   Versatile DNA/siRNA transfection reagent (Polyplus; Sartorius, Europe)
   were used to transfect CCA cells. Reverse Transcription Quantitative
   Polymerase Chain Reaction (RT-qPCR) and Western Blot (WB) were used to
   verify the overexpression of SLCO1B3 and CEACAM1.The cells were
   harvested and used for the subsequent experiment after 48 h of
   transfection.

Cell counting kit-8 (CCK-8)

   The RBE and HUCC-T1 cells (8000 cells/well) were cultivated in 96-well
   plates. After 48 h of transfection, 10 µl CCK-8 solution
   (MedChemExpress, Shanghai, China) was added to every well. A microplate
   reader (Bio-Tek Instruments, Winooski, VT, USA) was employed to detect
   the absorbance values at 450 nm after another 2 h of incubation at
   37 °C.

Cell apoptosis

   Annexin V-FITC/PI apoptosis detection kits (Vazyme, Shanghai, China)
   were used to detect cell apoptosis after 48 h of transfection. Cell
   apoptosis rate (%) = (advanced apoptotic cells + early apoptotic
   cells)/total cells × 100% was the formula used to calculate the rate of
   cell apoptosis.

Wound healing assay

   5 × 10^5 CCA cells were plated in 6-well plates and then transfected. A
   line wound was scratched by a 10-µl pipette tip. Photographs were taken
   at 0, 24 h, 48 h and 72 h after wounding.

Results

The altered bile acid profile and genetic variation of bile acid
metabolism-related genes in CCA

   In this study, serum from CCA patients(n = 13) and healthy
   individuals(n = 16) were collected to quantify 30 different bile acids.
   It was discovered that primary bile acids tended to rise, whereas
   secondary bile acids predisposed to fall in CCA patients (Fig. [79]2A).
   However, regardless of primary or secondary bile acids, serum levels of
   bile acids were higher in advanced CCA in comparison with early CCA
   (Fig. [80]2B). Time-dependent receiver operating characteristic (ROC)
   analyses also suggested that bile acids including TCA, GCA, TCDCA,
   NorCA, isoLCA, and βDCA had an excellent diagnostic performance for CCA
   (Fig. [81]2C).

Fig. 2.

   [82]Fig. 2
   [83]Open in a new tab

   The altered bile acid profile and genetic variation of bile
   acid-related genes in cholangiocarcinoma The altered bile acid profile
   in patients with CCA (A). Increased bile acids in patients with
   advanced cholangiocarcinoma(B). ROC curve for bile acids with the areas
   under the curve greater than 0.8 (C). Genetic variation of 82 BRGs in
   the FU-iCCA cohort (D). Positions of CNVs in BRGs on chromosomes (E)

   To analyze alterations in the bile acid profile from the genetic level
   in CCA, we included a total of 82 bile acid metabolism-related genes
   (BRGs) from the Molecular signatures database (MSigDB). Genetic
   mutations of BRGs were discovered in 21.34% (54/253) of the 253 CCA
   patients (Fig. [84]2D). Next, the somatic copy number variations (CNVs)
   of BRGs were explored, and universal CNVs were also identified in 82
   BRGs (Fig. [85]2E-F). All these results demonstrated the disorder of
   bile acid metabolism in CCA.

Consensus clustering identified two distinct bile acid metabolism-related
subtypes in CCA

   Considering the disordered bile acids metabolism identified above, the
   expression characteristics of 82 BRGs in CCA was analyzed. CCA patients
   were categorized according to the expression of 82 BRGs in the FU-iCCA
   cohort. Two resulting clusters, including C1 and C2, were defined
   (Fig. [86]3A-C). C1 was defined as the bile acid metabolism-active
   subtype, and C2 referred to the bile acid metabolism-inactive subtype
   (Fig. [87]3C). The reproducibility of bile acid-related subtypes in two
   GEO cohorts ([88]GSE89749 and [89]GSE26566) was also validated
   (Fig. [90]3A-E), and displayed high consistency with the FU-iCCA cohort
   according to an unsupervised Subclass Mapping method (SubMap)
   (Fig. [91]3D-E). These results demonstrated the heterogeneity of bile
   acid metabolism and identified two distinct bile acid
   metabolism-related subtypes in CCA.

Fig. 3.

   [92]Fig. 3
   [93]Open in a new tab

   Identification of two bile acid metabolism-related subtypes in
   cholangiocarcinoma Consensus clustering heatmap of two subtypes (k = 2)
   (A). Principal component analysis of the two subtypes (B). Differences
   in the expression of BRGs between the two subtypes (C). Subclass
   mapping of the FU-iCCA cohort and [94]GSE89749 cohort subtypes (D).
   Subclass mapping of the FU-iCCA cohort and [95]GSE26566 cohort subtypes
   (E). Bonferroni adjusted p-values indicate correlations between
   subtypes. Red represents significant similarities between subtypes
   (p < 0.05), while blue represents differences between subtypes
   (p > 0.05)

Correlation of bile acid metabolism-related subtypes with clinical and
prognostic characteristics in CCA

   Next, we analyzed the differences in prognostic and clinicopathological
   features of the two bile acid metabolism-related subtypes we identified
   (Fig. [96]4A-B). The C2 subtypes were associated with longer overall
   survival, earlier TNM stages, less vascular invasion, regional lymph
   node metastasis and perineural invasion in the FU-iCCA cohort
   (Fig. [97]4A, C). The [98]GSE89749 cohorts also displayed similar
   correlations (Fig. [99]4B, D). Notably, there were also differences in
   the histological and anatomical types between bile acid
   metabolism-related subtypes. Adenocarcinoma and intrahepatic CCA were
   the predominant types in bile acid metabolism-inactive subtype(C2),
   while papillary carcinoma and perihilar CCA were more common in
   subtypes with active bile acid metabolism(C1) (Fig. [100]4D-F).

Fig. 4.

   [101]Fig. 4
   [102]Open in a new tab

   Clinical and prognostic characteristics of bile acid-related subtypes
   in FU-iCCA and [103]GSE89749 cohorts Survival analyses between two bile
   acid-related subtypes in the FU-iCCA cohorts (A) and [104]GSE89749
   cohorts (B). Correlation between clinical characteristics and the CCA
   subtypes in the FU-iCCA cohorts (C) and the [105]GSE89749 cohorts (D).
   Percentage stacked bar charts show histological (E) and anatomical (F)
   differences in bile acid-related subtypes of CCA. Clinical data of
   [106]GSE26566 were not available

   In addition, we compared our classification with previously reported
   proteomic subtypes and integrative subtypes of CCA based on somatic
   mutations, copy number alterations, DNA methylation, and mRNA
   expression [[107]22, [108]23]. A comparable overlap between our bile
   acid metabolic subtypes and these previously reported CCA subtypes was
   found (Fig. [109]4C-D). In general, bile acid metabolism-inactive
   subtype(C2) had a better prognosis and exhibited more benign clinical
   characteristics than bile acid metabolism-active subtype(C1), which was
   consistent with our findings of lower serum bile acid concentrations in
   early CCA in comparison with advanced CCA (Fig. [110]2B).

Transcriptome analysis revealed that bile acid metabolism-related subtypes
were associated with inflammation and immune responses in CCA

   To further characterize these two bile acid metabolism-related subtypes
   of CCA, we performed pathway enrichment analysis. In the GSVA
   enrichment analysis, the C1 subtype was enriched for the major tumor
   metabolic processes and the key pathways involved in tumor development
   (Fig. [111]5A) [[112]32–[113]36]. Additionally, it was observed that
   inflammatory and immune signaling pathways are enriched in the C1
   subtype according to GSVA, GO, KEGG and GSEA enrichment analyses
   (Fig. [114]5A-D). Chemokines played a crucial role in the inflammatory
   and immune response [[115]37]. As depicted in Fig. [116]5E, the two
   bile acid metabolism-related subtypes exhibited distinct expressions of
   chemokines. Positive correlations between most chemokines, particularly
   CCL15 and CCL16, and bile acid-related genes were revealed
   (Fig. [117]5F). Supplementary Fig. [118]1 and Supplementary Fig. [119]2
   presented similar correlations between the bile acid-related subtypes
   and inflammation and immunity in two GEO cohorts. These results
   suggested that different bile acid metabolism-related subtypes of CCA
   exhibited different characteristics of inflammation and immune
   responses, which may contribute to their different clinical features.

Fig. 5.

   [120]Fig. 5
   [121]Open in a new tab

   Transcriptome analysis of bile acid-related subtypes GSVA enrichment
   analysis (A), GO enrichment analysis (B), KEGG pathway analysis (C) and
   GSVA enrichment analysis (D) between two bile acid-related subtypes.
   Distinct expression of chemokines between two bile acid-related
   subtypes (E). Spearman rank correlation between bile acid-related genes
   and chemokines (F)

Different immune infiltration of bile acid metabolism-related subtypes in CCA

   Given significant differences in immune enrichments between subtypes of
   CCA, the immune-related techniques reported before were employed to
   characterize the immune infiltration of bile acid-related subtypes. The
   ssGSEA and CIBERSORTx analysis both verified the differences in
   neutrophils, mast cells and NK cells between the two subtypes
   (Fig. [122]6A-B). The poor prognosis of the C1 subtype may be partially
   explained by both the increase in neutrophils and mast cells and the
   decrease in NK cells (Fig. [123]6C). Moreover, BRGs, particularly AQP9,
   were positively correlated with both mast cells and neutrophils
   (Fig. [124]6D-E). To verify these findings, the immune cell
   infiltration of each subtype across the two GEO cohorts was then
   explored, and similar results were obtained (Supplementary
   Fig. [125]3).

Fig. 6.

   [126]Fig. 6
   [127]Open in a new tab

   Immune infiltration of two bile acid-related subtypes in the FU-iCCA
   cohort Immune infiltration of two bile acid-related subtypes by ssGSEA
   (A) and Cibersortx (B). Kaplan–Meier survival analysis of the
   neutrophils, mast cells and NK cells in patients with CCA (C). Spearman
   rank correlation coefficient between bile acid-related genes and
   neutrophils (D), mast cells (E)

Comparisons of drug sensitivity in bile acid metabolism-related subtypes of
CCA

   The difference in sensitivity of several common drugs was found between
   the subtypes (Fig. [128]7A-B). In addition, using the immunophenoscore
   (IPS), scores of four different immunophenotypes (MHC-related
   molecules, effector cells, suppressive cells and checkpoint cells) were
   calculated. The four immunophenotypes were combined into a general IPS.
   The higher the IPS, the stronger the immunogenicity, which means they
   will respond to immunotherapy more favorably. CCA patients with subtype
   C2 were more likely to benefit from immunotherapy as subtype C2
   exhibited a greater IPS (Fig. [129]7C).

Fig. 7.

   [130]Fig. 7
   [131]Open in a new tab

   Comparisons of drug sensitivity in bile acid metabolism-related
   subtypes Drugs that are more sensitive to patients in subtype C2 (A).
   Drugs that are more sensitive to patients in subtype C1 (B). Different
   immunophenoscore between bile acid-related subtypes(C)

Single-cell RNA sequencing revealed that immunotherapy altered BRGs
expression in the tumor microenvironment

   Based on TISCH, the transcriptional profiles of 5761 cells from tumor
   tissues of 10 CCA patients in the [132]GSE102988 dataset were analyzed
   (Supplementary Table [133]1) (Fig. [134]8A-E). BRGs were expressed
   mainly in liver progenitor cells, followed by malignant cells. Also,
   they were modestly expressed in various immune cells (Fig. [135]8F-G).
   Immunotherapy altered BRGs expression differently in various cell
   types. In contrast to malignant cells and immune cells, immunotherapy
   decreased the expression of BRGs in stromal cells, primarily hepatic
   progenitor cells (Fig. [136]8H-I). The changes in the BRG’s expression
   of immune cells were more closely explored. Following immunotherapy,
   the expression of BRGs was raised in conventional CD4^+T cells,
   monocytes/macrophages, exhausted CD8^+T cells, and B cells, whereas it
   was decreased in CD8^+T cells (Fig. [137]8I). These findings confirmed
   a connection between immunotherapy and bile acid metabolism in CCA
   patients.

Fig. 8.

   [138]Fig. 8
   [139]Open in a new tab

   The alteration of BRGs expression induced by immunotherapy in the TISCH
   database. Cells are colored by cell clusters (A) or cell type (B) in a
   UMAP graphic. The marker gene expression level for each annotated cell
   types (C). Percentage of different cell types in CCA patients (D). The
   total number of cells in every cell-type (E). Distribution (F) and
   expression (G) of BRGs in different cells. Expression of BRGs in
   different cell types from the malignancy level (H) and the
   major-lineage level (I) following different treatments

Identification and verification of a bile acid metabolism-related signature
with prognostic significance

   Based on the developed BRGs expression classification, we next built a
   prognostic signature with the Cox proportional hazards model and Lasso
   regression, one of the machine learning algorithms (Fig. [140]9A).
   Firstly, univariate COX analysis identified 25 BRGs with prognostic
   significance in all BRGs. Then, we performed Lasso regression on the 25
   prognostic BRGs with 10-fold cross-validation and obtained 18 BRGs for
   multivariate COX analysis (Fig. [141]9B-C). Finally, multivariate COX
   analysis identified a two-gene (SLCO1B3 and CEACAM1) signature with
   best prognostic significance (Fig. [142]9D). The risk score was
   calculated with regression coefficients: Risk
   score = 0.012662*SLCO1B3 + 0.003870*CEACAM1. There was significantly
   shorter overall survival (OS) in patients with high scores
   (Fig. [143]9E). The areas under the curves (AUCs) for the predicted
   overall survival rates at 1-, 2- and 4-year were separately 0.681,
   0.679 and 0.676 (Fig. [144]9I). Consistently, patients with high scores
   in the GEO cohorts had remarkably shorter disease-free survival (DFS)
   and OS (Fig. [145]9F-H). The AUCs of the bile acid-related risk score
   in the GEO cohorts were also showed that this bile acid-related
   signature was a potent predictor of the outcomes of CCA patients
   (Fig. [146]9J-L).

Fig. 9.

   [147]Fig. 9
   [148]Open in a new tab

   Identification and verification of a bile acid-related signature with
   prognostic significance. Flow chart shows the process by which the bile
   acid-related signature with prognostic significance was identified (A).
   Cross-validation for optimal gene selection (B). The Lasso coefficient
   profiles (C). An overview of survival, risk scores and key genes (D).
   Survival analysis for overall survival based on the risk score in
   FU-iCCA cohort (E), [149]GSE89749 cohort (F) and [150]GSE107943 cohort
   (G). Survival analysis for disease-free survival based on risk score in
   [151]GSE107943 cohort (H). Time-dependent ROC curves for predicting
   overall survival rates in FU-iCCA cohort (I), [152]GSE89749 cohort (J)
   and [153]GSE107943 cohort (K). Time-dependent ROC curves for predicting
   disease-free survival rates in the [154]GSE107943 cohort (L)

   Moreover, we performed univariate and multivariate Cox analyses for the
   signature and important clinical indicators of CCA (Fig. [155]10A-B).
   It was suggested that the risk score (HR: 1.7, 95%CI: 1.1–2.6,
   P = 0.013), intrahepatic metastasis, vascular invasion, regional lymph
   node metastasis and distal metastasis were independent prognostic
   factors of CCA (Fig. [156]10B). Based on the identified independent
   prognostic markers, a comprehensive nomogram was established to
   accurately predict the 1-year, 2-year and 4-year OS probabilities in
   CCA patients (Fig. [157]10C). The AUCs were 0.821, 0.828 and 0.820 at
   1, 2 and 4 years separately, which demonstrated the reliability of the
   nomogram’s predictions (Fig. [158]10D). The calibration curves and
   decision curve analysis of the nomogram also demonstrated the
   reliability and clinical practicability of the comprehensive nomogram
   for CCA (Fig. [159]10E-F).

Fig. 10.

   [160]Fig. 10
   [161]Open in a new tab

   Development and evaluation of the nomogram Univariate regression (A)
   and multivariate regression (B) of clinicopathological indicators and
   risk score. A comprehensive nomogram to predict the survival
   probability of CCA patients at 1, 2, and 4 years (C). The
   time-dependent ROC curves (D), calibration curve (E) and decision curve
   analysis (F) of the nomogram

Expression and functional verification of identified prognostic genes

   To further verify the above findings, immunohistochemistry results of
   SLCO1B3 and CEACAM1 in human normal liver and CCA tissues from the
   Human Protein Atlas (HPA) were obtained [[162]38], which presented an
   increased expression of SLCO1B3 and CEACAM1 in CCA tissues in
   comparison with normal bile duct (Fig. [163]11A-B). In addition,
   SLCO1B3 and CEACAM1 were overexpressed respectively in CCA cells (RBE
   cells and HUCC-T1 cells) to validate their biological functions. It was
   found that overexpression of SLCO1B3 or CEACAM1 promoted the
   proliferation (RBE cells: NC 100% ± 4.87%; SLCO1B3 116.33% ± 1.06%,
   p = 0.0019; CEACAM1 121.33% ± 2.39%, p = 0.0004) (HUCC-T1 cells: NC
   100% ± 0.79%; SLCO1B3 111.66% ± 6.78%, p = 0.0267; CEACAM1 110.13% ±
   0.95%, p = 0.0470), migration(RBE cells: NC 43.85% ± 3.42% vs. SLCO1B3
   68.09% ± 3.06%, p = 0.0034; NC 40.80% ± 2.49% vs. CEACAM1 54.50% ±
   3.05%, p = 0.0171) (HUCC-T1 cells: NC 62.80% ± 2.41% vs. SLCO1B3 73.87%
   ± 1.78%, p = 0.0160; NC 61.78% ± 4.20% vs. CEACAM1 78.34% ± 1.77%,
   p = 0.0438) and inhibited apoptosis(RBE cells: NC 25.21% ± 1.51%;
   SLCO1B3 18.28% ± 1.47%, p = 0.0014; CEACAM1 18.36% ± 0.68%, p = 0.0015)
   (HUCC-T1 cells: NC 24.49% ± 1.24%; SLCO1B3 14.35% ± 0.88%, p < 0.0001;
   CEACAM1 15.54% ± 1.25%, p = 0.0002) of RBE cells and HUCC-T1 cells
   (Fig. [164]11C-F).

Fig. 11.

   [165]Fig. 11
   [166]Open in a new tab

   Expression and functional verification of identified prognostic genes
   Immunohistochemistry results of SLCO1B3 (A) and CEACAM1(B) in human
   normal liver tissue and cholangiocarcinoma tissue from the Human
   Protein Atlas. Overexpression of SLCO1B3 or CEACAM1 promotes the
   proliferation (C), reduces apoptosis rates (D-E), and promotes the
   migration (F) of RBE cells and HUCC-T1 cells

Discussion

   Cholangiocytes are continuously exposed to high concentrations of bile
   acids, and bile acid metabolism has a close correlation with CCA
   development [[167]10, [168]11]. Various effects of bile acids on the
   growth of CCA have been reported [[169]12, [170]13]. Previous studies
   on plasma bile acid in CCA focused on the opposite trend of conjugated
   and unconjugated bile acids, and the types of bile acids quantified
   were not abundant enough [[171]13, [172]39, [173]40]. In this study, 30
   bile acids in the serum of CCA patients were quantified. On the
   contrary to primary bile acids, we discovered that secondary bile acids
   decreased in CCA. It was suggested that the primary-to-secondary bile
   acid metabolism mediated by the gut microbiome participated in CCA
   occurrence [[174]41]. Ma and colleagues also reported that bile acids
   were used as messengers by the gut microbiota to regulate the
   infiltration of NKT cells and thus affect the anti-tumor immunity of
   liver cancer [[175]20]. However, compared with TNM stage I patients,
   stage II-IV patients showed an increase in both primary and secondary
   bile acids. This may imply that the total bile acid may contribute to
   the progression of CCA by promoting cell proliferation, enhancing
   inflammation and lessening FXR-dependent chemoprotective effects after
   the emergence of CCA [[176]11].

   As CCA has high heterogeneity, previous studies used unsupervised tumor
   clustering to stratify CCA patients according to genomic and
   transcriptomic data, thereby bringing meaningful results of various
   population-specific differences [[177]22, [178]23, [179]42, [180]43].
   Recently, analysis of the metabolic expression profile has emerged as
   an informative method to explore tumor heterogeneity as tumor
   metabolism has attracted increasing attention [[181]44–[182]46]. This
   research identified a classification of CCA according to the bile acid
   metabolism-related gene expression. Two bile acid metabolic subtypes
   and their correlation to clinical characteristics, signaling pathways,
   immune landscape, and therapeutic response was discovered.
   Additionally, we found that bile acid metabolic subtypes were
   associated with histological and anatomical type of CCA. This may
   suggest that different histological and anatomical types have different
   effects on bile acid metabolism, and it may also suggest that bile acid
   metabolism may affect the histological features and location of CCA. To
   the best of our knowledge, few studies have discussed the relationship
   between bile acid metabolism and pathological types, anatomical
   locations of CCA. The causal relationship and interaction between them
   are worthy of further investigation.

   Based on the identified bile acid metabolic subtypes, a prognostic
   signature of bile acid metabolic genes (SLCO1B3 and CEACAM1) and a
   corresponding nomogram that could better predict overall survival of
   CCA patients were established and verified. Compared with the
   previously reported prognostic signature of CCA, our nomogram showed
   better predictive sensitivity and specificity [[183]47–[184]50]. The
   identified prognostic signature included two bile acid
   metabolism-related genes, SLCO1B3 and CEACAM1, which were the high-risk
   genes in CCA. SLCO1B3 was a human liver-specific transporter, which
   mediated the uptake of different endogenous and exogenous substances,
   such as bile acids [[185]51]. It was recently found this gene was
   expressed in several types of cancers [[186]52]. SLCO1B3 has been
   reported to promote tumor progression in colorectal cancer. Elevated
   expression of SLCO1B3 was linked to lymph node metastasis, tumor
   invasion, diminished differentiation, advanced disease stage, reduced
   overall survival in colorectal cancer [[187]53]. CEACAM1, originally
   described as biliary glycoprotein (BGP1) in human hepatic bile, was a
   transmembrane glycoprotein expressed on epithelial cells, endothelial
   cells, and immune cells [[188]54]. In general, dysregulated CEACAM1
   expression was often observed during the malignant progression of
   various cancers [[189]54, [190]55]. According to Ortenberg et al.,
   overexpression of CEACAM1 promoted melanoma proliferation through a
   SOX2-dependent mechanism [[191]56]. Han et al. discovered that CEACAM1
   knockdown retarded the proliferation of HT29 colon cancer cells
   [[192]57]. This research further verified the connection between
   SLCO1B3/CEACAM1 and the CCA development. The high expression of both
   SLCO1B3 and CEACAM1 indicated a poor prognosis in CCA. Besides, these
   genes appeared to inhibit the apoptosis of CCA cells while promoting
   their proliferation and migration.

   This research still has some limitations. First of all, the
   classification and prognostic signature of bile acid metabolism were
   established on the basis of public data sets. Additional validation is
   needed in larger randomized controlled cohorts. Additionally, it
   remains unclear why bile acids or BRGs regulates biological phenotypes,
   such as malignant phenotype or immune infiltration of CCA, more
   molecular biological experiments and analyses should be performed to
   better learn about the role and mechanism of bile acid metabolism in
   the TME of CCA.

Conclusions

   In conclusion, this study demonstrated the altered bile acid metabolism
   in CCA and identified two distinct CCA subtypes of bile acid
   metabolism. Each of the two bile acid metabolic subtypes was correlated
   with different clinical features, immune landscapes and therapeutic
   responses of CCA. In addition, altered bile acid metabolism in the TME
   after immunotherapy was validated according to scRNA-seq analysis.
   These findings broaden the understanding of both the molecular subtypes
   and the heterogeneity of bile acid metabolism in CCA. Based on this, a
   bile acid metabolism-related signature for risk stratification and
   prognostic prediction of CCA patients was created and experimentally
   validated.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [193]12885_2024_13081_MOESM1_ESM.tif^ (7.2MB, tif)

   Supplementary Material 1: Supplementary Fig. 1 transcriptome analysis
   of bile acid-related subtypes in GSE89749 cohorts GSVA enrichment
   analysis(A), GO enrichment analysis(B), KEGG pathway analysis(C) and
   GSVA enrichment analysis(D) between two bile acid-related subtypes.
   Distinct expression of chemokines between two subtypes(E).Spearman rank
   correlation between BRGs and chemokines(F).
   [194]12885_2024_13081_MOESM2_ESM.tif^ (4MB, tif)

   Supplementary Material 2: Supplementary Fig. 2 transcriptome analysis
   of bile acid-related subtypes in GSE26566 cohorts GSVA enrichment
   analysis(A), GO enrichment analysis(B), KEGG pathway analysis(C) and
   GSVA enrichment analysis(D) between two bile acid-related subtypes.
   [195]12885_2024_13081_MOESM3_ESM.tif^ (5.7MB, tif)

   Supplementary Material 3: Supplementary Fig. 3 Immune infiltration of
   two bile acid-related subtypes in the GEO cohorts Immune infiltration
   of two bile acid-related subtypes by ssGSEA(A) and Cibersortx(B) in
   GSE89749 cohorts. Immune infiltration of two bile acid-related subtypes
   by ssGSEA(C) and Cibersortx(D) in GSE26566 cohorts.
   [196]12885_2024_13081_MOESM4_ESM.xlsx^ (594KB, xlsx)

   Supplementary Material 4: Supplementary Table 1 The clinical
   information of CCA patients obtained from Our, NODE,and GEO cohorts.
   [197]12885_2024_13081_MOESM5_ESM.xlsx^ (9.6KB, xlsx)

   Supplementary Material 5: Supplementary Table 2 Gene list of bile acid
   metabolism.

Acknowledgements