Shared genetic regulators and diagnostic biomarkers in gallbladder stone and cholangiocarcinoma: A comprehensive comparative transcriptomic analysis

Highlights

• •Comparative transcriptomics mapped shared signatures across GS and cholangiocarcinoma.
• •DEGs implicated IL-17 signaling and bile acid metabolism in both diseases.
• •LASSO identified a 3-gene core signature: CENPH, NRSN2, and SPDYC.
• •Core genes showed robust diagnostic performance across independent datasets.
• •CENPH knockdown suppressed cholangiocarcinoma cell growth and invasion.

Abstract

Gallbladder stone (GS) and cholangiocarcinoma (CHOL) are common biliary tract disorders that share certain clinical and pathological characteristics but exhibit distinct molecular hallmarks. While GS-related inflammation, cholestasis, and epithelial injury-repair responses may provide a biologically plausible link, the shared genetic mechanisms underlying the pathogenesis of both diseases remain largely unexplored. The aim of this study was to elucidate this shared genetic landscape using a comparative transcriptomic approach. We identified 671 and 5197 differentially expressed genes (DEGs) in the GS and CHOL datasets, respectively. Gene ontology and KEGG enrichment analysis of these DEGs revealed possible involvement in IL-17 signaling and bile acid metabolism. The overlap of the DEGs yielded 15 common genes, which were further refined to three core genes—CENPH, NRSN2, and SPDYC—using LASSO regression analysis. The expression levels of these core genes were validated across multiple datasets, and ROC analysis demonstrated that they exhibited robust diagnostic performance for distinguishing both GS and CHOL. The copy number variations and promoter methylation status of these core genes were also evaluated. Candidate drug compounds targeting these genes were predicted using the PubChem database, and a gene-compound interaction network was constructed to facilitate future therapeutic investigations. We further confirmed that CENPH was significantly upregulated in CHOL cell lines (HuCC-T1 and RBE) compared to normal human intrahepatic biliary epithelial cells. Subsequently, loss-of-function assays showed that CENPH was essential for the proliferation, migration, and invasion of CHOL cells. In conclusion, our findings provide new insights into the shared genetic regulators of GS and CHOL and highlight the potential for developing effective diagnostic tools and tailored therapeutic strategies.

Graphical abstract

Introduction

Bile tract diseases, including gallstones (GS), bile duct stones, cholangitis, and cholangiocarcinoma (CHOL), pose substantial health challenges globally due to their increasing incidence, and the high morbidity and mortality rates[[1], [2], [3]]. These pathological conditions are associated with chronic inflammation, biliary injury, and epithelial remodeling, and often cause severe abdominal pain and other complications[[4], [5], [6]]. Furthermore, cholangiocarcinoma, a malignant tumor originating from the biliary epithelium, is characterized by its aggressive nature, late-stage diagnosis, and poor prognosis[7,8]. It is crucial to elucidate the genetic basis and the molecular mechanisms involved in these diseases for the development of effective diagnostic tools and therapeutic strategies. High-throughput technologies have revolutionized our understanding of the complex biological processes and pathways involved in various diseases[[9], [10], [11]]. For example, identification of genes that are upregulated or downregulated under different conditions can provide new insights into their potential roles in disease pathogenesis[12,13]. This approach has been extensively utilized in uncovering novel biomarkers and therapeutic targets in various diseases, including GS and CHOL[14,15].

Although GS may not be a direct precursor for all CHOL subtypes, long-standing GS-related biliary injury, recurrent cholecystitis/cholangitis, cholestasis, and dysregulated epithelial repair may create a chronic inflammatory microenvironment that is conducive to cholangiocarcinogenesis. Studies have shown that persistent biliary inflammation, bile-acid dysregulation, oxidative and nitrosative stress, and cytokine-driven pathways – including IL-6/STAT3, TNF/NF-κB, and COX-2-associated signaling – promote cholangiocyte proliferation and survival, and trigger genomic instability and stromal remodeling. Consistent with this, epidemiological and meta-analytic studies have reported positive associations between GS and biliary tract malignancies, which may suggest shared transcriptomic regulators[7,[16], [17], [18], [19]]. However, the common genetic mechanisms and/or regulators of GS and CHOL remain largely unexplored.

The aim of this study was to elucidate the molecular mechanisms and pathological pathways that drive both GS and CHOL. To this end, we screened the differentially expressed genes (DEGs) of these diseases from available datasets, and identified the biological functions and pathways associated with the shared DEGs through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses[[20], [21], [22]]. Gene set enrichment analysis (GSEA) was performed to identify the molecular signatures that are coordinately upregulated or downregulated in the disease states[23,24]. To validate the clinical utility of the identified genes, their differential expression was confirmed across independent datasets, and their diagnostic potential was assessed through receiver operating characteristic (ROC) analysis[25,26]. Moreover, we explored the copy number variations (CNVs), mutations, and promoter methylation status of the core genes to gain further insights into the biological mechanisms[27,28], and screened for candidate compounds that can target these genes. Our findings not only enrich our understanding of the shared genetic basis of GS and CHOL, but also contribute towards developing precise and effective therapeutic strategies.

Methods

Identification and functional annotation of DEGs

Data was retrieved from the GS (8 healthy control and 12 GS samples), TCGA-CHOL (9 healthy control and 35 CHOL samples), and GSE107943 (27 control and 30 CHOL samples) datasets. The GS and TCGA-CHOL datasets were the discovery cohorts, and the GSE107943 dataset was used for core gene validation and ROC curve analysis. The DEGs between the GS/CHOL samples and the corresponding control samples were screened from the respective datasets, with absolute log fold change (logFC) > 1 and the P-value or adjusted P-value < 0.05 as the criteria. The volcano plots of the DEGs were generated using the ggVolcano and ggrepel packages in R, and the pheatmap package in R was used to plot heatmaps of the top 100 genes (in terms of P-value). GO and KEGG enrichment analyses were performed using the clusterProfiler package in R[29,30], and the criteria for enrichment were P-values < 0.05. The results of GO analysis were visualized using the circlize and ComplexHeatmap packages, and the GOplot package was used for KEGG results.

Identification of shared disease genes and core gene selection

The DEGs in both the GS and CHOL datasets were intersected to establish a set of common disease genes. LASSO regression was performed on these genes using the glmnet package in R to further refine the selection and identify the core shared genes. The regression analysis was conducted independently for the GS and CHOL datasets. To reduce overfitting, penalized logistic regression was fitted with 10-fold cross-validation using the cv.glmnet function (alpha = 1, family = "binomial"), and a fixed random seed (1234) was set before fold assignment to improve reproducibility. The optimal lambda was defined by the minimum mean cross-validated binomial deviance (lambda.min), as opposed to the area under the curve (AUC) in the same dataset. ROC analyses were performed only after feature selection to assess the discriminatory capacity of the genes, and were not used for model tuning. The core genes from the two datasets were intersected through Venn diagrams drawn using the Funrich software to identify the core shared genes.

Validation of core genes

The expression levels of the three core genes were analyzed in GS, TCGA-CHOL, and GSE107943 datasets, and the ggpubr package in R was used to prepare the box plots and violin plots. The diagnostic performance of each core gene was evaluated in the discovery and validation cohorts by plotting ROC curves using the pROC package in R. To reduce overinterpretation arising from limited sample size, the logFC and Benjamini-Hochberg false discovery rate (FDR) of CENPH, NRSN2, and SPDYC from the differential expression analyses were also considered to test significance.

Immune analysis

The relative abundance of 29 types of immune cells and immune processes in each sample from the GS and TCGA-CHOL datasets were determined by single-sample gene set enrichment analysis (ssGSEA) using the GSVA and GSEABase packages in R. The correlation between the expression levels of the three core genes and the relative abundance of the immune cell types and immune processes were also analyzed, and the results were visualized using the reshape2, ggpubr, and ggExtra packages in R.

Hallmark-GSEA

The GS and TCGA-CHOL cohorts were divided into the respective high and low expression groups for each of the core genes based on their median expression levels, and the DEGs between the groups were screened using the limma package in R. Thereafter, GSEA was implemented on the Hallmark gene sets using the clusterProfiler package in R, and those with nominal P-value < 0.05 and FDR q-value < 0.25 were considered significantly enriched. The normalized enrichment score (NES) was calculated as a measure of the direction and relative magnitude of enrichment.

Gene expression correlation, CNV, chromosomal location, and promoter methylation analyses

The correlation between the expression levels of the three core genes in the TCGA-CHOL dataset was analyzed, and the correlation matrix was visualized using the "corrplot" package in R. The CNV plots of the genes were constructed with the "ggplot2" package in R, and their chromosomal locations were presented in the form of a circus plot using the RCircos package in R. The potential interaction networks of proteins associated with the core genes were predicted using the GENEMANIA website (genemania.org/), and the enriched pathways of the interacting proteins were identified. The promoter methylation status of the three core genes in the CHOL and control samples was determined using the UALCAN web resource (https://ualcan.path.uab.edu/).

Mutation analysis and screening of candidate compounds

The mutation status of the three core genes in the respective high and low expression groups was determined using the SangerBox web tool (http://www.sangerbox.com/login.html). Waterfall plots were constructed after excluding samples lacking mutations. The candidate compounds associated with the core genes were retrieved from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/), and the gene-compound network was visualized using Cytoscape (version 3.9.1).

Cell culture and transfection

Immortalized normal human intrahepatic biliary epithelial cells (NhIBE), and human CHOL cell lines (HuCC-T1 and RBE) were procured from Cell Bank of Chinese Academy of Sciences. All cell lines were maintained in RPMI-1640 medium (350-000-CL, Wisent, China) supplemented with 10% FBS (04-001-1A-AUS, BI, Israel), penicillin (SV30010, 100 U/ml), and streptomycin (SV30010, 100 μg/ml; both antibiotics from HyClone). The cells were cultured at 37°C in a humidified incubator with 5% CO₂ to ensure logarithmic growth phase conditions.

The RBE cells were transfected with two custom-designed siRNA sequences targeting CENPH (synthesized by GIMA Corporation, China) using the PolyFast transfection reagent (MCE, USA, HY-K1014). Briefly, the cells were seeded in 6-well plates at the density of 3 × 10⁵ cells/well in 2 ml medium, and allowed to adhere. The siRNA-reagent complexes were prepared according to manufacturer guidelines, and incubated for 15 min at room temperature. Media was replaced 6h post-transfection, and the cells were cultured for 48h prior to the downstream analyses. The siRNA sequences were as follows: si-NC: sense: 5′- UCCUCCGAACGUGUCACGUTT - 3′, antisense: 5′ - ACGUGACACGUUCGGAGAATT - 3′; si-CENPH-1: sense: 5′- TAGTTATTTGTAGTGGGAT - 3′, antisense: 5′ - AGAACAAATTATGCAAGAA - 3′; si-CENPH-2: sense: 5′- TTGAGAAGAATGTTGACAT - 3′, antisense: 5′ - AGATTGATTTGGACAGTAT- 3′.

Real time quantitative PCR (RT-qPCR)

Suitably treated cells were harvested using trypsin (SH30042.01, HyClone, USA), washed twice with ice-cold PBS, and lysed in 1 mL of TRIzol reagent (9108, Takara, Japan). After incubation on ice for 5 min, the lysate was mixed with 200 μL of chloroform (67-66-3, SINOPHARM, China), vigorously vortexed for 15 s, and centrifuged at 12,000 × g for 15 min at 4°C. The aqueous phase was transferred to a new tube, mixed with an equal volume of isopropanol (67-63-0, SINOPHARM, China), and incubated at room temperature for 10 min. The mixture was centrifuged at 12,000 × g for 10 min at 4°C, and the supernatant was discarded. The RNA pellet was washed with 1 mL of 75% ethanol (prepared with anhydrous ethanol, 64-17-5, SINOPHARM, China, and DEPC-treated water), centrifuged at 7,500 × g for 5 min at 4°C, and air-dried for 5–10 min at room temperature. The dried RNA was dissolved in 20 μL of DEPC-treated water. RNA concentration and purity were determined using a NanoDrop One C spectrophotometer (Thermo Fisher, #AZY2124606, USA).

A total of 1 μg of RNA was reverse transcribed using the HiScript II 1st Strand cDNA Synthesis Kit (+gDNA wiper) (R212-01, Vazyme, China) following the manufacturer's protocol. Quantitative PCR was performed using SYBR GreenER qPCR SuperMix Universal (11762500, Thermo Fisher Scientific, USA) on a 7500 Real-Time PCR platform (Thermo Fisher, USA). According to the manufacturer's instructions, ROX Reference Dye (12223011, Thermo Fisher Scientific, USA) was added at 0.1 μL per 50 μL reaction to normalize the fluorescent signal. The cycling conditions were as follows: initial denaturation at 95°C for 30 s, followed by 40 cycles of 95°C for 5 s and 60°C for 34 s. CENPH expression levels were normalized to β-actin using the 2⁻ΔΔCt method. The primer sequences were as follows: β-actin: Forward: 5’ – CCTGGCACCCAGCACAAT - 3’, Reverse: 5’ – GGGCCGGACTCGTCATAC - 3’; CENPH: Forward: 5’ - TTCCAGAACCTTATTTTGGGG - 3’, Reverse: 5’- CTTCTCAAGCTGCAGAACAATT - 3’.

Western blotting

The cultured cells were harvested and homogenized in RIPA buffer (P0013B, Beyotime, China) supplemented with protease inhibitors (P1010, 1:100 v/v; Beyotime, China), and the samples were subjected to three 1-second ultrasonic pulses at 40% amplitude (Qsonica Q700). The lysates were incubated on ice for 30 min with intermittent vortexing, and centrifuged at 10,000 g for 15 min at 4°C to remove debris. The protein content in the supernatants was quantified with Bradford assay. Equal protein amounts were separated by SDS-PAGE using 10% bis-Tris gels (20 μg protein/lane) at constant 100V, and transferred to 0.45 μm PVDF membranes (IPVH00010, Millipore) by the semi-dry method. After 10-min rapid blocking (P0252, Beyotime, China), the membranes were washed thrice with TBST (0.1% Tween-20), and incubated overnight with rabbit anti-CENPH (1:1000, 12841-1-AP) and anti-β-actin (1:2000, 81115-1-RR; both Proteintech) antibodies at 4°C. The membranes were incubated with HRP-linked secondary antibody (1:5000, SA00001-2, Proteintech) for 90 min at room temperature. Protein bands were detected using an ECL reagent (P0018S, Beyotime, China).

Immunofluorescence

The culture medium was aspirated 48 h post-transfection, and the cells were washed three times with PBS. Following permeabilization with 0.3% Triton X-100 (ST797, Beyotime, China) in PBS for 25 min at room temperature, the cells were blocked with 5% normal goat serum (C0265, Beyotime, China) in PBS for 1 h at room temperature. The cells were then incubated overnight at 4°C with rabbit anti-CENPH primary antibody (1:200, 12841-1-AP, Proteintech). After overnight incubation, the cells were washed three times with PBS for 5 min each. Subsequently, the cells were incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG (H+L) secondary antibody (1:500, A11008, Invitrogen, USA) for 1 h at 25°C in the dark. Following three additional washes with PBS for 5 min each, nuclear counterstaining was performed with DAPI (C1005, Beyotime, China) for 10 min in the dark. The cells were then washed with PBS and mounted with antifade mounting medium (P0126, Beyotime, China) . Fluorescence images were captured using a Nikon Eclipse Ti2-E microscope equipped with a DS-Qi2 camera (NIS-Elements AR 5.21.03 software).

CCK8 assay

The transfected RBE cells were seeded in 96-well plates at the density of 6 × 10³ cells/well and cultured under standard conditions for 24, 48, 72, and 96 h. At the pre-determined time points, 200 μL CCK8 working solution (C0038, Beyotime, China) was added to each well, and the cells were incubated for 2 h. Triplicate wells were set up for each experimental group. The optical density was measured at 450 nm using a SpectraMax i3x microplate reader (Molecular Devices).

Transwell assay

Transwell inserts with 8 μm pore size polyester membranes (PTEP24H48, Millipore, USA) were used. Matrigel (356234, Corning, USA) was thawed overnight at 4°C. All subsequent operations were performed on ice. Matrigel was diluted 1:8 with ice-cold serum-free medium to a concentration of approximately 1.25 mg/mL, and 40 μL of the diluted Matrigel was added to each upper chamber. The inserts were then incubated at 37°C for 1 h to allow the Matrigel to solidify. After solidification, excess liquid was carefully aspirated, and the inserts were gently washed once with 50 μL of ice-cold serum-free medium.

The inserts were placed in 24-well plates containing 500 μL of complete medium with 10% FBS (serving as a chemoattractant) per well. The transfected cells were detached with trypsin, resuspended in serum-free medium, and loaded into the upper chambers at a density of 2 × 10⁴ cells/chamber in 200 μL of serum-free medium. Following 20 h of incubation at 37°C in a 5% CO₂ incubator, the non-invaded cells on the upper surface of the membrane were removed using PBS-moistened cotton swabs. The invaded cells on the lower surface were fixed with 4% paraformaldehyde (P0099, Beyotime, China) for 20 min, and then stained with 0.1% crystal violet (C0121, Beyotime, China) for 20 min at room temperature. The stained cells were washed three times with PBS until the washing solution became clear. The membranes were air-dried, and images were captured using a brightfield microscope (Nikon Eclipse Ts2, Japan). Quantification was performed by counting the stained cells in five randomly selected fields per membrane under 200 × magnification. Triplicate inserts were set up for each experimental group, and the experiment was independently repeated three times.

Wound healing assay

Twenty-four hours post-transfection, the confluent monolayers were longitudinally scratched using sterile 200 μl micropipette tips aligned with a straight-edge guide. After three washed with PBS to remove cellular debris, serum-free basal medium was added. The wound area was photographed at 0 h and 48 h of culture using phase-contrast microscopy. Migratory capacity was determined by calculating the percentage reduction in the denuded area using digital morphometric software.

Results

Identification of DEGs

The DEGs between GS/CHOL and corresponding normal tissues were screened in the GS and TCGA-CHOL datasets respectively, with absolute logFC > 1 and P-value < 0.05 as the criteria. A total of 671 DEGs were identified in the GS dataset, of which 233 were upregulated and 438 were downregulated (Fig. 1A). The heatmap with the top 100 is shown in Fig. 1B. In the TCGA-CHOL dataset, there were 5197 DEGs, including 2509 upregulated and 2688 downregulated genes (Fig. 1C). The heatmap of the top 100 DEGs is shown in Fig. 1D.

Fig. 1.

(A) Volcano plot showing the DEGs between the GS and control samples in the GS dataset (absolute logFC > 1 and P-value < 0.05). The upregulated genes (233) are shown in purple and the downregulated genes (438) in green. (B) Heatmap of the top 100 DEGs from the GS dataset based on P-value. (C) Volcano plot showing the DEGs between the CHOL and control samples in the TCGA-CHOL dataset (absolute logFC > 1 and adjusted P-value < 0.05). The upregulated genes (2509) are shown in purple and downregulated genes (2688) in green. (D) Heatmap of the top 100 DEGs from the TCGA-CHOL dataset based on P-value.

GO and KEGG enrichment analyses

The DEGs from both datasets were functionally annotated by GO and KEGG enrichment analyses. The top five enriched GO terms from the biological processes (BP), cellular components (CC) and molecular functions (MF) in the GS dataset are shown in Fig. 2A. The KEGG enrichment analysis showed that pathway related to IL-17 was significantly enriched in the 671 DEGs, suggesting a relationship between GS and immune responses (Fig. 2B). The top five BP, CC, and MF terms associated with the 5197 DEGs in the CHOL dataset are shown in Fig. 2C. In addition, pathways related to bile acid and fatty acid metabolism were highly enriched among these DEGs, indicating a relationship with biliary function (Fig. 2D).

Fig. 2.

(A) GO enrichment analysis of the 671 DEGs in the GS dataset (P-value < 0.05). The top five enriched terms of the BP, CC, and MF categories are shown. (B) KEGG enrichment analysis of the 671 DEGs in the GS dataset (P-value < 0.05). (C) GO enrichment analysis of the 5197 DEGs in the TCGA-CHOL dataset (P-value < 0.05). The top five enriched terms of the BP, CC, and MF categories are shown. (D) KEGG enrichment analysis of the 5197 DEGs in the TCGA-CHOL dataset (P-value < 0.05).

Identification of common disease genes and core genes selection

The intersection of DEGs from the GS and CHOL datasets yielded 15 common disease genes, which may regulate both conditions (Fig. 3A). LASSO regression with 10-fold cross-validation was then used to select the core disease genes. In the GS dataset, nine genes were retained at lambda = 0.0002239693, which corresponded to the minimum mean cross-validated deviance rather than the highest same-cohort AUC (Figs. 3B-C). Likewise, seven genes were retained in the CHOL dataset at lambda = 0.0003963939 on the basis of minimum cross-validated deviance (Figs. 3D-E). Thus, lambda selection was driven by internal cross-validation error, whereas ROC analysis was reserved for downstream descriptive evaluation. The intersection of these genes resulted in three core common disease genes, namely CENPH, NRSN2, and SPDYC (Fig. 3F).

Fig. 3.

(A) Intersection of the DEGs in GS and CHOL, resulting in 15 shared disease genes. (B-C) LASSO regression on the 15 shared disease genes in GS with 10-fold cross-validation to identify the core disease genes. The selected lambda (0.0002239693) corresponded to the minimum mean cross-validated deviance. (D-E) LASSO regression on the 15 shared disease genes in CHOL with 10-fold cross-validation to identify the core disease genes. The selected lambda (0.0003963939) corresponded to the minimum mean cross-validated deviance. (F) Intersection of the core genes in GS and CHOL datasets yielded CENPH, NRSN2, and SPDYC.

Validation of core genes

The expression levels of the core genes were validated across all three cohorts. In the GS dataset, both CENPH and NRSN2 were downregulated, while SPDYC was upregulated compared to the control samples (Figs. 4A-C). Significant differential expression of these genes was observed in the TCGA-CHOL (Figs. 4D-F) and GSE107943 (Figs. 4G-I) datasets. The diagnostic performance of the core genes was evaluated by plotting ROC curves for the GS (Figs. 5A-C), TCGA-CHOL (Figs. 5D-F), and GSE107943 (Figs. 5G-I) datasets.

Fig. 4.

(A-C) Box-violin plots showing differential expression of CENPH, NRSN2, and SPDYC between the control and GS samples in the GS dataset. (D-F) Box-violin plots showing the differential expression of the three core genes between control and CHOL samples in the CHOL dataset. (G-I) Box-violin plots showing the differential expression of the three core genes between control and CHOL samples in the GSE107943 dataset.

Fig. 5.

(A-C) ROC curves for CENPH (A), NRSN2 (B), and SPDYC (C) in the GS dataset. (D-F) ROC curves for CENPH, NRSN2, and SPDYC in the TCGA-CHOL dataset. (G-I) ROC curves for CENPH, NRSN2, and SPDYC in the GSE107943 dataset.

Immune analysis

The relative abundance of 29 immune cell types and immune processes were estimated for each sample in the GS and TCGA-CHOL datasets. Furthermore, the correlation between the expression levels of the three core genes (CENPH, NRSN2, SPDYC) and the relative abundance of these immune elements was also calculated. Significant correlations were observed in both the GS (Figs. 6A-C) and TCGA-CHOL (Figs. 6D-F) datasets.

Fig. 6.

The relative abundance of 29 types of immune cells and immune processes in each sample in the GS and TCGA-CHOL datasets was calculated using the ssGSEA algorithm. (A-C) Correlation between the expression of CENPH (A), NRSN2 (B), and SPDYC (C) and the abundance of 29 immune cells and immune processes in the GS dataset, with significant correlations marked in red. (D-F) Correlation between the expression of CENPH (D), NRSN2 (E), and SPDYC (F) and the abundance of 29 immune cells and immune processes in the CHOL dataset, with significant correlations marked in red.

Results of Hallmark-GSEA

The DEGs between the high and low expression groups for each of the core genes were screened in the GS and TCGA-CHOL datasets. GSEA was subsequently performed on the Hallmark gene sets to identify the enriched gene sets in the high and low expression groups in the GS (Figs. 7A-C) and TCGA-CHOL (Figs. 7D-F) datasets.

Fig. 7.

(A-C) HALLMARK-GSEA analysis of high and low expression groups for each gene in the GS dataset. (D-F) HALLMARK-GSEA analysis of high and low expression groups for each gene in the TCGA-CHOL dataset.

Genetic and epigenetic features of core genes

The correlation between the expression levels of the three core genes was analyzed in the TCGA-CHOL dataset, which revealed a significant inverse correlation between SPDYC and NRSN2 (Fig. 8A). CNV analysis of the three genes indicated that fewer copies of SPDYC and CENPH were present in most patients, whereas the NRSN2 gene showed similar frequency of amplification and deletion events (Fig. 8B). The chromosomal locations of the three core genes are shown in the circular plot in Fig. 8C. The potential protein interactions and enriched pathways for these genes were also predicted using GENEMANIA (Fig. 8D).

Fig. 8.

(A) Correlation between the expression levels of core genes in the TCGA-CHOL dataset. (B) CNVs in the SPDYC, CENPH, and NRSN2 genes. (C) Circular chromosome location diagram of the three core genes. (D) GENEMANIA prediction of the potential protein interactions and enriched pathways of the three core genes. (E-G) Promoter methylation status of the three core genes in the CHOL and control samples.

The promoter methylation status of the core genes was also investigated to gain insights into their regulatory mechanisms. Notably, only NRSN2 showed a significant difference in promoter methylation between CHOL and control samples (Figs. 8E-G). Taken together, the CNV and promoter methylation findings suggest that both genetic and epigenetic alterations may contribute to transcriptional dysregulation of the core genes in CHOL.

Mutation analysis and screening of candidate compounds

The mutation status of the core genes in patient groups demarcated by the median expression levels of CENPH (Fig. 9A), NRSN2 (Fig. 9B), and SPDYC (Fig. 9C) was analyzed. Samples without mutations were not included in the waterfall plots. Lastly, candidate compounds associated with the three core genes were retrieved from PubChem and visualized as a gene-compound network (Fig. 9D). From a biological perspective, these compounds may influence the cell-cycle, chromosomal alterations, or immune-metabolic pathways highlighted by the present enrichment and functional analyses.

Fig. 9.

(A-C) Samples are divided into high and low expression groups based on the median expression of CENPH (A), NRSN2 (B), and SPDYC (C), and mutation waterfall plots are drawn for each group. Samples without mutations are not shown. (D) Network diagram of core genes and therapeutic drugs predicted based on the PubChem website.

CENPH plays a pro-oncogenic role in cholangiocarcinoma cells

The potential role of centromere protein H (CENPH) in CHOL pathogenesis was further evaluated. CENPH transcript and protein levels were significantly elevated in the malignant cholangiocytes (HuCC-T1 and RBE cell lines) compared to that in the normal intrahepatic biliary epithelial (NhIBE) cells (p<0.001, Fig. 10A-B). Immunofluorescence staining corroborated enhanced CENPH expression in the neoplastic cells (Fig. 10C). To determine the biological relevance of CENPH in the cholangiocytes, we generated RBE cells with transient CENPH knockdown using dual validated siRNA sequences (Fig. 10D, p<0.001), and performed functional assays. CENPH silencing significantly reduced the viability (p < 0.001, Fig. 10E) and invasive capacity (p < 0.001, Fig. 10F) of RBE cells compared to the normal controls. Additionally, wound healing assays showed a marked reduction in cell migration following CENPH knockdown (p < 0.001, Fig. 10G). These findings suggest that CENPH may function as a critical oncogene in CHOL by promoting the proliferative, invasive, and migratory capacities of cholangiocytes.

Fig. 10.

(A) CENPH mRNA expression in biliary duct cancer cell lines and normal biliary duct epithelial cell lines. (B) Immunoblot showing CENPH protein expression in bile duct cancer cell lines and normal bile duct epithelial cell lines. (C) Representative images showing immunofluorescence staining of CENPH protein in bile duct cancer cell lines and normal bile duct epithelial cell lines. (D) Immunoblot showing CENPH protein expression in control and CENPH-knockdown cells. (E) Cell viability of control and CENPH-knockdown cells. (F) Transwell assay of control and CENPH-knockdown cells. (G) Wound healing assay of control and CENPH-knockdown cells.

Discussion

In this study, we identified the shared DEGs of two significant bile duct diseases – GS and CHOL – to explore the common genetic and molecular mechanisms. The three core genes, namely CENPH, NRSN2, and SPDYC, consistently exhibited differential expression, together with genetic alterations, and epigenetic features across the GS and CHOL datasets. Our findings are suggestive of shared pathological pathways, which aligns with previous studies noting overlapping features between GS and CHOL. These novel molecular insights can potentially pave the way for the development of innovative diagnostic and therapeutic strategies for these diseases.

The CENPH gene, which encodes the conserved centromere protein complex, has been previously implicated in various cancers[33,34]. The aberrant expression of CENPH is associated with genomic instability, a hallmark of cancer. In our study however, the pro-oncogenic activity of CENPH was experimentally validated only in CHOL cells, whereas its putative relevance to GS was mainly inferred on the basis of shared transcriptomic signals, and should therefore be interpreted more cautiously. NRSN2, a gene associated with neural growth and development, is relatively unexplored in the context of bile duct diseases[35,36]. SPDYC, also known as speedy/RINGO cell cycle regulator family member C, is known for its role in cell cycle progression and proliferation[37]. Our findings suggest the potential involvement of these genes in bile duct diseases, warranting future in-depth exploration into the underlying mechanisms. Furthermore, the results of ROC analysis showed that these core genes have significant diagnostic potential, and could serve as biomarkers for the early detection of GS and CHOL, thus enhancing the chances of early intervention and improved prognosis. The high expression of CENPH was confirmed in malignant cholangiocytes, and in vitro functional assays further showed that genetic silencing of CENPH inhibited the proliferation, invasion and migration of the cancer cells. These findings suggest that CENPH plays an oncogenic role in bile duct cancer cells, thus validating the results obtained from the CHOL dataset.

The results of enrichment analyses point to a potentially relevant cross-talk between the inflammatory IL-17 pathway and bile duct metabolism, which may converge to create a chronic inflammatory-cholestatic milieu. Bile acid accumulation can amplify epithelial stress and inflammatory signaling, while IL-17A has been shown to synergize with bile acid-induced inflammation under cholestatic conditions. Therefore, the enrichment of IL-17 signaling in GS and bile acid metabolism in CHOL may reflect a microenvironment where chronic GS-related inflammation and bile acid accumulation synergistically promote malignant transformation of the biliary epithelium[17,18,31,32].

The CNVs and promoter methylation status of the core genes in the GS and CHOL datasets were also explored to determine the possible basis of their differential expression. In addition, we identified several candidate compounds that can potentially target the core genes, thereby providing a starting point for future therapeutic exploration. These compounds should be regarded as exploratory candidates rather than validated target-drug pairs. Given that CENPH and SPDYC are closely linked to cell-cycle progression and chromosomal maintenance, cell-cycle-directed agents may represent the most rational compounds for future prioritization. For example, palbociclib, a CDK4/6 inhibitor, and dinaciclib, a pan-CDK inhibitor, have been clinically evaluated for other solid malignancies[38,39], and offer a solid framework to validate the drug candidates identified in this study.

Despite the encouraging results, this study has certain limitations that ought to be acknowledged. While our findings are backed by rigorous statistical analysis and validation across multiple datasets, experimental validation is required to confirm these results and elucidate the functional relevance of these genes. In addition, due to the modest sample size of the GS cohort and the limited number of control samples in TCGA-CHOL cohort, the feature-selection through LASSO regression should be interpreted as exploratory until confirmed in larger independent cohorts. The limited sample size of the discovery cohorts also reduces the precision of both differential-expression and ROC-based estimates. Furthermore, given the lack of serum CA19-9 and CEA measurements in the analyzed cohorts, the results of the ROC analysis merely indicate that the core genes can discriminate between GS/CHOL and normal samples at the transcriptomic level. It remains to be investigated whether the three-gene signature can outperform established clinical biomarkers. Additionally, patient-specific factors, including age, lifestyle, and comorbidities, which could potentially affect the expression of these genes, were not considered in this study. The CNV and promoter methylation results should also be interpreted cautiously. Although these alterations may help explain abnormal gene expression, we did not perform matched cis-action analyses to directly test whether copy number dosage or promoter methylation status determined transcript abundance on a per-sample basis. Accordingly, these genomic and epigenetic changes are better regarded as candidate regulatory contributors as opposed to proven direct drivers of core-gene expression in GS or CHOL. Moreover, since PubChem is principally a chemical repository, the compounds screened using the core genes will have to be validated in curated interaction resources such as DGIdb or CTD, and through functional experiments to establish direct targetability. Finally, while the pro-oncogenic role of CENPH was experimentally validated in CHOL cell models, we could not confirm a cross-disease regulatory function shared by GS and CHOL. Since the present study is primarily based on publicly available transcriptomic resources, larger prospective clinical cohorts are required to validate the robustness and clinical applicability of the identified biomarkers. In addition, functional and mechanistic studies of NRSN2 and SPDYC are also warranted, as these two genes remain less well characterized in bile duct diseases.

In conclusion, our findings not only corroborate previous observations regarding the genetic underpinnings of GS and CHOL, but also offer novel insights into the shared molecular mechanisms of GS and CHOL, which may have significant implications for the development of diagnostic tools and therapeutic strategies. While our results need to be validated experimentally, they nonetheless contribute to the growing body of knowledge regarding these complex diseases, paving the way for further research.

Ethical statement

All samples used in this study were approved by the Ethics Committee of The Affiliated Gaochun Hospital of Jiangsu University.

Funding

None.

CRediT authorship contribution statement

Chen Zhu: Writing – review & editing, Writing – original draft, Validation, Project administration, Methodology, Investigation, Formal analysis, Data curation. Kaige Huang: Writing – review & editing, Writing – original draft, Supervision, Resources. Xiaohui Zhang: Writing – review & editing, Writing – original draft, Resources, Project administration, Formal analysis, Data curation. Weiwei Shi: Visualization, Resources, Methodology, Formal analysis. Jian Fang: Writing – original draft, Visualization, Software, Project administration, Formal analysis. Senyan Hu: Writing – review & editing, Validation, Supervision, Formal analysis, Data curation, Conceptualization. Jiatao Zhang: Visualization, Supervision, Project administration, Funding acquisition, Data curation. Hongbo Jiang: Visualization, Resources, Investigation, Funding acquisition, Formal analysis. Dongdong Xie: Writing – original draft, Software, Methodology, Investigation, Formal analysis. Liangming Xu: Writing – review & editing, Supervision, Formal analysis, Data curation. Ying Sun: Writing – original draft, Supervision, Software, Methodology, Data curation, Conceptualization. Wenting Cheng: Writing – original draft, Visualization, Supervision, Data curation. Xueying Kong: Writing – review & editing, Software, Project administration, Formal analysis, Data curation. Xiaohu Yang: Writing – original draft, Visualization, Resources, Project administration, Formal analysis. Dong Xu: Writing – review & editing, Writing – original draft, Visualization, Supervision, Project administration, Investigation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability: Publicly available datasets analyzed in this study were obtained from The Cancer Genome Atlas (TCGA-CHOL) and the Gene Expression Omnibus (GEO; GSE107943) databases. The GS RNA-sequencing data is available from the corresponding author upon reasonable request.