SLCO1B3 is associated with poor survival and promotes bladder cancer progression through Wnt/β-catenin signaling

Abstract

Background

Bladder cancer possesses the highest incidence rates globally. Integrative bioinformatics analyses are necessary to develop predictive biomarkers and potential therapeutic targets. Absorption, distribution, metabolism, and excretion (ADME), the disposition of pharmaceutical compounds in organisms, exert a crucial effect on tumorigenesis.

Methods and Results

Herein, the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm identified a 6-ADME gene signature comprising CYP1A1, CYP3A5, SLC15A2, SLCO1B3, TPMT, and UGT1A1. Lower risk scores were significantly linked to higher disease-specific survival (DSS)/overall survival (OS) in patients with bladder cancer. The receiver operating characteristic (ROC) curve demonstrated that the risk score-based model exhibited a limited to moderate predictive performance in discriminating bladder cancer patients’ outcomes. Univariate and multivariate Cox risk regression analyses identified risk factors for bladder cancer patients’ prognosis: risk scores and grade. The nomogram model revealed that its prognostic outcome is reliable. In other words, the risk model could stratify patients with bladder cancer by different clinical characteristics. Moreover, among these 6 ADME key genes, SLCO1B3 was strongly associated with disease-free survival in bladder cancer patients according to GSE32894; however, its specific roles in bladder cancer have not yet been fully elucidated. In vitro, SLCO1B3 knockdown markedly inhibited cell proliferation, migration, and invasion in bladder cancer by regulating the Wnt/β-catenin signaling axis, which may correlate with its regulation of epithelial-mesenchymal transition (EMT)-related proteins.

Conclusions

A risk model based on 6 key ADME genes has been established using the LASSO algorithm and shows good prognostic potential for patients with bladder cancer. The risk score can also significantly distinguish bladder cancer samples across different clinical statuses, suggesting that the risk model may have diagnostic value for bladder cancer patients.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12920-026-02438-y.

Introduction

Bladder cancer is a common tumor worldwide, with the highest incidence in Europe, North America, and Northern Africa [1]. Bladder cancer has the highest cost of treatment and monitoring per patient, costing US$3.7 billion each year and bringing a huge medical burden [2]. About 90% of bladder cancers originate from the urothelium (also known as the transitional cell epithelium) [3]. These transitional cell carcinomas develop in two ways, resulting in two distinct carcinomas: “low-grade non-invasive” carcinoma (including 80% of urothelial carcinomas) and “high-grade invasive” carcinoma. Generally, surgery is an effective approach for treating non-invasive diseases; nevertheless, the recurrence rate is 70%, and the risk of progression to highly invasive disease is 10–20%. Despite current treatment strategies, patients with muscle-invasive bladder cancer remain at high risk of disease progression and cancer-specific mortality, with the reported 5-year overall survival rate remaining below 50% [3, 4]. Thus, integrative bioinformatics analyses developing predictive biomarkers and potential therapeutic targets are necessary.

Absorption, distribution, metabolism, and excretion (ADME) refer to the disposition of pharmaceutical compounds within organisms in pharmacokinetics and pharmacology [5–8]. Currently, the PharmaADME consortium classifies 298 ADME genes that encode phase I (functionalization) and phase II (conjugation) drug-metabolizing enzymes, transporters, and modifiers [5, 6]. Representative Phase I metabolic enzymes include oxidases, hydrolases, dehydrogenases, and deaminases. Cytochrome P450 (CYP), a superfamily of phase I drug-metabolism enzymes, includes many members (CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2E1, CYP3A4, and CYP3A5) [9]. Phase II enzymes are mainly composed of transferases, including UDP-glucuronosyltransferases (UGTs), glutathione transferases (GSTs), sulfotransferases (SULTs), N-acetyl transferases (NATs), and thiopurine methyltransferase (TPMT) [10]. Drug transporters include two main superfamilies [(ATP-binding cassette (ABC) and solute carrier (SLC)]. ABC transporter (such as ABCB1, ABCC2, and ABCG2) acts as an efflux transporter by squeezing drug molecules into the cell; SLC transporter (such as SLC15A2, SLC22A1, SLC22A2, SLC22A6, SLCO1B1, and SLCO1B3) acts as an influx transporter, contributing to small molecule uptake into cells [11]. Given the critical roles of these genes in multiple cancers [12–16], ADME genes may serve as biomarkers for bladder cancer prognosis and as therapeutic targets for bladder cancer treatment.

Herein, the datasets were obtained from the Gene Expression Omnibus (GEO) and TCGA-BLCA databases. A total of 32 ADME key genes were obtained from the PharmaADME database. A 6-ADME gene signature was constructed based on the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm. Cases from GSE32894, GSE13507, GSE48075, and TCGA-BLCA were used to analyze the correlation between ADME key genes/ 6-ADME gene signature and disease-specific survival (DSS)/overall survival (OS) in bladder cancer patients using the Kaplan–Meier estimator. The survival analysis was carried out using the two-tailed log-rank test. The analysis of the risk score model’s classification performance, measured by calibration and discrimination, was conducted using the receiver operating characteristic (ROC) curve. Gene Set Enrichment Analysis (GSEA) was performed to analyze the differences between high- and low-risk groups in GSE32894. Utilizing Metascape for pathway enrichment annotation and survival analysis, SLCO1B3 was identified as the subject of further investigation. Moreover, knockdown of SLCO1B3 markedly suppressed cell proliferation, migration, and invasion in bladder cancer cells. Taken together, our findings demonstrate that SLCO1B3 holds promise as a potential target for the clinical treatment of bladder cancer for the first time.

Materials and methods

Data source and ADME gene selection

Datasets GSE32894, GSE13507, and GSE48075 were obtained from the GEO database. GSE32894 contains 224 bladder cancer cases with DSS information. GSE13507 contains 110 bladder cancer carcinoma patients with OS information. GSE48075 contains 73 bladder carcinoma patients with OS information. The OS information of 406 bladder carcinoma patients was collected from The Cancer Genome Atlas Urothelial Bladder Carcinoma Collection (TCGA-BLCA). GSE32894 was used as the training cohort for prognostic model development, whereas GSE48075, TCGA-BLCA, and GSE13507 were used as independent external validation cohorts.

A total of 32 ADME key genes were collected from the PharmaADME (http://pharmaadme.org/) database (Table S1).

Identification of prognostic signature

Using GSE32894 as the training cohort, the coefficients for the 32 ADME genes associated with OS in bladder cancer patients were determined by the LASSO algorithm, and a 6-ADME gene signature was subsequently constructed. The prognostic performance of the model was further evaluated in three independent external validation cohorts (GSE48075, TCGA-BLCA, and GSE13507). The LASSO package in R was used to calculate each patient’s risk score based on the normalized expression of the candidate genes (Expi) and their corresponding regression coefficients (Coei) [17, 18] .

Herein, the formula is:

Survival analyses using the Kaplan–Meier method

Cases in GSE32894 were divided into high- and low-ADME gene expression groups using the median expression value as a dividing line. Kaplan-Meier analysis and a two-tailed log-rank test were used to assess whether the 6 ADME genes in GSE32894 were associated with disease-specific survival in bladder cancer patients.

The cases in GSE32894, GSE13507, GSE48075, and TCGA-BLCA were divided into high- and low-risk groups based on the median risk score. Patients with bladder cancer had their DSS and OS correlated with their risk score using the Kaplan-Meier technique and a two-tailed log-rank test.

In addition, to evaluate the robustness of the median risk score cutoff used for risk stratification, a separate bootstrap resampling analysis (1,000 iterations with replacement) was performed using the R package boot. The median risk score was recalculated in each bootstrap sample to estimate the empirical distribution, bias, and bias-corrected and accelerated (BCa) 95% confidence interval of the cutoff value.

Classification performance of the risk score model analyzed by the ROC curve

The R package survival ROC was used to generate ROC curves for 1-, 2-, 3-, 4-, and 5-year periods, which were then used for a predictive accuracy study based on GSE32894, GSE48075, TCGA-BLCA, and GSE13507. The univariate and multivariate Cox regression analyses were used to investigate the association between clinical risk markers and survival to develop genome-clinicopathologic nomograms. Markers, including age, gender, and grade, were used to create the genome-clinicopathologic nomograms. The markers were based on the 6-ADME gene signature risk score.

Genome-clinicopathologic nomogram

The model performance was assessed via a bootstrap method with 1000-iteration resampling using calibration and discrimination. The discrimination assessment, which is the capacity of the model to differentiate between patients who have survived and those who have died, was carried out by calculating the consistency index (C index). After that, the 1-, 3-, and 5-year survival nomogram calibration curves were plotted to assess how closely the predicted survival probabilities matched the actual probabilities.

GSEA for interpreting gene expression data

The differences between high-risk group and low-risk group in GSE32894 were analyzed using GSEA conducted in javaGSEA v. 3.0 based on Molecular Signatures Database v. 6.2. Within the high-risk group, poor survival-related factors were determined via searching C2 (curated gene sets), C5 (GO gene sets), and C6 (oncogenic signatures) for identifying the enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, biological activities, cellular elements, and molecular functions, as well as maladjusted oncogenic signatures. |NES| > 1 and FDR < 0.05 are indicative of statistical significance.

Bioinformatics analysis by online tools

The Gene Ontology (GO) analysis of the 6 ADME key genes was performed using Metascape (https://metascape.org/). The online tool Sangerbox (http://past20.sangerbox.com/) was used to predict SLCO1B3 expression in TCGA-BLCA. The online tool RNAInter (http://www.rna-society.org/rnainter/home.html) was used to predict the potential interaction between mRNA and proteins. Metascape (https://metascape.org/gp/index.html#/main/step1) was employed to analyze the GO and KEGG of those interaction factors.

Clinical sample collection

Tumor tissue samples were surgically removed and collected from bladder cancer cases undergoing surgical excision at Xiangya Hospital. A total of 15 cases of primary bladder urothelial carcinoma were harvested, including tumorous and adjacent tissues. Inclusion criteria were: (1) complete clinical data and follow-up records, (2) postoperative pathology showing T1-T4 stage urothelial carcinoma, and (3) imaging and pathology showing no signs of distant metastasis.

Cell lines and cultivation

A normal human urethral epithelial cell line, SV-HUC-1 (CRL-9520), and two bladder cancer cell lines, RT4 (HTB-2) and T24 (HTB-4), were procured from ATCC (Manassas, USA). SV-HUC-1 cells were cultured in F-12 K medium, and T24 and RT4 cells were cultured in McCoy’s 5 A medium containing 10% fetal bovine serum (FBS). All cell types were cultured at 37 °C in a humidified atmosphere containing 5% CO₂.

Quantitative real time PCR (qRT-PCR)

Cell and tissue samples were collected, and Trizol (Invitrogen, Carlsbad, USA) was employed to extract total RNA. A reverse transcription kit (TaKaRa, Tokyo, Japan) was used according to the kit protocol. The LightCycler 480 fluorescence quantitative PCR instrument (Roche Diagnostics, Indianapolis, USA) was used to measure gene expression, with reaction conditions following the instructions of the fluorescence quantitative PCR kit (SYBR Green Mix, Roche Diagnostics). The cycling conditions were set as previously described [19]. Each quantitative PCR reaction was carried out three times. β-actin was used as the internal control. The 2^−ΔΔCt method, where ΔΔCt= Experimental group (Ct target gene - Ct internal reference) - Control group (Ct target gene - Ct internal reference), was employed to conduct data analysis. Detailed sequences of each gene and its internal reference amplification primers are provided. The primers utilized in the present study are listed in Table S2.

Cell transfection and treatment

Versatile DNA reagent jetPRIME^® (101000046, PolyPlus, Illkirch, France) was used for transfection as directed by the manufacturer. The short Hairpin RNA (shRNA) sequences targeting SLCO1B3 were synthesized and annealed by Shanghai GenePharma and inserted into the plv-shRNA2 plasmid; the sequences are listed in Table S2. When cells in the six-well plate grew to a confluence of 70 ~ 80%, transfection was carried out in accordance with the instructions. Functional experiments were conducted, and cell RNA and total protein were isolated 48 h following transfection. Cell groups were designated according to the corresponding shRNA constructs. To further investigate the involvement of Wnt/β-catenin signaling, cells were treated with the pathway activator CHIR99021 (4 µM; S1263, Selleck Chemicals, Houston, USA) for 24 h [20].

Immunoblotting

To yield protein samples, RIPA lysis buffer (Beyotime, Shanghai, China) was used to lyse tissues or cells; samples were heated for 5 min in a boiling water bath to denature proteins. The protein was subjected to electrophoresis at 80 V for 30 min and then switched to 120 V for 1–2 h until the bromophenol blue indicator reached the bottom edge. The transfer was performed in an ice bath at a transfer current of 220 mA for 120 min. The membrane was then placed in a blocking solution for 60 min at room temperature (RT), followed by an overnight incubation at 4 °C with primary antibodies, and a 1-h incubation at RT with secondary antibodies the next day. An enhanced chemiluminescence (ECL) substrate (Merck Millipore, Billerica, USA) was employed to visualize protein signals. GAPDH was used as a loading control. The gray intensity was calculated using ImageJ software (NIH, Bethesda, USA). The antibodies used are provided in Table S3.

Histological analysis

After routine fixation and embedding of tissues, sections were prepared and stained with hematoxylin and eosin (H&E) according to the protocol of the H&E staining kit (Beyotime). SLCO1B3 expression in tissue samples was determined through immunohistochemical (IHC) staining. After the sections underwent antigen retrieval, endogenous peroxidase blocking, and closure of nonspecific binding sites, they were incubated overnight in a moist chamber at 4 °C with the primary antibody against SLCO1B3. Following overnight incubation, the PV-9000 Universal Two-Step Detection Kit (Zhongshan Golden Bridge Biotechnology Co, Beijing, China) and DAB color reagent were used to complete subsequent steps. The antibodies utilized in the IHC staining are presented in Table S3.

Scratch migration assay

Cells with different transfections were evenly spread in the wells of a six-well plate. Cells were grown to 100% confluence in each well, vertically scratched in the middle of the well using a sterile 200 µL pipette tip, and then incubated at 37 °C in a low-serum medium in a 5% CO₂ incubator. Photographs were taken for reference and record. At 0 h and 24 h, a microscope was used to observe and photograph the same location of the scratch. The changes in the recorded scratch width were statistically analyzed.

Transwell invasion assay

Cells with different transfections were diluted to the same concentration using a culture medium without FBS. Matrigel-coated (1:8, #356234, Corning Incorporated, Corning, USA) sterile Transwell chambers were planted into the wells of a 24-well plate. Subsequently, 100 µL of the equally concentrated cell suspension was introduced into the top chamber, while 800 µL of complete medium supplemented with 10% FBS was introduced into the bottom chamber. Next, the chambers were inserted into a cell incubator. After 18 h, the Transwell chambers were removed, and cells were stained with crystal violet for 20 min. Photographs were taken of five randomly selected fields, and cell counts were statistically analyzed using ImageJ, with counts recorded for each field of view.

Cell Counting Kit-8 (CCK-8) detects cell proliferation

Cells from different transfections were prepared as single-cell suspensions, counted, and plated into a 96-well plate (5,000 cells/well) with 100 µL of culture medium in each well. After adding 10 µL of CCK-8 solution (ab228554, Abcam, Cambridge, UK) to each well, cell viability was evaluated at different time points, and the plate was returned to the incubator for a 2-h incubation. The absorbance was then determined at 450 nm.

Statistical analyses

SPSS (version 22.0) and the R programming language (version 3.5.1; https://www.r-project.org/) were used for all statistical analyses. P < 0.05 was chosen as the threshold of statistical significance. Clinical values that were absent from the list were removed, and if the value of any parameter was missing, the whole specimen was omitted from the study. From the date of diagnosis till the date of death, OS was determined. DSS measured the proportion of patients who did not experience death from a certain condition within a given trial or treatment group. All continuous variables are presented as mean ± standard deviation (SD). The Student’s t-test was used to compare two groups. For comparisons involving more than two groups, one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was applied. Survival analysis between high- and low-risk groups was conducted in R using the Kaplan-Meier technique and a two-tailed log-rank test.

Results

ADME gene-based risk model established using the LASSO algorithm

To establish the risk model comprising ADME key genes, LASSO Cox regression analysis was performed on 32 core ADME genes obtained from Pharma ADME (http://pharmaadme.org/) and listed in Table S1, using data from 224 cases in GSE32894. Based on the minimum criteria, a 6-ADME key gene risk signature was obtained (Fig. 1A) using the risk score formula. The 6 ADME key genes are: CYP1A1, CYP3A5, SLC15A2, SLCO1B3, TPMT, and UGT1A1 (Fig. 1B).

Fig. 1.

Absorption, distribution, metabolism, and excretion (ADME) gene-based risk model established using the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm (A-B) The LASSO algorithm was performed to calculate the coefficients of 32 core ADME genes. Coefficient of each of the 6 genes that form the signature was shown

Cases in GSE32894 were divided into high- and low-expression groups using the median of the 6 genes mentioned above as the cutoff; the Kaplan–Meier method and a two-sided log-rank test were used to assess the association between the 6 ADME key genes and DSS in bladder cancer patients in GSE32894. Expression of CYP3A5, SLCO1B3, and UGT1A1 was significantly correlated with bladder cancer patients’ DSS (Fig. S1A, B, and C), respectively, whereas CYP1A1, SLC15A2, and TPMT were not correlated with patients’ DSS (Fig. S1D, E, and F), suggesting that the 6-gene signature has better prognostic potential than a single gene. To further evaluate the robustness of the median risk score used for patient stratification, bootstrap resampling (1,000 iterations) was performed. The original median cutoff value was − 1.463, with a bootstrap-estimated bias of only − 0.016, indicating minimal systematic deviation. The empirical distribution of the bootstrap-estimated median values was centered on the original cutoff, suggesting good stability (Fig. S2A). Moreover, the normal Q–Q plot demonstrated that the bootstrap estimates approximately followed a normal distribution. The bias-corrected and accelerated (BCa) 95% confidence interval ranged from − 2.065 to -0.951, further supporting the robustness of the selected cutoff value (Fig. S2B).

Correlation of the risk score with the bladder cancer patients’ DSS or OS

Next, the predictive performance and generalizability of the established model were subsequently evaluated using three independent external validation cohorts (GSE48075, TCGA-BLCA, and GSE13507). The risk scores for cases in GSE32894, GSE48075, TCGA-BLCA, and GSE13507 were calculated using the formula described above. Using the median risk score, cases were classified into high- and low-risk groups. The Kaplan–Meier method with a two-sided log-rank test was used to assess the correlation between risk score and DSS/OS in patients with bladder cancer. According to GSE32894 (Fig. 2A-B), GSE48075 (Fig. 2C-D), TCGA-BLCA (Fig. 2E-F), and GSE13507 (Fig. 2G-H), lower risk scores were associated with better DSS/OS outcomes, suggesting that the 6-gene signature has prognostic value in bladder cancer.

Fig. 2.

Correlation of the risk score with the bladder cancer patients’ disease-specific survival (DSS) or overall survival (OS) Cases in GSE32894 (A-B), GSE48075 (C-D), TCGA-BLCA (E-F), and GSE13507 (G-H) were assigned into high- and low-risk score groups, respectively, using the median risk score as cut-off. The correlation between risk scores and DSS/OS in bladder cancer patients was analyzed using the Kaplan–Meier method with a two-sided log-rank test

Model classification performance verification

Subsequently, the predictive efficiency of the survival model was examined using the ROC curve [21]. The area under the curve (AUC) for DSS/OS from GSE32894 (Fig. S3A), GSE48075 (Fig. S3B), TCGA-BLCA (Fig. S3C), and GSE13507 (Fig. S3D) was 0.61 (DSS, 5-year), 0.71 (OS, 5-year), 0.53 (OS, 5-year), and 0.57 (OS, 5-year). The ROC curve indicated that the risk score-based model exhibited a limited to moderate discriminative ability.

Next, univariate and multivariate Cox risk regression analyses were carried out to identify bladder cancer patient DSS-associated factors. As shown in Fig. S4, grade and risk score were significantly associated with DSS in bladder cancer patients.

Nomogram model was subsequently used to evaluate the 1-, 3-, and 5-year prognosis of bladder cancer patients based on GSE32894, which included age, gender, grade, and risk score. The consistency index of the risk score for GSE32894 is 0.85 (Fig. S5), indicating that the model’s prognostic performance is reliable.

Diagnostic potential of risk scores

To evaluate the prognostic ability of the risk score signature, the correlation of the risk score with patients’ age, clinical grade, clinical response to treatment, survival outcome, TNM stage, and other clinical characteristics was analyzed using cases from TCGA-BLCA. The cases from TCGA-BLCA were assigned into different groups, including age (> 55 vs. < = 55, Fig. S6A), grade (low-grade vs. high-grade, Fig. S6B), TNM-T stages (T0/1/2 vs. T3/4, Fig. S6C), TNM-M stages (M0 vs. M1/MX, Fig. S6D), primary therapy outcome (stable disease vs. progressive disease, Fig. S6E), primary lymph node (no vs. yes, Fig. S6F), vital status (alive vs. dead, Fig. S6G), and diagnosis subtype (papillary vs. non-papillary, Fig. S6H), risk scores for cases within paired groups were markedly different. Moreover, according to GSE32894, when cases were assigned to different grades (low-grade vs. high-grade, Fig. S7A) or TNM-T stages (T0/1/2 vs. T3/4, Fig. S7B), the risk scores across groups differed significantly. According to GSE13507, when cases were grouped by tumor grade (low-grade vs. high-grade, Fig. S7C), invasion status (muscle-invasive vs. non-muscle-invasive, Fig. S7D), and progression status (stable disease vs. progressive disease, Fig. S7E), the risk scores across groups differed significantly. In other words, the risk model could stratify patients with bladder cancer based on various clinical characteristics.

The differences between high- and low-risk groups in GSE32894 were analyzed based on GSEA

Finally, GSEA was used for analyzing differences between high- and low-risk groups in GSE32894. As shown in our results, remarkably higher scores were observed in the bladder cancer-specific gene set in the high-risk group (Fig. S8A), the gluconeogenesis process (Fig. S8B-C), and mitochondrial composition and activity pathways (Fig. S8D-E).

SLCO1B3 regulates bladder cancer cell proliferation and epithelial-mesenchymal transition (EMT)

Next, a specific single gene associated with bladder cancer was analyzed using these 6 core ADME genes. The pathway enrichment analysis indicates that the 6 ADME genes were mainly linked to cellular responses to xenobiotic stimuli, xenobiotic catabolic process, and biological oxidation pathways (Fig. 3A). Subsequently, using the median risk score value as a cutoff, the bladder cancer patients were classified into a high-risk group and a low-risk group. SLCO1B3, CYP1A1, and SLC15A2 were significantly upregulated, whereas UGT1A1 and CYP3A5 were significantly downregulated in the high-risk group (Fig. 3B). The correlation between the six risk-associated genes and prognosis was analyzed based on GSE32894, and UGT1A1, CYP3A5, and SLCO1B3 were dramatically related to bladder cancer patients’ DSS in GSE32894 at the single-gene level (P < 0.05, Fig. S1A, B, and C). Considering SLCO1B3 is highly expressed in bladder cancer and differs from UGT1A1 or CYP3A5, therefore, the specific role and underlying mechanism of SLCO1B3 was further investigated. The results of functional enrichment analysis of SLCO1B3-related genes showed that SLCO1B3 was significantly associated with multiple signaling pathways, including the Wnt signaling pathway and the MAPK cascade (Fig. 3C-D).

Fig. 3.

Bladder cancer prognostic model based on 6 ADME core genes to screen single genes associated with bladder cancer (A) Pathway enrichment analysis of the 6 ADME core genes by Metascape (https://metascape.org/). B Expression levels of the 6 ADME core genes in GSE32894. C-D Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment annotation performed on SLCO1B3

Clinical bladder carcinoma and adjacent para-cancerous specimens were harvested and evaluated for the expression of SLCO1B3. As expected, the expression level of SLCO1B3 mRNA and protein levels were significantly increased in bladder cancer tissue samples compared with adjacent para-cancerous samples (Fig. 4A-D). In addition, compared with the normal human urethral epithelial cell line SV-HUC-1, SLCO1B3 mRNA and protein expressions was significantly upregulated in two bladder carcinoma cell lines, RT4 and T24 (Fig. 4E-F).

Fig. 4.

SLCO1B3 is overexpressed in bladder cancer tissues and cell lines (A) The mRNA expression levels of SLCO1B3 were determined in bladder cancer and adjacent non-cancerous tissues using qRT-PCR. B Histopathological alterations in bladder cancer and adjacent non-cancerous tissues were examined using hematoxylin and eosin (H&E) staining. C-D The protein levels of SLCO1B3 in bladder cancer and adjacent non-cancerous tissues were examined using Immunohistochemical staining (IHC staining) and Immunoblotting. E-F The mRNA and protein expression levels of SLCO1B3 were determined in a normal human urethral epithelial cell line SV-HUC-1 and two bladder cancer cell lines, RT4 and T24 using qRT-PCR and Immunoblotting. Data are analyzed using unpaired Student’s t-test for (C, D) or one-way ANOVA followed Tukey’s post hoc test for (E, F). * P < 0.05, ** P < 0.01, *** P < 0.001, compared with adjacent non-cancerous tissues or SV-HUC-1

Furthermore, SLCO1B3 was knocked down in RT4 and T24 cells, and its specific effects on cellular phenotypes were investigated. The shRNA sequences targeting SLCO1B3 were transfected into RT4 and T24 cells, and the effects were confirmed by Immunoblotting (Fig. 5A). Functional experiments showed that SLCO1B3 knockdown significantly suppressed cell proliferation, migration, and invasion in bladder cancer cells (Fig. 5B-D). Furthermore, EMT-related proteins were assessed due to their important roles in cancer cell migration and invasion. The results showed that SLCO1B3 knockdown significantly increased E-cadherin expression, whereas it reduced the expression of N-cadherin, snail, and vimentin (Fig. 5E).

Fig. 5.

Knocking down SLCO1B3 inhibits the proliferation, migration, and invasion of bladder cancer cells (A) SLCO1B3 knockdown was achieved in RT4 and T24 cells by transducing shRNA targeting SLCO1B3 (sh-SLCO1B3#1 and sh-SLCO1B3#2); transduction efficiency was verified using Immunoblotting. Then, RT4 and T24 cells were transduced with sh-SLCO1B3#1 and sh-SLCO1B3#2 and examined for the cell viability using Cell Counting Kit-8 (CCK-8) assay (B); cell migration using scratch wound healing assay (C); cell invasion using Transwell (D); the protein levels of E-cadherin, N-cadherin, Snail, and vimentin using Immunoblotting (E). Data are analyzed using one-way ANOVA followed Tukey’s post hoc test. ** P < 0.01, compared with sh-NC group

SLCO1B3 promotes bladder cancer progression via activation of the Wnt/β-catenin signaling pathway

Considering that the Wnt/β-catenin signaling pathway plays a central role in regulating tumor proliferation, EMT, and metastatic progression in bladder cancer [22, 23], we further investigated whether SLCO1B3 exerts its oncogenic function through this pathway. SLCO1B3 knockdown significantly inhibited Wnt/β-catenin signaling activity in RT4 and T24 bladder cancer cells, as demonstrated by decreased expression of active β-catenin, c-Myc, and Cyclin D1 (Fig. 6A). Functionally, SLCO1B3 silencing markedly suppressed bladder cancer cell proliferation (Fig. 6B), migration (Fig. 6C), and invasion (Fig. 6D). Notably, activation of Wnt/β-catenin signaling using CHIR99021 partially reversed these inhibitory effects, restoring malignant phenotypes in both cell lines (Fig. 6B-D). In addition, CHIR99021 treatment also attenuated the EMT-suppressive effects induced by SLCO1B3 knockdown, as evidenced by decreased E-cadherin and increased N-cadherin, Snail, and Vimentin expression (Fig. 6E). Consistently, CHIR99021 treatment restored downstream signaling activity, as shown by re-elevated β-catenin, c-Myc, and Cyclin D1 levels (Fig. 6F). These findings suggested that SLCO1B3 promoted bladder cancer progression, at least in part, through activation of the Wnt/β-catenin signaling axis and subsequent regulation of EMT processes.

Fig. 6.

SLCO1B3 regulates bladder cancer progression via the Wnt/β-catenin signaling pathway (A) Expression levels of Wnt/β-catenin signaling pathway–related proteins, including active β-catenin, c-Myc, and Cyclin D1, in RT4 and T24 bladder cancer cells following SLCO1B3 knockdown were detected by Immunoblotting. Next, RT4 and T24 cells were transduced with sh-SLCO1B3#1 and/or treated with Wnt/β-catenin pathway activator CHIR99021 (4 µM) for 24 h and examined for the cell viability using CCK-8 assay (B); cell migration using scratch wound healing assay (C); cell invasion using Transwell (D); the protein levels of E-cadherin, N-cadherin, Snail, and vimentin using Immunoblotting (E); the protein levels of active β-catenin, c-Myc, and Cyclin D1 using Immunoblotting (F). Data are analyzed using one-way ANOVA followed Tukey’s post hoc test. ** P < 0.01, compared with sh-NC+ DMSO group; ## P < 0.01, compared with sh-SLCO1B3 + CHIR99021 group

Discussion

Herein, the LASSO algorithm identified a 6-ADME gene signature comprising CYP1A1, CYP3A5, SLC15A2, SLCO1B3, TPMT, and UGT1A1. CYP3A5, SLCO1B3, and UGT1A1 were shown to be dramatically linked to the DSS of patients with bladder cancer. In addition, lower risk scores were significantly linked to higher DSS/OS in patients with bladder cancer. The ROC curve indicated that the risk score-based model exhibited a limited to moderate discriminative ability. Univariate and multivariate Cox risk regression analyses showed that risk scores and grade were identified as the risk factors for bladder cancer patients’ prognosis. The nomogram model revealed that its prognostic outcome is reliable. Moreover, the risk scores across groups differed significantly. In other words, the risk model could stratify patients with bladder cancer based on various clinical characteristics. As revealed by GSEA analysis, the high-risk group showed considerably higher scores in bladder cancer-specific gene sets, gluconeogenesis process, and mitochondrial composition and activity pathways.

With advances in high-throughput screening of ADME characteristics in drug discovery, in silico ADME analysis and prediction play crucial roles, helping pharmaceutical companies select candidate drugs before expensive clinical trials [24–26]. In particular, machine learning methods link ADME genes to molecular characteristics and establish complex structure-property relationships across multiple molecular structures and mechanisms, demonstrating the enormous potential of using ADME genes as biomarkers and therapeutic targets [27, 28]. Herein, 32 ADME key genes were obtained from the Pharma ADME database and used for integrative bioinformatics analyses to identify potential biomarkers for bladder cancer prognosis. Recently, integrated profile analysis for bladder cancer gene expression and the construction of gene signatures associated with bladder cancer prognosis have aroused extensive attention. For example, as reported by Chen et al. [29], m⁶A RNA methylation modulators contribute significantly to bladder cancer progression and influence its prognosis, as identified by the LASSO Cox regression model. Cao et al. [30] established an EMT-associated gene signature using the LASSO Cox regression model in bladder carcinoma. Herein, using the LASSO algorithm, a 6-ADME key gene signature comprising CYP1A1, CYP3A5, SLC15A2, SLCO1B3, TPMT, and UGT1A1 was identified.

Reportedly, CYP1A1 mRNA expression was overexpressed in 65% of bladder carcinomas [31]. CYP1A1 gene polymorphisms may be strongly associated with an increased risk of bladder cancer [32]. SNPs in NR1I2 and its target genes (CYP3A5 and ABCB1) have been found to exert a genetically significant effect on temsirolimus pharmacokinetics and toxicity in patients with bladder cancer [33]. In addition, whole-genome analysis of patients revealed that SLC15A2 genomic variation was strongly associated with an abnormal response to sorafenib treatment [34]. SLCO1B3 is expressed extranhepatically in various cancer cells and tissues and plays an emerging role in cancer resistance [35]. TPMT genetic polymorphism, as a well-established pharmacogenetic example, spans from molecular genetics to clinical diagnostics for individualized thiopurine therapy (such as azathioprine, mercaptopurine, and thioguanine) [36]. UGT1A1 gene polymorphisms are strongly associated with the onset and prognosis of non-small cell lung cancer [37], as well as with the curative effect and toxicity of preoperative chemoradiotherapy with irinotecan [38].

In this study, only CYP3A5, SLCO1B3, and UGT1A1 were significantly associated with DSS in patients with bladder cancer according to GSE32894. However, in GSE32894, GSE13507, GSE48075, and TCGA-BLCA, lower risk scores were notably associated with higher DSS/OS in patients with bladder cancer, indicating that the risk score model has better predictive efficiency. The satisfactory predictive efficiency of the 6-ADME key gene signature was subsequently revealed by the ROC curve. Moreover, as shown by univariate and multivariate Cox regression analyses, risk scores and grade were prognostic factors for patients with bladder cancer. As shown by the nomogram model, the prognostic outcome is reliable. Moreover, the risk model could stratify bladder cancer patients by clinical characteristics, suggesting it may have diagnostic value. Lastly, according to the GSEA analysis, the high-risk group showed notably elevated scores in the bladder cancer-specific gene set, gluconeogenesis, and mitochondrial composition and activity pathways.

The SLCO1B3-encoded protein OATP1B3 (organic anion transporting polypeptide 1B3) was initially identified in the liver as a transmembrane transport protein. Subsequent studies have shown that SLCO1B3 is also expressed in multiple solid tumor tissue samples, including prostate cancer [39] and breast cancer [40], and it is either tumor-suppressing or tumor-promoting in various malignancies. Herein, abnormally high SLCO1B3 expression was observed in bladder carcinoma and was associated with poor patient prognosis. Knockdown of SLCO1B3 expression could significantly inhibit the malignant phenotype of bladder carcinoma. Myometrial invasion and metastasis of bladder cancer cells rely heavily on EMT, a crucial stage in which epithelial cells acquire invasiveness [41–43]. Snail, N-cadherin, and vimentin, markers of the mesenchymal lineage, are upregulated during this transition, whereas E-cadherin, Zonula occludens-1 (ZO-1), and occludin, indicators of the epithelial lineage, are downregulated [43, 44]. In this study, knockdown of SLCO1B3 dramatically increased E-cadherin levels but reduced snail, N-cadherin, and vimentin, suggesting that the aforementioned functions of SLCO1B3 in bladder cancer aggressiveness are related to its regulatory role in the expression of key EMT genes. Moreover, SLCO1B3 promoted bladder cancer progression, at least in part, through activation of the Wnt/β-catenin signaling axis and subsequent regulation of EMT processes.

The limitations of this study are also worthy of attention. First, this study is based on retrospective datasets from GEO and TCGA; therefore, inherent selection bias and inter-dataset heterogeneity may affect the robustness and generalizability of the prognostic model. Second, although the 6-ADME gene signature demonstrated statistically significant survival stratification, its overall predictive performance remains limited, as reflected by relatively low AUC values, indicating only modest clinical applicability at present. Third, mechanistic validation of SLCO1B3 has been performed mainly in vitro, and in vivo evidence remains lacking. In this study, the Wnt/β-catenin signaling axis was studied to elucidate the potential mechanism of SLCO1B3; however, it is acknowledged that other signaling pathways, such as PI3K/Akt and MAPK, which are well-established in bladder cancer progression, may also be involved and warrant further investigation. In addition, the relatively small number of clinical samples further limits the translational robustness of our findings. Future studies should therefore include larger, multicenter, prospective cohorts, as well as multi-omics integration, to improve model stability and predictive accuracy. Moreover, additional in vivo experiments and comprehensive mechanistic studies are needed further to clarify the role of SLCO1B3 in bladder cancer and evaluate its potential as a therapeutic target.

Taken together, a risk model based on 6 key ADME genes has been established using the LASSO algorithm and has good prognostic value for patients with bladder carcinoma. It is important to mention that the risk score can also significantly distinguish bladder cancer samples in different clinical statuses, suggesting that the risk model might have certain diagnostic value for patients with bladder cancer. Moreover, we first demonstrated the effects and mechanisms of SLCO1B3 in bladder cancer aggressiveness, and targeting SLCO1B3 holds promise as a potential treatment for bladder cancer.

Authors’ contributions

M.X. and X.S. contributed to experimental design and supervising the whole experimental process; Ku., X.Z. and M.C. were involved in the experimental conducting; L.Q. and W.H. contributed to the data analysis and manuscript preparation. All the authors read, revised and approved the final manuscript.

Funding

This study was supported by the National Natural Science Foundation of China (No. 81802561) and the Natural Science Foundation of Hunan Province, China (No. 2019JJ50977 and 2023JJ30881).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study involving human participants was approved by the Ethics Committee of Xiangya Hospital, Central South University, in accordance with the Declaration of Helsinki (latest revision). Written informed consent was obtained from all participants before inclusion in the study. All procedures were conducted in compliance with relevant institutional, national, and international guidelines and regulations.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.