Transcriptome data for IC/BPS patients were obtained from the Gene Expression Omnibus (GEO) database, encompassing GSE11783 (15), GSE28242 (16), and GSE57560 (17). Key information about these datasets is summarized in Table 1 . Due to differences in sequencing platforms, methodologies, and experimental designs across the datasets, we applied the “Combat” function to adjust for batch effects arising from these variations. Before applying “Combat,” we first preprocessed the data by removing outliers. The method was then used to correct for batch effects, ensuring that subsequent analyses more accurately reflected biological variations rather than experimental design differences. Following established protocols, the datasets were integrated using the “sva” package in R software, effectively eliminating batch effects for subsequent analysis (18, 19).

DEGs between IC/BPS samples and normal samples were identified using the “limma” package in the combined dataset (20). A threshold of P-value < 0.05 and |log2FC| > 1 was applied to screen for DEGs. Adjusted P-value < 0.05 and FDR (Q-value) < 0.25 were set as the cut-off criteria. The expression patterns of these DEGs in both patient groups were then visualized using the “pheatmap” package.

To further investigate the biological role of these DEGs, we performed functional enrichment analyses of “clusterProfiler” packages (P-value < 0.05), including the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and genes Genomes (KEGG) (21). Additionally, Gene Set Enrichment Analysis (GSEA) was undertaken in both patient groups, utilizing the “clusterProfiler” and “enrichplot” packages. Detailed descriptions of these methods can be found in previous studies (22, 23).

To further analyze potential pathogenic genes in IC/BPS patients, we used the “WGCNA” and “limma” R packages to construct gene co-expression networks and analyze correlations between gene modules and specific phenotypes (20, 24). The key gene of the relevant module is defined as the first principal component gene of each module and is adopted as a representative of all genes in each module (Module membership > 0.80 and P < 0.05).

To pinpoint the key pathogenic genes, we employed three machine-learning algorithms—random forest (RF) (25), loss absolute shrinkage and selection operator (LASSO) (26) and support vector machine -recursive feature elimination (SVM-RFE) (27). The analyses for SVM-RFE, RF, and LASSO regression were conducted using the R packages “glmnet,” “kernlab,” and “randomForest” (28–30). Genes situated in the overlapping regions of the three algorithms were deemed crucial in IC/BPS diagnosis. Subsequently, a diagnostic nomogram model was constructed based on these identified genes. To further evaluate the predictive model’s capability in distinguishing between normal and IC/BPS tissue, we conducted decision curve analysis (DCA), calibration curve, and clinical impact curve (CIC) using the R packages “pROC” and “rmda” (31, 32). Additionally, we assessed the studies based on receiver operating characteristic (ROC) curves, calculating the area under the curve (AUC) to measure the predictive power of the key genes. Detailed methodologies are referenced in several prior studies (33).

Immune cells play a crucial role in the crosstalk that influences disease progression and treatment response. To assess each immune cell across all samples, we utilized the ssGSEA algorithm, incorporating the “GSVA,” “GSEABase,” and “limma” packages. This analysis was based on 28 immune cell gene sets, allowing us to scrutinize and compare the differences between normal and IC/BPS tissues (34).

Between January 2022 and August 2023, we collected 8 bladder tissue samples from individuals diagnosed with interstitial cystitis/bladder pain syndrome (IC/BPS) and 10 bladder tissue samples from unaffected controls (UC), following previously established methods (11, 35). Cystoscopy was performed on all IC/BPS patients, and biopsies were taken from tissue suspected to contain Hunner’s lesions. UC samples were obtained from patients with benign bladder tumors, with biopsies confirming the absence of inflammatory changes. Laboratory test results, including hemoglobin (Hb), white blood cell (WBC) count, neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR), and absolute lymphocyte count (ALC), were collected within 12 hours prior to biopsy. Additionally, clinical data on age, sex, history of urinary surgery, diabetes, hypertension, smoking, and alcohol consumption were recorded. The classification of samples into IC/BPS or UC groups was independently verified by two pathologists.

The specific immunohistochemical staining methods employed in this study were referenced from our previous studies (22, 36). All tissues were transformed into pathological sections and then processed using an IHC kit as per the manufacturer’s instructions (cat. SP-9001; ZSGB-BIO, China). The primary antibodies utilized in this study included PLAC8 (cat. 12284-1-AP, Proteintech, China), CXCL10 (cat. DF6417, Affinity, USA), c-Kit (CD117) (cat. AF6153, Affinity, USA), SDC1 (CD138) (cat. DF6367, Affinity, USA), CD163 (cat. DF8235, Affinity, USA), CD20 (cat. DF13319, Affinity, USA), CD14 (cat. DF13278, Affinity, USA), FucT-IV (CD15) (cat. DF8547, Affinity, USA), and CD3D (cat. DF6370, Affinity, USA). Two senior pathologists independently evaluated the IHC staining results. The percentage of positive cells was categorized as follows: 1 = 0-25%, 2 = 26%-50%, 3 = 51%-75%, and 4 = 76%-100%. Staining intensity was defined as 0 = no staining; 1 = weak staining; 2 = medium staining; 3 = strong staining. The final IHC score was derived by multiplying the staining percentage and staining intensity. Immune cell staining was assessed based on the number of positive cells within the field of vision.

Normal human urothelial cell line (SV-HUC-1) was purchased from Procell Life Science & Technology Co., Ltd. (China) using Ham’s F-12 (F-12K) medium modified by Kaighn (Gibco, USA) for cultivation. SV-HUC-1 cell lines were treated with lipopolysaccharide (LPS, 10ug/L, MedChemExpress) and adenosine triphosphate (ATP, 2.5 mM, MedChemExpress) to induce inflammatory damage. We overexpressed PLAC8 using PB-EF1a-FLAG-PLAC8-IRES-Hygromycin vector, and selected PB-EF1a-3FLAG-GFP-IRES-Hygromycin vector as a negative control group. These vectors were purchased from Sigma-Aldrich according to the manufacturer’s instructions.

Total proteins were extracted from cells using RIPA lysis buffers of protease and phosphatase inhibitors (Beyotime Biotechnology, China). Pierce BCA protein detection kit was used to determine the protein concentration (Beyotime Biotechnology, China). Protein samples were isolated by sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to PVDF membranes (Beyotime Biotechnology, China). Incubate overnight with primary antibodies, including PLAC8 (1:1000; Proteintech, China), Flag (1:1000; Sigma-Aldrich, USA), α-tubulin (1:10000, Sigma-Aldrich, USA), AKT (1:1000; Proteintech, China), p-AKT (1:1000; Proteintech, China), mTOR (1:1000; Proteintech, China), p-mTOR (1:1000; Proteintech, China), and PI3K (1:1000; Proteintech, China). The PVDF membrane was then incubated with the horseradish peroxide-coupled secondary antibody (Beyotime Biotechnology, China). Protein bands were detected using an enhanced chemiluminescence kit (Thermo Fisher, USA).

Cell proliferation ability was detected using cell counting kit-8 (Beyotime Biotechnology, China). 5000 cells were mixed and inoculated into a 96-well plate incubator for 24h. The absorbance value of 450 nm was measured according to the manufacturer’s instructions.

SV-HUC-1 cells were inoculated into the 6-well plate, and subsequent experiments were carried out when the cell confluence reached 100%. Wounds were formed with a 200 μL pipette suction and cell migration was assessed with photographs taken every 12 hours.

The statistical analysis was conducted using R software (version 4.2.1), and data processing followed established literature protocols. The unpaired Student’s t test was used for statistical comparison between the two groups. The Kruskal-Wallis test is used for statistical comparison of non-normally distributed data. A significance level of P-value < 0.05 was considered statistically significant.

We acquired three IC/BPS datasets, GSE11783, GSE28242, and GSE57560, from the GEO database, comprising 14 normal samples and 31 IC/BPS samples. Our data analysis process is shown in Figure 1 . A total of 87 DEGs were identified, comprising 43 up-regulated genes and 44 down-regulated genes ( Figure 2A ; Supplementary Table 1 ). To elucidate the biological functions of these DEGs, we conducted functional enrichment analysis. The results of GO analysis indicated that these DEGs were significantly enriched in processes such as extracellular matrix organization, connective tissue development, collagen fibril organization, collagen-containing extracellular matrix, myofibril, contractile fiber, extracellular matrix structural constituent, peptidase regulator activity, and endopeptidase regulator activity ( Figure 2B ; Supplementary Table 2 ). These biological functions involve the changes of extracellular matrix tissue, connective tissue development, collagen fiber tissue and other factors, which are one of the pathological features of IC/BPS. Furthermore, KEGG analysis unveiled the predominant involvement of DEGs in pathways related to the IL-17 signaling pathway, AGE-RAGE signaling pathway in diabetic complications, and TGF-beta signaling pathway ( Figure 2C and Supplementary Table 3 ). Abnormal activation of these signaling pathways is a key step in the pathogenesis of interstitial cystitis, exacerbating bladder inflammation and injury by promoting local inflammatory responses and infiltration of immune cells. DO analysis revealed associations with various medical conditions, including cell type benign neoplasm, renal cell carcinoma, myopathy, esophageal cancer, cerebrovascular disease, esophageal carcinoma, urinary bladder cancer, aortic aneurysm, aortic disease, endometriosis, cholestasis, and prostate carcinoma ( Figure 2D ; Supplementary Table 4 ). GSEA was performed on both normal and IC/BPS samples to explore signaling pathways ( Supplementary Table 5 ). In the control group, the top three terms included arachidonic acid metabolism, linoleic acid metabolism, and ribosome ( Figure 2E ). In the treatment group, the top five terms comprised cell cycle, chemokine signaling pathway, complement and coagulation cascades, cytokine-cytokine receptor interaction, and oocyte meiosis ( Figure 2F ). The interaction and abnormal activation of these signaling pathways together drive the development of chronic inflammation, fibrosis, and bladder dysfunction of IC/BPS, providing potential targets for treatment.

Subsequently, we delved into the identification of key pathogenic genes in IC/BPS through the WGCNA method. The IC/BPS dataset was stratified into two distinct groups—comprising the control group (normal samples) and the treatment group (IC/BPS samples)—with no discernible outliers ( Figure 3A ). Employing a scale-free R² value of 0.95, we ascertained the soft thresholding capability of WGCNA, ultimately selecting 6 as the soft thresholding power β ( Figure 3B ). Subsequent to employing average link hierarchy clustering and evaluating soft threshold capability, we identified 9 modules, with a particular focus on the turquoise, blue, greenyellow, and tan modules due to their significant associations with IC/BPS ( Figures 3C–E ). The correlation between these selected modules and IC/BPS is visually represented in the scatter plot ( Figures 3F–I ). A comprehensive total of 2612 genes exhibited significant associations with IC/BPS in terms of gene significance and functional modules ( Supplementary Table 6 ).

In order to enhance the precision of data processing, we employed a Venn diagram to identify the overlapping genes between DEGs and key module genes in IC/BPS patients. This analysis revealed a total of 82 genes within the overlapping regions ( Figure 4A ; Supplementary Table 7 ). Subsequently, for the identification of potential diagnostic markers in IC/BPS patients, we applied three machine-learning algorithms to pinpoint crucial genes. Initially, the SVM-RFE algorithm identified 16 key genes in IC/BPS patients ( Figure 4B ). Following this, the RF algorithm, coupled with the distinctive characteristics of IC/BPS patients, selected 12 key genes ( Figures 4C, D ). Lastly, utilizing LASSO regression analysis, we identified 7 key genes from statistically significant univariate variables in IC/BPS patients ( Figures 4E, F ).

The integration of the predictions from these three machine-learning algorithms highlighted three key genes: PLAC8, S100A8, and PPBP ( Figure 5A ). To assess the clinical significance of these key genes, we constructed a nomogram model for diagnosing IC/BPS based on these predictive genes ( Figure 5B ). Decision curve analysis (DCA), calibration curve, and clinical impact curve (CIC) results demonstrated a favorable clinical benefit of our nomogram model ( Figures 5C–E ). Additionally, ROC curve analysis indicated that PLAC8 (AUC: 0.887), S100A8 (AUC: 0.818), and PPBP (AUC: 0.871) may serve as valuable biomarkers for IC/BPS patients ( Figures 5F–H ). Furthermore, the mRNA expression levels of PLAC8, S100A8, and PPBP were significantly upregulated in IC/BPS patients compared to normal patients ( Figures 5I–K ).

We employed the ssGSEA algorithm to assess the landscape of immune cell infiltration in IC/BPS compared to normal samples. Our findings revealed a notable increase in the infiltration levels of 28 types of immune cells in the majority of IC/BPS samples as opposed to normal samples ( Figure 6A ). Specifically, the infiltration levels of 23 types of immune cells in IC/BPS samples were significantly higher than those in normal samples, encompassing Activated B cells, Activated CD4 T cells, Activated CD8 T cells, Activated dendritic cells, Eosinophils, Gamma delta T cells, Immature B cells, Immature dendritic cells, MDSCs, Macrophages, Mast cells, Monocytes, Natural killer T cells, Natural killer cells, Neutrophils, Plasmacytoid dendritic cells, Regulatory T cells, T follicular helper cells, Type 1 T helper cells, Type 2 T helper cells, Effector memory CD4 T cells, Memory B cells, and Central memory CD4 T cells ( Figure 6B ).

Furthermore, our analysis indicated significant positive associations between three key genes and multiple immune cells in IC/BPS patients, particularly PLAC8 ( Figure 6C ). To validate the role of these key genes and immune cells in IC/BPS patients, we gathered 18 tissue samples (10 UC samples and 8 IC/BPS samples) for subsequent analysis, confirmed through cystoscopy and tissue biopsy ( Figure 6D ). There was no significant difference in clinical features between the two groups ( Table 2 ). In IC/BPS tissues, we observed significantly upregulated expression of PLAC8, CXCL10, CD14 (Monocytes), CD15 (Neutrophils), and CD117 (Mast cells) compared to UC tissues ( Figures 7A–D ). The relative proportions of each immune cell marker’s protein expression in the UC and IC/BPS cohorts exhibited distinctions, with a notable increase in mast cells and monocytes in IC/BPS tissues ( Figure 7E ).

PLAC8 protein expression levels were significantly upregulated in inflammatory cell models induced by LPS and ATP compared to controls ( Figure 8A ). Following transfection, PLAC8 was overexpressed in SV-HUC-1 cells, as confirmed by western blot analysis, compared to control cells ( Figure 8B ). We observed that the expression of key proteins in the AKT/mTOR/PI3K signaling pathway—such as AKT, mTOR, and PI3K—was inhibited in PLAC8-overexpressing cells compared to the control group, including in the inflammatory cell models ( Figure 8C ). PLAC8 overexpression significantly enhanced the proliferative capacity of SV-HUC-1 cells compared to untreated cells ( Figure 8D ). However, a wound healing assay revealed that PLAC8 had no significant effect on the migration ability of SV-HUC-1 cells, including in the inflammatory cell models ( Figure 8E ).