Work

Key role of KRT6A in interstitial cystitis: Machine learning-identified feature genes for bladder pain, immunity and fibrosis

Year: 2025

Type: article

Abstract: To the Editor: Interstitial cystitis/bladder pain syndrome (IC/BPS) is a chronic inflammatory condition characterized by persistent pelvic pain, urinary urgency, frequency, and bladder fibrosis, which significantly diminishes the quality of life of patients.[1] Despite its global prevalence of approximately 300 cases per 100,000 females, the condition remains poorly recognized and underdiagnosed in China, with reported incidences ranging from 21.8 to 100.0 per 100,000 individuals. Currently, IC/BPS diagnosis primarily relies on symptom evaluation and exclusion of bacterial infection, leading to frequent misdiagnoses.[2] While therapies such as antihistamines, amitriptyline, and pentosan polysulfate provide partial relief, their efficacy is limited, and adverse effects remain common.[3] Although the precise pathogenesis of IC/BPS remains unclear, accumulating evidence have suggested that immune dysregulation, chronic inflammation, and urothelial dysfunction may play pivotal roles.[4] Understanding the correlation between molecular mechanisms and clinical manifestations such as pain, fibrosis, and immune responses is crucial. The identification of phenotype-related genes associated with immune cell infiltration may uncover potential biomarkers, advancing diagnostic accuracy and targeted therapies. Recently, machine learning (ML) methods have provided powerful tools for identifying disease-associated biomarkers and gene expression signatures.[5] Utilizing ML algorithms to analyze transcriptomic data can reveal underlying molecular patterns, enhancing our understanding of complex diseases such as IC/BPS. In this study, we employed ML-based approaches to identify the clinical phenotype-related genes in IC/BPS, verified their roles in modulating immune responses, and validated our key findings using in vitro experiments. Gene expression data for IC/BPS patients and controls were obtained from publicly available GEO datasets (GSE11783, GSE11839, GSE28242, and GSE57560). ML algorithms, including RF, ENET, LASSO, Ridge Regression, GBM, XGBoost, and DT, were evaluated using leave-one-out cross-validation, and key feature genes were determined using SHapley Additive Explanation (SHAP). Phenotype-related genes associated with inflammation, pain, and fibrosis were sourced from public databases. Immune cell infiltration was analyzed by cell type identification by estimating the relative subsets of RNA transcripts (CIBERSORT) and single-sample gene set enrichment analysis (ssGSEA). Weighted gene co-expression network analysis (WGCNA) identified clinically significant gene modules with KRT6A as a risk factor. In vitro, SV-HUC-1 bladder epithelial cells were cultured under inflammatory conditions using TNF-α (Thermo Fisher Scientific, Waltham, MA, USA). KRT6A knockdown was performed using siRNA. Cell migration and invasion were assessed using wound healing and Transwell assays, respectively. Western blotting and real-time quantitative reverse transcription polymerase chain reaction were used to measure the protein and gene expression levels of KRT6A and fibrosis- and pyroptosis-related markers. Primers sequence of relevant experiments can be seen in Supplementary Table 1, https://links.lww.com/CM9/C530. Proteomic analysis was used to identify differentially expressed proteins. Each vitro assay was repeated independently three times, with three technical replicates per condition. The human samples used in this study have received ethical approval from the National Natural Science Foundation of China (No.81970658), and the informed consent of patients is guaranteed. Immunofluorescence staining of bladder tissues from patients with IC/BPS (n = 3) and adjacent noncancerous bladder tissues from patients with bladder cancer (n = 3) was used to assess the expression of KRT6A and complement-related proteins. Statistical analyses were performed using R software version 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria). Continuous data were expressed as the mean ± standard deviation (SD), with group differences assessed by Student’s t-test or Wilcoxon test. Pearson’s correlation analysis was used to evaluate gene correlations, with statistical significance defined as P <0.05. Dataset GSE11783 consistently achieved high accuracy (>75%) across all models. Despite their small sample sizes, datasets GSE28242 and GSE57560 achieved prediction accuracies of 0.77 and 0.81 with XGBoost, GBM, and RF algorithms, revealing XGBoost as the optimal method with highest mean accuracy [Figure 1A]. After removing batch effects using combat methods and eliminating the unrobust dataset (GSE11839), data from 26 IC/BPS samples and 19 control samples were combined for following analysis. SHAP analysis identified the top 20 feature genes based on their mean importance. Hence, the upregulated gene KRT6A potentially serves as an IC/BPS risk factor [Supplementary Figure 1, https://links.lww.com/CM9/C530].Figure 1: Key role of KRT6A in IC/BPS: insights from integrated ML, bioinformatics, and functional analyses. (A) Classification prediction of each ML algorithm for each dataset. Combination 1 includes all datasets while combination 2 includes three datasets (GSE11839 eliminated). (B) Protein-protein interaction network of clinical-phenotype-related DEG. The depth of nodes is represented by size, while the thickness of the dashed lines indicates the credibility of interactions. (C) KEGG (left) and GO (right) analysis of proteomic differentially expressed proteins of si-KRT6A and control group. (D) Immunofluorescence staining of IC/BPS and bladder mucosa control samples was performed under confocal microscopy at 20× magnification. DEGs: Differentially expressed genes; GO: Gene ontology; IC/BPS: Interstitial cystitis/bladder pain syndrome; KEGG: Kyoto Encyclopedia of Genes and Genomes; ML: Machine learning; NA: Not applicable.Differentially expressed genes (DEGs) were identified using limma package (P <0.05, |log FC| >1.5), yielding 73 DEGs (45 upregulated and 28 downregulated genes). Intersections with the immune-, fibrosis-, and pain-related databases identified 47 phenotype-related DEGs. Protein-protein interaction network analysis highlighted core genes significantly associated with pain, immunity, and fibrosis, of which KRT6A belonged to the pain and fibrosis phenotype cluster [Figure 1B]. Pearson correlation analyses revealed significant associations between feature genes and complement-related factors in IC/BPS. Notably, complement components such as C3, C5AR1, and CFB were markedly upregulated, while key complement inhibitors including CD46, CFD, and CPN1 were significantly downregulated (all P <0.05), suggesting a state of excessive complement activation. Among feature genes, KRT6A was negatively correlated with complement inhibitors CD46 (r = –0.52, P <0.001), CFD (r = –0.57, P <0.001), and CPN1 (r = –0.21, P <0.001), and positively correlated with CFB (r = 0.53, P <0.001), implicating its potential contribution to local inflammatory responses in IC/BPS. CIBERSORT and ssGSEA analyses provided comprehensive immune profiles of IC/BPS. ssGSEA revealed an increased infiltration of immune cells, except natural killer cells, immature dendritic cells, and T helper cell 17 (Th17) cells (all P <0.05). CIBERSORT analyses revealed significant increases in the numbers of neutrophils, CD4+ memory cells, and M0 macrophages, whereas reductions in those of CD8+ T cell, plasma, and resting mast cells (all P <0.05). Correlation analysis indicated that the KRT6A cluster was associated with macrophage differentiation (r = 0.47, P <0.001) and eosinophil infiltration (r = 0.56, P <0.001). WGCNA was used to construct a gene network with 11 distinct modules. The dark green module exhibited the strongest correlation with IC/BPS (r = 0.85, P <0.01) and included key pathways related to inflammation, pyroptosis, and fibrosis. The central hub genes identified within the module, including NFKB1, TLR4, CASP1, TGFB1, and KRT6A (all P <0.01), were strongly associated with chronic inflammation, immune activation, and fibroblast proliferation in IC/BPS [Supplementary Figure 2, https://links.lww.com/CM9/C530]. These robust correlations indicated that KRT6A is intricately involved in the inflammatory signaling cascades and fibrosis-related processes mediated by these key regulators, highlighting its pivotal role in the pathological mechanisms underlying IC/BPS. Functional assays confirmed the pivotal role of KRT6A in urothelial cell behavior. Quantitative PCR analysis showed that KRT6A mRNA expression was reduced by 77.0% (P <0.001) following siRNA-mediated knockdown, and western blotting confirmed a 25.9% reduction in protein expression (P <0.01). Wound healing assays showed reduced closure (si-KRT6A: 18.8% vs. control: 67.3%, P <0.001), and Transwell assays revealed fewer migrated cells (72 ± 11 vs. 148 ± 17, P <0.001). Western blotting demonstrated decreased E-cadherin (–21.8%, P <0.01), β-catenin (–29.9%, P <0.01), vimentin (–59.6%, P <0.01), and GSDMD (–62.1%, P <0.01), indicating weakened adhesion and reduced pyroptosis [Supplementary Figure 3, https://links.lww.com/CM9/C530]. Proteomic analysis using data-independent acquisition identified 230 differentially expressed proteins (DEPs, P <0.05, |log FC| >2) between the si-KRT6A and control groups. Subsequent KEGG and GO pathway analyses revealed that these DEPs were significantly enriched in pathways associated with extracellular matrix remodeling, complement activation, and negative regulation of blood coagulation (all P <0.001) [Figure 1C]. These alterations implied that elevated KRT6A expression under inflammatory conditions may disrupt urothelial barrier integrity, exacerbate chronic inflammation, and promote fibrosis during IC/BPS pathogenesis. Immunofluorescence staining of bladder tissues from IC/BPS patients and controls revealed significantly elevated expression of KRT6A, C3, and C5 [Figure 1D]. Quantification showed a 182.3% increase in KRT6A intensity in the basal urothelium of IC/BPS tissues (P <0.01). C3 and C5 levels in the submucosa were also markedly elevated by 240.4% and 287.5%, respectively (both P <0.001). These quantitative data confirmed the elevated KRT6A expression and complement activation, suggesting that increased KRT6A expression plays a pivotal role in initiating or sustaining complement-driven inflammatory responses. Based on these findings, we propose a pathological mechanism in which sustained epithelial KRT6A upregulation during chronic inflammation promotes apoptosis and disrupts cellular adhesion, weakening the urothelial barrier. Concurrently, KRT6A suppresses coagulation and complement inhibitory factors, thereby sustaining classical complement activation and contributing to the scattered hemorrhagic lesions observed on cystoscopic examination. Additionally, the compromised epithelial barrier may allow the leakage of urinary components into the subepithelial tissues, causing neural sensitization and persistent bladder pain. In summary, this study identified and validated KRT6A as a critical biomarker in IC/BPS using machine learning-based transcriptomic analyses combined with functional validation. Elevated KRT6A expression was closely associated with immune cell infiltration, complement pathway activation, pyroptosis, and impaired urothelial barrier function, thus providing valuable diagnostic and therapeutic insights into IC/BPS pathogenesis. Although further mechanistic investigations are necessary, the role of KRT6A in modulating immune responses and fibrosis establishes it as a promising target for future studies on IC/BPS progression and management. Acknowledgments We thank the authors of the GSE11783, GSE11839, GSE28242 and GSE57560 datasets for their contribution, and we thank the TCellSI tools (https://guolab.wchscu.cn/TCellSI/#!/) for providing the platform for data analysis. Funding This study was supported by grants from the National Nature Science Foundation (No. 81970658) and the Natural Science Foundation of Shanghai (No. 20ZR1434100). Conflicts of interest None.

Source: Chinese Medical Journal

DOI: 10.1097/cm9.0000000000003728