Pan-Cancer Analysis of CLDN3 and Its Contribution to 5-FU Resistance in Colorectal Cancer.

1
Background
Cancer is a major global public health issue with far‐reaching consequences for human health. In 2022, there were 20 million new cancer cases and 10 million cancer‐related deaths [1]. Alarmingly, these figures are expected to increase by 77% over the next 30 years compared to 2020 levels [1]. Despite significant advances in cancer diagnosis and treatment, the prognosis for many cancer types remains poor, primarily because of recurrence and metastasis. A factor contributing to this poor prognosis is epithelial‐mesenchymal transition (EMT), a process whereby tumour cells gain increased invasiveness, allowing them to infiltrate deeper tissues via the mucosa [2]. This invasive process is characterised by the disruption of various intercellular junctions, including tight junctions (TJs), adherens junctions, desmosomes and gap junctions between epithelial cells [3]. Among these, dysfunction of tight junctions plays a particularly critical role in cancer progression, although the precise molecular mechanisms underlying this process remain poorly understood [4].
Claudins are a family of tight junction proteins whose abnormal alterations contribute to the loss of cell polarity, cytoskeletal remodelling and changes in cell morphology [5]. CLDN3, a member of this family, is located on chromosome 7q11.23 and encodes a transmembrane protein consisting of 220 amino acids [6]. Research has shown that the CLDN3 protein comprises two extracellular loops (ECL1 and ECL2), an intracellular loop formed by four transmembrane domains and N‐ and C‐terminal sequences [5]. Additionally, studies have demonstrated that CLDN3 is closely associated with the development of several malignant tumours, including prostate, liver, lung, pancreas cancers and glioblastoma multiforme [7, 8, 9, 10, 11]. CLDN3 can activate multiple tumourigenic signalling pathways that influence malignant phenotypes, such as cell proliferation, migration, apoptosis and resistance to chemotherapy [12]. Although extensive evidence from cell and animal studies supports the significant role of CLDN3 in malignant tumours, comprehensive pan‐cancer analysis of the CLDN3 gene using clinical data is still lacking. Whereas correlational studies cannot imply causation, they provide a landscape of associations that can guide specific mechanistic validations, such as those we perform here for CRC.
TRIM28, also referred to as Kruppel‐associated box protein 1 (KAP1) or transcription intermediary factor 1β (TIF1β), is an E3 ubiquitin ligase [13]. Research has shown that TRIM28 plays a critical role in regulating protein stability and significantly contributes to tumourigenesis and cancer progression [14]. Specifically, TRIM28 mediates the polyubiquitination and subsequent degradation of various substrates in collaboration with melanoma‐associated antigen‐coding genes (MAGE) [15]. These substrates include important proteins such as AMP‐activated protein kinase (AMPK) and RING finger LIM domain interacting protein (RLIM) [16]. Moreover, TRIM28 prevents transcription‐induced DNA breaks and is crucial for the SUMOylation of the DNA replication factor, proliferating cell nuclear antigen (PCNA) [17]. Additionally, TRIM28 catalyses the SUMOylation of programed cell death ligand 1 (PD‐L1) target proteins [18]. Through these mechanisms, TRIM28 functions as an oncogene, promoting the malignant progression of cancer through interactions with other oncogenes [19]. However, despite its well‐documented involvement in various cancers, the precise role of TRIM28 in colorectal cancer (CRC) and the underlying mechanisms remain unclear.
In this study, we performed a comprehensive pan‐cancer analysis of CLDN3 using publicly available data to investigate its role in cancer. The analysis covered various aspects, including expression profiles, prognosis, survival outcomes, DNA methylation, genetic and epigenetic characteristics, immune infiltration, single‐cell analysis, drug sensitivity and enrichment analysis. To further validate the role of CLDN3 in CRC, we investigated how alterations in its expression affect the resistance of CRC cells to 5‐FU. Moreover, we discovered that TRIM28 directly interacts with CLDN3, promoting its SUMOylation, which delays CLDN3 degradation and facilitates tumour growth and metastasis. Targeting the TRIM28/CLDN3 axis may represent a promising therapeutic strategy for the diagnosis and treatment of CRC.

2
Materials and Methods
2.1
Expression Analysis of CLDN3
CLDN3 mRNA expression in normal tissues was obtained from the human protein atlas (HPA) (http://www.proteinatlas.org) [20], based on RNA‐seq data. CLDN3 protein expression was detected in 44 human tissues, with immunohistochemical data from both normal and cancerous tissues sourced from the ‘Human Pathology’ module of HPA. Additionally, immunofluorescence localisation images of CLDN3 were retrieved from the ‘SUBCELL’ module. To further investigate the differential expression of CLDN3, data from the ‘Gene_DE’ module of TIMER2 (http://timer.cistrome.org/) [21] were used to compare expression levels between various cancerous tissues and adjacent non‐cancerous tissues. For tumour entries without corresponding normal tissue, we used the ‘Expression DIY’ module of the GEPIA2 server (http://gepia2.cancer‐pku.cn/) [22] to plot boxplots comparing CLDN3 expression in tumours and adjacent normal tissues. The analysis was conducted with a |Log2FC| cutoff = 1, a p‐value cutoff = 0.05 and matching TCGA normal and GTEx data. Finally, CLDN3 protein expression was analysed using the CPTAC (clinical proteomic tumour analysis consortium) dataset available on the UALCAN website (http://ualcan.path.uab.edu/analysis‐prot.html) [23]. Multi‐omics data from TCGA and CPTAC were preprocessed using standard pipelines. Transcriptomic data were log2‐transformed and normalised using the TMM method to ensure cross‐sample comparability, following established protocols for pan‐cancer reproducibility [24, 25, 26].

2.2
Diagnosis and Prognostic Analysis
The ‘clinical module’ of TISIDB (http://cis.hku.hk/TISIDB/) was employed to investigate the association between CLDN3 expression and overall survival (OS) across various human cancer types [27]. Subsequently, ‘Survival analysis’ module in GEPIA2 was utilised to analyse the OS of CLDN3 in all TCGA tumours, using median expression levels (50% high and 50% low) as thresholds.
To enhance the precision of the analysis, the Kaplan–Meier (KM) plotter (http://kmplot.com/analysis/index.php?p=background) was used with an automated optimal cut‐off value to correlate CLDN3 expression with OS across different tumour types [28]. Furthermore, TCGA database data were analysed using the pROC package in R to assess the diagnostic potential of CLDN3 in cancer. The AUC value ranges from 0.5 to 1 [29].

2.3
Genetic and Epigenetic Alteration Analysis
The cBioPortal of TCGA PanCancer atlas studies (http://www.cbioportal.org) was employed to examine the genetic and epigenetic alterations of CLDN3 across various cancer types [30]. The alteration frequency of CLDN3 across different tumours was summarised in the ‘Cancer types summary’ module. Using the ‘Plots’ module, general mutation counts of CLDN3 in various TCGA cancer types were then described. Additionally, the gene set cancer analysis (GSCA) database (http://bioinfo.life.hust.edu.cn/GSCA/#/) [31] provided detailed copy number variation (CNV) data for CLDN3 and its associated mRNA expression levels. To further investigate the relationship between CLDN3 expression and cancer features, we analysed data from the TCGA database. The correlation between tumour mutational burden (TMB), microsatellite instability (MSI) and CLDN3 expression was evaluated using the ‘maftools’ R package.

2.4
Methylation Analysis of CLDN3
The correlation between CLDN3 methylation and mRNA expression in specific cancers was summarised using the GSCA website. Survival differences (DFI, DSS, OS and PFS) between cancer patients with high and low CLDN3 methylation levels were further analysed using the ‘Mutation’ module of GSCA.

2.5
CLDN3 and Tumour‐Immune System
The TISIDB dataset was used to explore the correlation between tumour‐infiltrating lymphocyte (TIL) abundance and CLDN3 expression in the ‘Lymphocyte’ module. In addition, the ‘Immunomodulator’ module was employed to examine the relationship between three types of immunomodulators (inhibitors, stimulators and MHC molecules) and CLDN3 expression. The ‘Chemokine’ module was subsequently utilised to assess the correlation between chemokines (or their receptors) and CLDN3 levels. Furthermore, TISIDB was applied to evaluate whether significant differences in CLDN3 expression or mutations exist between immune therapy responders and non‐responders (e.g., PD‐1, PD‐L1 and CTLA‐4).

2.6
CLDN3 and Immune Infiltration in Cancer
The correlation between CLDN3 and immune cell infiltration was evaluated using the ‘gene’ module from the TIMER 2.0 database.

2.7
Enrichment Analysis of CLDN3‐Related Partners
Fifty proteins interacting with CLDN3 were identified from the STRING database (https://string‐db.org) [32]. In addition, 100 genes associated with CLDN3 in both TCGA tumour and normal tissues, were obtained using the ‘Similar Gene Detection’ module of GEPIA2. The overlapping genes between STRING and GEPIA2 were then selected using the VENNY tool (Venny 2.1.0, http://liuxiaoyuyuan.cn/). KEGG pathways linked to these genes were retrieved via SANGERBOX, and functional analysis of CLDN3‐related genes was performed using Metascape (http://metascape.org) [33]. Lastly, the relationship between activated signalling pathways and CLDN3‐related genes was investigated using GSCALite.

2.8
Single‐Cell Analysis of CLDN3 Expression
The correlation between CLDN3 expression and tumour biology was examined using the CancerSEA [34]. Additionally, the TISHI2 (http://tisch.comp‐genomics.org/home/) [35] was employed to analyse the expression of CLDN3 from different colon cancer cell datasets. The box plots depicting the promoter methylation levels of CLDN3 across different clinical features were obtained through UALCAN. For single‐cell data, strict quality control filters were applied to remove low‐quality cells (mitochondrial genes > 5%) following established protocols.

2.9
Drug Sensitivity
Using the CellMiner online database (https://discover.nci.nih.gov/cellminer/home.do) [36], we analysed the correlation between drug response and gene expression levels. Based on these results, we investigated the association between CLDN3 expression and drug IC50 (half‐maximal inhibitory concentration) using the CTRP [37] and GDSC [38] databases.

2.10
Experiments in CRC Cells
Detailed information and procedures for the molecular biological experiments are described in Supporting Information S1.

2.11
Statistical Analysis
All statistical analyses were performed using SPSS software, version 20.0 (IBM, Chicago, IL, USA). Data are reported as the mean ± standard error of the mean (SEM). For comparisons between two groups, Student's t‐test was employed when the assumption of homogeneity of variances was satisfied; otherwise, the non‐parametric Mann–Whitney U test was used. For comparisons involving more than two groups, one‐way analysis of variance (ANOVA) was conducted if variances were homogeneous, whereas the Kruskal–Wallis test was applied when this assumption was violated. A p‐value < 0.05 was considered statistically significant.

3
Results
3.1
Analysis of CLDN3 Expression in Normal and Cancer Tissues
To investigate the expression levels of CLDN3 in human normal and cancer tissues, we analysed data from the HPA database. The results indicated that CLDN3 expression is tissue‐specific (Figure S1A–D). Notably, CLDN3 exhibited significantly higher mRNA expression in the small intestine, colon and duodenum (Figure 1A). The CLDN3 protein is widely expressed at moderate to high levels in over half of normal tissues (Figure 1B). Further analysis using the TIMER2.0 database showed that CLDN3 is upregulated in most cancer types, including BRCA, CHOL, ESCA, KIRP, LUAD, PRAD, STAD and UCEC, whereas it is downregulated in KICH and LUCS (Figure 1C).
CPTAC samples indicated abnormal expression of CLDN3 in various cancers, with protein‐upregulation observed in BRCA, COAD, UCEC, LUAD and OV (Figure 1D,G). The GEPIA2 database analysis corroborated these findings, showing upregulation of CLDN3 in COAD, THYM, UCS, OV and READ and downregulation in SKCM and SARC (Figure 1E). The GSCA database further confirmed that CLDN3 mRNA expression in THCA, BRCA, PRAD, STAD, KIRP and LUAD tumour tissues (Figure S3A). Immunohistochemistry results from the HPA database demonstrated significantly higher CLDN3 expression in LIHC, COAD and READ compared to healthy tissues (Figure 1F). Furthermore, immunofluorescence data suggest that CLDN3 is primarily localised to the cell membrane (Figure 1H). CLDN3 expression is also closely associated with tumour pathological staging (Figure S2B–D) and grade (Figure S3D). Subtype analysis further reveals a correlation between CLDN3 expression and the immune and molecular subtypes of cancers (Figure S3B,C). In summary, CLDN3 is upregulated in most cancers, suggesting its potential role in cancer diagnosis.

3.2
Diagnosis and Survival Analysis of CLDN3 in Cancers
CLDN3 is aberrantly expressed across various cancer tissues and cell types. To assess its potential as a prognostic biomarker, we performed bioinformatics analyses to explore the relationship between CLDN3 expression and cancer outcomes. Data from the TISIDB, GEPIA2 and GSCA database respectively indicated that CLDN3 expression correlates with overall survival (OS) in ESCA, LUSC, MESO, READ and SARC (Figure 2A–C). We also assessed the diagnostic utility of CLDN3 across multiple cancers using ROC curves. The area under the curve (AUC) values for various cancers were as follows: CESC (AUC = 0.64), DLBC (AUC = 0.679), ESCA (AUC = 0.703), MESO (AUC = 0.65), PAAD (AUC = 0.628), SARC (AUC = 0.665), THCA (AUC = 0.933) and UCS (AUC = 0.892) (Figure 2D). These findings highlight the potential of CLDN3 as a diagnostic biomarker for these cancers.

3.3
Epigenetic Features of CLDN3 in Cancers
Gene expression is intricately regulated by epigenetic mechanisms. To further explore this relationship, we analysed the genetic alterations and epigenetic characteristics of CLDN3. Analysis using cBioPortal revealed that the highest mutation frequency of CLDN3 occurs in gastric cancer patients, with gene amplification being the predominant alteration type (Figure 3A). The overall mutation count of CLDN3 across various cancer types is presented in Figure 3B, which identifies 14 missense mutation sites (Figure 3C).
Somatic copy number variations (CNVs), frequently implicated in tumourigenesis and progression, are prevalent across numerous human malignancies. Our analysis demonstrated that CLDN3 CNVs are present in all 33 cancer types within the TCGA dataset (Figure 3D). To investigate the potential genomic factors associated with CLDN3 expression, we assessed the correlation between CNV and CLDN3 expression levels. A significant positive correlation was observed in PRAD, ACC, STAD etc (Figure 3E).
Tumour mutational burden (TMB) and microsatellite instability (MSI) are critical factors influencing patient prognosis and therapeutic response. To examine the association between CLDN3 expression and TMB/MSI, we conducted a pan‐cancer analysis of TCGA data using the R programming language. Our findings revealed a positive correlation between CLDN3 expression and MSI in THCA, TGCT, STAD, SARC, LUAD, LIHC and ESCA, whereas a negative correlation was observed in UCS, UCEC, READ and COAD (Figure 3F). Additionally, CLDN3 expression showed a positive correlation with TMB in THYM, TGCT, STAD, PRAD, LUAD, KIRP, BRCA, BLCA, HNSC and ESCA, and a negative correlation in UCEC, READ, KICH and COAD (Figure 3F).

3.4
Methylation Analysis of CLDN3 Across Various Cancer Types
DNA methylation is a crucial epigenetic mechanism that plays a key role in regulating cancer progression. To investigate the methylation status of CLDN3 in various cancers, we utilised the GSCA database. Methylation differences were observed between cancer types and normal samples (Figure 4A). As shown in Figure 4B, DNA methylation levels were negatively correlated with mRNA expression across tumour types, particularly in ESCA, STAD, CESC, KIRP and LUAD.
LUSC patients with CLDN3 methylation demonstrated significantly better overall survival (OS), disease‐specific survival (DSS) and progression‐free survival (PFS). Similarly, PRAD patients with CLDN3 methylation showed improved progression‐free survival PFS and disease‐free interval (DFI) (Figure 4C).

3.5
Interactions Between Tumour‐Immune System and CLDN3
The interaction between tumours and the immune system plays a critical role in cancer initiation, progression and treatment. Therefore, understanding tumour‐immune cell interactions is essential for predicting responses to immunotherapy and identifying new therapeutic targets. Using the TISIDB dataset, we identified significant correlations between CLDN3 and various immune‐related factors, including tumour‐infiltrating lymphocytes (TILs), immune inhibitors, immune stimulators, MHC molecules, chemokines and receptors (Figure 5A). Our analysis further revealed a significant mutation difference in CLDN3 between responders and non‐responders to anti‐PDL‐1 treatment in melanoma (Figure 5B). However, no such differences were found between responders and non‐responders (Figure 5B).

3.6
Interactions Between CLDN3 and Immune Infiltration
Tumour tissue comprises not only tumour cells but also stromal cells, tumour‐associated fibroblasts, immune cells and other components. Using the TIMER 2.0 database, we explored the potential role of CLDN3 in immune cell infiltration. As illustrated in Figure 6A,B, CLDN3 expression shows a positive correlation with the bioinformatically inferred infiltration abundance of regulatory T cells (Tregs) in most tumour types. The top six tumours exhibiting the strongest correlations are ESCA (Rho = 0.545, p = 2.52e‐15), UCS (Rho = 0.323, p = 1.82e‐02), UVM (Rho = 0.312, p = 5.67e‐03), LUAD (Rho = 0.29, p = 5.29e‐11), HNSC (Rho = 0.28, p = 7.90e‐03) and UCEC (Rho = 0.258, p = 1.54e‐02). Furthermore, the expression of CLDN3 is negatively correlated with the abundance of MAST cell infiltration. The top six tumours are TGCT (Rho = −0.752, p = 4.6e‐28), READ (Rho = −0.372, p = 3.07e‐04), BRCA‐Her2 (Rho = −0.285, p = 1.53e‐02), BRCA‐LumA (Rho = −0.264, p = 1.08e‐09), HNSC‐HPV+ (Rho = −0.247, p = 1.95e‐02) and BRCA‐LumB (Rho = −0.235, p = 1.06e‐03) (Figure 6C,D).

3.7
Identification of CLDN3‐Related Genes and Their Associated Biological Functions
Fifty proteins interacting with CLDN3 were identified using the STRING database and established a protein–protein interaction (PPI) network (Figure 7A), KEGG and Metascape analysis suggests that they may be associated with tight junctions, cell adhesion molecules (CAMs), pathogenic Escherichia coli infection, leucocyte transendothelial migration, hepatitis C and so on (Figure 7B,C). GEPIA2 tool was utilised to identify the top 100 genes positively correlated with CLDN3 expression. We examined the co‐expression of CLDN3 with the top 10 genes identified via GEPIA2 and top 10 genes from the protein–protein interaction (PPI) analysis across various cancers using the TIMER2 database. The analysis showed that, in most cancers, CLDN3 shows a significant positive correlation with these 20 genes (Figure 7D,E). The intersection of the 50 genes from the PPI analysis and the 100 genes from GEPIA2 revealed 11 common genes: CDX1, CDX2, CLDN4, CLDN7, EPCAM, KRT18, KRT8, MARVELD2, MARVELD3, OCLN and TJP3 (Figure 7F). Finally, the GSCALITE tool was employed to analyse the co‐functionality of these 11 genes with CLDN3 across various cancers. The results indicated that these genes are correlated with the cancer‐related pathways of apoptosis, cell cycle, DNA damage, EMT, hormone AR and hormone ER (Figure 7G,H).

3.8
Validation of CLDN3 Expression and Function in CRC Cells
The above results indicate that CLDN3 is significantly overexpressed in multiple malignancies, including colorectal, ovarian, gastric and lung cancers. Especially in colorectal cancer, CLDN3 shows high expression, significant CNV alterations and reduced DNA methylation, indicating a strong potential role in driving its malignant biological characteristics. Single‐cell transcriptomics offers valuable insights into the functional roles of candidate molecules at the resolution of individual cells. In CRC, CLDN3 expression is positively associated with angiogenesis, apoptosis, differentiation, inflammation and hypoxia, whereas it is negatively correlated with cell cycle and DNA repair (Figure 8A). The TISCH2 database analysis revealed a significant increase in CLDN3 expression in both regulatory epithelial and malignant cells across these datasets (Figure 8C). For a more detailed single‐cell analysis, we selected the CRC_GSE146771‐Smartseq2 dataset and utilised the TISCH2 tool to compare CLDN3 expression across various cells. The results indicated that CLDN3 was predominantly expressed in malignant cells and plasma (Figure 8B,D). Furthermore, promoter methylation analysis via UALCAN revealed that the promoter methylation level of CLDN3 in COAD tumour tissues was significantly lower than in adjacent normal tissues (Figure 4D). Analysis of individual cancer stages and tumour grades indicated that lower promoter methylation levels were associated with higher tumour stage and grade (Figure 8E).
To further investigate its role, we conducted a series of experiments to determine whether CLDN3 knockdown could suppress malignant cellular behaviours of CRC cells. Initially, western blot and RT‐qPCR analyses results showed that CLDN3‐shRNA3 exhibited the highest knockdown efficiency. Stably CLDN3‐knockdown and ‐overexpression CRC cell lines (HCT116 and SW480) were successfully established (Figure 8F,G). Subsequent assays, including CCK‐8, EdU and colony formation, confirmed that CLDN3 knockdown significantly inhibited cell proliferation (Figure 8H–J). Additionally, the wound healing assay revealed that CLDN3 knockdown impaired cell migration (Figure 8K). Finally, flow cytometry analysis indicated that CLDN3 knockdown significantly increased the proportion of cells in the G1 phase while decreasing the number in the S/G2 phases, thereby inhibiting cell proliferation (Figure 8L).

3.9
CLDN3 Enhances the Chemoresistance to 5‐FU in CRC Cells
To investigate the therapeutic potential of CLDN3, pharmacological efficacy and mRNA expression data were retrieved from CellMiner to evaluate the correlation between drug sensitivity and CLDN3 expression. The analysis revealed a positive association between CLDN3 expression and IC50 of several compounds, including 8‐chloro‐adenosine, Acetalax, AMG‐900, Bisacodyl, Virap's active ingredient, CUDC‐305, Elesclomol, Fluorouracil, Fulvestrant, Linsitinib and SR16157 (Figure 9A). These findings may offer valuable personalised treatment strategies for cancer patients with elevated CLDN3 expression. Moreover, similar results were obtained from analyses using CTRP and GDSC data on the GSCA platform, which also demonstrated a correlation between increased CLDN3 expression and resistance to various small‐molecule compounds (Figure 9B,C). Notably, the expression of CLDN3 is correlated with the IC50 of 5‐fluorouracil (5‐FU), which is an important chemotherapy drug for colorectal cancer. The IC50 values of 5‐FU for HCT116 and SW480 cells were first determined using the CCK‐8 assay (Figure 9D). Flow cytometry analysis revealed that CLDN3 overexpression partially inhibited apoptosis of CRC cells (Figure 9E). Specifically, the apoptosis induced by 5‐FU was significantly reduced in CLDN3‐overexpressing cells compared to the control group, suggesting resistance to 5‐FU (Figure 9E). We used a protein array to screen the signalling pathways activated by CLDN3 in CRC. The results suggest that the p38 signalling pathway is significantly correlated with CLDN3 expression (Figure 9F), and pre‐treatment with the p38/MAPK inhibitor SB 203580 in CLDN3‐overexpressing HCT116 and SW480 cells significantly enhanced their chemoresensitivity to 5‐FU (Figure 9G). Collectively, these findings suggest that aberrant CLDN3 expression may contribute to the resistance against 5‐FU chemotherapy in CRC.

3.10
Trim28 Is an Interacting Protein of CLDN3
CLDN3 is associated with the proliferation, metastasis and chemoresistance of CRC cells. Previous research has shown that protein stability is regulated by diverse post‐translational modifications, including ubiquitination, sumoylation and acetylation. A comprehensive understanding of the regulatory mechanisms underlying CLDN3 protein stability necessitates the identification of its interacting proteins. To address this, we utilised a combination of protein silver staining assays and mass spectrometry to investigate CLDN3‐associated proteins in colorectal cancer cells. This approach led to the identification of 975 CLDN3‐binding proteins, 56 of which are implicated in the regulation of cellular ubiquitination. Notably, several ubiquitin‐like molecules and their ligases, such as Trim28, SUMO1 and UBE2M, were identified as key interactors (Figure 10A). KEGG pathway enrichment analysis indicated that CLDN3‐binding proteins are predominantly associated with key signalling, including EGFR, ribosome and ErbB signalling (Figure 10B). Among these interacting proteins, TRIM28 emerged as a particularly significant candidate. Data from the GEPIA2 database further revealed a strong positive correlation between CLDN3 and TRIM28 expression across multiple malignancies, especially in cancer types such as KICH, COAD, UCEC, CHOL and PCPG (Figure 10C). Additionally, TIMER2.0 database showed positive correlations between CLDN3 and TRIM28 expression across most tumour types including READ, BRCA and head and neck cancer (HNSC) (Figure 10D). Collectively, these findings suggest that CLDN3 and TRIM28 may synergistically contribute to cancer development and progression.

3.11
Trim28 Catalysed the SUMOylation of CLDN3 to Regulate Its Stability
We performed immunoprecipitation to confirm the potential interaction between CLDN3 and TRIM28. The results showed that both CLDN3 and TRIM28 antibodies successfully co‐precipitated endogenous TRIM28 and CLDN3, providing strong evidence for their interaction (Figure 11A). Further experiments involved co‐transfecting GFP‐TRIM28 and HA‐CLDN3 overexpression plasmids into HEK293T cells, followed by immunoprecipitation with anti‐HA and anti‐GFP antibodies. The results confirmed that GFP‐TRIM28 and HA‐CLDN3 proteins co‐precipitated, further supporting the existence of a protein‐protein interaction between CLDN3 and TRIM28 (Figure 11B).
To examine the intracellular localisation of CLDN3 and TRIM28, we employed immunofluorescence techniques and captured their fluorescence signals with a laser confocal microscope. The results confirmed that the two proteins co‐localise in CRC cells (Figure 11C). Research has demonstrated that TRIM28 functions as an E3 ligase responsible for the SUMOylation of proteins. Based on this, we hypothesise that TRIM28 may regulate CLDN3 stability by promoting its SUMOylation. To investigate this function, we first constructed a Flag‐tagged SUMO1 plasmid and co‐transfected it with a GFP‐TRIM28 plasmid into HEK293T cells. Immunoprecipitation was then performed to detect the SUMOylation of endogenous CLDN3. The results revealed that, compared to the control group, multiple distinct protein bands appeared alongside the main CLDN3 band. These additional bands were detected by both CLDN3 and Flag antibodies, indicating that TRIM28 expression promotes the SUMOylation of endogenous CLDN3 (Figure 11D). In parallel, we co‐transfected HA‐CLDN3, GFP‐TRIM28 and Flag‐SUMO1 plasmids into HEK293T cells and performed immunoprecipitation to detect the SUMOylation of exogenous CLDN3. The results revealed multiple protein bands above the HA‐CLDN3 band, which were detected by both HA and Flag antibodies. These findings suggest that exogenous CLDN3 undergoes SUMOylation catalysed by TRIM28 (Figure 11E).
Previous studies have demonstrated that SUMOylation enhances protein stability. Based on above findings, we hypothesised that TRIM28 might stabilise CLDN3 through SUMOylation. To test this hypothesis, we downregulated TRIM28 expression in HCT116 cells using RNA interference (Figure 11F), while simultaneously inhibiting new CLDN3 synthesis with cycloheximide (CHX). Western blot analysis showed significantly increased degradation of CLDN3 in the TRIM28 knockdown group compared to the control, suggesting that TRIM28 downregulation reduces CLDN3 stability (Figure 11G). Collectively, these results indicate that TRIM28‐catalysed SUMOylation of CLDN3, plays a key role in regulating its stability.

3.12
SUMOylation of CLDN3 at Lys156 Regulates Its Protein Stability and Promotes CRC Progression
To identify potential SUMOylation sites on CLDN3, we initially performed bioinformatic analysis, which identified Lys156 (K156) and Lys190 (K190) as high‐probability candidates (Figure 12A). To validate these predictions, we constructed expression vectors for wild‐type CLDN3 (CLDN3 WT) and its site‐specific mutants, CLDN3K156A and CLDN3K190A. Western blot analysis revealed that the K156A mutation, but not the K190A mutation, significantly reduced the SUMOylation of CLDN3 (Figure 12B), identifying K156 as the primary SUMOylation site. We further examined the relationship between SUMOylation and protein stability of CLDN3. Immunoblotting demonstrated that CLDN3K156A mutant markedly reduced the protein stability of CLDN3 (Figure 12C), suggesting that SUMOylation at K156 is essential for maintaining CLDN3 levels. Finally, colony formation and transwell assays showed that while ectopic expression of CLDN3 WT effectively restored cell proliferation, migration and invasion, these rescue effects were largely abrogated by the K156A mutation (Figure 12D,E). Collectively, these findings indicate that SUMOylation at K156 is a critical post‐translational modification required for CLDN3 stability and its oncogenic functions in CRC.

4
Discussion
Claudins are transmembrane proteins that are integral to the cytoskeleton and play a crucial role in the formation of tight junctions (TJs). These proteins exhibit tissue‐ and cell‐specific expression, and their dysregulation can impair the structure and function of TJs [39]. Such disruptions in cell polarity and intercellular barriers are considered critical factors in the initiation and progression of cancer. Aberrant expression of Claudin family members has been reported in various cancers, including liver, breast, osteosarcoma, colorectal, cervical and prostate cancers [40, 41, 42, 43, 44, 45]. Altered Claudin expression influences cancer cell processes, such as proliferation, metastasis, invasion and apoptosis [46, 47].
CLDN3 functions as a receptor for Clostridium perfringens enterotoxin (CEP) and represents a potential target for cancer therapies [48]. Human monoclonal antibodies targeting CLDN3 may offer both diagnostic and therapeutic benefits for CLDN3‐expressing tumours [49]. As a barrier‐forming protein, CLDN3 has been recognised as both a biomarker and a therapeutic target in various cancers. For instance, Rangel demonstrated that CLDN3 expression is upregulated in ovarian cancer [50]. Moreover, increased CLDN3 expression is associated with higher tumour grades in clear cell renal carcinoma and poorer overall survival in patients [51]. Consistent with these findings, our study confirms that CLDN3 is predominantly expressed in several cancers. In specific cancer types, its expression correlates with tumour grades, subtypes and stages. Notably, elevated CLDN3 expression is strongly linked to poor prognosis in oesophageal carcinoma (ESCA), lung squamous cell carcinoma (LUSC), mesothelioma (MESO), rectal adenocarcinoma (READ) and sarcoma (SARC), although further research is required to fully assess its diagnostic and therapeutic potential.
Numerous studies have demonstrated a correlation between genetic mutations and various tumour characteristics, including tumourigenesis, progression, metastasis, invasion and drug response [52, 53]. In this study, we investigated the genetic and epigenetic characteristics of CLDN3 across different cancer types. The most prevalent mutations identified in CLDN3 are missense mutations. Although the functional impact of these specific mutations requires experimental verification, their presence suggests potential genomic instability in CLDN3. Notably, the highest frequency of CLDN3 mutations was observed in stomach adenocarcinoma (STAD). These mutations may confer a growth advantage and enhance metastatic potential. Copy number variations (CNVs) play central roles in oncogenesis and cancer therapy [54]. In particular, our analysis demonstrated that CLDN3 CNVs are present in all 33 cancer types, which may contribute to changes in gene transcription and mRNA stability. Tumour mutational burden (TMB) and microsatellite instability (MSI) have been established as reliable biomarkers for immunotherapy across various cancers [55]. We also identified significant correlations between CLDN3 expression and both MSI and TMB across multiple cancer types. Moreover, DNA methylation, a key epigenetic mechanism, is involved in biological processes such as cancer initiation and progression [56]. We found that CLDN3 methylation is negatively correlated with expression in the majority of cancers, suggesting that DNA methylation may partially regulate CLDN3 dysregulation. Our analysis suggests that genomic alterations and promoter hypomethylation are potential contributors to CLDN3 dysregulation. However, the lack of correlation in certain cancer types suggests a multi‐layered regulatory network may also be involved in regulating protein stability. In conclusion, genetic variations in CLDN3 represent a potential therapeutic target in cancer treatment. Future research should further explore the role of CLDN3 genetic and epigenetic features in tumour progression and prognosis, which could help inform the development of personalised treatment strategies for cancer patients.
The tumour microenvironment (TME), which consists of immune cells, tumour cells and stromal cells, plays a crucial role in cancer development, metastasis and resistance to therapy [57]. Immune suppression within the TME is generally regarded as detrimental, indicating that modulating the TME could improve cancer prognosis. In gastric cancer, dysregulated CLDN3 expression alters MHC‐I expression, thus affecting immunogenicity [58]. Our findings demonstrate that CLDN3 expression correlates with tumour‐infiltrating lymphocytes (TILs), immune suppressors, immune activators, MHC molecules, chemokines and their respective receptors. Mast cells, typically associated with allergies and inflammation, also play a pivotal role in tumour biology and the TME. These cells release chemokines and cytokines that recruit immune cells into the TME, alter immune cell function and promote cancer progression [59]. Notably, we found that CLDN3 expression is positively correlated with T‐cell regulatory infiltration and negatively correlated with mast cell infiltration across various cancer types, suggesting the potential role of CLDN3 in regulating the TME. Yet these estimates are derived solely from computational deconvolution tools that are sensitive to tumour purity, stromal variation and algorithmic limitations, orthogonal immunological validation is needed to confirm these in silico findings.
Dysregulated expression of oncogenes is closely linked to the efficacy of cancer therapeutics. For example, GLIPR1 regulates cisplatin resistance in human lung cancer cells [60], whereas ERCC1 inhibition increases cisplatin resistance in gastric cancer cells [61]. In ovarian cancer, overexpression of CLDN3 suppresses CTR1 expression, thereby conferring cisplatin resistance [62]. Our study explores the correlation between CLDN3 expression and drug sensitivity, using data from the GDSC and CTRP. The findings provide valuable insights for the clinical implementation of personalised therapies. In lung squamous cell carcinoma, CLDN3 activates the WNT pathway, which modulates epithelial‐mesenchymal transition (EMT) and migration/invasion in squamous cell carcinoma (SqCC) cells [10]. In contrast, silencing CLDN3 inhibits proliferation, metastasis and invasion in glioblastoma multiforme [7]. To further investigate the mechanisms by which CLDN3 influences cancer progression, we integrated co‐expression networks and analysed CLDN3‐associated gene sets. Using STRING, we identified CLDN3‐associated and interacting proteins from GEPIA2, along with CLDN3‐related proteins. Combining results from protein‐protein interaction (PPI) networks and GSCA, we performed gene set enrichment analysis (GSEA), revealing that CLDN3 affects several tumour signalling pathways, including those involved in apoptosis, cell cycle, hormone receptor (AR), RAS/MAPK and receptor tyrosine kinase (RTK) signalling. A study also demonstrated that downstream signals regulating the RTK pathway influence CLDN3 protein levels [63].
Although above pan‐cancer findings of CLDN3 suggest a broad functional involvement, we specifically experimentally validated its mechanistic role in promoting 5‐FU resistance in colorectal cancer (CRC). We observed that CLDN3 knockdown impedes CRC cell proliferation and migration, whereas inducing cell cycle arrest at the G2/M phase. Furthermore, the suppression of CLDN3 significantly increased apoptosis in these cells. Previous studies have identified the p38/MAPK signalling pathway as a key regulator of chemotherapy resistance in tumour cells. Inhibition of this pathway has been shown to significantly enhance the sensitivity of CRC cells to 5‐FU [64]. Similarly, recent research has demonstrated that p38/MAPK inhibitors enhance the sensitivity of liver cancer cells to 5‐FU [65]. Based on these results, we pre‐treated CLDN3‐overexpressing CRC cells with p38/MAPK pathway inhibitors, results showed that inactivated‐p38 signalling enhances the sensitivity of these cells to 5‐FU. Therefore, targeting CLDN3 may provide novel therapeutic strategies for tumour‐chemoresistance, such as reducing the stability of CLDN3 to promote its degradation. In this study, we employed co‐immunoprecipitation (Co‐IP) to identify CLDN3‐interacting proteins and discovered that TRIM28 is one key interactors for CLDN3. TRIM28, a member of the tripartite motif‐containing protein (TRIM) family, plays a crucial role in tumourigenesis by regulating transcription, epigenetic modifications and post‐translational modifications in eukaryotic cells. Recent studies have shown that TRIM28 mediates protein SUMOylation and functions as a key E3 ligase in this process [66]. Our study confirms that TRIM28 mediates the SUMOylation of CLDN3, and disrupting TRIM28 expression accelerates CLDN3 degradation. Targeting TRIM28 to prevent CLDN3 SUMOylation could promote its degradation, thereby inhibiting tumour growth and metastasis. TRIM28 is a pleiotropic regulator with multiple downstream targets. Although our Co‐IP and protein stability assays confirm a direct interaction, studies utilising site‐directed mutagenesis of CLDN3 SUMO‐acceptor sites are warranted to unequivocally prove the dependence of CLDN3 stability on this specific modification. In this project, our findings showed that SUMOylation at K156 is a critical post‐translational modification required for CLDN3 stability and its oncogenic functions in CRC. Future investigations could significantly benefit from emerging deep learning frameworks. Recent advances in graph neural networks and transformer‐based models offer powerful tools for integrating transcriptomic, epigenetic and spatial single‐cell datasets [67]. Incorporating such methodologies could help systematically characterise the CLDN3‐centred regulatory network and refine predictions of TRIM28‐mediated SUMOylation sites with higher precision.
However, some of the findings in this study rely heavily on a single database, which means some results lack cross‐validation. Although bioinformatics analyses provide initial insights into the role of CLDN3 in cancer progression, additional in vivo and in vitro experiments are essential to corroborate these observations. Results showed that CLDN3 may act as a functional driver across malignancies, but several datasets presented involve opposing associations across tumour types. It suggests that its function may be context‐specific depending on the tumour microenvironment and tissue origin. Furthermore, the study does not include an independent in‐house clinical validation cohort, that prospective clinical validation (integrating immunohistochemistry or RNA expression analysis) is a priority for our future work. The specific mechanisms underlying CLDN3 dysregulation and its contribution to 5‐FU resistance require further investigation. These studies are critical for developing improved treatment strategies for 5‐FU‐resistant patients in clinical practice.

5
Conclusion
Overall, our study is the first to perform a comprehensive pan‐cancer analysis of the CLDN3 gene, delivering important preliminary insights into its role in colorectal cancer proliferation, migration, invasion and drug resistance. These findings provide a valuable foundation for advancing targeted therapeutic approaches against tumours.

Author Contributions

Xi Zeng: conceptualization, data curation, formal analysis, investigation, methodology, validation, writing – original draft, writing – review and editing. Lu Zhang: conceptualization, data curation, formal analysis, writing – original draft, writing – review and editing. Qing Chen: data curation, writing – review and editing. Yan Zeng: data curation, formal analysis, investigation, validation, writing – review and editing. Guanglei Yang: data curation, formal analysis, writing – review and editing. Dan Feng: conceptualization, formal analysis, funding acquisition, investigation, validation, writing – review and editing. Bin Han: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, writing – original draft, writing – review and editing.

Funding
This research was funded by Nanchong City and School Cooperation Project (Grant Nos. 22SXQT0346 and 20SXQT0101); Affiliated Hospital of North Sichuan Medical College Plan Projects (Grant Nos. 2021YS009 and 2022JB006); Anti‐infective Agent Creation Engineering Research Centre of Sichuan Province Fund (Grant No. KGR202404).

Consent
The authors have nothing to report.

Conflicts of Interest
The authors declare no conflicts of interest.

Supporting information

Supporting Information S1

Figure S1: The expression and Isoform of CLDN3 in human tissues.

Figure S2: Expression of CLDN3 in different pathological stages of cancers.

Figure S3: CLDN3 mRNA expression in different cancer subtypes and grades.