Landscape of cancer associated EpCAM mutations: molecular modeling, predictive insights and impact on patient survival

Abstract

Background

EpCAM (epithelial cell adhesion molecule) is a key regulator of epithelial cell–cell adhesion, signal transduction, tissue regeneration, and serves as a stem cell marker. It is frequently overexpressed in epithelial cancers and is linked to tumor progression, survival, and metastasis. However, the functional impact of EpCAM mutations in cancer remains poorly understood.

Methods

To investigate the role of EpCAM mutations, we performed a comprehensive analysis of cancer cohorts from multiple genomic datasets, identifying novel somatic EpCAM mutations across diverse epithelial cancers. Using bioinformatics tools (SIFT, PolyPhen-2, Mutation Assessor) and molecular modeling, we assessed the potential impact of these mutations. Further, homology modeling and all-atom molecular dynamics (MD) simulations were conducted to evaluate structural changes. From an analysis of 300 studies comprising 300,300 cancer samples, we identified 160 recurrent somatic mutations across epithelial cancers. Of these, seven mutations most frequently associated with lung cancer were further validated through molecular dynamics simulations, evaluation of ERK signaling activity, and assessment of sensitivity to the MEK inhibitor Trametinib.

Results

Our findings revealed that cancer-associated mutations, particularly in the TY-1 and RCD regions, induce structural instability in EpCAM, leading to altered functional properties. Patient cohort analyses indicated that EpCAM mutations correlate with reduced survival rates in colon and hepatocellular carcinoma and contribute to early tumor progression in lung cancer. Moreover, introducing these mutations into lung cancer cells enhanced their sensitivity to MEK inhibitors, suggesting a potential therapeutic vulnerability.

Conclusion

This study provides novel insights into the structural and functional consequences of EpCAM mutations in cancer, demonstrating their association with reduced survival, tumor progression, and drug sensitivity. These findings highlight EpCAM as a promising therapeutic target in epithelial cancers.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-025-14455-8.

Introduction

EpCAM, a transmembrane signaling protein identified as a tumor-associated antigen that is overexpressed in various epithelial cancers [1, 2]. Beyond its role in mediating cell–cell adhesion, EpCAM is known as a signaling molecule implicated in cancer cell proliferation, invasion, migration, differentiation, and immune evasion [3]. Its role in cancer biology and consistent expression across several cancer types make EpCAM a key biomarker and therapeutic target. In normal tissue EpCAM is found on the basolateral surfaces of epithelial cells in tissues such as the skin, gastrointestinal tract, respiratory tract, lung, and reproductive organs. EpCAM is highly expressed during embryogenesis, where it facilitates cell proliferation and differentiation [4, 5].

In a well-characterized phenotype of EpCAM mutations, autosomal recessive mutations in the EpCAM gene are associated with congenital tufting enteropathy (CTE) [6, 7]. The prevalence of germline MLH1 hyper-methylation and EpCAM deletions is notably high among genetically confirmed cases of Lynch syndrome [6, 8]. This epigenetic silencing predisposes individuals to Lynch syndrome, an inherited condition linked to colorectal, endometrial, and other cancers. Moreover, the loss of EpCAM function has been linked to phenotypic alterations, including epithelial-to-mesenchymal transition (EMT) [9], while gain-of-function has been implicated in enhanced oncogenic signaling [10]. EpCAM plays a crucial role in regulating key cancer-associated signaling pathways, including AP-1, NF-kB, Wnt/β-catenin, PI3K/AKT, and MAPK/ERK [9, 11–15]. By stabilizing β-catenin and promoting its nuclear translocation, EpCAM drives the transcription of Wnt target genes [16]. Similarly, its interactions with components of the PI3K/AKT pathway promote tumor cell survival and resistance to apoptosis. EpCAM mutations have also been shown to dysregulate ERK signaling, contributing to drug resistance, a major challenge in cancer therapy. For instance, dysregulated ERK signaling can be linked to reduced sensitivity to targeted therapies, highlighting the clinical implications of EpCAM dysfunction [9]. Its expression on the surface of cancer cells has facilitated its use as a therapeutic target in antibody–drug conjugates [17]. However, mutations in EpCAM are now being recognized as factors influencing the effectiveness of these therapeutic interventions. For example, altered EpCAM expression and function can be linked to enhanced drug resistance, primarily due to changes in cancer cell signaling or survival mechanisms. Our recent studies have demonstrated that cancer-associated EpCAM mutations resulted in the loss of its function, altered localization, and promotion of epithelial-to-mesenchymal transition (EMT), thereby facilitating tumor metastasis [18, 19].

EpCAM, also known as CD326, consists of distinct structural domains that serve specific biological functions. The extracellular domain (EpEx) comprises several key regions: an N-terminal domain (NTD) responsible for mediating protein–protein interactions, particularly homophilic adhesion; a thyroglobulin-like domain (TY-1), which contains cysteine-rich motifs involved in proteolytic regulation and structural stability; and a C-terminal domain located near the membrane, which is essential for signaling and proteolytic cleavage. Within this C-terminal region lies a specialized structural feature known as the ridge on the C-terminal domain (RCD), spanning residues 227–239. EpCAM is anchored to the plasma membrane by a single transmembrane domain (TMD) composed of a hydrophobic alpha helix. The intracellular domain (EpICD) is a short cytoplasmic tail that, following proteolytic cleavage, is released and translocates to the nucleus, where it participates in intracellular signaling pathways.

In the present study, for the first time, we investigate the landscape of EpCAM mutations across major cancers of epithelial origin. Using a bioinformatics approach, we screened for the possible damaging mutations. In addition to two previously experimentally validated mutations, six newly identified high-frequency mutations in advanced-stage lung cancer were predicted through bioinformatics analyses, employing homology modeling and molecular dynamics (MD) simulations predicting significant structural disruption to the EpCAM protein structure. Recent molecular dynamics studies on gene mutations have provided valuable insights into protein function, stability, and interactions, uncovering mechanisms that are challenging to observe experimentally. These findings highlight how mutations can lead to the loss or gain of gene activity, offering a deeper understanding of their impact on biological processes [20–22]. Through in-silico analyses and experimental validation, we demonstrate that certain EpCAM mutations result in a loss of structural stability, thereby disrupting ERK signaling. Furthermore, the co-occurrence of EpCAM mutations with known oncogenes reveals potential therapeutic vulnerabilities. Detecting these mutations early in disease progression, particularly in the genomic era, could enhance patient survival predictions and guide the development of targeted therapeutic strategies.

Materials and methods

Cell culture and reagents

All studied cell lines were obtained from the American Type Culture Collection (ATCC, Rockville, MD, USA). A549 (ATCC#CCL-185), H1299 (ATCC#CCRL-5803), H460 (ATCC#HTB-177), H226 (ATCC#HTB-177) and H23 (ATCC#CRL-5800). Phoenix-AMPHO (ATCC#CRL-3213) cells cultured in DMEM. NSCLC cells were cultured in RPMI-1640. Both culture media were supplemented with 10% FBS, 2 mM glutamine (Gibco #25030081), 1% penicillin/streptomycin, and incubated at 37°C in an atmosphere of 95% air and 5% CO₂. The MAPK inhibitor Trametinib was purchased from Selleck Chemicals (Houston, TX, USA). Cell culture was monitored for mycoplasma routinely.

EpCAM mutation constructs

The EpCAM nucleotide sequence was accessed with the NCBI reference sequence NM_002354. EpCAM deletion mutants were generated based on the genomic data sets (Data S2A). To the wild-type EpCAM single amino acids change was made by custom gene synthesis of EpCAM-D92V, EpCAM-E137A, EpCAM-Y186C, EpCAM-Q204E, EpCAM-N111K, EpCAM-P84S, and EpCAM-Y215S. EpCAM mutant constructs were generated using synthetic gene fragments from Integrated DNA Technologies (IDT, Coralville, IA) as described before [18]. For example, C66Y EpCAM was generated as a G-block fragment (197G > A; substitution position 197, G→A). The DNA sequence was sub-cloned into both pcDNA3 (HindIII-XbaI restriction sites) or the retroviral vector pBABE-puro (BamHI-SalI restriction after mutating BamHI sites).

Retroviral transduction

In a six well plate, Phoenix-AMPHO packaging cells were transfected when nearly confluent with 2.5 μg of pBABE-Puro-EpCAM constructs using FuGENE-HD (Promega#E2311). Forty-eight hours post transfection, viral supernatants were collected, filtered through 0.45 μm filters, and then added to the cells in media containing 8 μg/mL protamine sulfate. After one-two successive retroviral infections, cells were grown for 48 h and selected in puromycin for 2 weeks.

SRE-ERK dual luciferase assay

SRE reporter assay was performed as described earlier [14]. Briefly, cells were cultured in 12-well plate for 24 h. DNA constructs 400 ng-SRE-Luc (Promega# E1340) together with internal control plasmid 50 ng-pRLSV40-Luc (Promega# E2231) was transfected using FuGENE-HD (Promega # E2311). After 12 h post transfection cells were serum-starved for 12 h, treated with or without 20% FBS overnight. Cells were lysed, and the dual luciferase assay was performed using the Dual-Luciferase Reporter assay system (Promega #E1910). Mean values of luciferase activity relative to untreated and were calculated from triplicate wells. The experiments were repeated three times to ensure consistency. Mean ± SD of three technical replicates. * p < 0.05, ** p < 0.001, two-tailed Student’s t-test.

Trametinib treatment and cell viability assay

EpCAM transduced or transfected cells 5.0 × 10³ were seeded into 96-well plates and cultured in a 5% CO₂ incubator for 24 h. Trametinib (Selleck Chemicals #S2673) was prepared in DMSO and applied at concentrations ranging from 100 to 200 nM. Cell survival was assessed 72 h later using the Cell Titer-Glo (Promega #G7570) according to the manufacturer's instructions. Luminescence was recorded by SpectraMax-i3 (Molecular Devices). Cell death was monitored using Annexin-V staining (#V13241, ThermoFisher) as recommended by the manufacturer.

EpCAM mutation dataset

EpCAM mutation status and sample related information were collected from the cBioPortal for Cancer Genomics (https://www.cbioportal.org). EpCAM mutation data (see data S2B) was acquired from AACR-GENIE (https://genie.cbioportal.org) and the Catalogue Of Somatic Mutations In Cancer (COSMIC, https://cancer.sanger.ac.uk/cosmic) [23, 24]. All compiled data is included in the supplementary data S2A. The mutation data was restricted to somatic mutations where the protein was altered in single amino acid. Overlapping Samples from GENIE v.13 and cBioPortal were removed. For TMB distributions, tumors with TMB between 5 and 600 were grouped in increments of 5 or 100 for 10 different cancer types. EpCAM mutation per group was identified as 1, 2–4, 5–9, 10–19, 20–29, 30–50 and increment of 100 till 600 (see Fig. 2D). Cell line EpCAM expression was accessed through the publicly available Cancer Cell Line Encyclopedia (CCLE, https://portals.broadinstitute.org/ccle) at the Broad Institute (Data S2C). EpCAM expression from NCI-60 cell line panel (GSE32474) was accessed, data was plotted using GraphPad. EpCAM expression in 70 NSCLC tumor lines GSE32989 was accessed and analyzed using GENE-E (https://broadinstitute.org/GENE-E/).

Fig. 2.

Identification of cancer-associated EpCAM mutations. A Publicly available datasets, including cBioPortal, AACR-GENIE, and COSMIC, were analyzed to identify mutations across various cancer cohorts. B Mutation frequency was calculated as the number of EpCAM mutations per 1000 samples across the combined datasets. C Over 200 new missense mutations were identified within the EpCAM gene. D The distribution of mutations by base changes and the overall mutation types were characterized

Molecular modeling of EpCAM

The 3D structure of human EpCAM wild type (EpCAM_WT) was built using homology modelling with the help of Modeller 10.2 software [25]. The crystal structure of human was used (PDB ID: 4MZV) (https://www.rcsb.org/) as a template to build full length 3D model of EpCAM_WT and generated total 500 different conformations of EpCAM_WT. The 3D structure of EpCAM_WT was selected based on DOPE score (-26,725.0 kcal/Mol). The Structural refinement of the modelled EpCAM_WT structure was done by performing all atom MD simulations in explicit solvent using GROMACS 2021.5. The force filed amberff99SBildn was used to generate topology files and TIP3P model for solvation with periodic boundary conditions. The required number of counterions were added to neutralize the system. Steric clashes or bad contacts raised during homology modelling were relaxed by performing energy minimization with Steepest-Descent method followed by Conjugate-Gradient. Canonical ensembles NVT and NPT were used to equilibrate the system for 1 ns. Further, unrestrained MD simulation was performed for the period of 1 µs to get detailed insights to the structural stability and evolution of diverse conformations over the potential energy surface. The long-range electrostatic interactions were treated with PME and LINCS algorithm to constraint the H-bonds. The coordinates and energies were recorded at every 2 fs and 200 ps respectively. A modified Berendsen thermostat and Parrinello-Rahman algorithm was used to maintain constant temperature (300 K) and pressure (1 bar) during the simulation. The structure of EpCAM_WT having least energy near the global state was extracted from the MD trajectory and then used further to generate homology models of mutant EpCAM structures for EpCAM_D92V, EpCAM_E137A, EpCAM_Y186C, EpCAM_Q204E, EpCAM_N111K, EpCAM_P84S, and EpCAM_Y215S). The MD simulations for the mutants were also performed using the similar protocol adopted for EpCAM_WT simulation. The trajectories obtained were checked for quality and detailed structural analysis was performed using in build Gromacs tools and other necessary packages such as DSSP [26], CPPTRAJ [27] wherever required. The plots were generated using Grace 5.1.25 and quality images were prepared using UCSF Chimera 1.15 [28].

Statistical analysis

All experiments were performed at least three times in triplicate. All statistical analyses were performed using GraphPad Prism 9.4 (GraphPad, La Jolla, CA). Numerical data are presented as mean ± sd. Single comparisons were performed by unpaired Student’s t tests and multiple comparisons were performed by ANOVA. P-values < 0.05 were considered to be statistically significant. Kaplan–Meier survival analysis was performed based on EpCAM mutation and the relevant data of patient populations. GraphPad Prism was used to analyze and plot the graphs.

Results

EpCAM expression is restricted to epithelial type cells

To assess overall expressions of EpCAM at RNA and protein levels in normal tissues, two data sets were analysed. Results highlights significant expression of EpCAM RNA and protein within the gastrointestinal tract and other tissues (Fig. 1, A-B). Conversely, EpCAM expression is largely absent in non-epithelial tissues such as connective tissue, muscle, and hematopoietic cells under normal physiological conditions. Notably, EpCAM expression is frequently elevated in epithelial-origin malignancies, underscoring its relevance as a diagnostic and therapeutic target. For example, elevated EpCAM levels are consistently observed in adenocarcinomas of the breast [29], colon [30], oesophagus [31], pancreas [32], ovary, and prostate [33], as well as in squamous cell carcinomas of the lung, and cervix [34]. Detailed analysis of a lung adenocarcinoma cohort demonstrates upregulation of EpCAM in tumors (Fig. 1, C), with consistent expression across different disease stages and consistently in EGFR, KRAS oncogene-driven tumors (Fig. 1, D). In hepatocellular carcinoma (HCC), EpCAM is typically absent in mature hepatocytes but is expressed in hepatic progenitor cells and certain tumor subsets, suggesting its involvement in tumor initiation and cellular dedifferentiation [35]. Cancer Cell Line Encyclopedia (CCLE) [36] and NCI-60 [37] datasets further reveal that EpCAM expression is restricted to cancer cell lines derived from epithelial origin tissues such as the colon-gastrointestinal tract, ovary, prostate, breast, lung, and kidney (Figure S1 A-B). This tissue-specific expression pattern reinforces the utility of EpCAM as a marker for epithelial malignancies and a target for precision oncology.

Fig. 1.

EpCAM expression is predominantly observed in epithelial cells and related tissues. A-B, RNA and protein expression levels of EpCAM, based on data from Protein Atlas (www.proteinatlas.org), and GTEx Portal (https://gtexportal.org/) are shown, with expression ranked from high to low. C EpCAM expression is elevated in tumor tissues compared to normal tissues, as analyzed using two independent tumor-normal paired datasets: GSE18842 and HSE1007. D Analysis of EpCAM expression across different stages of NSCLC and in relation to EGFR, KRAS, and ALK oncogenic alterations. Microarray datasets GSE43580 and GSE31210 were normalized and visualized using GraphPad. These results highlight that EpCAM expression is largely confined to normal epithelial tissues, including the gastrointestinal tract, thyroid, kidney, pancreas, breast, and lung, and is significantly upregulated in cancer tissues

Comprehensive analysis of EpCAM mutations

Expanding upon the insights gained from our previous studies demonstrating that cancer-associated EpCAM mutations play critical roles as tumor suppressor or tumor promoter [18, 19]. Some EpCAM mutations result in the loss of its membrane localization, secretion, and binding to CTSL, suggesting its role as a tumor suppressor [18]. On the other hand, EpCAM mutations in the LDL domain in cancer cells driven by RAS signalling promoted invasion, migration and sensitized to drugs [19]. To further elucidate the functional implications of EpCAM mutations in cancer, we employed advanced predictive and experimental techniques, paving the way for a deeper understanding of the role of these mutations in pathology. Four different cohorts cBioPortal, COSMIC, AACR-GENIE and FMI data (Foundation Medicine-NCI) were accessed (Fig. 2A). Building on our previous reports and updated TCGA data, we undertook a systematic analysis to catalogue and understand the role of novel EpCAM mutations in cancer. By analyzing over 300 studies comprising 300,300 samples, we determined an overall EpCAM mutation frequency across all cancer types (Fig. 2B). These cancer-associated mutations were distributed as missense (79%), nonsense (6%), splice (7%), frameshift (4%), deletion (2%), and fusion mutations. Notably, MSH2-EpCAM fusion were recorded in samples from patients with esophageal adenocarcinoma, squamous cell carcinoma, and stomach adenocarcinoma (Data S2A). This is consistent with prior reports where exon 9 deletions in EpCAM led to MSH2 promoter hyper-methylation and silencing [38]. Additionally, we identified EpCAM-ABCG8 and NUP42-EpCAM fusions in high-grade uterine and ovarian carcinomas (Data S2A). Missense EpCAM mutations were also identified in commonly used CCLE cell lines (Data S2C). While mutations were distributed across all nine exons of EpCAM, no specific hotspot regions were identified. Among the 160 unique protein-altering mutations identified in initial screening, the highest mutation frequencies were found in lung cancer (29 cases) and skin cancer (22 cases), followed by colorectal cancer (13 cases) and uterine cancer (13 cases). Interestingly, colorectal cancer a cancer type known to overexpress EpCAM displayed a higher prevalence of splice-related mutations. Of particular note, EpCAM mutations was lowest from 1,346 pancreatic cancer samples analysed (Data S2A).

To examine the co-occurrence of somatic EpCAM mutations across varying tumor mutation burden (TMB) groups, we categorized the data into two major groups: low TMB (1–100) and high TMB (> 100) (Fig. 2D, see methods). The total number of somatic EpCAM mutations was then tabulated from each TMB category. An initial search from the GENIE cohort [23] across various cancer types revealed mutation frequency of EpCAM at 0.5%, with 889 EpCAM mutations including recurring from 1,783,034 samples (Data S2B). As depicted in Fig. 2D, majority of EpCAM mutations in lung cancer were primarily found in samples with fewer than 100 TMB group. For breast, prostate and pancreatic cancer, they were mostly restricted to around 15 TMB group. In contrast, EpCAM mutations in colon and uterine cancers were spread across all TMB groups (5–600 TMB see methods). The restriction of EpCAM mutations to samples with relatively low TMB in lung and breast cancers suggests that these mutations may play an important role in the early stage of tumorigenesis. In contrast, the widespread distribution of EpCAM mutations in colon and uterine cancers indicates a more pervasive involvement in both early and late stages of tumor progression.

Structural stability analysis of previously characterized EpCAM mutations

We utilized two experimentally validated damaging mutations, C66Y and L240A as a proof of concept to validate our strategy [18, 19]. Using MD simulation to ensure the quality of the obtained trajectories, we performed a quality check by plotting the potential energy, temperature, and pressure throughout the dynamics. The values for EpCAM-L240A were -127,562 kJ/mol for potential energy, 1.53331 K for temperature, and 6.06631 bar for pressure, while for EpCAM-C66Y, the corresponding values were -111,969 kJ/mol, 0.231919 K, and 7.42474 bar. These values exhibited minimal fluctuations during the 1 µs simulation, with temperature and pressure maintained at 300 K and 1 bar, respectively. To gain deeper structural and functional insights, we compared the time-dependent evolution of RMSD values with EpCAM-WT (Fig. 3,B) EpCAM-L240A displayed a steady increase in RMSD throughout the 1 µs simulation, indicating significant conformational changes compared to its initial structure. Specifically, the starting structure of the TY-1 loop in EpCAM-L240A, formed by residues Ala63 to Arg138, was extended, but during the MD simulation, it moved toward both the N-terminal and C-terminal regions, adopting a more compact globular shape. Additionally, the ridge on the C-terminal domain (RCD) exhibited a closing movement. In contrast, the TY-1 loop of the EpCAM-C66Y mutant showed a closing movement toward the TYD region (Fig. 3A). The steady decrease in the radius of gyration (Rg) of EpCAM-L240A further supports its compact folding compared to EpCAM-WT, as depicted in Fig. 3C.

Fig. 3.

Molecular modeling of phenotype-characterized cancer-associated EpCAM mutations. A Significant local conformational changes observed in wild type EpCAM and its mutants, L240A and C66Y. Conformational changes in the EpCAM protein and its mutants were analyzed during molecular dynamics simulations. The original EpCAM structure (PDB ID: 4MZV) is shown in yellow with the TY-loop highlighted in green, while the average predicted structure is depicted in blue with the TY-1 loop highlighted in magenta. B Graphical representation of backbone conformations, C Radius of gyration, D Residual fluctuations, and E Solvent-accessible surface area (SASA). F Probability distribution function plots for EpCAM-WT, EpCAM-L240A and EpCAM-C66Y. G Analysis of intra-protein hydrogen bonds for EpCAM-WT (black), EpCAM-L240A ​ (cyan), and EpCAM-C66Y​(orange). H Conformational changes in EpCAM Protein structures during MD simulation. Superimposed figures of EpCAM_WT with EpCAM_L240A and EpCAM_WT with EpCAM_C66Y. I Principal components Analysis of EpCAM_WT, EpCAM mutants L240A and C66Y. J Dynamic cross correlation matrix (DCCM) of backbone atoms of EpCAM-WT, EpCAM-L240A and EpCAM-C66Y. DCCM maps, obtained from 1 μs trajectories, contain contours of different colours. Correlated motion and anti-correlated motions are represented by amber and blue colours, respectively. The Black coloured box indicated the correlation between the TY-loop and RCD regions

The RMSD values for EpCAM-WT and EpCAM-C66Y exhibit stable dynamic behavior throughout the simulation period (Fig. 3B). In contrast, the residual fluctuations for EpCAM-WT are significantly lower than those observed in the two mutants, EpCAM-L240A and EpCAM-C66Y. The TY-1 loop in both mutants shows considerable dynamics, with RMS fluctuations reaching up to 8.5 Å. EpCAM-C66Y exhibits partial unfolding and refolding during the simulation (Fig. 3D). Residues from the hydrophobic core of EpCAM-L240A become buried within the protein's interior, making them less exposed to the solvent (Fig. 3E). Meanwhile, EpCAM-C66Y and EpCAM-WT display similar trends in their SASA values, suggesting stable dynamics. The distribution of the backbone RMSD values indicates that EpCAM-WT and EpCAM- L240A have probability of maintaining RMSD values 0.6 and 0.8 Å respectively, while EpCAM- C66Y shows a maximum RMSD value of around 1.8 Å (Fig. 3F). The TY-1 loop of EpCAM-C66Y also exhibits flexibility, although its RCD region (ridge on CD) undergoes significant conformational changes, leading to a closure movement at the C-terminal domain (Fig. 3A). In contrast, the TY-1 loop in EpCAM-WT does not show any such closure movement in the RCD, CTD, or TY-loop regions, although structural transitions within the TY-1 loop are observed during the MD simulation (Fig. 3A).

Intramolecular hydrogen bonds (H-bonds), essential for stabilizing protein structure, were analyzed for EpCAM-WT, EpCAM-L240A, and EpCAM-C66Y. The maximum number of H-bonds formed in each variant is shown in Fig. 3G. The MD simulation indicated that EpCAM-L240A undergoes significant conformational changes during the simulation (Fig. 3H). To quantify these movements, DynDom was used to identify the fixed (residues 30–73 and 95–265) and moving (residues 74–94) domains in EpCAM-L240A. The moving domain shows a rotational movement of 108.7 degrees, a translational movement of -7.7 Å, and a closure movement of 95.0%. Bending region analysis reveals that residues from the TY-1 loop, specifically Lys70-Gly75 and Gly93-Asp98, serve as hinges facilitating this large rotational and closure movement (see Figure S3A,B). In contrast, DynDom did not identify significant dynamic movements in EpCAM-WT or EpCAM-C66Y, as their conformational changes were too subtle for detection (data not shown). A 2D projection of the principal components, PC1 and PC2, indicates that EpCAM-L240A occupies a larger conformational space in the free energy landscape, whereas EpCAM-WT forms a compact cluster, occupying less space. EpCAM-C66Y shows fewer diverse conformations compared to EpCAM-L240A (Fig. 3I). The cosine content values for EpCAM-WT, EpCAM-L240A, and EpCAM-C66Y are 0.78, 0.92, and 0.91, respectively, indicating good sampling across the conformational space.

The dynamic cross-correlation matrix (DCCM) was generated using representative frames from the MD simulations to explore the concerted motion of the individual domains of EpCAM. Figure 3J displays the DCCM plots for EpCAM-WT, EpCAM-L240A, and EpCAM-C66Y. The diagonal amber line indicates strong self-correlation of individual residues, while the amplitude of positive to negative correlations is represented by a gradient from amber to blue. In EpCAM-WT, the dynamic TY-1 loop region shows a moderately negative correlation with the N-terminal domain (NTD), while this negative correlation diminishes in both EpCAM-L240A and EpCAM-C66Y. Additionally, the amplitude of positive correlation is higher in EpCAM-C66Y than in EpCAM-L240A. Both mutants exhibit increased positive correlation between the RCD region and the TY-1 loop during the simulation. Overall, the amplitude of negative correlations observed in EpCAM-WT decreases substantially in both mutants. This suggests that both mutants experience positive cooperative motion, particularly in the RCD, NTD, and TY-1 loop regions, indicating significant local conformational changes that could influence EpCAM dimerization and protein–protein interactions.

In conclusion, the molecular modeling study reveals that the EpCAM-L240A mutant exhibits highly dynamic behavior, particularly at the TY-1 loop and RCD regions, making it relatively less stable compared to EpCAM-WT. Although EpCAM-C66Y undergoes moderate conformational changes, it significantly alters the native conformation of the TY-1 loop. We propose that the TY-1 loop plays a crucial role in the dimerization of human EpCAM, and recent studies suggest that the TY-1 loop/domain of EpCAM binds and inhibits the protease CTSL (Cathepsin-L) [18]. The substantial structural changes, particularly in the TY-1 loop/domain, may contribute loss of CTSL binding to promote tumor cell invasion and metastasis. Altogether, MD simulation and previously reported functional data suggest that both EpCAM-L240A and EpCAM-C66Y mutations exhibited a loss of EpCAM structure and thereby function.

Bioinformatics analysis and Molecular modeling of lung cancer-associated EpCAM mutations

To systematically catalogue and assess large list of damaging mutations in EpCAM, first we employed three computational tools sorting intolerant from tolerant (SIFT), polymorphism phenotyping (PolyPhen) and mutation assessor [39–42]. The results were normalized from the three computational tools on a scale from 0 to 1, with zero representing the wild-type EpCAM sequence. By combining the outputs from all three tools, we were able to generate a comprehensive mutational impact score (Fig. 4A, see Data S4). After initial screening, we selected additional six high frequency, high-scoring mutations from lung cancer for further validation. This approach offered a valuable framework for screening and prioritizing mutations for functional validation. To assess the quality of the obtained trajectories following molecular dynamics simulations, first we plotted the potential energy, temperature, and pressure throughout the dynamics. The values for the mutants EpCAM were:

• EpCAMD92V: (-104,671 kJ/mol), (0.12362 K), (-5.72833 bar)

• EpCAME137A: (-100,359 kJ/mol), (0.244 K), (4.62956 bar)

• EpCAMY186C: (-118,137 kJ/mol), (0.100088 K), (-0.119114 bar)

• EpCAMQ204E: (-100,348 kJ/mol), (0.0841861 K), (1.89489 bar)

• EpCAMP84S: (-99,460.7 kJ/mol), (0.050235 K), (4.04775 bar)

• EpCAMY215S: (-101,241 kJ/mol), (0.289105 K), (-2.47406 bar)

Fig. 4.

Bioinformatics and MD simulation-based analysis of EpCAM mutations. A Computational tools, including SIFT (Sorting Intolerant From Tolerant), PolyPhen (Polymorphism Phenotyping), and Mutation Assessor, were used to evaluate structural alterations for each mutation. Scores ranged from 0 (least damaging) to 1 (most damaging). The combined scoring provides a comprehensive view of the most damaging mutations, as shown in the figure. Molecular dynamics (MD) simulations were performed for seven lung cancer-associated mutations to characterize various structural and dynamic parameters. B Principal Component Analysis (PCA) showing 2D projection correlation maps from MD simulations for the wild type (WT) EpCAM and its mutant structures. C Probability distribution functions of WT EpCAM (black) and mutants: D92V ( turquoise), E137A (orange), Y186C (blue), Q204E (brown), P84S ( gray), and Y215S ( (violet). D Root Mean Square (RMS) fluctuation spectrum of the TY loop replace and remove inset. E–F Structural changes observed during MD simulations by superimposing the WT EpCAM and mutant structures, specifically comparing WT EpCAM (black) and the N111K mutant. G Wild type EpCAM (black in colour) and N111K (brown in colour) RMSD, RMSF,RMS, Rg, SASA and H Hydrogen bonds

These mutants showed minimal fluctuations during the 1 µs simulation, with temperature and pressure maintained at 300 K and 1 bar, respectively. Molecular dynamics simulations were employed to estimate the structural and conformational changes of the proteins. To gain deeper structural and functional insights into the novel point mutants of EpCAM, including those related to lung cancer, we also compared the time-dependent evolution of the RMSD values for EpCAM mutants with those of EpCAM-WT.

The RMSD profile for the backbone of EpCAM-WT and its six mutants was generated for all atoms from the initial structure to assess the impact of mutations on the stability of the protein structure (see Figure S4A). The MD simulation of EpCAM-WT exhibited a steady RMSD of approximately 4.5 Å throughout the simulation. When compared to WT-EpCAM, EpCAM-N111K exhibited RMS fluctuations at TY loop (Fig. 4D) and RMSD around 4.5 Å (Fig. 4G), showing similar trends to the wild type from the start to 700 ns. After 700 ns, the RMSD decreased to approximately 3 Å at 850 ns, before rising again to 5 Å at 1 ms Overall, MD simulation and superimposing WT-EpCAM, EpCAM-N111K demonstrated structural changes at TY loop (Fig. 4E,F). In contrast, EpCAM-D92V showed fluctuations between 50 and 200 ns, with a decrease in RMSD from 220 to 400 ns. After that, the RMSD remained relatively stable at around 4.5 Å until 850 ns, after which some fluctuations occurred between 850 and 900 ns, as shown in Figure S4A. The TY-1 loop of EpCAM-D92V remained flexible during the simulation. For the mutant EpCAM-E137A, the RMSD increased to approximately 6.3 Å, indicating structural instability as observed in Figure S4A. EpCAMY-186C exhibited similar RMSD values to EpCAM-WT up to the last 1000 ns. However, the mutant EpCAM-Q204E displayed a notably unstable structure, with RMSD values remaining stable up to 500 ns, after it increased to 8 Å at 1 ms, suggesting a loss of structural stability due to the mutation at position Q204E. EpCAM-P84S showed distinct deviations, ranging from approximately 2.5 to 6 Å. Initially, its RMSD was lower than that of the wild type from 50 to 650 ns, after which it steadily increased to 6 Å. EpCAM-Y215S required some time to achieve a relatively stable structure as observed in Figure S4A.

Next, we analyzed how cancer-associated mutations affect the dynamic behavior of various regions in EpCAM, including the N-terminal domain (NTD), C-terminal domain (CTD), ridge on CD (RCD), stalk and C-terminal domain (SCD), and importantly, the TY-1 loop. This was done by calculating the root mean square fluctuation (RMSF) of the Cα atoms for all the mutants and comparing them to the EpCAM-WT. The residual fluctuations in EpCAM-WT were significantly lower compared to the six mutants particularly when observing the dynamics of the TY-1 loop (residues 63–138) and its interactions with different protein sites. The TY-1 loop in all mutants EpCAM-P84S, EpCAM-Q204E, EpCAM-D92V, EpCAM-E137A, EpCAM-N111K, EpCAM-Y186C, and EpCAM-Y215S was highly dynamic, with RMS fluctuations reaching up to 9 Å, 8 Å, 7.5 Å, 7 Å, 6 Å, 5.5 Å, and 5 Å, respectively (Fig. 4G and S4A). EpCAM-P84S exhibited partial unfolding and folding, with the highest fluctuations occurring in the TY-1 loop residues between ~ 75–100, as illustrated in Figures S4A. Residues 25–75, which are part of the NTD, showed some motion in EpCAM-D92V, EpCAM-E137A, and EpCAM-P84S mutants. Additionally, significant fluctuations were observed in the C-terminal domain, particularly in Q204E of the last amino acids, with fluctuations reaching up to 19 Å, as seen in Figure S4A. Notably, the most pronounced fluctuations occurred in the TY-1 domain (residues 63–138) in all mutants. The EpCAM-Q204E mutant exhibited some additional fluctuation in the RCD region (residues 227–239) compared to the wild-type, as shown in Figure S4A. The SCD region (residues 202–205) remained relatively stable during the simulation across all mutants. From Figure S4A, we can conclude that the residual fluctuations mainly occur in the NTD (residues ~ 24–62), the TY-1 loop (~ 75–100), and the last two residues (264–265), with higher fluctuations observed in EpCAM-D92V, particularly in the TY-1 loop region (63–138). EpCAM-E137A exhibited significant fluctuations mainly at positions 80–95 of the TY-1 loop. EpCAM-WT demonstrated a compact structure with an RMSD range of 18.5–19.5 Å throughout the simulation, while the mutants showed a gradual loss of compactness. The mutants EpCAM-Q204E and EpCAM-P84S showed continuous folding and unfolding, with Rg values ranging from 19.5–20.5 Å and 19–20.0 Å, respectively, as seen in Figure, S4BThe remaining mutants EpCAM-D92V, EpCAM-E137A, EpCAM-N111K, EpCAM-Y186C, and EpCAM-Y215S showed less compactness, with Rg values ranging from 18.5–20.5 Å, 18.5–20.0 Å, 18.5–20.0 Å, 18.5–19.5 Å, and 19.0–20.0 Å, respectively. In terms of solvent accessibility, residues from the hydrophobic core of the EpCAM-Y186C, EpCAM-E137A, and EpCAM-P84S mutants were buried inside the protein, reducing their exposure to the solvent environment, as shown in Figures S4B In contrast, EpCAM-D92V, EpCAM-Q204E, EpCAM-N111K, and EpCAM-Y215S exhibited stable dynamics, with similar trends in their solvent-accessible surface area (SASA) values.

Additionally, we analyzed the intramolecular hydrogen bonding interactions in EpCAM-WT and its mutants N111K. The maximum number of H-bonding interactions is depicted in Fig. 4H. Considering the conformational changes observed from the dynamics of the mutant, we compared the 3D structures with the wild-type EpCAM structure. Wild-type EpCAM displayed an open conformation of the TY-1 loop, but this conformation was altered. Specifically, EpCAM-D92V exhibited changes in the TY-1 loop structure, which slightly moved toward the C-terminal domain (CTD) binding site. Additionally, the RCD region (His227-Gln239) shifted toward the CTD compared to the wild-type EpCAM. In a study it was proposed that EpCAM has two binding sites for the TY-1 loop: the TYD and CTD [43]. Our study found that in the EpCAM-E137A mutant, the TY-1 loop residue that binds to the CTD in wild-type EpCAM binds to the TYD site, while the TY-1 loop that normally binds to the TYD site interacts with the CTD in this mutant (see Figure S4A left panel). In EpCAM-Q204E, EpCAM-N111K, and EpCAM-Y215S, the TY-1 loop shifted toward the TYD site, resulting in a closed conformation of the TYD. In the other mutants, no significant movement was observed.

Figure S5, illustrates the time evolution of the secondary structure along the protein chain, showing the stability of the N-terminal α-helix and fluctuations in the central β-sheet. Hydrogen bond analysis further indicated that the α-helical regions maintained a stable network of hydrogen bonds, while the β-sheet regions experienced notable disruption in hydrogen bonding after 150 ns, correlating with the observed structural transition. To examine the structural plasticity during MD simulation, we used DSSP to track secondary structural elements in both wild-type EpCAM and its mutants. Figure S5 shows the secondary structure changes over time. Both the wild-type and mutant proteins exhibited coils, β-sheets, β-bridges, bends, turns, and various types of helices. However, the mutants showed a considerable decrease in the formation of β-sheets and bends compared to wild-type EpCAM. We performed essential dynamics (ED) analysis to study the correlated motions of the wild-type and mutant EpCAM proteins. The eigenvalues from this analysis indicated that major fluctuations in the system were confined to the first two eigenvectors for both the wild type and the mutants. As seen in Fig. 4B, the mutant proteins covered a larger space along the first and second principal components (PC1 and PC2) compared to the EpCAM-WT. The probability distribution for the EpCAM-WT remained within a 6 Å threshold, while the mutant structures EpCAM-Q204E, EpCAM-Y215S, EpCAM-E137A, EpCAM-D92V, EpCAM-P84S, EpCAM-N111K, and EpCAM-Y186C showed maximum alterations of 9 Å, 7 Å, 7 Å, 7 Å, 6 Å, 6 Å, and 6 Å, respectively, (Fig. 4C). In the EpCAM-WT, both correlated and anti-correlated motions were observed, while in the mutant EpCAM-Q204E, EpCAM-E137A, and EpCAM-Y215S, mostly anti-correlated motion was observed (Figure S4C). In summary, molecular modeling reveals that mutants EpCAM-Q204E, EpCAM-E137A, EpCAM-Y215S, EpCAM-N111K, and EpCAM-D92V exhibit significant dynamic behavior, particularly in the TY-1 loop and RCD regions during MD simulations, which leads to a less stable structure compared to EpCAM-WT. In contrast, the mutants EpCAM-P84S and EpCAM-Y186C show moderate conformational changes. The significant structural changes observed in the TY-1 domain across all mutants would likely disrupt the structure and function of EpCAM, influencing its role in cellular processes.

EpCAM mutations predicts poor survival in colon and hepatocellular cancer

EpCAM frequently overexpressed in colon cancer which promote cancer cell proliferation, migration, and resistance to apoptosis [1]. EpCAM is also considered a marker of cancer stem cells in colon cancer, where its expression potentially influences the self-renewal capability of these stem cells, contributing to tumor relapse and chemo-resistance [13, 17]. Clinically, the overexpression of EpCAM in colon cancer is associated with poor survival outcomes, likely due to its role in promoting invasiveness and metastasis. High levels of EpCAM have been correlated with advanced disease stage and worse prognosis in colon cancer patients. In hepatocellular carcinoma (HCC), EpCAM overexpression is similarly linked to poor prognosis, with elevated levels associated with tumor progression, vascular invasion, and metastasis [35, 44]. EpCAM expression in HCC is also correlated with shorter survival and a higher likelihood of recurrence post-surgery. To investigate the role of these newly identified EpCAM mutations, a cohort of patients with EpCAM mutations alongside other driver mutations was analyzed. As shown in Figs. 5A-C, somatic mutations in EpCAM coexist with mutations in key oncogenes such as KRAS, BRAF, p53, and EGFR. In the group with altered EpCAM expression, the survival was significantly worse, with a median survival of 42.78 months, compared to 62.25 months in the unaltered group (Fig. 5B). As seen in Fig. 5D,E, in HCC, EpCAM mutations were observed in conjunction with mutations in p53, MDM2, DNMT1, and KEAP1 where the overall survival for patients with EpCAM mutations was 9.86 months, compared to the unaltered group (40.45 months). These findings highlight the critical role of EpCAM mutations in cancer progression, reinforcing their involvement in tumorigenesis and their potential utility as prognostic markers.

Fig. 5.

Recurrence of EpCAM mutations in oncogenic driver mutations of colon and Hepatocellular cancers. A EpCAM mutations were analyzed in colon cancer patients using TCGA and cBioPortal datasets. B Patient data from the same datasets were examined for overall survival. The Kaplan–Meier plot shows that patients with EpCAM mutations or alterations have worse survival outcomes compared to those in the unaltered group. C EpCAM mutations were frequently observed alongside driver mutations in genes such as KRAS, BRAF, TP53, and EGFR. D-E EpCAM mutations in hepatocellular cancers demonstrate worst survival

The prevalence of EpCAM mutations in NSCLC may hold significant prognostic value

Building on our previous study demonstrating that RAS can stabilize EpCAM expression [19], we hypothesized that somatic mutations in the EpCAM, frequently co-occurring with driver oncogenes such as KRAS, EGFR, TP53, and BRAF, could possess prognostic significance. To investigate, first, we analyzed two cohorts (Fig. 6A, Table S1), and as expected, most EpCAM mutations were found in conjunction with mutations in TP53, KRAS, EGFR, STK11, BRAF, and PTEN (Fig. 6D). The GENEI-v16 cohort [23], which includes data from 26,851 patients, revealed 284 missense mutations (data S2B). Surprisingly, overall survival did not significantly differ between patients with EpCAM mutations and those without, across both cohorts (see Fig. 6B-C, Table S2, Figure S6A). This analysis suggests that EpCAM mutations alone may not be a significant determinant of survival in these populations. To assess this unexpected observations, we revisited the tumor mutation burden (TMB) Fig. 2D. Interestingly, EpCAM mutations in lung cancer were predominantly found in samples with a low TMB. This suggests that these alterations may play a role in the early stages of tumor development rather than in its progression; however, incorporating additional patient data and analyses could provide deeper insights into their precise temporal and clinical significance. Despite non-significant p-value due to small sample size, the observed divergence in survival curves indicates a potentially meaningful biological association that merits further investigation. Expanding the patient cohort through multi-center collaborations could provide the statistical power needed to clarify the relevance of this finding and strengthen the overall conclusions of this new investigation.

Fig. 6.

High frequency of EpCAM mutations in NSCLC suggests roles in tumorigenesis and drug sensitivity. A-B Analysis of metastatic lung cancer patients indicates that EpCAM mutations do not significantly alter overall survival. C-D The GENIE cohort dataset reveals that NSCLC has the highest frequency of EpCAM mutations. These mutations are predominantly associated with cancers harboring mutations in TP53, KRAS, EGFR, KEAP1, and LKB1. E Stable EpCAM mutations were expressed in A549 cells. SRE reporter assay (measuring ERK activity, see methods) shows that introducing EpCAM mutations increased ERK activity. F In loss-of-function models, NSCLC cell lines (A549, H1299, H460, H226, and H23) expressing EpCAM mutations demonstrate increased cell death. G Stable A549 cells treated with Trametinib for 72 h. For treatment, cell lines were treated with 200–500-nM Trametinib for 72 h. Cell death was monitored by annexin-V staining. H. Structural disruption of EpCAM due to somatic mutation drives aberrant ERK pathway activation, resulting in enhanced cellular proliferation. Pharmacologic inhibition of MEK presents a rational therapeutic strategy to counteract this oncogenic signaling cascade

To delineate the roles of EpCAM expression vs mutation, we first assessed EpCAM expression across a panel of lung cancer cell lines, focusing on epithelial characteristics. As shown in Figure S6B, EpCAM expression was predominantly observed in cells expressing epithelial markers established in NSCLC [45]. The panel revealed that the cell lines A549, H1299, H460, H226, and H23 showed undetectable levels of EpCAM (also tested in the lab, data not shown). Building on our prior report that EpCAM can influence ERK1/2 signaling [9]. We used retroviral constructs to transduce A549 cells with EpCAM-WT and seven EpCAM mutations characterized in Fig. 4. Serum stimulation of the cells showed a marked decrease in ERK activity in EpCAM-WT cells compared to those with EpCAM mutations (Fig. 6E). These results suggests that EpCAM mutation increased ERK signaling. To investigate the role of EpCAM mutations and drug targeting, in drug resistance, we transfected A549, H1299, H460, H226, and H23 cells, which harbor KRAS, p53, or LKB1 mutations (Fig. 6F). Twenty-four hours post-transfection, the cells were treated with 200–500 nM Trametinib for 72 h, followed by a cell viability assay. The results revealed that EpCAM mutations differentially activated ERK1/2 signaling, sensitizing cells to drug-induced apoptosis (Fig. 6 F,G). Notably, when correlating ERK activation with apoptotic sensitivity, we observed that EpCAM mutants exhibiting higher sensitivity to apoptosis were associated with altered localization and stabilization [18]. These findings suggest that EpCAM mutations may enhance responsiveness to MEK inhibitors. In conclusion, EpCAM mutations promote increased ERK signaling and enhance the sensitivity of lung cancer cells to MEK inhibitors, highlighting their potential as therapeutic targets to improve treatment outcomes.

Discussion

Cancer-related mutations that regulate critical cellular processes, including proliferation, apoptosis, DNA repair, and cell signaling, are classified as oncogenes, tumor suppressor genes, or DNA repair genes. While a single genetic mutation is rarely sufficient to cause cancer, the accumulation of multiple genetic changes both in relation to one another and over time creates opportunities to detect cancers at much earlier stages [46]. Membrane proteins play a crucial role in facilitating intercellular communication and transmitting signals within cells by mediating protein interactions and regulating downstream cellular processes. Due to their pivotal functions, membrane proteins are often key targets for drug development and therapeutic interventions. As a tumor antigen, EpCAM overexpression is linked to several cancer cells related signaling [9, 11, 13–16, 19]. The goal of our study was to catalog cancer-related mutations of EpCAM, characterize their features, and highlight their significance to the research community for further investigation.

In this study, through the integration of multiple cohort datasets to catalog all EpCAM mutations, along with in-silico tools, advanced molecular modeling, and experimental validation, we reached several key conclusions. First, our study leveraged newly curated genomic datasets to identify several novel and recurrent EpCAM mutations across multiple cancer cohorts. Despite the relatively low frequency of somatic EpCAM mutations (ranging from 0.1–0.6%), our findings underscore the biological significance of these mutations in epithelial cancers. Second, TMB and survival analyses suggest that EpCAM mutation status could serve as an important marker for prognostic evaluations, personalized therapy strategies, and suggests role of EpCAM mutations in early or late stage of tumorigenesis. Third, for the first time, we demonstrated that mutations in the EpCAM gene destabilize its protein structure, particularly affecting the TY-1 loop and the RCD region. Homology modeling and MD simulations highlighted how these structural destabilizations compromise stability of EpCAM and dynamics, which are critical for its biological functions. Fourth, EpCAM mutations may potentially disrupt protein–protein interactions with essential cellular proteins, including CTSL [18], claudin [47], EGFR [48], PKC [49] occludins, catenins, MUC1, CD44 and CD63 [50–52]. EpCAM has been recognized as a novel component of tetraspanin-enriched microdomains (TEMs), forming a primary complex with the tetraspanin CD9 [53, 54]. These interactions are critical for maintaining cellular integrity, adhesion, and signaling processes. Loss of efficient binding due to structural instability could lead to dysfunctional cellular signaling pathways and tumor-promoting mechanisms. Fifth, our data provide preliminary evidence that some EpCAM mutations, such as C66Y and L240A localized in cytosol instead of membrane can potentially interfere with proto-oncogenes EGFR or KRAS signaling [19]. Finally six, Structural impairments caused by EpCAM mutations may explain the partial success of past and potential future antibody therapies targeting EpCAM [55].

In an overall study design, we performed comprehensive analysis of integrated data from major cancer genomics resources, including cBioPortal, COSMIC, AACR-GENIE, and Foundation Medicine-NCI (FMI) (Fig. 2A-C). We identified 160 new somatic mutations in addition to our previous work [18] The data suggests with missense mutations being the most common (79%), followed by nonsense mutations (6%), splice site mutations (7%), frameshift mutations (4%), deletions (2%), and fusion mutations (Figure S1C). Notably, MSH2-EpCAM fusions were identified in esophageal adenocarcinoma, squamous cell carcinoma, and stomach adenocarcinoma, aligning with previous reports of exon 9 deletions in EpCAM leading to MSH2 promoter hyper-methylation and gene silencing [56]. We also identified EpCAM-ABCG8 and NUP42-EpCAM fusions in high-grade uterine and ovarian carcinomas. In terms of mutation prevalence, EpCAM mutations were most common in lung cancer, skin cancer, and colorectal cancer, with pancreatic cancer also showing a notable concentration of mutations. (Fig. 2B, Figure S2).

For molecular dynamic simulation of EpCAM mutants, the cis-dimer crystal structure PDB: 4MGV [33] guided us to focus key regions for possible for the loss of function. The N-terminal domain of EpCAM, a critical region containing functional domains such as the EGF-like domain, RCD and TY-1 domain. Our previous work demonstrating how EpCAM mutations at C66Y and L240A altered the cellular phenotype in cancer cells was validated. MD simulations revealed significant structural destabilization in these mutants. For example, EpCAM-L240A exhibited heightened dynamic instability, especially in the TY-1 loop and RCD regions, resulting in a less stable structure compared to the EpCAM-WT. On the other hand, EpCAM-C66Y, while undergoing moderate conformational changes, significantly disrupted the native conformation of the TY-1 loop (Fig. 3). Next, we focused on studying seven lung cancer-associated mutations we discovered (Fig. 4, S4, S5). Results revealed that these mutations induced significant structural fluctuations, particularly in the TY-1 loop and the RCD region. Among the analyzed mutants, EpCAM-Q204E and EpCAM-P84S showed the most pronounced instability, characterized by higher root mean square deviation (RMSD) values, indicating substantial loss of structural integrity. These mutations also caused marked conformational changes in the TY-1 loop. Specifically, the structural shift observed in EpCAM-Q204E likely hinders its interaction network, undermining its role in tumor suppression and facilitating cell migration. Similarly, the RCD region a possible hotspot for maintaining EpCAM’s structural stability and protein-binding capacity, exhibited notable disruptions in mutant EpCAM-D92V. We predict that structural changes in RCD region could alter interactions with key molecular partners, such as claudins, occludins, and catenins, which are crucial for maintaining epithelial integrity and preventing tumor dissemination [3, 57]. Interestingly, while mutations EpCAM-Y186C and EpCAM-P84S demonstrated more moderate conformational changes, their altered dynamics were still sufficient to affect EpCAM’s function. For example, the ability of EpCAM to inhibit cathepsin-L through-TY-1 loop may be impaired even by subtler alterations in structure [18]. These findings pave the way for future investigations aimed at experimentally validating these mutations' effects on EpCAM's biochemical activities and downstream signaling.

Building on our previous study, we postulate that some EpCAM mutations, in conjunction with RAS, EGFR oncogenes could have detrimental effects on cancer cell growth and contribute to drug resistance [19]. Figure 5A data of colon cancer cohort shows that somatic mutations in EpCAM often coexist with mutations in key oncogenes such as KRAS, BRAF, TP53, and EGFR. Patients with altered/mutation EpCAM exhibited significantly worse survival outcomes in colon cancer A similar trend was observed in hepatocellular carcinoma (HCC), where EpCAM mutations co-occurred with mutations in TP53, MDM2, DNMT1, and KEAP1 (Fig. 5B), and patients with EpCAM mutations exhibited significantly worse overall survival compared to those with unaltered EpCAM expression. For lung cancer cohort, highest frequency of EpCAM mutations was observed in non-small cell lung cancer (NSCLC). Somatic mutations in the EpCAM gene often co-occurred with mutations in key oncogenes, such as KRAS, EGFR, TP53, and BRAF (Fig. 6, S6). Among over 1,000 identified mutations in the EpCAM gene, 283 were somatic mutations (Fig. S2). Surprisingly overall survival did not differ significantly between patients with EpCAM mutations and those without across both cohorts (Fig. 6B-C). This finding prompted further investigation into the potential role of tumor mutation burden (TMB) and the co-occurrence of EpCAM mutations. Our results suggest that EpCAM mutations, when found in conjunction with a lower TMB, may play a more prominent role in the early stages of tumor development rather than in later-stage disease progression.

To validate our in-silico and MD simulations findings, we conducted an experimental approach to support the data. EpCAM is known to alter ERK signaling [9]. ERK reporter assay indicated that cells with EpCAM mutations exhibited increased ERK1/2 activity, which was sensitized by Trametinib treatment (Fig. 6E, F). Further analysis revealed that cells transfected with EpCAM mutations showed a correlation with Trametinib-induced apoptosis and ERK activity (Fig. 6G). In such cases, aberrant activation of the ERK pathway in tumors may foster an immunosuppressive tumor microenvironment, thereby promoting immune resistance and facilitating immune evasion. In conclusion, EpCAM mutations in cancer cells may lead to context-dependent alterations in protein structure and stability. Mutations in the functional domain of EpCAM, which play critical roles in cell signaling, can serve as valuable indicators for biomarker development and drug discovery. Our findings demonstrate that specific EpCAM mutations lead to structural destabilization and altered signaling, localization, which can be directly linked to patient outcomes and therapeutic response. These mutation-driven alterations can be integrated into biomarker pipelines to stratify patients based on risk or treatment sensitivity. Furthermore, the enhanced sensitivity of certain EpCAM mutations to MEK inhibitors highlights their potential as actionable targets in drug development, supporting personalized and mutation-guided cancer therapies.

Conclusion

Our study highlights the functional and structural consequences of EpCAM mutations in epithelial cancers, revealing their role in tumor progression, reduced patient survival, and altered drug sensitivity. The findings demonstrate that mutations in TY-1 and RCD regions destabilize EpCAM, potentially disrupting its adhesive and signaling functions. EpCAM-mutant lung cancer cells exhibit increased sensitivity to MEK inhibitors, suggesting a novel therapeutic avenue. These results emphasize EpCAM mutations as potential biomarkers for cancer prognosis and treatment strategies, positioning EpCAM as a promising therapeutic target in epithelial malignancies.

Abbreviations

CCLE

Cancer Cell Line Encyclopedia

CTD

C-terminal domain

CTSL

Cathepsin-L

DCCM

Dynamic cross-correlation matrix

DSSP

Defined secondary structure protein analysis

EpCAM

Epithelial cell adhesion molecule

H-bonds

Hydrogen bonds

HCC

Hepatocellular carcinoma

MD

Molecular dynamics

NSCLC

Non-small cell lung cancer

NTD

N-terminal domain

PolyPhen

Polymorphism phenotyping

RCD

Ridge on the C-terminal domain

RMSD

Root mean square deviation

RMSF

Root mean square fluctuation

SASA

Solvent accessible surface area

SIFT

Sorting intolerant from tolerant

TMB

Tumor mutation burden

TYD

Type-1A-like domain

Authors’ contributions

NVS, KDS and WEG were responsible for study conception, design, and supervision manuscript writing. NVS performed the in vitro experiments. ABL generated EpCAM mutation data and analyzed. PSD and KS performed Homology modeling and MD simulation. TPB, RMB supervised and write manuscript. All authors have read and approved the manuscript.

Funding

St Joseph Hospital Foundation supported the research work.

Data availability

The datasets used and/or analyzed during the current study are available as data S2A, S2B and S2C in supplementary.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

Authors declare no conflict of interest.