Computational and Structure-Guided engineering of colicin E1 and E2 for EMT -driven cadherin selectivity in cancer

Graphical abstract

Highlights

• •First EMT-aware computational design strategy targeting cadherin switching in metastasis.
• •Discovers intrinsic EMT-state–dependent cadherin selectivity of bacterial colicins.
• •Establishes stability-driven colicin engineering for EMT-specific cadherin targeting.
• •Identifies Colicin E2 (A579E) as a novel cadherin–selective anti-metastatic lead.
• •Provides a new in silico paradigm for EMT-guided biologic cancer therapy design.

Abstract

Epithelial–mesenchymal transition (EMT) is a fundamental driver of cancer invasion, metastasis, and therapeutic resistance, mediated by dynamic switching between E-cadherin (CDH1) and N-cadherin (CDH2). Although cadherins are central regulators of EMT, current therapeutic strategies rarely exploit EMT-state–specific cadherin dependencies. In this study, we present a structure-driven, EMT-aware computational framework to repurpose and engineer colicins as cadherin-selective anti-metastatic biologics. Pan-cancer transcriptomic profiling across thirty-three tumor types revealed distinct CDH1- and CDH2-dominant EMT landscapes, providing a rational basis for receptor-informed targeting. stereo chemically validated structures of Colicin E1 and Colicin E2 were subjected to stability-guided mutagenesis using FoldX, followed by protein–protein docking and atomistic molecular dynamics simulations with E- and N-cadherin. Docking analyses demonstrated intrinsic cadherin preferences, with Colicin E2 exhibiting stronger affinity for CDH1 and Colicin E1 favouring CDH2, mirroring EMT-associated cadherin switching. Molecular dynamics simulations further confirmed these trends, revealing stable complex formation, reduced backbone deviation, sustained interfacial contacts, and distinct residue-level flexibility profiles in preferred colicin–cadherin pairs. Structure-guided mutations significantly enhanced binding stability and specificity, identifying Colicin E2 mutant A579E as the most optimized variant, characterized by improved docking scores, reduced conformational fluctuations, and persistent E-cadherin engagement throughout simulations. Compared with prior EMT-targeting approaches, this work uniquely integrates transcriptomic context with structure-guided protein engineering to achieve cadherin-state selectivity rather than broad inhibition. Collectively, this study establishes a novel computational paradigm for EMT-state-guided biologic design and provides a predictive foundation for future experimental validation, cadherin-resolved cancer stratification, and translational development of precision anti-metastatic protein therapeutics in diverse solid tumor contexts clinically.

1. Introduction

Escherichia coli (E. coli) is a Gram-negative bacterium recognized both as a commensal organism and a versatile pathogen.[1] Beyond its significance in infectious disease, E. coli has served as a foundational model in molecular biology due to its well-characterized plasmid systems. Among these, the ColE1 plasmid has played a pivotal role in the development of cloning and expression vectors. Originally encoding colicin E1, the plasmid is a small, high-copy-number circular DNA molecule that replicates efficiently—up to 40 copies per cell—and is easily manipulated and transferred through transformation.[2], [3] The cea gene within ColE1 encodes colicin E1, and although modern expression vectors typically replace this gene with selectable markers such as bla (ampicillin resistance), the ColE1 backbone remains central to genetic engineering and synthetic biology[4].

In natural environments, colicins function as antimicrobial toxins produced by E. coli to eliminate competing bacterial strains during stress, particularly under the SOS response. Colicins are high-molecular-mass proteins (40–80 kDa) characterized by a simple genetic organization composed of three core genes: a cytotoxic colicin gene, an immunity protein gene, and a lysis gene that mediates toxin release through controlled cell lysis.[5], [6]) Once internalized, colicins can exert cytotoxic effects through distinct mechanisms, including membrane pore formation (colicin E1) and RNA degradation (colicin E2), ultimately leading to disruption of cellular homeostasis. These effects may influence key EMT-related signalling pathways, contributing to inhibition of tumor cell proliferation, migration, invasion, and metastasis. Structurally, colicins share a conserved three-domain architecture: an N-terminal translocation domain, a central receptor-binding domain, and a C-terminal cytotoxic domain—each required for efficient import and toxicity.[7] Colicin E1 operates as a pore-forming toxin, inserting into the target cell’s inner membrane to create voltage-dependent ion channels. This leads to proton influx, disruption of the proton motive force (PMF), ATP depletion, and bactericidal death. Remarkably, a single molecule of colicin E1 is sufficient to kill a susceptible bacterial cell.[8] In contrast, colicin E2 functions as a nonspecific DNase; once internalized into the cytoplasm, it cleaves genomic DNA, leading to rapid chromosome degradation and cell death.[9]These potent mechanisms highlight the evolutionary efficiency and specificity of colicin-mediated killing.

Recent research has extended the relevance of colicins beyond microbiology, demonstrating that E-type colicins also possess selective cytotoxicity against cancer cells. Colicin E1, for example, exhibits strong inhibitory effects on HT-29 colon cancer cells while sparing normal epithelial cells This selectivity may be driven by unique lipid compositions, altered membrane potential, receptor overexpression, and enhanced oxidative susceptibility in malignant cells. As a result, E-type colicins have gained attention as potential anticancer peptides with intrinsic tumor-targeting capabilities[10], [11].

Parallel to these discoveries, the Epithelial-to-Mesenchymal Transition (EMT) has emerged as a critical biological process underlying cancer progression, metastasis, and therapeutic resistance. EMT involves the loss of epithelial polarity and cell–cell adhesion, accompanied by acquisition of mesenchymal traits such as motility and invasiveness.[12], [13], [14] A defining event in EMT is the cadherin switch, marked by downregulation of E-cadherin (CDH1) and upregulation of N-cadherin (CDH2).[15], [16] Loss of E-cadherin disrupts adherens junctions and promotes β-catenin nuclear translocation, activating oncogenic Wnt signalling. Conversely, increased N-cadherin stabilizes cell migration, supports angiogenesis, and enhances cancer cell survival via interactions with FGFR and PDGFR.[17], [18] Given its central role in metastasis, EMT—particularly its cadherin regulators—has become an attractive therapeutic target. Bioinformatics studies using transcriptomics, interaction network mapping, and pathway enrichment analyses have further revealed the complexity of EMT regulation across cancer types.[19], [20], [21] These multi-omic insights emphasize the need to identify molecules that can modulate cadherin-dependent signalling, potentially reversing EMT to preventing metastasis.

Recent evidence suggests that colicins may interact with eukaryotic membranes and receptors[22]raising the intriguing possibility that they may influence cadherin-mediated signalling. Because cadherins govern EMT, targeting these proteins represents a compelling strategy for precision oncology. The ability of bacterial peptides such as colicin E1 and E2 to selectively impact malignant cells positions them as promising candidates for anti-metastatic therapy. Importantly, structure-guided mutagenesis of colicin E1 and E2 offers an opportunity to enhance their receptor specificity, therapeutic potency, and tumour selectivity. Assessing the stability and conformational dynamics of wild-type versus mutant colicins through molecular modelling, docking, and molecular dynamics simulations can help identify variants with improved binding affinity for cancer-associated cadherins for disruptive properties (Fig. 1). Such engineered mutations could ultimately produce more stable, more precise, and more effective colicin-based therapeutic agents capable of targeting EMT-driven cancers.

Fig. 1.

This figure compares the wild-type colicin (Col-Wt) with engineered mutants of Col-E1 and Col-E2, highlighting changes in residue-level interactions, cadherin binding, membrane engagement, and apoptosis induction. Mutant profiles show enhanced binding strength, improved pore-forming ability, and greater tumor-selective cytotoxicity compared to the wild type. Overall, the schematic illustrates how specific mutations shift colicin activity from a natural antimicrobial role toward potent anticancer functionality.

Statement of Significance.

Aspect	Statement
Problem or Issue	Metastasis and therapy resistance in cancer are driven by epithelial–mesenchymal transition (EMT), yet there are limited computational strategies that exploit EMT-specific molecular features, such as cadherin switching, to guide rational therapeutic design. Existing biologics rarely account for EMT heterogeneity or receptor-state specificity, limiting their precision and translational potential.
What is Already Known	Colicins are potent bacterial toxins with well-defined structures and mechanisms of action. E-type colicins have shown selective cytotoxicity toward cancer cells, while EMT-associated cadherin dysregulation (CDH1/CDH2) is a validated hallmark of metastasis. However, their intersection has not been systematically explored.
What This Paper Adds	This study introduces an EMT-aware, structure-guided computational framework that integrates pan-cancer transcriptomics, protein engineering, docking, and molecular dynamics to design cadherin-selective colicin variants with enhanced stability and specificity.

2. Application of cadherin patterns in Colicin therapy design

2.1. STRING network

The STRING protein–protein interaction (PPI) analysis, performed using STRING v12.0,[23] revealed contrasting interaction signatures for Colicin E1 and Colicin E2, highlighting their distinct mechanisms for crossing the bacterial envelope. The Colicin E1 network is dominated by a TolC-centered cluster, where TolC acts as the major interaction hub linked to multidrug efflux components such as AcrA, AcrB, AcrD, AcrF, MacA, MacB, EmrA, EmrB, MdtA, and MdtF.[24] This clustering suggests that Colicin E1 hijacks TolC-dependent efflux machinery to facilitate its translocation through the outer membrane. In contrast, the Colicin E2 network is organized around the Tol–Pal complex, with TolA forming strong connections with TolQ, TolR, TolB, and Pal, supported by additional interactions with OmpF, YbgC, YbgF, Skp, ZapC, and LptC.[25] These patterns emphasize that Colicin E2 relies on energy-coupled Tol–Pal trans-envelope pathways for import. Together, these networks demonstrate a mechanistic divergence where Colicin E1 primarily exploits efflux-associated channels, while Colicin E2 utilizes structural trans-envelope complexes for entry(Fig. 2).

Fig. 2.

The comparative string interaction networks of bacterial and human adhesion-related proteins. (a) shows the densely interconnected efflux transport system associated with TolC, highlighting key components such as AcrB, MacA, and EmrB involved in membrane permeability and toxin export. (b) depicts the TolA-associated membrane translocation machinery, including OmpF, Pal, Skp, and Lpp, which together mediate colicin uptake across the bacterial envelope. (c) presents the human cadherin interaction network, where CDH1 and CDH2 form central nodes linked to catenins (CTNNB1, CTNNA1, CTNND1) and structural regulators (JUP, VCL, PSEN1), reflecting the adhesion and EMT-related signaling hub relevant to cancer progression.

Interestingly, the bacterial transport strategies utilized by colicins E1 and E2 show conceptual parallels with cancer cell biology, particularly in membrane permeability, receptor-mediated uptake, and adhesion protein dynamics. Just as TolC- and TolA-dependent pathways regulate selective membrane passage in bacteria, cancer cells modulate transport channels and cadherin expression—most notably E-cadherin (CDH1) and N-cadherin (CDH2)—to influence invasion, metastasis, and therapeutic responsiveness. Emerging evidence that colicins can interact with eukaryotic membranes and cadherins highlights a mechanistic convergence between microbial toxin entry and tumor cell adhesion remodelling.[25], [26] The STRING protein–protein interaction (PPI) network further illustrates the functional landscape of CDH1 and CDH2, revealing their roles as central nodes within the cadherin–catenin complex through interactions with CTNNB1 (β-catenin), CTNNA1 (α-catenin), CTNND1 (p120-catenin), and JUP (plakoglobin). Additional partners such as VCL, PSEN1, and CDH3 underscore the interconnected signalling architecture that governs epithelial stability and EMT progression. The dense connectivity reflects overlapping yet distinct contributions of CDH1 and CDH2 to cell adhesion and migration, with CDH1 preserving epithelial integrity while CDH2 drives mesenchymal transition and metastatic behaviour. Together, these comparisons suggest that understanding how colicin E1 and E2 navigate bacterial envelope barriers and recognize specific receptors may inform the development of colicin-derived anticancer peptides capable of targeting cadherin-mediated adhesion, disrupting tumor membrane structure, or selectively penetrating the tumor microenvironment.

2.2. Structural quality assessment of Colicin E1 and Colicin E2

The three-dimensional structures of Colicin E1 and Colicin E2 were retrieved from the RCSB Protein Data Bank (PDB IDs: 2I88 and 3U43, respectively) using the PDB advanced search interface Coordinates were downloaded in PDB format and inspected in PyMOL v2.x (Schrödinger LLC) for completeness, chain integrity, and overall fold visualization.[27], [28] Unresolved regions were observed in the functional domains used for downstream modelling. Stereochemical quality was evaluated with PROCHECK via the EMBL-EBI structure validation server, which calculates backbone ϕ/ψ distributions, side-chain χ1 geometry, and overall G-factors relative to high-resolution crystallographic standards.Ramachandran plots and associated statistics (percentage of residues in most favoured, additionally allowed, generously allowed, and disallowed regions) were exported directly from PROCHECK and used to assess model reliability [29].

The structural comparison of Colicin E1 (PDB: 2I88) and Colicin E2 (PDB: 3U43) reveals clear architectural distinctions that align with their divergent mechanisms of translocation and cytotoxicity. Colicin E1 adopts a compact α-helical architecture; however, quality assessments revealed moderate structural strain, with a subset of residues populating additionally allowed or disallowed regions of conformational space, indicative of increased backbone flexibility and localized instability. Such dynamic features may enable transient receptor interactions but could limit the persistence of cadherin engagement within the mechanically and chemically heterogeneous tumor microenvironment. In contrast, Colicin E2 displays a more ordered and well-aligned structural organization, with a higher proportion of residues occupying favored regions of the Ramachandran map and minimal stereochemical outliers, reflecting superior backbone geometry and reduced angular strain. Residue-wise validation further confirmed uniform structural stability across the Colicin E2 fold. This enhanced stereochemical integrity suggests a rigid and energetically favorable scaffold capable of sustaining stable receptor binding under physiological conditions. Collectively, these structure validation results indicate that although both wild-type colicins are structurally competent, Colicin E2 exhibits superior conformational reliability and stability, supporting its preferential selection as a foundational scaffold for cadherin-targeted cancer therapy and for subsequent mutational or structure-guided optimization aimed at enhancing anticancer efficacy while minimizing nonspecific interactions(Fig. 3).

Fig. 3.

Structural Comparison and Stereochemical Validation of Colicin E1 and Colicin E2 (a) The three-dimensional structure of wild-type Colicin E1 displays a compact α-helical fold with localized deviation markers, reflecting moderate conformational strain within the protein scaffold. (b) Ramachandran analysis of Colicin E1 indicates that most residues occupy favoured and additionally allowed regions, with a small fraction in disallowed regions, suggesting increased backbone flexibility. (c) In contrast, the wild-type Colicin E2 structure exhibits a more extended and well-aligned architecture with fewer deviation markers, indicative of improved structural organization. (d) Consistent with this, the Ramachandran plot for Colicin E2 shows a higher proportion of residues clustered in favoured regions and minimal outliers, confirming superior stereochemical quality.

2.3. Pan-cancer expression profiling of CDH1 and CDH2 across human malignancies

Pan-cancer transcriptomic analysis of TCGA datasets reveals a robust and biologically coherent dichotomy in CDH1 (E-cadherin) and CDH2 (N-cadherin) expression across human malignancies, capturing the classical cadherin switch that underpins epithelial–mesenchymal transition (EMT) and metastatic progression. Epithelial-derived tumors such as BRCA, COAD, LUAD, STAD, and ESCA retain high CDH1 expression, reflecting preserved cell–cell adhesion and epithelial identity, whereas renal cancers, gliomas, sarcomas, melanoma, and EMT-driven carcinomas exhibit marked CDH1 suppression concomitant with elevated CDH2 levels, indicative of mesenchymal reprogramming and invasive potential. These reciprocal expression landscapes highlight pronounced cadherin heterogeneity both across and within tumor lineages, underscoring the necessity of receptor-informed therapeutic targeting. Within this framework, the CDH1/CDH2 axis provides a rational molecular basis for colicin-based precision oncology: tumors with high CDH1 expression present stable epithelial cadherin interfaces suitable for engagement by wild-type or epithelial-optimized colicin variants, whereas CDH1-low/CDH2-high EMT-shifted cancers represent ideal candidates for engineered colicins designed to selectively recognize mesenchymal cadherin states. Importantly, this cadherin-resolved stratification aligns directly with our in silico docking and molecular dynamics findings, which demonstrate distinct binding preferences and stability profiles of wild-type versus engineered colicins toward CDH1 and CDH2(Fig. 4). Together, integration of TCGA pan-cancer cadherin expression data with structure-guided colicin modelling establishes CDH1 and CDH2 as functionally actionable biomarkers, enabling rational selection, optimization, and deployment of colicin-derived therapeutics tailored to tumor EMT status and metastatic behaviour.

Fig. 4.

Expression landscape and prognostic significance of EMT-linked cadherins across cancers. (a–b) Transcript per million (TPM) profiles illustrate the differential expression patterns of CDH1 (a) and CDH2 (b) across multiple cancer types, highlighting tumor-specific variability and opposing expression trends consistent with EMT dynamics. (c–d) Boxplot analyses compare CDH1 (c) and CDH2 (d) expression between tumor and corresponding normal tissues, revealing significant deregulation and context-dependent cadherin switching across malignancies. (e–f) Kaplan–Meier overall survival analyses demonstrate the prognostic relevance of cadherin expression, where high CDH1 expression (e) associates with improved survival, while elevated CDH2 expression (f) correlates with poorer outcomes, as supported by hazard ratios and log-rank statistics.(g) Pan-cancer heatmap summarizing differential expression scores of CDH1 and CDH2 across cancer cohorts further underscores their inverse regulation and clinical relevance in EMT-driven tumor progression.

3. Methodology

3.1. Structural Preparation

The wild-type structures of Colicin E1 and Colicin E2 were retrieved from the RCSB PDB database.[30] For Colicin E1, we used PDB ID 6WXI (the nano disc-embedded TolC–Colicin E1 fragment) RCSB PDB; for Colicin E2, we used PDB ID 3U43, which corresponds to the colicin E2 DNase domain complex as a starting structural template RCSB PDB. Where full-length models were incomplete (e.g. missing loops or terminal segments), missing regions were built using homology modelling (Swiss-Model) to reconstruct full-length sequences suitable for computational work.[31]) All initial models were visually inspected using PyMOL to confirm correct chain continuity, residue numbering, and to ensure absence of gross structural distortions.[32] Prior to any mutational modelling or docking, each wild-type colicin model was refined using the Repair PDB module of FoldX v5.0, which optimizes side-chain rotamers, rebuilds missing atoms, corrects incomplete residues, and removes steric clashes and unfavourable contacts that could distort energetic calculations.[33] Hydrogens were added and protonation states adjusted where needed. This produced energetically minimized, stereo chemically validated wild-type colicin structures; these served as the uniform baseline for all subsequent mutation modelling, binding energy calculations, and docking simulations, thus preventing spurious artifacts from structural flaws in initial models [34].

3.2. Mutation Generation and stability Prediction

Point mutations were generated using the Build Model command in FoldX, which systematically constructs rotamer-optimized mutant structures by sampling side-chain conformations within the local environment of the substituted residue. For each mutation, five independent Build Model runs (number Of Runs = 5) were performed to ensure statistical reliability and reduce noise arising from rotamer variability. FoldX computes the total energy for both the mutant and the repaired wild-type reference structure, producing Raw, Difference, and Average energy output files.[33]Mutation-specific stability changes (ΔΔG_fold) were extracted from the Average file, calculated as:

$$\mathrm{\Delta }\mathrm{\Delta }Gfold=\mathrm{\Delta }Gmutant-\mathrm{\Delta }Gwild-type$$

Negative ΔG fold values indicate stabilizing mutations, whereas positive values reflect destabilization. This procedure ensures that reported ΔΔG values represent converged, energetically minimized predictions, independent of stochastic fluctuations from single rotamer sampling.

3.3. Complex stability evaluation using FoldX Analyse complex

To quantify the effect of each point mutation on cadherin–colicin interactions, we employed the Analyse Complex function in FoldX, which calculates the total interaction energy between the colicin variant and its cadherin partner. For every mutant–cadherin complex generated, FoldX first computed the folding energy of each individual chain and the inter-chain binding energy (ΔΔG_bind) of the complex. Mutation-induced changes in binding affinity (ΔΔG_bind) were then obtained by subtracting the wild-type interaction energy from the corresponding mutant value, with positive ΔΔG_bind indicating weakened binding and negative values reflecting enhanced interaction. In addition to overall binding free energy, FoldX provides detailed energetic decomposition, including electrostatics, van der Waals forces, solvation, hydrogen bonding, and entropic contributions, allowing us to pinpoint how each substitution perturbs interface stability. This unified workflow enabled a consistent, structure-based assessment of how single-site mutations modulate the strength and physio-chemical determinants of cadherin–colicin binding.[35] .

3.4. Structure-Based molecular docking of colicins with cadherins (CDH1/CDH2)

Protein–protein docking analyses were performed using HADDOCK 2.4, an integrative and information-driven docking platform that incorporates experimental data, bioinformatics-derived restraints, and flexible refinement to generate accurate macromolecular complex predictions[36])The extracellular domain structures of human E-cadherin (CDH1) and N-cadherin (CDH2) were retrieved from UniProt (The UniProt Consortium, 2023) and prepared using Auto Dock Tools 1.5.7, where missing hydrogens were added, Gasteiger–Marsili charges were assigned, and bond orders were standardized to ensure correct topology.[37], [38] Each receptor and ligand structure underwent pre-docking energy minimization using the GROMOS96 force field implemented in Swiss-PdbViewer (Deep View) to optimize backbone geometry, refine side-chain conformations, and eliminate steric clashes, thereby providing energetically favourable starting coordinates for subsequent docking. Binding pockets and surface-accessible cavities on cadherins were predicted using the geometry-based FPocket algorithm, which identifies ligandable regions through Voronoi tessellation and alpha-sphere clustering; top-ranked pockets were selected based on pocket score, volume, polarity, and hydrophobicity, and residues forming these pockets were designated as active interaction residues for HADDOCK.[36], [39] Wild-type and FoldX-engineered colicin variants, generated through Build Model and validated via structural repair and minimization were used as ligands for docking against the cadherin active sites.[33], [37] .

Docking simulations followed the standard HADDOCK workflow consisting of rigid-body docking, semi-flexible simulated annealing, and explicit solvent refinement. A total of 1,000 initial rigid-body complexes were generated, followed by refinement of the top 200 structures during the semi-flexible stage, and final selection of the best 50 models refined in explicit TIP3P water. Complexes were scored using the HADDOCK weighted scoring function, which integrates van der Waals energy, electrostatic energy, desolvation contributions, and ambiguous interaction restraints, and were clustered based on interface RMSD using a 0.6 Å cutoff to identify the most consistent binding modes. Interface analysis of the refined complexes was performed using BIOVIA Discovery Studio Visualizer for 3D mapping of molecular contacts, while LigPlot + was used to generate 2D diagrams illustrating hydrogen bonds, salt bridges, and hydrophobic interactions between colicin variants and cadherins Buried surface area, hydrogen bond persistence, and residue-level interaction profiles were quantified from the HADDOCK output to characterize the stability and specificity of the predicted interfaces, enabling a detailed understanding of colicin–cadherin binding behaviour [32], [36], [40].

3.5. Molecular dynamics simulation of Colicin–Cadherin interactions

Molecular dynamics (MD) simulations were performed to investigate the conformational stability and intermolecular interaction dynamics of Colicin E1 and Colicin E2, including their wild-type and mutant variants, in complex with E-cadherin (CDH1) and N-cadherin (CDH2). All simulations were executed using the Desmond MD engine[41]within the Schrödinger Suite. Each protein–protein complex was parameterized using the OPLS4 force field, which offers improved accuracy for biomolecular systems.[42]Systems were solvated in an orthorhombic box using the TIP3P water model[43]with a 10 Å buffer, neutralized with appropriate Na⁺ and Cl⁻ counterions, and ion strength was adjusted to maintain electrostatic balance for both wild-type and mutant constructs. Long-range electrostatics were computed using the Particle Mesh Ewald (PME) algorithm[44]with a 9 Å cutoff for electrostatic and van der Waals interactions. After initial minimization and equilibration, all systems were simulated under NPT ensemble conditions at 300 K and 1 atm, regulated by the Nosé–Hoover thermostat (Nosé, 1984; Hoover, 1985) and Martyna–Tobias–Klein barostat.[45]Production trajectories were run for 25 ns using a 2-fs integration timestep, and structural snapshots were saved every 10 ps. Identical simulation protocols were applied across all Colicin E1 and E2 wild-type and mutant complexes to ensure rigorous comparison of their dynamic behaviour, enabling high-confidence evaluation of mutation-induced effects on cadherin binding stability.

4. Result and discussion

The computational framework applied in this study provides a robust validation of colicin–cadherin interactions, accurately reflecting biological patterns observed across EMT-associated CDH1/CDH2 expression landscapes. STRING and stereochemical analyses confirmed the structural and mechanistic divergence of Colicin E1 and E2; while docking and MD simulations further demonstrated their differential affinity toward epithelial (CDH1) and mesenchymal (CDH2) cadherins. These trends align closely with the TCGA pan-cancer cadherin profiles, reinforcing that our in-silico pipeline reliably models EMT-driven receptor switching. Wild-type colicins exhibited distinct binding biases—E2 favouring CDH1 and E1 favouring CDH2—whereas engineered mutants enhanced stability and strengthened interface complementarity for their target cadherins. Together, these findings validate the predictive power of our approach and highlight engineered colicins as promising candidates for cadherin-guided, EMT-informed anticancer therapy.

4.1. Wild-Type docking validation results

Docking and binding-affinity analyses revealed distinct and biologically coherent patterns of cadherin specificity between wild-type Colicin E1 and Colicin E2. Colicin E2 exhibited the strongest interaction with CDH1, supported by the lowest HADDOCK score (−119.3 ± 16.2), minimal RMSD (1.0 ± 0.7 Å), large buried surface area (2858.2 ± 87.2 Å2), and favourable thermodynamics (ΔG =  − 11.6 kcal/mol; Kd = 3.1 × 10⁻⁹ M), together indicating a highly stable, tightly packed epithelial-cadherin interface enriched with extensive charged–charged, charged–apolar, and apolar–apolar contacts. In contrast, Colicin E1 showed a pronounced preference for CDH2, displaying a more favourable HADDOCK score (−63.0 ± 7.0) and the strongest overall binding free energy (ΔG =  − 13.7 kcal/mol; Kd = 8.7 × 10⁻11 M), driven by dominant electrostatic contributions and an interaction network reflecting the polar-rich and flexible architecture of N-cadherin. Although its RMSD was higher (66.3 ± 1.1 Å), the interaction remained energetically robust, consistent with CDH2′s dynamic role in EMT-associated adhesion. Colicin E2 binding to CDH2 and Colicin E1 binding to CDH1 showed intermediate or weaker stability, respectively, further underscored by their Z-scores and interaction fingerprints. Together, these results establish a clear intrinsic specificity: Colicin E2 preferentially targets epithelial cadherin CDH1, while Colicin E1 exhibits stronger affinity toward mesenchymal cadherin CDH2, mirroring EMT-associated cadherin switching in cancer and providing a rational foundation for designing cadherin-selective colicin therapeutics.

4.2. Comparison of docking results in the Wild-Type model with CDH1 and CDH2

Comparative docking and binding-energy assessments revealed a clear and biologically coherent pattern of cadherin selectivity between wild-type Colicin E1 and Colicin E2. Colicin E2 displayed the strongest affinity toward CDH1, characterized by an exceptionally favourable HADDOCK score (−119.3 ± 16.2), minimal structural deviation (RMSD 1.0 ± 0.7 Å), a large buried interface (2858.2 ± 87.2 Å,[2]and nanomolar binding free energy (ΔG =  − 11.6 kcal/mol; Kd = 3.1 × 10⁻⁹ M). The COL E2–CDH1 interface was dominated by robust hydrogen-bonding pairs (e.g., Arg156–Asp408, His150–Glu363), electrostatic complementarity, and hydrophobic packing, collectively supporting a highly ordered epithelial-cadherin–colicin assembly (Table 1) (Table 2).

Table 1.

Docking Results of Wild-Type Colicin E1 and E2 with CDH1 and CDH2.

Parameter	Colicin E1 – CDH1	Colicin E1 −CDH2	Colicin E2 – CDH1	Colicin E2 – CDH2
HADDOCK Score	−59.5 ± 14.0	−63.0 ± 7.0	−119.3 ± 16.2	−99.6 ± 16.2
Cluster Size	15	10	15	6
RMSD (Å)	2.2 ± 1.6	66.3 ± 1.1	1.0 ± 0.7	26.6 ± 0.2
Van der Waals Energy	−77.2 ± 7.3	−39.3 ± 14.5	−86.6 ± 3.3	−52.4 ± 9.8
Electrostatic Energy	−496.4 ± 71.4	−362.6 ± 83.0	−355.6 ± 37.1	−469.3 ± 71.2
Desolvation Energy	10.2 ± 11.0	8.6 ± 10.7	1.0 ± 4.6	23.5 ± 4.5
Restraints Violation Energy	1068.4 ± 142.0	402.5 ± 82.8	373.8 ± 164.9	231.2 ± 57.3
Buried Surface Area (Å2)	3426.1 ± 84.0	1973.1 ± 182.9	2858.2 ± 87.2	2437.9 ± 171.9
Z-Score	−1.8	−1.5	−2.7	−1.8
ΔG (kcal/mol)	−7.7	−13.7	−11.6	−10.9
Kd (M)	2.2 × 10⁻6	8.7 × 10⁻11	3.1 × 10⁻9	9.5 × 10⁻9
ICs charged–charged	16	27	22	21
ICs charged–polar	20	36	7	22
ICs charged–apolar	16	42	42	33
ICs polar–polar	1	10	4	6
ICs polar–apolar	3	21	8	12
ICs apolar–apolar	10	12	17	9
NIS charged	27.86	28.57	27.16	26.4
NIS apolar	42.72	41.97	42.36	43.87

Table 2.

Molecular Interaction Landscape of Colicin E1 E2 (wild type) Complexes with E-Cadherin (CDH1) and N-Cadherin (CDH2), Highlighting Key Hydrogen Bonds, Hydrophobic Contacts, and Electrostatic Interfaces.

Type of Interaction		Residue 1 (Chain A)	Residue 2 (Chain B)	Interaction Type
Colicin E1	CDH1	Gln255 (A)	Gln80 (B)	Hydrogen Bond
		Asp254 (A)	Arg157 (B)	Hydrogen Bond
		Arg150 (A)	Asn342 (B)	Hydrogen Bond
		Lys179 (A)	Ser368 (B)	Hydrogen Bond
	CDH2	Arg154 (A)	Asp350 (B)	Hydrogen Bond
		Arg151 (A)	Ser343 (B)	Multiple Hydrogen Bonds
		Asn144 (A)	Lys335 (B)	Hydrogen Bond
		Ser145 (A)	Lys158 (B)	Hydrogen Bond
Colicin E2	CDH1	Arg156 (A)	Asp408 (B)	Hydrogen Bond
		His150 (A)	Glu363 (B)	Hydrogen Bond
		Arg178 (A)	Glu161 (B)	Hydrogen Bond
	CDH2	Arg92 (A)	Asp576 (B)	Hydrogen Bond
		Arg154 (A)	Pro464 (B)	Hydrogen Bond
		Arg87 (A)	Glu309 (B)	Hydrogen Bond
		Arg65 (A)	Asp294 (B)	Hydrogen Bond
		Arg194 (A)	Gln308 (B)	Multiple Hydrogen Bonds
Colicin E1	CDH1	Thr174 (A)	Val170 (B)	Hydrophobic Interaction
		Val150 (A)	Leu164 (B)	Hydrophobic Interaction
	CDH2	Leu140 (A)	Val167 (B)	Hydrophobic Interaction
		Tyr144 (A)	Ile170 (B)	Hydrophobic Interaction
Colicin E2	CDH1	Ile149 (A)	Leu144 (B)	Hydrophobic Interaction
		Phe154 (A)	Ile170 (B)	Hydrophobic Interaction
	CDH2	Leu187 (A)	Val466 (B)	Hydrophobic Interaction
		Phe176 (A)	Leu240 (B)	Hydrophobic Interaction
		Val77 (A)	Ile240 (B)	Hydrophobic Interaction
		Tyr85 (A)	Ile215 (B)	Hydrophobic Interaction
Colicin E1	CDH1	Arg150 (A)	Asp183 (B)	Electrostatic Interaction
		Lys179 (A)	Asp150 (B)	Charge Interaction
	CDH2	Lys335 (B)	Asp145 (A)	Charge Stabilization
		Lys158 (B)	Glu167 (A)	Electrostatic Interaction
Colicin E2	CDH1	Arg178 (A)	Glu161 (B)	Electrostatic Interaction
		Lys362 (B)	Arg145 (A)	Polar–Electrostatic Interaction
	CDH2	Arg154 (A)	Glu309 (B)	Charge Interaction
		Arg173 (A)	Gly23 (B)	Polar Interaction
		Arg182 (A)	Asp576 (B)	Electrostatic Interaction

Conversely, Colicin E1 exhibited a pronounced preference for CDH2, supported by its most favourable binding free energy (ΔG =  − 13.7 kcal/mol; Kd = 8.7 × 10⁻¹¹ M) and a stable docking pose reinforced by a dense polar–electrostatic network, including Arg154–Asp350, Arg151–Ser343, and Lys335–Asp145. Hydrophobic contacts (Leu140–Val167, Tyr144–Ile170) further stabilized the interaction, reflecting the flexible, polar-rich architecture characteristic of N-cadherin(Fig. 5). In contrast, Colicin E1–CDH1 and Colicin E2–CDH2 complexes displayed modestly stable interactions (Table 2), with reduced interface area, fewer persistent hydrogen bonds, and less favourable energetics, consistent with their higher Z-scores and weaker interaction fingerprints.

Fig. 5.

This figure presents the comparative docking and MD-refined interaction landscapes of colicin variants with multiple cadherin targets. In (a), COL E1 shows a stable binding interface with CDH1, supported by persistent hydrogen bonds and hydrophobic contacts during MD. (b) Interaction with the CDH2 variant reveals altered residue accessibility and distinct MD-driven rearrangements, indicating mutation-dependent modulation of binding. (c) COL E2 forms a well-stabilized complex with CDH1, maintaining a dense hydrogen-bond network and strong surface complementarity throughout the simulation. (d) COL E2 interaction with CDH2 displays dynamic hydrogen-bond formation and hydrophobic stabilization, accompanied by subtle conformational adjustments in CDH2. (e) Two-dimensional interaction diagrams summarize key residue-level contacts—hydrogen bonds, hydrophobic interactions, and water-mediated bridges—highlighting shared and complex-specific binding hotspots across all wild-type systems.

Together, these integrated structural and energetic signatures reveal a definitive specificity axis: Colicin E2 preferentially engages epithelial cadherin CDH1, whereas Colicin E1 exhibits stronger affinity for mesenchymal cadherin CDH2. This duality mirrors cadherin switching during epithelial–mesenchymal transition (EMT) and provides a mechanistic foundation for designing cadherin-selective colicin therapeutics with relevance to cancer progression and metastasis.

4.3. Validation and docking evaluation of mutant models of Colicin E1 & E2 with CDH1 and CDH2

Computational saturation mutagenesis of Colicin E1 and E2 revealed mutation-specific effects on protein stability, enabling rational selection of variants for downstream structural docking. Several Colicin E1 substitutions, including NA338L, SA107F, KA366L, and KA158L, exhibited negative values, indicating stabilizing or mildly stabilizing effects, whereas highly destabilizing variants such as LA184A and LA90E were excluded. Similarly, stability profiling of Colicin E2 identified a set of structurally permissive mutations—A579A/E, A297A/E, A536A/E, and A572A/E—all exhibiting small negative values, supporting their suitability for interaction studies. Docking analysis demonstrated that Colicin E1 mutants display distinct receptor preferences, with NA338L and KA366L showing exceptionally strong binding to CDH1 (HADDOCK scores − 108.8 and − 55.2, respectively), supported by large buried surface areas (>2800 Å[2]and favourable electrostatic contributions. Conversely, SA107F and KA158L preferentially interacted with CDH2, achieving docking scores between − 72.6 and − 79.5 and forming enriched charged–apolar and apolar–apolar contacts, suggesting a receptor-specific mode of interaction(Table 3). Across all E1 variants, CDH1 complexes exhibited denser polar–polar and charged–charged interactions, whereas CDH2 complexes were dominated by hydrophobic contributions, reflecting divergent cadherin-binding mechanisms.

Table 3.

Validation parameter of mutant models of colicin E1 with CDH1 and CDH2.

Parameter	NA338L		SA107F		KA366L		KA158L
Target binding	(CDH1)	(CDH2)	(CDH1)	(CDH2)	(CDH1)	(CDH2)	(CDH1)	(CDH2)
HADDOCK score	−108.8 ± 26.3	−85.5 ± 9.1	−34.7 ± 18.5	−75.2 ± 8.1	−55.2 ± 14.6	−72.6 ± 15.5	−45.8 ± 21.4	−79.5 ± 9.7
Cluster size	4	4	6	19	9	5	7	7
RMSD (Å)	1.1 ± 0.7	9.3 ± 0.7	19.0 ± 0.0	31.8 ± 0.3	21.3 ± 0.4	31.2 ± 0.3	7.6 ± 0.3	36.8 ± 0.2
Van der Waals energy	−112.0 ± 23.9	−58.6 ± 9.8	−63.7 ± 8.4	−54.4 ± 5.0	−65.8 ± 14.4	−45.9 ± 6.7	−67.5 ± 9.6	−57.6 ± 4.0
Electrostatic energy	−478.9 ± 21.6	−372.4 ± 58.6	−405.2 ± 130.1	−285.3 ± 70.6	−471.5 ± 86.0	−346.2 ± 51.1	−484.9 ± 99.0	−345.3 ± 33.5
Desolvation energy	7.6 ± 4.4	6.1 ± 6.9	30.6 ± 8.5	12.6 ± 8.2	20.5 ± 2.4	5.0 ± 1.9	21.2 ± 4.5	23.6 ± 2.0
Restraints violation energy	913.7 ± 195.7	415.1 ± 67.8	794.3 ± 102.2	236.8 ± 18.0	843.8 ± 134.8	374.9 ± 36.9	975.1 ± 212.7	235.1 ± 64.1
Buried Surface Area (Å2)	3831.5 ± 307.8	1849.5 ± 56.5	2464.0 ± 413.3	1995.6 ± 78.9	2801.8 ± 154.4	1633.8 ± 130.7	2891.3 ± 389.5	2328.1 ± 47.5
Z-Score	−2.3	−1.9	−1.9	−1.4	−1.8	−1.0	−1.4	−1.4
ΔG (kcal/mol)	−13.7	−8.5	−10.3	−10.0	−13.0	−8.1	−12.6	−12.3
Kd (M)	8.7 × 10⁻11	5.5 × 10⁻7	2.9 × 10⁻8	4.9 × 10⁻8	2.9 × 10⁻10	1.1 × 10⁻6	5.9 × 10⁻10	1.0 × 10⁻9
ICs charged–charged	27	18	25	14	21	18	20	16
ICs charged–polar	36	14	27	20	14	15	14	28
ICs charged–apolar	42	22	25	19	37	18	42	32
ICs polar–polar	10	2	2	5	5	2	7	2
ICs polar–apolar	21	4	7	16	18	4	16	17
ICs apolar–apolar	12	5	4	5	4	10	6	9
NIS charged (%)	28.57	27.98	28.46	28.04	28.42	27.89	28.56	27.98
NIS apolar (%)	41.97	42.63	41.79	42.73	41.79	42.66	41.59	42.77

Colicin E2 mutants showed an equally striking mutation-dependent shift in cadherin recognition. A572E, A536A/E, and A297A/E emerged as the strongest CDH1 binders, characterized by large buried surface areas (1700–2800 Å,[2]low RMSD values, and deeply negative electrostatic energies (−321 to − 515 kcal/mol), resulting in high-affinity ΔG values in the range of − 10.9 to − 12.9 kcal/mol. In contrast, A579A/E and A297A/E exhibited the most favourable CDH2 affinities, with nanomolar Kd values (1.5 × 10⁻⁹ to 7.5 × 10⁻⁸ M) and strong contributions from hydrophobic and charged–apolar interactions, indicating that subtle alterations in local side-chain chemistry can rewire cadherin-binding specificity. Structural mapping of both Colicin families confirmed that stabilizing mutations induce reorientation of key loop and helical regions, improve hydrogen-bond geometry, and enhance the exposure of hydrophobic patches that facilitate cadherin insertion. Notably, E1–CDH1 complexes formed compact and deeply embedded interfaces, whereas E1–CDH2 complexes displayed more extended interaction topologies, consistent with their distinct energetic signatures(Table 4). For Colicin E2, mutations at positions 536 and 572 reinforced backbone stability and promoted favourable repositioning of the receptor-contacting helices, explaining their enhanced ΔG and Kd values.

Table 4.

Validation parameter of mutant models of colicin E2 with CDH1 and CDH2.

Parameter	A579A		A579E		A297A		A297E		A536A		A536E		A572A		A572E
Target binding	(CDH1)	(CDH2)	(CDH1)	(CDH2)	(CDH1)	(CDH2)	(CDH1)	(CDH2)	(CDH1)	(CDH2)	(CDH1)	(CDH2)	(CDH1)	(CDH2)	(CDH1)	(CDH2)
HADDOCK score	5	—	6	—	2	—	9	—	4	—	7	—	8	—	1	—
Cluster size	−17.4 ± 9.5	−52.6 ± 10.2	−9.2 ± 24.7	−67.4 ± 12.9	–33.5 ± 16.6	−59.5 ± 8.2	0.5 ± 23.7	−72.3 ± 9.3	−31.8 ± 27.3	−58.5 ± 21.8	−58.6 ± 8.5	−60.3 ± 5.9	−18.6 ± 18.2	−59.1 ± 2.8	−12.7 ± 17.3	−93.5 ± 31.1
RMSD (Å)	5	4	4	4	6	4	4	5	5	4	7	9	4	5	6	5
Van der Waals energy	25.1 ± 0.4	43.3 ± 0.2	13.5 ± 0.4	48.5 ± 0.4	43.1 ± 0.1	54.9 ± 0.6	21.4 ± 0.4	5.8 ± 0.1	2.4 ± 1.4	42.3 ± 0.5	42.6 ± 0.2	26.5 ± 1.1	28.7 ± 1.2	13.4 ± 0.2	19.2 ± 0.9	1.7 ± 1.2
Electrostatic energy	−54.8 ± 12.1	−58.7 ± 8.3	−52.0 ± 8.6	−24.8 ± 14.6	−66.1 ± 14.5	−35.0 ± 4.6	−59.7 ± 12.3	−65.9 ± 4.8	−79.0 ± 8.8	−71.4 ± 16.5	−69.7 ± 18.6	−61.5 ± 6.0	−47.8 ± 7.1	−37.5 ± 4.6	−59.3 ± 10.3	−63.7 ± 12.7
Desolvation energy	−385.4 ± 70.9	−258.3 ± 70.1	−348.0 ± 88.4	−515.3 ± 32.3	−389.0 ± 42.3	−336.4 ± 52.6	−207.1 ± 50.7	−278.4 ± 48.7	–232.4 ± 69.8	−221.6 ± 57.5	−364.9 ± 46.4	−216.1 ± 8.6	−378.6 ± 67.0	−321.1 ± 37.3	−267.5 ± 42.6	−398.3 ± 91.8
Restraints violation energy	22.5 ± 5.6	21.6 ± 3.9	16.6 ± 4.2	23.2 ± 5.5	22.2 ± 3.3	5.1 ± 2.4	18.6 ± 7.2	12.0 ± 7.2	2.2 ± 2.9	13.9 ± 4.3	21.0 ± 3.6	16.9 ± 3.8	17.1 ± 4.2	6.1 ± 1.8	14.1 ± 3.5	28.6 ± 4.3
Buried Surface Area (Å[2]	920.7 ± 137.9	362.3 ± 72.6	957.3 ± 50.6	372.8 ± 33.6	881.6 ± 51.8	376.1 ± 37.7	830.6 ± 106.4	372.6 ± 41.7	914.7 ± 108.6	432.6 ± 37.4	630.1 ± 98.2	275.1 ± 42.8	878.8 ± 123.2	365.5 ± 52.2	859.9 ± 114.4	212.7 ± 82.5
Z-Score	2856.1 ± 312.8	2118.2 ± 280.7	1934.2 ± 259.0	2111.3 ± 315.7	2825.7 ± 375.0	1694.3 ± 154.7	1863.2 ± 247.3	2303.2 ± 210.9	2599.2 ± 141.7	2696.5 ± 449.5	3106.3 ± 228.4	2147.6 ± 21.8	1965.7 ± 220.3	1712.0 ± 124.6	2139.6 ± 192.0	2280.4 ± 482.4
ΔG (kcal/mol)	5	13_1	6	14_1	2	7_1	9	16_1	4	1_1	3	5_1	8	8_1	1	8_1
Kd (M)	−10.0	−9.7	−10.0	−9.1	−9.6	−12.0	−10.8	−11.8	−12.9	−9.2	−10.7	−9.6	−11.5	−10.9	−8.7	−10.9
ICs charged–charged	4.7e-8	7.5e-8	4.7e-8	2.2e-7	8.7e-8	1.5e-9	1.1e-8	2.1e-9	3.7e-10	1.8e-7	1.5e-8	9.1e-8	3.7e-9	9.7e-9	4.1e-7	9.7e-9
ICs charged–polar	11	16	15	14	14	9	11	15	11	9	11	18	19	14	10	14
ICs charged–apolar	22	18	28	17	28	26	16	20	20	17	26	10	20	31	24	31
ICs polar–polar	39	21	30	20	25	30	39	33	51	26	32	19	30	44	24	44
ICs polar–apolar	11	5	5	0	14	5	8	16	7	7	13	0	6	11	6	11
ICs apolar–apolar	13	13	10	7	19	22	14	27	17	13	21	8	16	14	10	14
NIS charged (%)	10	15	8	3	14	10	13	18	10	9	17	6	8	4	10	4
NIS apolar (%)	27.27	26.87	27.34	26.82	26.96	26.80	27.19	27.09	27.37	26.89	27.09	26.91	27.20	26.75	27.43	26.75

Collectively, these results establish a clear divergence in receptor preference between Colicin E1 and E2 mutants: Colicin E1 Favors CDH1 through electrostatic-dominant interactions, whereas Colicin E2 preferentially engages CDH2 via hydrophobic complementarity. Integrating stability, docking energetics, and structural interaction mapping enabled the identification of lead variants with optimal cadherin-binding properties. The strongest CDH1 binders—E1-NA338L, E1-KA366L, E2-A572E, E2-A536A, and E2-A297 represent high-affinity candidates for targeting epithelial cadherin-rich cancer phenotypes. Conversely, the top CDH2 binders E1-SA107F, E1-KA158L, E2-A579A/E, and E2-A297E (Fig. 6) are predicted to preferentially target N-cadherin–dominant, mesenchymal or invasive tumor contexts. These findings together highlight a powerful mutational strategy for re-engineering Colicin–cadherin interactions and provide a mechanistic foundation for developing precision cadherin-targeted antimicrobial or anticancer therapeutics.

Fig. 6.

Sequence-level comparison of wild-type and mutant colicin variants. Multiple sequence alignment of wild-type colicins and their engineered mutants (a) highlights site-specific amino acid substitutions, marked in distinct colors, demonstrating targeted residue modifications while preserving overall sequence architecture. The positional mutation map for colicin E1 (b) illustrates that mutations are strategically localized to selected regions rather than randomly distributed, indicating rational design aimed at functionally relevant sites. A detailed residue-level alignment (c) further confirms the precise nature of these substitutions, with arrows denoting mutation positions and showing minimal perturbation of conserved domains. Consistently, the mutation distribution plot for colicin E2 variants (d) reveals analogous, site-specific residue changes across the sequence length. Collectively, these analyses confirm the controlled and rational engineering of colicin variants, ensuring modulation of functional properties while maintaining global sequence integrity suitable for downstream structural and functional studies.

4.4. MD analysis of Wild-Type vs engineered Colicin E1 and E2

Comprehensive molecular dynamics simulations were performed to delineate the conformational stability and residue-level dynamics of wild-type Colicin E1 and Colicin E2 in complex with the epithelial (CDH1) and mesenchymal (CDH2) cadherins. Backbone RMSD trajectories demonstrate that both colicins undergo rapid equilibration following initial relaxation, confirming the overall stability of the simulated complexes. However, Colicin E1 consistently exhibits higher RMSD values and broader temporal fluctuations, particularly in the CDH2-bound system, reflecting increased global conformational plasticity. In contrast, Colicin E2 displays lower RMSD amplitudes and earlier convergence, indicative of a more rigid and structurally constrained interaction with both cadherins.Residue-wise RMSF profiling reveals marked differences in local flexibility between the two colicins. Colicin E1 shows elevated fluctuations across multiple regions, including surface-exposed loops and terminal segments, suggesting enhanced dynamic adaptability at the protein–cadherin interface. These fluctuations are accentuated in the CDH2-bound complex, consistent with the greater structural heterogeneity observed in the RMSD analysis. By comparison, Colicin E2 maintains uniformly low RMSF values across most residues, highlighting a globally stabilized fold with limited local rearrangements during cadherin engagement.

Time-resolved RMSF heatmaps further substantiate these observations by capturing the persistence and spatial distribution of residue mobility throughout the simulation timeframe. Colicin E1–cadherin complexes exhibit sustained regions of higher mobility, indicating dynamic rearrangements that are maintained over time rather than transient in nature. Conversely, Colicin E2–cadherin complexes display a narrow fluctuation spectrum with minimal temporal drift, reflecting a dynamically robust interaction interface.Together, these results identify fundamental differences in the dynamic behavior of wild-type Colicin E1 and Colicin E2 upon cadherin binding. While Colicin E1 retains greater conformational flexibility that may facilitate adaptive or transient interactions, Colicin E2 forms more dynamically stable and conformationally restrained complexes with both CDH1 and CDH2. This enhanced dynamic stability of Colicin E2 provides a mechanistic basis for its potentially superior engagement with cadherin targets and supports its prioritization for downstream structure-guided optimization and functional validation in cadherin-associated disease contexts.

The comparative molecular dynamics evaluation of engineered colicin variants reveals clear stability- and function-relevant differences in their interactions with E-cadherin (CDH1) and N-cadherin (CDH2). The Colicin E1 mutant exhibits progressively increasing backbone deviations and intermittent large fluctuations during simulation, accompanied by elevated residue-level flexibility across loop and terminal regions in both cadherin-bound states. This dynamic behavior suggests pronounced conformational plasticity, which may undermine persistent cadherin engagement and limit its effectiveness in suppressing metastasis-associated signaling. In contrast, Colicin E2–derived mutants display markedly improved equilibration and reduced structural drift. Mutant-1 shows partial stabilization relative to Colicin E1, yet retains localized high-mobility regions proximal to cadherin-interacting interfaces, indicating incomplete dynamic restraint. Strikingly, Colicin E2 Mutant-2 demonstrates the lowest RMSD values with early convergence and uniformly minimal RMSF across the protein, particularly at cadherin-contact regions, reflecting a rigid, well-organized binding architecture. The combined reduction in global deviation and local flexibility in Colicin E2 Mutant-2 supports sustained and stable cadherin inhibition, a property that is mechanistically consistent with effective blockade of EMT-driven cell migration and metastatic progression. Overall, these integrated RMSD–RMSF profiles identify Colicin E2 Mutant-2 as the most structurally optimized and therapeutically promising candidate for targeting cadherin-mediated cancer proliferation and metastasis in precision oncology applications.

4.5. A novel cadherin- targeted protein selection approach for precision oncology

The integrated docking, stability profiling, and molecular dynamics analyses collectively demonstrate that engineered colicin variants markedly outperform their wild-type counterparts in terms of cadherin specificity, binding stability, and dynamic restraint—key prerequisites for translational anticancer applications. While wild-type Colicin E1 and E2 already display biologically meaningful cadherin preferences that mirror EMT-associated cadherin switching (E1–CDH2 and E2–CDH1), their elevated conformational flexibility and context-dependent stability limit their suitability as therapeutic scaffolds. In contrast, rationally engineered mutants exhibit enhanced structural robustness, optimized interface complementarity, and sustained cadherin engagement, positioning them as superior candidates for selective cancer targeting.

Among wild-type proteins, Colicin E2 forms more dynamically stable complexes than Colicin E1, as evidenced by lower RMSD amplitudes, uniform RMSF profiles, and persistent interaction networks with both CDH1 and CDH2. This inherent rigidity likely underlies its stronger epithelial cadherin affinity and suggests a natural predisposition for targeting epithelial-like tumor states. However, the engineered Colicin E2 Mutant-2 (A579E) emerges as the most therapeutically optimal construct across all analyses. These variant combines the lowest RMSD with early convergence, uniformly suppressed residue-level fluctuations, particularly at cadherin-contacting helices and loops, and highly favourable docking energetics and buried surface areas(Fig. 8). These features indicate a rigid, long-lived receptor-bound conformation capable of maintaining sustained cadherin blockade a critical requirement for inhibiting EMT-driven migration, invasion, and metastatic dissemination.

Fig. 8.

(a–b) RMSD analysis shows that both wild-type Colicin E1 and Colicin E2 form stable complexes with CDH1 and CDH2 during the simulations. However, Colicin E1 exhibits higher RMSD values and larger fluctuations, especially in the CDH2-bound state, indicating greater overall structural flexibility. In contrast, Colicin E2 shows lower RMSD values and stabilizes earlier, suggesting a more stable protein–cadherin interaction. (c) Residue-level RMSF analysis further supports this observation, with Colicin E1 displaying higher flexibility in several regions, particularly in loop and terminal segments, whereas Colicin E2 remains comparatively rigid across both CDH1 and CDH2 complexes. (d) RMSF heatmaps illustrate these trends over time, revealing sustained regions of higher mobility in Colicin E1 complexes, while Colicin E2 exhibits consistently low and uniform fluctuations. Overall, these results indicate that wild-type Colicin E2 forms more stable and less flexible complexes with cadherins than Colicin E1.

Engineered Colicin E1 variants, while significantly improved relative to the wild type, retain greater conformational plasticity. Mutants such as NA338L and KA366L demonstrate strong CDH1 affinity driven by electrostatic dominance, whereas SA107F and KA158L preferentially stabilize CDH2 hydrophobic and charged–apolar interactions. This suggests that Colicin E1 mutants may be better suited for context-specific targeting of mesenchymal or invasive tumor phenotypes, albeit with a narrower dynamic stability window than Colicin E2 mutants. Importantly, none of the engineered variants exhibit destabilizing physicochemical profiles (Instability Index < 40, preserved half-life, neutral GRAVY scores), supporting their experimental tractability(Fig. 7&9).

Fig. 7.

The structural docking analyses of Colicin E1 and E2 mutants with CDH1 and CDH2 reveal mutation-dependent differences in receptor engagement and interface complementarity. (a)For Colicin E1, mutants NA338L, SA107F, KA366L, and KA158L exhibited distinct interaction architectures with CDH1, where stabilizing substitutions enhanced hydrogen-bond density and hydrophobic packing, as reflected in the corresponding 2D interaction maps. (b)When docked with CDH2, these same mutants displayed altered binding orientations and expanded electrostatic and van der Waals contacts, indicating receptor-specific adaptability. (c) In Colicin E2, mutants such as A579A, A579E, A297A, A297E, A536A/E, and A572A/E formed diverse interaction networks with CDH1, with stabilizing mutations showing enriched interfacial bonding consistent with their favorable ΔΔG profiles. (d) Docking with CDH2 revealed further remodeling of binding poses, where several mutants achieved deeper interfacial embedding and extensive non-covalent interactions, demonstrating mutation-specific tuning of cadherin affinity. Together, these integrated 3D and 2D interaction analyses highlight how targeted mutagenesis modulates cadherin recognition and provides a structural basis for selecting high-affinity colicin variants.

From a translational perspective, the engineered colicins particularly Colicin E2 Mutant-2 offer a mechanistic advantage over wild-type proteins by enabling selective engagement of cadherins that are aberrantly expressed in cancer cells but tightly regulated in normal epithelia. This selectivity provides a plausible explanation for reduced off-target toxicity: normal cells maintain intact cadherin junctions and lower receptor accessibility, whereas tumor cells undergoing EMT expose cadherin interfaces that are structurally permissive to colicin binding (Fig. 9). These features directly support in vitro validation strategies, including cadherin-high versus cadherin-low cell line comparisons, EMT-reversal assays (E-cadherin restoration, N-cadherin suppression), migration and invasion inhibition assays, and minimal cytotoxicity in normal epithelial controls (Table 5).

Fig. 9.

Comparative molecular dynamics analysis of mutant colicins in complex with E-cadherin (CDH1) and N-cadherin (CDH2). (a) RMSD profiles show that the Colicin E1 mutant undergoes progressive deviation and pronounced fluctuations, indicating reduced global stability, whereas Colicin E2 mutants display improved equilibration, with Mutant-2 exhibiting the lowest and most consistent RMSD values. (b) Residue-wise RMSF analysis reveals elevated flexibility in the Colicin E1 mutant (NA338L) and localized high-mobility regions in Colicin E2 Mutant-1(A579A), while Colicin E2 Mutant-2(A579E) maintains uniformly low fluctuations, particularly at cadherin-binding regions. (c) Time-resolved secondary structure element (SSE) analysis shows that the introduced mutations preserve the overall α-helical and β-strand architecture of Colicin E1 and E2 throughout the simulation. Minor, localized fluctuations in loop regions indicate mutation-dependent flexibility without disrupting the global secondary structure integrity.

Table 5.

Computational Cytotoxicity and Stability Profiles of Wild-Type and Engineered Colicin E1 and E2: This table summarizes in silico–predicted physicochemical and stability parameters of wild-type and engineered Colicin E1 and Colicin E2 variants, including molecular weight, aromaticity, instability index, isoelectric point, aliphatic index, GRAVY score, and predicted half-life across mammalian, yeast, and E. coli systems. These parameters are widely used indicators of protein stability, membrane interaction propensity, and bioavailability, which collectively influence cytotoxic selectivity and therapeutic feasibility. The comparable molecular weights and preserved half-life across variants indicate that engineered mutations do not compromise protein integrity, while favourable instability indices (<40) and negative GRAVY values suggest stable, soluble proteins with reduced nonspecific cytotoxicity, supporting their suitability for targeted cadherin-mediated anticancer applications.

ID	Molecular Weight	Aromaticity	Instability Index	Isoelectric Point	Aliphatic Index	GRAVY	Half Life Mammalian	Half Life Yeast	Half Life E. coli
Colicin E1
Wild type	57279.164	0.054	29.076	9.306	83.008	−0.595	30 h	>20 h	>10 h
5 M	57278.219	0.054	29.076	9.306	83.755	−0.581	30 h	>20 h	>10 h
10 M	57339.26	0.056	28.707	9.306	83.008	−0.588	30 h	>20 h	>10 h
12 M	57264.149	0.054	29.076	9.27	83.755	−0.58	30 h	>20 h	>10 h
13 M	57264.149	0.054	29.076	9.27	83.755	−0.58	30 h	>20 h	>10 h
Colicin E2
Wild type	61628.49	0.052	36.577	9.017	62.513	−0.749	30 h	>20 h	>10 h
13 M	61584.48	0.052	36.245	9.104	62.685	−0.74	30 h	>20 h	>10 h
14 M	61627.51	0.052	36.245	9.104	62.513	−0.749	30 h	>20 h	>10 h
18 M	61627.51	0.052	36.835	9.104	62.513	−0.749	30 h	>20 h	>10 h
20 M	61627.51	0.052	36.908	9.104	62.513	−0.749	30 h	>20 h	>10 h
27 M	61585.47	0.052	36.577	9.017	62.685	−0.74	30 h	>20 h	>10 h
29 M	61598.47	0.052	37.054	9.017	62.685	−0.745	30 h	>20 h	>10 h
31 M	61642.52	0.052	36.55	9.017	62.685	−0.745	30 h	>20 h	>10 h
32 M	61642.52	0.052	36.577	9.017	62.685	−0.745	30 h	>20 h	>10 h

our findings support an alternative, receptor-associated mechanism of interaction. Structural and docking analyses revealed strong binding affinities of colicin E1 and E2 toward CDH1 and CDH2, which are key regulators of epithelial–mesenchymal transition (EMT) and are frequently dysregulated in tumor cells. Given that cadherins are transmembrane proteins involved in adhesion and intracellular signaling, their engagement by colicins may facilitate receptor-mediated endocytosis, including clathrin-mediated and caveolae-dependent pathways, which are well-established mechanisms for protein internalization in eukaryotic systems.[46] Following internalization, colicins may exert cytotoxic effects analogous to their known biochemical functions: colicin E1, a pore-forming toxin, can disrupt membrane integrity and ion homeostasis, whereas colicin E2 exhibits nuclease activity capable of inducing nucleic acid damage and subsequent apoptosis.[5] In addition to direct cytotoxicity, cadherin engagement may influence downstream signaling pathways associated with EMT, including β-catenin, Wnt, and PI3K/Akt signaling, thereby promoting epithelial characteristics (increased CDH1 expression) and suppressing mesenchymal traits (reduced CDH2 expression), ultimately limiting tumor cell migration and invasion.[47] Notably, similar mechanisms have been described for antimicrobial peptides and bacteriocins, which preferentially target cancer cells through membrane interactions and intracellular disruption.[48]Collectively, these findings support a biologically plausible model in which colicins interact with cadherin interfaces, undergo endocytic internalization, and exert intracellular anticancer effects (Fig. 11), thereby providing a mechanistic basis for their role in EMT modulation and tumor suppression (Fig. 10).

Fig. 11.

Overview of the experimental validation and translational workflow for structure-guided colicin targeting of cadherins in EMT: This figure provides a comprehensive overview of a multi-phase strategy integrating computational modelling with experimental validation to explore colicin–cadherin interactions in EMT-associated cancer. The workflow begins with in silico docking and structural analysis to identify key binding residues, followed by mutagenesis and binding validation. It then progresses through colicin identification and purification from clinical isolates, structural modification and stability assessment of optimized variants, and functional evaluation in cell-based assays assessing viability, migration, apoptosis, and EMT marker expression. Finally, the approach extends to in vivo validation in mouse models to evaluate therapeutic efficacy and tumor response. Overall, the figure illustrates a complete bench-to-bedside pipeline linking computational predictions to biological and translational outcomes.

Fig. 10.

Mechanistic model of colicin E1 and E2 targeting cadherins to inhibit EMT and tumor progression: This schematic depicts the receptor-associated mechanism by which colicin E1 (targeting CDH1) and colicin E2 (targeting CDH2) bind cadherins, undergo endocytic internalization, and translocate into the cytosol. Colicin E1 induces membrane pore formation, while colicin E2 mediates RNA degradation, leading to apoptosis and cell cycle arrest. These effects modulate key EMT signaling pathways (β-catenin, Wnt, PI3K/Akt, MAPK), resulting in suppression of migration, invasion, and metastasis, and highlighting the selective anti-cancer potential of colicin-based targeting.

The inhibition of epithelial–mesenchymal transition (EMT) and metastatic potential by colicin can be mechanistically explained through its interaction with cadherin-mediated signalling pathways. Our structural analyses demonstrate that colicin E1 and E2 exhibit strong binding affinity toward CDH1 and CDH2, which are central regulators of EMT. EMT is characterized by the loss of epithelial adhesion (mediated by CDH1) and the gain of mesenchymal features (associated with CDH2), a process often referred to as “cadherin switching,” which enhances tumor cell migration and invasion. Binding of colicin to cadherin extracellular domains may stabilize E-cadherin–mediated adherens junctions, thereby reinforcing cell–cell adhesion and preventing dissociation of β-catenin from the membrane complex. This stabilization limits β-catenin nuclear translocation and subsequent transcription of EMT-promoting genes. Concurrently, interaction with N-cadherin may interfere with pro-migratory signalling pathways, including PI3K/Akt and Wnt signalling, which are critical for maintaining mesenchymal phenotypes and invasive behaviour. In addition, internalized colicins may further enhance EMT suppression through intracellular stress responses, including membrane perturbation (colicin E1) and nucleic acid damage (colicin E2), ultimately leading to apoptosis of highly migratory cancer cells. Together, these effects contribute to the reversal or inhibition of EMT, characterized by increased CDH1 expression, decreased CDH2 expression, reduced cellular motility, and diminished metastatic potential(Fig. 10). Thus, colicin–cadherin interaction provides a plausible mechanistic link between molecular binding events and functional suppression of EMT and tumor progression.

A key limitation of this study is the potential lack of selectivity, as CDH1 (E-cadherin) is expressed in both normal and tumor epithelial cells. However, tumor susceptibility to colicin may depend on differences in cellular context rather than expression alone. In normal tissues, E-cadherin is tightly organized within stable adherens junctions and is less accessible,[1], [2] whereas in tumor cells it is often mislocalized and dynamically remodeled during epithelial–mesenchymal transition (EMT), increasing its surface exposure.[3], [4] Additionally, cancer cells exhibit enhanced endocytic activity and membrane alterations that may facilitate colicin uptake,[5] along with greater sensitivity to membrane disruption and translational inhibition.[6]Thus, the interaction is likely context-dependent rather than absolutely selective, and further in vitro and in vivo studies are required to validate this differential susceptibility.

Importantly, our findings are consistent with and extend previous studies demonstrating that bacteriocins and antimicrobial peptides can interact with eukaryotic membranes and receptors, inducing selective cytotoxicity in cancer cells while sparing normal tissue. Prior reports on colicin pore formation, membrane translocation, and receptor-mediated uptake primarily focused on bacterial systems. however, emerging evidence supports their broader bioactivity in mammalian contexts. From a biotechnological perspective, the engineered colicin variants identified in this study represent promising candidates for recombinant expression and downstream functional validation. The stability-enhancing mutations and favourable physicochemical properties observed here support their feasibility for heterologous expression in microbial systems. Moreover, the computational pipeline presented integrating transcriptomics, structure-guided mutagenesis, docking, and molecular dynamics offers a scalable strategy for rapid screening and optimization of bioactive proteins prior to experimental validation, reducing time and cost in biotherapeutic development. Our work advances this paradigm by providing the first comprehensive EMT-informed, cadherin-resolved, mutant-guided framework for repurposing colicins as precision anticancer agents.

In summary, while wild-type Colicin E1 and E2 establish the biological feasibility of cadherin targeting, engineered Colicin E2 particularly Mutant-2 represents the most suitable therapeutic lead, combining superior stability, receptor specificity, and dynamic restraint. These properties are mechanistically aligned with effective EMT suppression and metastasis inhibition, providing a strong rationale for downstream in vitro and in vivo validation and positioning engineered colicins as a novel class of cadherin-targeted anticancer biologics.

5. Conclusion

Our study establishes a comprehensive, EMT-informed framework for repurposing colicins as precision cadherin-targeted anticancer proteins. Through integrated docking, stability profiling, and molecular dynamics analyses, we demonstrate that engineered colicin variants significantly outperform wild-type proteins in cadherin specificity, binding stability, and dynamic restraint. The intrinsic preference of wild-type Colicin E2 for CDH1 and Colicin E1 for CDH2 mirrors EMT-associated cadherin switching, validating cadherins as rational therapeutic entry points. Importantly, structure-guided mutagenesis enabled fine-tuning of this specificity, yielding optimized variants with enhanced interface complementarity and reduced conformational flexibility. Among all candidates, Colicin E2 Mutant-2 (A579E) emerged as the most promising therapeutic lead, combining superior stereochemical integrity, sustained cadherin engagement, and favourable physicochemical properties predictive of low off-target cytotoxicity. These features are mechanistically aligned with effective suppression of EMT-driven migration, invasion, and metastasis.

6. Experimental validation strategies for Structure-Guided predictions

To strengthen the biological and clinical relevance of the structure-guided computational findings, a series of complementary experimental approaches should be incorporated to validate the predicted interactions between colicin E1/E2 and cadherins (CDH1 and CDH2). While molecular docking and dynamics simulations provide valuable insights into binding affinity and structural stability, these models do not fully capture the complexity of cellular and in vivo environments. Initial validation can be achieved through biophysical interaction assays such as surface plasmon resonance (SPR), isothermal titration calorimetry (ITC), and co-immunoprecipitation (Co-IP), which can quantitatively confirm direct binding and affinity between colicins and cadherins.[49]Furthermore, site-directed mutagenesis targeting key residues identified in silico can be employed to assess their functional importance in binding specificity and interaction stability.

At the cellular level, in vitro functional assays including MTT for viability, wound healing (scratch) assays for migration, and flow cytometry for apoptosis and cell cycle analysis will help determine the phenotypic consequences of colicin–cadherin interactions. Additionally, gene and protein expression analyses such as qPCR and Western blotting for epithelial–mesenchymal transition (EMT) markers (E-cadherin, N-cadherin, vimentin) will provide mechanistic insights into how these interactions influence EMT dynamics.[50] To further validate spatial and molecular interactions, immunofluorescence and confocal microscopy can be used to visualize colicin localization and its co-localization with cadherins, supporting receptor-mediated engagement and potential internalization pathways. These approaches may also help elucidate whether clathrin-mediated or caveolae-dependent endocytosis is involved [46].

Importantly, in vivo validation using xenograft or syngeneic mouse models is essential to assess therapeutic relevance, including tumor growth inhibition, tissue-specific targeting, and systemic toxicity. Histological analyses and immunohistochemistry can further confirm modulation of EMT markers within tumor tissues. It is also critical to acknowledge that cellular and in vivo environments introduce additional layers of complexity that are not fully represented in computational models. Factors such as the presence of competing proteins, nucleic acids, membrane dynamics, post-translational modifications, and tumor microenvironment conditions can significantly influence protein behavior and binding outcomes.[51]Therefore, discrepancies between in silico predictions and experimental observations should be anticipated and interpreted within this broader biological context. Collectively, integrating these experimental strategies will provide robust validation of the computational predictions and enhance the translational potential of colicin-based targeting approaches in EMT-associated cancers.

7. Future Perspectives and translational Outlook

The therapeutic relevance of targeting epithelial–mesenchymal transition (EMT) in cancer has been firmly established, as EMT drives loss of epithelial adhesion, acquisition of migratory phenotypes, therapy resistance, and metastatic dissemination through coordinated suppression of E-cadherin (CDH1) and induction of N-cadherin (CDH2) signalling. Seminal studies have demonstrated that cadherin switching is not merely a biomarker but a functional driver of metastasis, regulating β-catenin signalling, cytoskeletal remodelling, and pro-survival pathways in tumor cells.[52]Consequently, several cadherin-directed strategies including peptide antagonists such as ADH-1 (Exherin) and antibody-based approaches have been explored, yet their clinical translation has been limited by insufficient selectivity, transient efficacy, and off-target toxicity.

In parallel, an emerging body of work has shown that bacteriocins and colicins, traditionally studied as bacterial toxins, can interact with eukaryotic membranes and surface receptors, exhibiting selective cytotoxicity toward cancer cells while sparing normal tissue.[12] More recent computational and experimental studies including our previous work on Colicin A and Colicin N demonstrated that colicins can bind E-cadherin and N-cadherin, perturb cadherin-mediated adhesion, and potentially influence EMT-associated signalling pathways. However, these studies primarily established biological possibility using wild-type proteins and did not address the structural instability, context-dependent binding, and EMT-state specificity required for therapeutic development [22].

The present study substantially extends this literature by integrating EMT-informed cadherin expression profiling, stability-guided mutagenesis, and long-timescale molecular dynamics simulations to systematically engineer colicin variants optimized for cancer-relevant cadherin states. Unlike prior EMT-targeting strategies that broadly inhibit cadherins irrespective of tumor context, our data demonstrate that cadherin specificity can be rationally rewired through minimal amino-acid substitutions, producing colicin variants that preferentially engage CDH1-dominant epithelial tumors or CDH2-enriched mesenchymal tumors. This approach directly addresses a major limitation of earlier EMT-targeting therapeutics, which often fail to account for intratumoral and intertumoral EMT heterogeneity.

Importantly, our findings align with and mechanistically extend large-scale transcriptomic analyses from TCGA and related studies showing that cadherin expression varies widely across cancer types and progression stages.[53] By mapping colicin binding preferences onto these EMT-defined cadherin landscapes, the engineered Colicin E2 mutants particularly Mutant-2 (A579E) emerge as structurally optimized candidates capable of sustained cadherin engagement, reduced conformational drift, and minimized nonspecific cytotoxicity. This represents a clear conceptual advance over earlier bacteriocin studies, which lacked EMT resolution and mutation-driven optimization.

From a translational standpoint, engineered colicins may function as cadherin-modulating biologics rather than indiscriminate cytotoxins, selectively disrupting EMT-driven signalling, restoring epithelial characteristics, and sensitizing tumors to chemotherapy or immunotherapy. This concept is consistent with prior EMT-reversal strategies targeting transcription factors (e.g., Snail, ZEB1) signalling nodes (TGF-β, Wnt/β-catenin), but offers a distinct advantage by acting directly at the tumor cell surface, where cadherin accessibility is highest during EMT.[54]Moreover, the preserved physicochemical stability and predicted half-life of the engineered variants suggest compatibility with advanced delivery platforms, including nanoparticle- or exosome-based systems, an avenue increasingly explored for protein therapeutics but not yet applied to colicins.

8. Limitation of the study

A key limitation of the present study is the lack of in vivo validation to assess both the potential immunogenicity and selectivity of colicins. As bacterial proteins, colicins may be recognized as foreign by the human immune system and could induce immune responses upon systemic administration, while their immunological safety profile remains unestablished under physiological conditions. In addition, selectivity remains a concern, as CDH1 (E-cadherin) is expressed in both normal and tumor epithelial cells. However, tumor susceptibility may depend on differences in cellular context rather than expression alone. In normal tissues, E-cadherin is tightly organized within stable adherens junctions and is less accessible, whereas in tumor cells it is often mislocalized and dynamically remodeled during epithelial–mesenchymal transition (EMT), increasing its surface exposure. Furthermore, cancer cells exhibit enhanced endocytic activity and membrane alterations that may facilitate colicin uptake, along with increased vulnerability to membrane disruption and translational inhibition. To address these limitations, future studies will focus on in vivo evaluation of immune responses, biodistribution, and therapeutic efficacy. Additionally, we are exploring an exosome-based delivery system for colicin, which may reduce immunogenicity by shielding antigenic epitopes and improving targeted delivery. Nevertheless, these strategies require rigorous experimental validation, and both immunogenicity and selectivity should be considered key limitations of the current study.

Data Availability Statement

All data analysed during this study are included in this published article. Publicly available datasets used in this study were obtained from The Cancer Genome Atlas (TCGA), the GEPIA2 and UCSC Xena platforms. Structural data were retrieved from the Protein Data Bank (PDB). Additional computational data supporting the findings of this study are available from the first and corresponding author upon request.

Funding sources

The authors declare that financial support was received for the research and publication of this article. The study was supported by the “Chancellor Research Fellowship” for Ph.D. – SRM-MHS (Sanction No: CRFM003), which significantly facilitated the successful execution of the research work. The authors also gratefully acknowledge the support provided by SRM Medical College Hospital and Research Centre, Faculty of Medicine and Health Sciences, SRM Institute of Science and Technology, in the form of reimbursement of the article processing charges.

CRediT authorship contribution statement

Poornima Baskar Vimala: Writing – original draft, Project administration, Methodology, Funding acquisition. Leela kagithakara Vajravelu: Writing – original draft, Validation, Project administration, Investigation. Vishnu Priya Panneerselvam: Methodology, Investigation, Formal analysis, Data curation. Dakshina M Nair: Supervision, Software, Resources. Rahul Harikumar Lathakumari: Visualization, Methodology, Data curation. Jayaprakash Thulukanam: Validation, Supervision, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.