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

• 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. 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.