Mapping metastatic potential through cytoskeletal remodeling: A mechanobiology-based multi-cancer analysis

Metastasis remains the leading cause of cancer-related mortality, yet current clinical tools lack sensitivity for early detection of metastatic potential. While gene expression assays have improved prognostic stratification in select cancers, they often overlook dynamic cellular behaviors such as cytoskeletal remodeling and mechanotransduction. We curated a panel of 11 mechanobiology-related genes involved in actin cytoskeleton regulation and evaluated their predictive value using multi-modal data from the DepMap database, including gene expression, copy number variation, CRISPR, and RNAi screens. Analyses were conducted across 351 cell lines from breast, pancreatic, and central nervous system (CNS) tumors, stratified by tumor type and MetMap-derived metastatic potential. Gene expression and CNV emerged as the most informative parameters for distinguishing metastatic potential. Among the candidate genes, FSCN1 consistently demonstrated strong associations across tissues, classification schemes, and correlation with canonical metastasis markers. MYH9, ECM1, ANXA2, and EZR also showed robust performance, particularly in cumulative multi-gene signatures. Correlation analyses revealed significant positive correlations with invasion, EMT, stemness markers (e.g., MMP2, VIM, CD44), and inverse correlations with epithelial markers (e.g., CDH1, EPCAM), reinforcing the biological coherence of the panel. Our study introduces a previously unrecognized, mechanobiologically driven gene panel - FSCN1, MYH9, ECM1, ANXA2, and EZR - as a cross-cutting biomarker signature of metastatic potential, applicable across multiple cancer types. Unlike conventional markers that often reflect tumor-specific traits, this panel captures the mechanical and structural determinants of cell invasiveness. The robustness of these genes across diverse molecular assays and their alignment with core metastatic pathways underscore their translational relevance.