Abstract

Background Epithelial–mesenchymal transition (EMT) represents a critical process that facilitates metastatic dissemination and immune evasion in colorectal cancer (CRC); however, the molecular factors that connect EMT to modifications in the immune microenvironment remain poorly elucidated. In this investigation, we identify AJUBA as an essential regulator that mediates the association between EMT and immune modulation in CRC. Methods By integrating multi-cohort transcriptomic datasets (The Cancer Genome Atlas (TCGA)-CRC, GSE18105, GSE22598, GSE89076, and GSE110224) with single-cell RNA-seq data (GSE132465), we applied machine learning and deep learning methodologies to comprehensively identify EMT-associated genes demonstrating prognostic significance. AJUBA validation was performed at mRNA and protein levels in a cohort of 90 CRC patient samples using quantitative PCR, Western blotting, and immunohistochemical (IHC) approaches. Functional analyses involved siRNA-mediated knockdown experiments, coupled with evaluations of cell proliferation (CCK-8 assay), migration and invasion (transwell assay), clonogenic capacity (colony formation assay), and in vivo tumor growth in xenograft models. Immune infiltration was assessed via ssGSEA and CIBERSORT algorithms, and spatial transcriptomics data (GSE225857) were used to delineate AJUBA expression within tumor microdomains. Results Across multiple CRC cohorts, AJUBA exhibited marked upregulation and showed distinct enrichment in epithelial cells with activated EMT characteristics. Spatial transcriptomic profiling demonstrated AJUBA colocalization with cancer-associated fibroblasts (CAFs) within immune-excluded niches. Enhanced AJUBA expression exhibited a positive correlation with heightened infiltration of M2 macrophages and activation of VEGF/NOTCH signaling cascades. In vivo, AJUBA knockdown led to suppressed tumor growth, reduced Ki-67 proliferation indices, and diminished M2 macrophage abundance. Clinically, elevated