Integrative single-cell and machine learning analyses identify MACC1 as a regulator of TGF-β-associated immune evasion in colorectal cancer

To investigate the cellular heterogeneity within the tumor microenvironment of colorectal cancer (CRC) and identify potential biomarkers associated with malignant progression and immunosuppressive features, aiming to provide novel targets for precise therapy. Single-cell sequencing data of CRC were integrated to construct a microenvironmental atlas, with malignant cells identified via inferCNV. The TGF-β phenotype was dissected using multidimensional scoring, single-cell trajectories, and hdWGCNA. Seven machine learning algorithms were introduced to construct a predictive model, followed by cross-cohort validation across 24 independent external cohorts. The immune microenvironment was evaluated using deconvolution algorithms, and in vitro proliferation, invasion, and Western blot mechanistic assays were conducted using an shRNA-mediated knockdown model. Malignant cells exhibited exceptionally high TGF-β activity, which was significantly correlated with cell stemness. An optimized LASSO model filtered 10 candidate genes. Pan-cohort validation and survival analysis identified MACC1 as a robust biomarker associated with poor prognosis. Further analysis revealed that high MACC1 expression is accompanied by profound immunosuppressive microenvironmental features. In vitro validation demonstrated that silencing MACC1 attenuated the proliferative and invasive capacities of CRC cell lines (DLD-1/HCT-15) and triggered the synchronous downregulation of downstream TGF-β pathway proteins (p-SMAD2/3). MACC1 is highly associated with CRC malignant phenotypes and immune-related alterations, with its oncogenic effects potentially involving modulation of the TGF-β signaling axis. MACC1 may serve as a promising prognostic biomarker and potential therapeutic target for CRC.