Integrative multi-omics analysis identifies a CMA-associated heterogeneity risk score and a cDCs-based immune score for robust prognostic stratification in colon cancer with single-center and experimental validation.

BACKGROUND: Chaperone-mediated autophagy (CMA) plays an important role in tumor progression and remodeling of the tumor immune microenvironment. However, its functional heterogeneity, immune associations, and clinical significance in colon cancer remain unclear.

METHODS: Single-cell and bulk transcriptomic data were integrated to characterize CMA-related features in colon cancer. At the single-cell level, CMA activity was assessed across cell types, and differentially expressed genes between CMA-high and CMA-low groups were identified in myeloid subpopulations. Robust candidates were screened by recurrence frequency. At the bulk level, TCGA-COAD was used as the primary training cohort, and GSE17538 and GSE38832 for cross-cohort performance evaluation. CMA activity was quantified by ssGSEA, and candidates were further refined by WGCNA and tumor-normal differential expression analysis. An ensemble machine learning framework incorporating 101 algorithm combinations was used to construct a dual prognostic system consisting of the Risk Score and the Immune Risk Score. Prognostic performance was evaluated by Kaplan-Meier analysis, time-dependent ROC curves, and Cox regression. The Immune Risk Score was additionally evaluated in an independent single-center transcriptome cohort from Liaoning Central Hospital. MAPKAPK3 was identified as a key functional gene and validated in vitro. Drug sensitivity was predicted using pRRophetic and CGP2016.

RESULTS: CMA activity showed marked intercellular heterogeneity and was predominantly enriched in myeloid cells. Frequency-based screening, WGCNA, and tumor-normal differential expression analysis identified robust CMA-related candidates. The Risk Score showed favorable prognostic performance and generalizability across cohorts. The Risk Score remained an independent prognostic factor, whereas the Immune Risk Score functioned as an integrated prognostic score combining clinicopathologic and immune microenvironmental information. Immune analyses revealed consistent differences in regulatory T cells and resting dendritic cells across risk groups, suggesting an association between CMA-related risk states and an immunosuppressive microenvironment. The three-variable clinicomicroenvironmental model showed good predictive performance. MAPKAPK3 overexpression promoted proliferation, migration, and invasion of colon cancer cells, providing preliminary gain-of-function evidence for its tumor-promoting role.

CONCLUSIONS: This study revealed CMA-related heterogeneity and immune microenvironmental features in colon cancer and established a robust dual prognostic system. MAPKAPK3 may serve as a key functional gene associated with tumor progression and microenvironment remodeling within the CMA-related prognostic framework.