Abstract
BACKGROUND: Intratumor heterogeneity (ITH), a critical driver of tumor evolution and immune evasion, remains inadequately characterized at the transcriptomic level in colorectal cancer (CRC), and its clinical implications are not yet fully understood. METHODS: We integrated transcriptomic datasets from TCGA-COAD/READ and two independent GEO cohorts (GSE40967 and GSE87211) to develop an RNA-seq-based ITH score using the DEPTH2 algorithm and to con0struct an ITH-related gene (ITRG) prognostic model. A unified cutoff value of 0.64 was established to stratify patients into high- and low-ITH groups. Using 52 survival-associated ITRGs, we generated a nine-gene prognostic signature selected from 101 distinct combinations of feature selection techniques and modeling algorithms and validated its performance in two external datasets. The tumor microenvironment and potential responsiveness to immune checkpoint inhibitors were evaluated using ssGSEA, ESTIMATE, and TIDE algorithms. SHAP analysis, together with in vitro and in vivo experiments, was employed to identify and functionally validate key regulatory genes. RESULTS: Patients in the high-ITH group had markedly poorer overall survival (OS) than those in the low-ITH group. The ITH score correlated strongly with aggressive clinical features, including T3/4 invasion depth, N1/2 nodal status, and AJCC stage III. The nine-gene prognostic signature demonstrated consistent predictive capability across the TCGA training cohort and both GEO validation cohorts. In TCGA, the model yielded time-dependent AUCs of 0.669, 0.664, and 0.645 for 1-, 3-, and 5-year OS, respectively, and retained its status as an independent prognostic indicator in multivariate Cox regression analysis; in GSE40967 and GSE87211, the corresponding AUCs ranged from 0.544-0.573 and 0.648-0.744. The high-risk subgroup was characterized by a stromal immune phenotype enriched with cancer-associated fibroblasts and macrophages, elevated expression of multiple immune checkpoints, increased TIDE scores, and higher tumor mutational burden. SHAP analysis identified IL20RB as the top risk-associated gene, whose knockdown significantly suppressed CRC cell proliferation, migration, invasion, and tumorigenicity in vitro and in vivo. CONCLUSION: This study introduces and validates a transcriptomic ITH score and a nine-gene ITRG-based prognostic model that delineate the immune landscape and enables effective survival stratification in CRC, complementing the limitations of current staging systems. Additionally, IL20RB is highlighted as a promising therapeutic target, supporting the development of personalized immuno-targeted combination therapies in CRC.