Decoding colorectal cancer chemotherapy and immunotherapy resistance: revealing the role of NOL3 using single-cell RNA sequencing and machine learning

The high heterogeneity of colorectal cancer (CRC) complicates accurate prognosis prediction. Hypoxia can affect cell death mechanisms, leading to resistance to many antitumor therapies and potentially causing relapse. Programmed cell death (PCD) forms are also crucial in regulating cancer progression. Given this, the prognostic value of the cross-talk between hypoxic conditions and PCD mechanisms in CRC merits deeper exploration. Our methodology involved a large-scale analysis of four multicenter cohorts (n = 1294) utilizing a reliable and efficient data processing pipeline to explore the associations between hypoxia and 18 PCD signatures. Thirteen widely used machine learning algorithms were applied to identify cell death risk features associated with hypoxia, and NOL3 was selected as a key prognostic marker for CRC. We further explored the role of NOL3 in regulating cell proliferation, migration, and apoptosis. Drug sensitivity analysis confirmed its involvement in chemotherapy resistance, and the impact of NOL3 on the efficacy of Daporinad was evaluated using MTT assays. Additionally, Western blot analysis revealed NOL3's influence on epithelial-mesenchymal transition (EMT) and hypoxia signaling pathways. Our study further examined the involvement of NOL3 in immune cell infiltration patterns, microenvironmental characteristics of tumors, and metabolic pathway activities. An optimal predictive model was constructed by combining hypoxia-associated genes and PCD-related genes, demonstrating superior prognostic performance. The top three machine learning models-XGBoost, RR, and RF-achieved an average AUC value of 0.86. NOL3 demonstrated significant predictive performance for CRC prognosis. The reduction of NOL3 expression significantly inhibited the proliferation and migratory capacity of CRC cells, while simultaneously promoting apoptotic cell death. High NOL3 expression was associated with resistance to multiple chemotherapy drugs. Assessment of cell viability via MTT assay demonstrated that NOL3 knockdown potentiated the cytotoxic effects of Daporinad in CRC cells Additionally, NOL3 knockdown suppressed EMT and HIF signaling pathways. High NOL3 expression correlated with an immune-desert tumor microenvironment. Metabolic analysis indicated that elevated NOL3 expression in CRC cells promotes glycolysis. Further single-cell sequencing analysis identified NOL3 as a potential predictive marker for immune therapy non-response. CellChat analysis suggested that High_NOL3_Epithelial may interact with various immune cells through signaling pathways such as EGF, SAA, and TWEAK, contributing to immune therapy resistance. We established a composite prognostic model by combining features from hypoxia and PCD gene sets and validated NOL3 as a dual biomarker, offering significant clinical implications for prognostication and the tailoring of immunotherapy and chemotherapy in CRC patients.