Abstract Background Thyroid carcinoma (TC) presents a rising global incidence, with a subset of cases progressing aggressively despite standard therapies. The tumor microenvironment (TME), particularly the functional state of CD8⁺ T cells, is crucial in disease progression, yet a systematic, high-resolution understanding of the cellular interactions associated with immune exhaustion and tumor evolution in TC remains limited. Methods To address this, we performed an integrated single-cell and spatial transcriptomic analysis of human TC samples. Our approach combined scRNA-seq data processing, cell communication inference, spatial co-localization validation, and machine learning-based prognostic modeling. Result We identified a distinct tumor cell subcluster (Trem/T01) characterized by APOE expression and a CD8⁺ exhausted T cell (Tex) subset. Computational and spatial analyses revealed a potential APOE (from Trem)- NCF1 (from CD8⁺ Tex) interaction axis, with these cell types demonstrating significant spatial proximity in tumor tissues. Furthermore, we derived a prognostic gene signature from this network, constructing a robust risk stratification model validated across independent cohorts. Conclusion This study uniquely delineates a specific tumor-immune interactive niche in TC, providing insights into potential mechanisms into immune evasion via the APOE-NCF1 axis and delivering a translatable framework for patient prognostication.
Su et al. (2026) studied this question.