Feed

Thyroid cancer ranges from indolent differentiated tumors to metastatic, radioiodine-refractory and anaplastic disease. Conventional histology and molecular classification define major risk groups but do not resolve the cell states and spatial heterogeneity that accompany progression. This review integrates recent single-cell RNA sequencing, spatial transcriptomics and validation studies across localized papillary thyroid carcinoma (PTC), metastatic or radioiodine-refractory differentiated thyroid cancer (DTC), poorly differentiated thyroid carcinoma (PDTC) and anaplastic thyroid carcinoma (ATC). Current evidence indicates that progression is accompanied by changes in malignant epithelial states, stromal regions and immune patterns. Malignant epithelial cells shift from follicular-like programs toward partial EMT-like, dedifferentiation-like and anaplastic states; stromal changes include invasive borders, stiff peritumoral matrix and CAF/ECM-rich poorly differentiated regions; and immune patterns differ between progressive PTC, indolent lymphoid-organized tumors and myeloid-rich ATC. Functionally supported examples, including POSTN-IL-4 signaling, CCL20/CXCL5 macrophage-tumor reciprocal signals and SIGLEC15-associated checkpoint signaling, show how these technologies can generate therapeutic hypotheses. Yet most datasets are cross-sectional, and many ligand-receptor interactions remain computational candidates. We use the available evidence to relate epithelial plasticity, genomic context, stromal regions, immune remodeling and candidate cell-cell interactions across disease states, while separating observed associations from established stepwise tumor evolution. This synthesis highlights mechanisms and therapeutic hypotheses that require functional and clinical validation.