Developing an epithelial signature for prognosis of papillary thyroid carcinoma.

Papillary thyroid carcinoma (PTC) shows heterogeneous clinical behavior, and current clinicopathologic factors do not fully capture epithelial aggressiveness or tumor microenvironment (TME) remodeling. We integrated single-cell RNA sequencing (scRNA-seq) datasets from normal thyroid tissue, primary PTC, and lymph-node metastases to construct a cross-stage cellular atlas and to characterize epithelial and T/NK cell heterogeneity. Guided by the malignant epithelial program identified at single-cell resolution, we developed a four-gene risk signature in the TCGA-THCA cohort using Cox and LASSO regression. We further evaluated pathway activity, immune and stromal features, T cell receptor (TCR) diversity, and exploratory in silico drug-response patterns, and we performed small interfering RNA (siRNA)-mediated knockdown of TMEM45A and STC1 in PTC cell lines. The integrated atlas revealed stage-related remodeling of epithelial, immune, and stromal compartments and identified a malignant epithelial state enriched for proliferation-, stress-, and migration-related programs. The four-gene signature was associated with overall survival stratification in TCGA-THCA after adjustment for age, sex, and stage. The high-risk group showed reduced immune infiltration, stronger stromal and epithelial-mesenchymal transition (EMT) signals, lower TCR diversity, and distinct predicted drug-response patterns. In vitro, TMEM45A or STC1 silencing suppressed proliferation and migration and altered EMT-related markers in PTC cells. These findings indicate that a malignant epithelial program defined by single-cell analysis can be translated into a four-gene signature associated with epithelial aggressiveness and TME remodeling in PTC, although its clinical utility requires validation in independent cohorts.