A SERPINA1-associated SOX4–FN1 epithelial program linked to lymphatic endothelial responses in aggressive papillary thyroid carcinoma

Background Lymph node metastasis (LNM) complicates risk assessment and treatment planning in papillary thyroid carcinoma (PTC). How tumor epithelial programs relate to the lymphatic microenvironment remains unclear. Methods We integrated single-cell RNA sequencing and spatial transcriptomic data with bulk cohorts to characterize epithelial states and their clinical associations. Bulk cohorts came from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO). Deconvolution, pathway, regulatory, and cell–cell communication analyses examined epithelial–lymphatic relationships; machine learning was used to construct an LNM classifier. We tested key findings using perturbation and rescue experiments, lymphatic endothelial cell (LEC) assays, xenografts, and human tissue analyses. Results An atlas of 221,187 cells from 46 thyroid specimens identified an extracellular matrix-associated THY5 epithelial state. THY5 was enriched in pooled cells from aggressive tumors, although its sample-level abundance did not differ significantly between tumor groups. Bulk analyses linked higher estimated THY5 abundance to adverse clinicopathological features and immune exclusion. An exploratory 11-gene classifier incorporating SERPINA1 and FN1 discriminated lymph node status. Areas under the receiver operating characteristic curve were 0.810 in the TCGA internal evaluation set and 0.768 and 0.800 in two GEO cohorts. Communication and spatial analyses implicated FN1-related epithelial–lymphatic interactions and local THY5–LEC co-occurrence. Perturbation experiments supported functional relationships among SERPINA1, SOX4, and FN1. SOX4 re-expression increased FN1 mRNA and protein levels relative to SERPINA1 knockdown alone. FN1 re-expression in SERPINA1-depleted tumor cells counteracted reductions in LEC markers, tube formation, and migration. SERPINA1 knockdown also suppressed xenograft growth and reduced FN1 staining. Conclusion Our findings support functional links among SERPINA1-associated SOX4–FN1 expression, malignant epithelial behavior and LEC responses in PTC, highlighting the potential of this molecular program for investigating epithelial–lymphatic interactions. The 11-gene LNM classifier discriminated lymph node status across the evaluated cohorts, providing an exploratory framework for molecularly informed risk assessment.