Background: Pancreatic ductal adenocarcinoma (PDAC) has dismal survival, and vascular invasion is strongly associated with dissemination and poor outcomes; however, its molecular basis remains unclear. Methods: Transcriptomic and clinical data from The Cancer Genome Atlas-Pancreatic Adenocarcinoma (TCGA-PAAD) cohort were integrated with Genotype-Tissue Expression (GTEx) normal pancreas data. Vascular invasion-associated candidate genes were identified using subgroup-specific differential expression filtering. A three-gene prognostic signature was constructed using Cox and least absolute shrinkage and selection operator (LASSO) regression and validated in an independent PACA-AU cohort. Nucleoporin 35 (NUP35) was functionally evaluated by shRNA knockdown, phenotypic assays, endothelial assays using conditioned medium, and Western blotting of extracellular signal-regulated kinase (ERK)-vascular endothelial growth factor A (VEGFA) signaling. Results: We identified 172 vascular invasion-associated candidate genes and developed a three-gene model comprising NUP35, GMNN, and KLK11. The model stratified TCGA patients into risk groups with significantly different overall survival (OS; log-rank p = 0.014) and good predictive performance, with areas under the receiver operating characteristic curve (AUC) of 0.659, 0.722, and 0.796 for 1-, 3-, and 5-year OS, respectively. Consistent trends were observed in PACA-AU. Risk group transcriptomes were enriched for proliferative and tumor progression programs. Among the signature genes, NUP35 was prioritized because higher NUP35 expression was associated with poorer survival and positively correlated with VEGFA expression. NUP35 knockdown suppressed malignant phenotypes, reduced endothelial migration and tube formation, and decreased phosphorylated ERK and VEGFA without altering total ERK levels. Conclusions: A vascular invasion-related three-gene signature enables prognostic stratification in PDAC. NUP35 is associated with malignant and pro-angiogenic phenotypes and may regulate angiogenic activity partly through ERK-VEGFA signaling, supporting its potential as a prognostic biomarker and candidate therapeutic vulnerability.
Sun et al. (Sat,) studied this question.