Transcription factor activity profiling uncovers prognostic subtypes and a novel oncogenic role for PRDM4 in NSMP endometrial carcinoma

BACKGROUND: Among the four molecular subtypes of endometrial carcinoma (EC), the no specific molecular profile (NSMP) group is the largest, comprising roughly 40-50% of cases. Patients within this subtype follow markedly different clinical courses, yet no reliable tools exist to separate those who will do well from those at risk of recurrence. Because transcription factors (TFs) sit at the top of gene regulatory hierarchies, their activity patterns may capture biological differences that conventional expression signatures miss. METHODS: We analyzed transcriptomic, mutation, and clinical data from TCGA-UCEC. An EC-specific regulatory network was built with ARACNe-AP, and per-sample TF protein activity was estimated using VIPER. Consensus clustering on the activity matrix defined novel subtypes within the NSMP population. We characterized these subtypes through pathway, immune infiltration, and mutation analyses. A LASSO-Cox model based on TF activities was trained and tested in an independent GEO cohort, then combined with clinical variables into a nomogram. The top-ranked master regulator, PRDM4, was investigated by immunohistochemistry, shRNA knockdown, overexpression experiments, and chromatin immunoprecipitation (ChIP-qPCR). RESULTS: Three subtypes (C1-C3) emerged with clearly separated survival curves (log-rank P < 0.0001). C1 tumors showed active immune signaling, frequent PTEN/PIK3CA mutations, and good outcomes. C3 tumors were enriched for epithelial-mesenchymal transition and hypoxia programs, carried an immunosuppressive microenvironment dominated by regulatory T cells and M2 macrophages, harbored more TP53 mutations, and had the worst prognosis. A nine-TF risk score achieved 3-year AUC values of 0.775 (training) and 0.579 (validation). The accompanying nomogram showed good calibration and decision-curve performance. PRDM4 was the master regulator most strongly linked to the aggressive C3 phenotype. In functional assays, PRDM4 drove proliferation, migration, and invasion while suppressing apoptosis. ChIP-qPCR confirmed direct binding of PRDM4 to the LATS2, VEGFA, and ITGA5 promoters. CONCLUSIONS: TF activity profiling splits NSMP EC into prognostically meaningful groups and yields a workable risk model. PRDM4 functions as an oncogenic transcription factor in this context, offering a candidate biomarker and potential therapeutic target for high-risk NSMP patients.