Abstract
BACKGROUND: Risk stratification and personalized care are crucial in managing osteosarcoma due to its complexity and heterogeneity. However, current prognostic prediction using clinical variables has limited accuracy. Thus, this study aimed to explore potential molecular biomarkers to improve prognostic assessment. METHODS: High-throughput inhibitor screening of 150 compounds with broad targeting properties was performed and indicated a direction towards super-enhancers (SEs). Bulk RNA-seq, scRNA-seq, and immunohistochemistry (IHC) were used to investigate SE-associated gene expression profiles in osteosarcoma cells and patient tissue specimens. Data of 212 osteosarcoma patients who received standard treatment were collected and randomized into training and validation groups for retrospective analysis. Prognostic signatures and nomograms for overall survival (OS) and lung metastasis-free survival (LMFS) were developed using Cox regression analyses. The discriminatory power, calibration, and clinical value of nomograms were evaluated. RESULTS: High-throughput inhibitor screening showed that SEs significantly contribute to the oncogenic transcriptional output in osteosarcoma. Based on this finding, focus was given to 10 SE-associated genes with distinct characteristics and potential oncogenic function. With multi-omics approaches, the hyperexpression of these genes was observed in tumor cell subclusters of patient specimens, which were consistently correlated with poor outcomes and rapid metastasis, and the majority of these identified SE-associated genes were confirmed as independent risk factors for poor outcomes. Two molecular signatures were then developed to predict survival and occurrence of lung metastasis: the SE-derived OS-signature (comprising LACTB, CEP55, SRSF3, TCF7L2, and FOXP1) and the SE-derived LMFS-signature (comprising SRSF3, TCF7L2, FOXP1, and APOLD1). Both signatures significantly improved prognostic accuracy beyond conventional clinical factors. CONCLUSIONS: Oncogenic transcription driven by SEs exhibit strong associations with osteosarcoma outcomes. The SE-derived signatures developed in this study hold promise as prognostic biomarkers for predicting OS and LMFS in patients undergoing standard treatments. Integrative prognostic models that combine conventional clinical factors with these SE-derived signatures demonstrate substantially improved accuracy, and have the potential to facilitate patient counseling and individualized management.