Abstract
BACKGROUND: Osteosarcoma is the most common primary malignant bone tumor, with high invasiveness and metastatic potential and a poor prognosis in patients with metastatic cancer. Despite the rapid advancements in genomics in recent years that provided new perspectives for studying the molecular mechanisms of osteosarcoma, the understanding of its tumor heterogeneity and evolutionary mutation process remains limited. METHODS: In this study, whole-exome evolutionary profiling was performed on data from the TARGET database representing 61 osteosarcoma cases. Subclonal architectures were reconstructed to characterize mutational trajectories. Differential mutation analysis was used to identify candidate metastasis-associated mutations. These features were used to build a metastasis-prediction classifier, which was cross-validated and tested on an independent external cohort. Finally, Suppes' probabilistic theory of causality was integrated with cohort data to infer high-frequency evolutionary paths linked to metastasis. RESULTS: A linear evolutionary trajectory was observed in 62% of patients, indicating sequential clonal expansion. Eight key mutations were closely associated with metastatic progression. The classifier achieved 83% accuracy in cross-validation and maintained robust performance on the external validation set. Through causal inference, distinct evolutionary routes underpinning metastasis were uncovered, with ATRX mutations frequently occurring as early events that reshaped clonal dynamics and facilitated tumor spread. CONCLUSIONS: In this study, the dynamic evolutionary landscape of osteosarcoma metastasis was delineated, an early metastasis classification model was constructed, and the impact of early clonal ATRX mutations on metastasis initiation were highlighted. These findings offer potential avenues for the early diagnosis and risk assessment of osteosarcoma.