Abstract Introduction: Osteosarcoma, the most common primary bone tumor in dogs, has a guarded prognosis. A major hurdle in developing more effective therapies is the lack of osteosarcoma-specific biomarkers to predict risk, prognosis, or therapeutic response. Exosomes are secreted microvesicles emerging as powerful diagnostic tools. However, wide clinical use of exosomes is precluded by the difficulty in identifying diseased cargo from the vastly larger background of normal cargo. We developed a method to distinguish tumor-derived exosome cargo from normal background, allowing for identification of an osteosarcoma-specific gene signature. Methods: Serum exosomes were enriched from osteosarcoma xenografts and control mice. Enriched genes associated with canine osteosarcoma were identified with RNA-sequencing. From identified candidates, we defined an osteosarcoma-associated gene signature. qRT-PCR amplification validated the gene signature in serum exosomes from clinical canine cases. Machine learning algorithms classified patients into disease groups based on this gene signature. Results: We identified an osteosarcoma-associated signature consisting of five mRNAs (SKA2, NEU1, PAF1, PSMG2, and NOB1). Serum exosomes were isolated from dogs in clinical groups, including “healthy,” “osteosarcoma,” “other bone tumor,” or “non-neoplastic disease.” Machine learning classified samples, with 82% and 86% of untrained samples predicted as osteosarcoma by CN2 and RF models, respectively. Post-treatment samples “misclassified” as non-osteosarcoma were associated with longer remissions, potentially from a lack of remaining osteosarcoma cells. Conclusions: We identified a gene signature associated with canine osteosarcoma. This gene signature was validated by qRT-PCR with serum exosomes from patients with osteosarcoma, as well as used to train artificial intelligence to correctly classify canine patients according to disease group with up to 93.8% accuracy. Clinical Significance: This study combines a bioinformatics approach to biomarker discovery with machine learning to correctly identify osteosarcoma in canine patients. These results set the stage for future discoveries to inform cancer risk, diagnosis, prognosis, and response to therapy. Citation Format: Kelly M. Makielski, Alicia J. Donnelly, Milcah C. Scott, Hirotaka Tomiyasu, Ali Khammanivong, William C. Kisseberth, Jaime F. Modiano. Validation of an exosomal osteosarcoma-associated gene signature in dogs with osteosarcoma [abstract]. In: Proceedings of the AACR Special Conference on Advances in Liquid Biopsies; Jan 13-16, 2020; Miami, FL. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(11_Suppl):Abstract nr B59.
Read full abstract