Identifying osteosarcoma signatures in cell free DNA using 5-hydroxymethylcytosine profiling.

Abstract

10022 Background: Osteosarcoma is the most common primary bone tumor in children and adolescents. Prognostic biomarkers are limited, and post-treatment surveillance relies heavily on radiologic imaging. Capture and analysis of cell-free DNA (cfDNA) from the bloodstream is an emerging technique that could be used to provide additional prognostic information and track residual disease following treatment. 5-hydroxymethylcytosine (5-hmC) is an epigenetic marker of open chromatin and active gene expression, and tumor specific patterns of global 5-hmC deposition are identifiable in several cancers and vary according to disease burden. We profiled 5-hmC using plasma-derived cfDNA from osteosarcoma patients and healthy children to assess its utility as a biomarker of disease status. Methods: cfDNA was isolated from patients diagnosed with osteosarcoma and well child controls. Nano-hmC-Seal, a technique developed to isolate 5-hmC-modified DNA fragments, was used to generate sequencing libraries of cfDNA. Genomic DNA from osteosarcoma cell lines was serially diluted with healthy donor-derived cfDNA. DESeq2 was used to identify genes with high 5-hmC deposition in osteosarcoma cell line DNA relative to serial dilutions. In patient cfDNA samples, the relative abundance of 5-hmC deposition on these osteosarcoma genes was determined using gene set variation analysis (GSVA). GSVA returns a score ranging from -1 to 1, with positive values indicating a relative abundance of 5-hmC deposition on osteosarcoma genes. Results: The cohort consisted of 33 healthy children and 17 patients with osteosarcoma including 53 serially collected samples from these patients. 139 osteosarcoma genes were identified and used for GSVA analysis. Patients with osteosarcoma had significantly higher GSVA scores than well children (0.26 vs -0.38; p < 0.001). Hierarchal clustering identified two clusters of samples from children with osteosarcoma. Cluster 1 contained samples with scores correlating with osteosarcoma-derived DNA compared to Cluster 2 (0.14 vs -0.18; p < 0.001). Furthermore, Cluster 1 contained 17 of 23 samples from patients with active disease on clinical imaging (sensitivity = 74%, specificity = 57%). High 5-hmC deposition on MYC was readily identified in one sample obtained at diagnosis from the only patient with a MYC amplified tumor. Additional samples from this patient during treatment showed decreasing 5-hmC deposition on MYC, consistent with a treated tumor shedding less DNA into the blood. Additionally, decreased disease burden following surgical resection was identifiable in 5-hmC profiles (presurgical GSVA score 0.22 vs postsurgical -0.08; p = 0.004). Conclusions: Osteosarcoma 5-hmC signatures can identify patients with active disease. Monitoring cfDNA in osteosarcoma using 5-hmC profiling is a promising technique that warrants further investigation with larger patient cohorts.