Molecular Landscape of Canine Osteosarcoma

Abstract

Osteosarcoma affects about 2.8% of dogs with spontaneous cancer and its occurrence is significantly higher than observed in humans. The prognosis of this disease is relatively poor, where the survival rate of ~45% of dogs with bone cancer is one year. With the recent advancement in characterizing mutational landscapes of various cancer types, personalized medicine has become increasingly important for cancer treatment. The purpose of this study is to characterize the mutation and expression profile of canine osteosarcoma. Using whole exome capture (Agilent Sure Select Canine V2) we sequenced 26 primary tumors and matched normal samples, with coupled treatment and outcome data. These samples were previously analyzed for gene expression using Affymetrix Canine 2.0 microarrays.The BWA tool was used to map Illumina reads to the CanFam3.1 genome and short variants were called and annotated using Mutect2 and SnpEff tools, respectively. Additionally, copy number variations were called and annotated using Sequenza and GISTIC tools. The total number of somatic variants identified across 26 samples ranged from 148 to 3,785 The protein coding mutations per Megabase ranged from 0.25 to 11.92. Among the protein coding somatic variants identified as SNVs and INDELS, an average of 80% were missense mutations (range 64% to 92%). The top recurrent mutations were observed in TP53 (82% of the samples), SETD2 (22%). We have also identified 56 known cancer genes with copy number variations (amplification or deletion) in at least 58% of the samples. When compared to pediatric osteosarcoma, there was an overlap of 38 genes (eg: TFEB, MYC) that were recurrently amplified and 69 genes (eg: RB1, MSH3) were recurrently deleted in both canine and human osteosarcoma.Using the microarray data, we have identified 585 differentially expressed genes between normal bone and osteosarcoma samples. Using GeneSet Expression Analysis (GSEA) we observed down‐regulation of myogenesis genes and up‐regulation of extracellular matrix genes in tumor samples. Additionally, we have quantified T‐cell and macrophage infiltration in the tumors via immunohistochemistry. Across 20 samples, the median percentage of tumor area with CD3+ and MAC387 positive cells was 0.14% and 0.21%, respectively. There was no significant correlation between disease‐free interval (DFI) and median dichotomized T‐cell or macrophage infiltration. However, Kaplan‐Meier analysis revealed that dogs with TP53 missense mutations had significantly longer DFIs than those with TP53 wildtype/null mutations. Further, correlation of gene expression and DFI can provide us with prognostic biomarkers for osteosarcoma in dogs.Support or Funding InformationAnschutz Foundation and Morris Animal Foundation