Abstract 6047: Identification of a 13 gene signature to predict survival in localized osteosarcoma

Abstract Osteosarcoma (OS) is the most prevalent bone tumor in pediatric patients. Regimens of neoadjuvant chemotherapy have improved survival of OS patients greatly, however the 5-year survival rate for localized OS is 75% with a 30-50% recurrence rate. We sought to identify genes which could predict chemo-response and survival in localized OS. The TARGET OS RNA-seq dataset was utilized to identify genes and pathways associated with localized patient relapse and survival. We identified 478 differentially expressed genes with a 1.5 FC and FDR < 0.05 common to overall survival and relapse We further performed string analysis to generate a protein-protein interaction network followed by hub analysis with Cytohubba using betweenness centrality and radiality measures. Combining the top 10 hub genes from these two methods resulted in a total of 13 genes: MYOM2, VEGFA, VCAM1, EGFR, MUC1, IHH, GLI1, GPC3, IGF2, GRIA1, GNG12, GNGT1 and C3. These 13 genes were used to stratify localized patients in the TARGET dataset into high-risk and low-risk tertiles. The low-risk group had 100% overall survival while the high-risk group had 44% 5-year survival (p=2e-4). We also found a significant correlation between the 13 genes and time to death in localized patients (p=0.04). Additionally, there was a significant difference in expression of the 13 genes between alive and deceased patients (p=2e-5) and patients who relapsed (p=1.5e-4). Overall, these data suggest that these 13 genes could predict relapse and overall survival in OS patients with localized disease in the TARGET cohort. We performed Weighted Gene Correlation Network Analysis (WGCNA) on the 478 overlapping genes and identified five modules, with our 13 genes split across these modules. All modules were also significantly correlated with vital status suggesting that the genes in our signature represent distinct sub-groups with possibly separate mechanisms. Over-representation analysis was performed for each module and while each module did have distinct pathways, there were 65 pathways which overlapped between 3 of the modules. Of particular interest was Hedgehog signaling, with 2 of our 13 genes, IHH and GLI1, key to Hedgehog signaling, and a Hedgehog pathway inhibitor, Gant-58, scored high in reversing the 478 gene signature as determined using Connectivity Map (Broad Institute). We tested Gant-58 against two PDX OS models. Gant-58 did not inhibit a non-relapsed, chemo sensitive localized PDX-derived cell line, but showed potent activity towards a recurrent localized PDX with elevated IHH and GLI1 expression (p<0.0001). In summary, we identified 13 genes that predict overall survival and relapse in localized OS patients. The 13 genes represent distinct modules of co-expressing genes that significantly correlate with survival. Furthermore, preliminary data indicate Hedgehog pathway has a key role in survival and recurrence of localized OS patients. Citation Format: Tajhal D. Patel, Kshiti Dholakia, Tanmay R. Gandhi, Rupa S. Kanchi, Sandra L. Grimm, Chenlian Fu, Jason T. Yustein, Cristian Coarfa. Identification of a 13 gene signature to predict survival in localized osteosarcoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6047.