Digital image analysis workflows for evaluation of cell behavior and tumor microenvironment to aid therapeutic assessment in high-risk neuroblastoma.

Digital pathology and artificial intelligence are promising emerging tools in precision oncology as they provide more robust and reproducible analysis of histologic, morphologic and topologic characteristics of tumor cells and the surrounding microenvironment. This study aims to develop digital image analysis workflows for therapeutic assessment in preclinical in vivo models. For this purpose, we generated pipelines that enable automatic detection and quantification of vitronectin and αvβ3 in heterotopic high-risk neuroblastoma xenografts, demonstrating that digital analysis workflows can be used to provide robust detection of vitronectin secretion and αvβ3 expression by malignant neuroblasts and to evaluate the possibility of combining traditional chemotherapy (etoposide) with extracellular matrix-targeted therapies (cilengitide). Digital image analysis added evidence for the relevance of territorial vitronectin as a therapeutic target in neuroblastoma, since its expression is modified after treatment, with a mean percentage of 60.44% in combined therapy tumors vs 45.08% in control ones. In addition, the present study revealed the efficacy of cilengitide for reducing αvβ3 expression, with a mean αvβ3 positivity of 34.17% in cilengitide treated material vs 66.14% in control and with less tumor growth when combined with etoposide, with a final mean volume of 0.04 cm in combined therapy vs 1.45 cm in control. The results of this work highlight the importance of extracellular matrix-focused therapies in preclinical studies to improve therapeutic assessment for high-risk neuroblastoma patients.