Multiscale modeling of the spatial structure of stem cells in neuroblastoma patient-derived tumoroids reveals a...

Neuroblastomas are heterogeneous pediatric tumors of the sympathetic nervous system for which treatments are still limited. Fundamental and applied approaches have been enabled thanks to the generation of patient-derived tumoroids (PDT), ex vivo 3D structures used as avatars of the original tumor. We generated neuroblastoma PDT and quantified the spatial distribution of CD133+ cancer stem cells using immunohistochemistry. We observed that those cells tend to aggregate in the PDT. In order to better understand the set of rules needed for generating such structures, we implemented a multiscale agent-based neuroblastoma tumoroid model. Model rules specify single cell's fate based on their intracellular content, which dynamically evolves according to a stochastic gene regulatory network. The state of this network can be modulated by cell-to-cell signaling through neighbor cell fate decisions and, possibly, spatial location. We first observed that in the absence of any spatial rules for inter-cellular interactions, no spatial structure emerged. The addition of simple rules (signaling by cell-to-cell contact or differential cell adhesion) only marginally improved the quantitative agreement to the experimental dataset. In sharp contrast, the addition of short-range pro-stem cell diffusive signaling among stem cells produced very realistic 3D PDT-like structures. This works highlights the power of our multiscale approach to discard too simplistic rules and to propose a minimal set of hypotheses required to reproduce qualitatively and quantitatively experimentally observed spatial structures. In the case of neuroblastoma-derived PDT, short-range spatial diffusion of stem-to-stem cell signaling proved to play a key role in successfully reconstructing the spatial structure.