Professor University of Minnesota-Twin Cities Minneapolis, Minnesota, United States
Introduction:: Glioblastoma is a fast-growing, lethal malignant brain tumor characterized by diffuse infiltration into the surrounding brain tissue. The standard of care for the disease has remained relatively constant since the approval of temozolomide chemotherapy nearly two decades ago, giving patients limited treatment options. Advances in genomic characterization methods have led to the distinction of three molecular subtypes in glioblastoma: proneural, classical, and mesenchymal. These subtypes have distinct cell migration and immune response behaviors and as a result, may require different treatment regimens to effectively eradicate the disease. Adequate in vitro models are challenging to develop and in vivo experiments require significant amounts of time and funding. Computational models can generate fast, high-throughput simulations of tumor growth and development. This work utilizes a three-dimensional Brownian dynamics-based tumor simulator capable of modeling cancer cell migration and proliferation behaviors as well as cancer cell-cytotoxic T cell interactions. A combination of in vitro and in vivo data from experiments using a genetically engineered mouse model that recapitulates glioblastoma proneural and mesenchymal subtypes is used to parameterize this model. This data-driven approach enables the simulation of therapies on tumors of different subtypes using experimentally-informed parameter values.
Materials and Methods:: Cancer cells are approximated as rigid spheres that are able to migrate, grow and divide (if necessary) at each time step. No cell overlapping is permitted. Cell migration is approximated as a 3D random walk. Each cell grows as a single sphere following an exponential growth rate. A cell will grow until it reaches a maximum size, after which it divides into two daughter cells directly next to each other with equal radii, such that volume and the cell’s original center of mass are conserved. The parameter values for cell migration speed and proliferation rate were determined from in vitro assays using cell lines derived from a genetically-engineered mouse model of glioblastoma that recapitulates human subtypes.
Cytotoxic CD8+ T cells are approximated as spheres. They are introduced into the simulation once a certain threshold number of cancer cells is reached. T cell migration is confined to the tumor volume, which is defined by calculating the farthest distance of any cell from the origin. To determine the initial location of a T cell, random coordinates within the tumor volume are chosen.
Cancer cell and T cell health are both approximated using a “hit points'' system to represent sublethal hits. When a T cell delivers a cytotoxic hit to a cancer cell, both the T cell and the cancer cell lose one hit point. Once a cell’s hit point value has been depleted to 0, it is removed from the simulation, representing cell death (cancer cells) or exhaustion (CD8+ T cells).
Results, Conclusions, and Discussions:: The simulator is parameterized using an NRAS-driven model model and a PDGFB-driven mouse model, which recapitulate human mesenchymal and proneural subtypes of glioblastoma, respectively. NRAS-driven tumors have faster cell migration speed and increased levels of CD8+ T cell infiltration. Simulated NRAS-driven tumors are more diffuse and have greater tumor volumes than PDGFB-driven tumors, which have cancer cells tightly packed in the center of the tumor volume.
Little is known at present about how cancer cells and CD8+ T cells interact in vivo in glioblastoma. Simulated parameters guiding T cell behavior include T cell migration speed, time of initial T cell infiltration, T cell injection rate, and number of sublethal hits given before reaching exhaustion (T cell hit points). These parameters are varied to predict effective immunotherapy treatment strategies.
This simulator can be used to visualize tumor growth over time while capturing single cell dynamics. By parameterizing the model with migration and immune cell infiltration data from a subtype-specific murine model of glioblastoma, this system can be used to make fast predictions and guide the direction of future in vivo experiments. This model will be used for larger scale simulations of full murine tumor development (4-7 weeks simulated time) to gain further insight into in vivo tumor progression. It can also be used to simulate various immunotherapy strategies in combination with chemotherapy, which could guide future in vivo experiments.
Acknowledgements (Optional): : This project was supported by the National Institutes of Health under award numbers U54CA2210190, P01CA254849, and U54CA268069.
References (Optional): : 1.Shamsan et al., bioRxiv, 2022
