Microscopic spatial organization of immune cells and stochastic fluctuations regulate patient response to immune checkpoint inhibitor drugs in melanoma

Recent innovations have enabled measurements of protein expressions in single cells as well as their spatial locations in the tumor microenviroment (TME). These multiscale datasets provide a unique opportunity to glean mechanisms and develop predictive models regarding how different cells interact in space to generate diverse responses and affect disease progression in the context of immune checkpoint inhibition (ICI) therapy. Utilizing publicly available imaging mass cytometry (IMC) data of melanomas of patients who later underwent ICI therapy, we use densities and pair correlations to extract spatial relationships between cells of interest and formulate hypotheses about their mechanistic relationships. We develop an interacting cell system to simulate and test various mechanisms of interplay between immune and cancer cells. By observing the spatial configurations generated by the spatial kinetic Monte Carlo simulation, we discover a unique spatial feature wherein exhausted CD8+ T cells line the periphery of the tumor mass and block activated CD8+ T cells from accessing and killing tumor cells. We find corroborative evidence for this fencing occurring in other publicly available melanoma data. Furthermore, our simulations show that despite the stochasticity, tumor progression fate may be predicted at early times in some cases.