Computational modeling of breast cancer metastases predicts response to dual checkpoint inhibition and reveals spatially-informed treatments
In advanced breast cancers, the highly suppressive tumor microenvironment contributes to poor responses to immune checkpoint inhibition. The degree of suppression varies by site of disease, which can affect metastatic progression. To investigate this phenomenon, we have developed an agent-based model of cell-cell interactions within the tumor microenvironment, specifically tuned to in vivo data from murine breast-to-lung metastases. The model includes nine functional cell types (cancer cells, M0-, M1-, and M2-like macrophages, CD4+ T cells, T regulatory cells [Tregs], CD8+ T cells, NK cells, and MDSC-like cells). We incorporated a novel continuum model to describe exhaustion of CD8+ T and NK cells, where their exhaustion is impacted by cell-cell interactions in the TME via direct and indirect interactions. We also modeled the effect of the immune checkpoint inhibitors (ICI) anti-PD-1 and anti-CTLA-4, termed "dual ICIs". We used this model to simulate the spatiotemporal trajectories of 19 metastases without treatment and with dual ICIs. The model predicted that dual ICIs lead to statistically significant reductions in cancer cell proliferation in seven of the metastases (37%), and spatial analyses of these metastases indicate they have structural similarities at multiple scales of tumor organization. Additionally, spatial analysis suggests that high colocalization of CD8+ T cells with Tregs could prevent dual ICIs from being effective, and simulation of Treg blockade combined with dual ICIs led to 15 total metastases having a statistically significant treatment effect (80% total). Thus, our model provides a highly resolved spatiotemporal platform for studying the cell-cell interactions that can drive resistance to ICIs in diverse TME settings and identifies potential therapeutic avenues to enhance the efficacy of emerging ICI therapies.