An in vivo model to evaluate donor-dependent cytokine release in response to single-agent or combination immune-oncology therapies.

Abstract

3114 Background: Although immune-oncology therapies such as checkpoint inhibitor, bi-specific antibody and CAR-T cell therapies are successfully used for cancer therapy, they can have very severe adverse effects such as cytokine release syndrome (CRS). The animal models and in vitro human PBMC assays presently in use do not reliably predict CRS in patients. Currently, the only widely accepted predictors of CRS are cancer burden and therapeutic dose. Despite this, most pre-clinical assays that evaluate CRS do not incorporate cancer cells and the safety of drug combinations has not been widely explored. A predictive assay that identifies patient/cancer/therapy combinations at risk for developing CRS upfront in addition to treatment efficacy would improve the safety of immune-oncology drug development. Methods: We have developed sensitive in vivo humanized mouse models for quantitating CRS that are rapid, reproducible and able to show variation among PBMC donors. The NSG mouse and its derivatives are engrafted with cancer cells and human PBMCs. Mice are then dosed with checkpoint inhibitors or bi-specific antibodies as a single therapy or in combination. Cytokine release is evaluated 2-6 hours post dosing. This assay can be modified to also evaluate efficacy by using luciferase labeled cancer cells and monitoring tumor burden using the Xenogen IVIS imaging system. Results: For all therapy groups, each cytokine tested (including human IFN-γ, IL-2, IL-6, IL-10 and TNF) was upregulated 2-6 hours after drug treatment, but different PBMC donors had various cytokines release levels. Cytokine release levels correlated with a dose response, PBMC engraftment levels and tumor burden. We can demonstrate additive and synergistic cytokine release in the combination treated groups and compare efficacy versus single agents. Our in vivo method was able to determine CRS missed in the in vitro testing method. Conclusions: We have developed a rapid, sensitive and reproducible novel in vivo PBMC humanized mouse model that can differentiate human PBMC donors based on individual safety response to single agent and combination therapeutics of immune checkpoint inhibitors and bispecific T-cell-engaging antibodies. Additionally, this assay can utilize luciferase labelled cell lines to measure treatment efficacy. Using this assay, we can potentially evaluate both cytokine release and efficacy of current immune-oncology therapies as single agents and in combination. This assay has immediate utility in current and future drug development.