Rational design of chimeric antigen receptor T cells against glypican 3 decouples toxicity from therapeutic efficacy.

Abstract

Chimeric antigen receptor (CAR) T cell therapy has yielded impressive clinical results in hematological malignancies and is a promising approach for solid tumor treatment. However, toxicity, including cytokine-release syndrome (CRS) and neurotoxicity, is a concern hampering its broader use. In selecting a lead CAR-T candidate against the oncofetal antigen glypican 3 (GPC3), we compared CARs bearing a low- and high-affinity single-chain variable fragment (scFv) binding to a similar epitope and cross-reactive with murine GPC3. Where the high-affinity CAR-T cells were toxic in vivo, the low-affinity CAR maintained cytotoxic function against antigen-positive tumor cells but did not show toxicity against normal tissues. High-affinity CAR-induced toxicity was caused by on-target, off-tumor binding, based on the observation that higher doses of the high-affinity CAR-T caused toxicity in non-tumor-bearing mice and accumulated in organs with low expression of GPC3. To explore another layer of controlling CAR-T toxicity, we developed a means to target and eliminate CAR-T cells using anti-TNF-α antibody therapy after CAR-T infusion. The antibody was shown to function by eliminating early antigen-activated, but not all, CAR-T cells, allowing a margin where the toxic response could be effectively decoupled from antitumor efficacy with only a minor loss in tumor control. By exploring additional traits of the CAR-T cells after activation, we identified a mechanism whereby we could use approved therapeutics and apply them as an exogenous kill switch that eliminated early activated CAR-T following antigen engagement in vivo. By combining the reduced-affinity CAR with this exogenous control mechanism, we provide evidence that we can modulate and control CAR-mediated toxicity.