Using protein geometry to optimize cytotoxicity and the cytokine window of a ROR1 specific T cell engager.

T cell engaging bispecific antibodies have shown clinical proof of concept for hematologic malignancies. Still, cytokine release syndrome, neurotoxicity, and on-target-off-tumor toxicity, especially in the solid tumor setting, represent major obstacles. Second generation TCEs have been described that decouple cytotoxicity from cytokine release by reducing the apparent binding affinity for CD3 and/or the TAA but the results of such engineering have generally led only to reduced maximum induction of cytokine release and often at the expense of maximum cytotoxicity. Using ROR1 as our model TAA and highly modular camelid nanobodies, we describe the engineering of a next generation decoupled TCE that incorporates a "cytokine window" defined as a dose range in which maximal killing is reached but cytokine release may be modulated from very low for safety to nearly that induced by first generation TCEs. This latter attribute supports pro-inflammatory anti-tumor activity including bystander killing and can potentially be used by clinicians to safely titrate patient dose to that which mediates maximum efficacy that is postulated as greater than that possible using standard second generation approaches. We used a combined method of optimizing TCE mediated synaptic distance and apparent affinity tuning of the TAA binding arms to generate a relatively long but persistent synapse that supports a wide cytokine window, potent killing and a reduced propensity towards immune exhaustion. Importantly, this next generation TCE induced significant tumor growth inhibition but unlike a first-generation non-decoupled benchmark TCE that induced lethal CRS, no signs of adverse events were observed.