TmPSMA-02: A CD2 endodomain containing double armored PSMA CAR T with enhanced efficacy and lower immune toxicity.

Abstract

158 Background: Early clinical responses to anti-PSMA Chimeric Antigen Receptor T cells (CARs) have been observed but much like the leukemia experience, significant toxicity from macrophage activation (MAS) has also been observed Building on our experience with a first in-human armoured CAR (TmPSMA-01) targeting PSMA (with a 41BB and CD3 endodomain) coupled with a dominant-negative TGFBR2 module, where we observed anti-tumor effects but also immune-mediated toxicity, we designed a CAR with additional armour, a PD1-CD28 switch, to overcome the immune suppressive microenvironment and further improve efficacy. We successfully built tricistronic constructs that expresses the CAR, a dnTGFBR2 module and a switch receptor for PD1 and CD28. We compared the 41BB endodomain and the CD2 domain, which has shown preliminary effects reducing cytokine release. We hypothesized that these changes would increase efficacy in immune-suppressive solid tumors while reducing excessive cytokine release and untoward immune toxicity. Methods: CAR constructs were synthesized into a self-inactivating lentiviral backbone and CAR T cells manufactured comprising a control TmPSMA-01 construct and two primary constructs: PSMA-CD2z:dnTGFBR:PD1-CD28 and PSMA-41BBz:dnTGFBR:PD1-CD28. Expression of the CAR as well as T cell markers were determined by flow cytometry, and CAR affinity for target by Lumicks. CAR activation and cytotoxicity were assessed by ELISPOT, xCelligence and Celligo standard assays. A co-culture macrophage activation system was applied to each construct to assess the potential of each construct to induce MAS-like responses. Results: A significant difference was seen in the MFI of the CAR, with the CD2 based CAR consistently ̃50% lower than the 41BB and this correlated to a faster off-rate from target cells measured by Lumicks. There was no difference in exhaustion or memory markers, and cytotoxicity across a range of E:T ratios was equivalent between the CD2 and 41BB constructs. Preliminary data indicated the CD2 endodomain preserved T cell activation but with lower cytokine release. The CD2 constructs consistently produced significantly less total IL2 and IFNg after exposure to targets, yet had similar numbers of ELISPOT positive cells for IL2 and IFNg indicating less cytokine production per kill. This also resulted in significantly less macrophage based cytokines such as IL6 and IL8 in the coculture model. Using a cell line knocked out for PDL1/2 and then reengineered for constitutive PDL1/2 expression, we demonstrated improved speed of killing in the xCelligence system dependent on the PD1:CD28 switch but no increase in cytokine release. Conclusions: A CD2 based endodomain tricistronic CAR targeting PSMA (TmPSMA-02) maintained cytotoxicity, memory profile and activation while having a significantly reduced risk of inducing macrophage activation.