Development of a Regulated, Optimized Mesothelin Chimeric Antigen Receptor (CAR) for the Treatment of Mesothelin Positive Cancers

Introduction: Mesothelin (MSLN) is an attractive target for chimeric antigen receptor T cell (CAR-T) immunotherapy as it is highly expressed on mesothelioma, ovarian, lung cancer, and some forms of leukemia, including approximately 30% of acute myeloid leukemia (AML); however, MSLN directed CAR-T cells are prone to exhaustion, and soluble proteins, such as shed MSLN and MUC16 can inhibit anti-tumor efficacy. To address these deficiencies, we developed a novel, fully-human MSLN-specific CAR using our OutSpacer™ technology to optimize the immune synapse geometry. Further, this MSLN CAR is expressed using a novel regulatory response element to dynamically express the CAR molecule to enhance CAR-T potency and durability.

Methods: A set of fully-human single domain binders (VHH) were produced and screened against a spacer library to identify high-performing CARs. The spacer library is composed of human extracellular polypeptides selected for optimal biophysical properties (e.g., flexibility), enabling systematic sampling to identify CAR constructs with optimal immune synapse geometry. More than 300 MSLN-directed CAR candidates were evaluated in vitro using 2D or 3D (spheroid) culture assays to measure tumor killing and cytokine production against several MSLN ⁺ cell lines that express different levels of the target protein. These functional assays were also performed in the presence of recombinant MSLN (10ug/mL) or MUC16 (1ug/mL) to assess potential inhibition by shed target or ligand. Top performing MSLN CARs were subsequently tested in NSG mice (2e6, 4e6 and 8e6 cells/animal; 2 donors) subcutaneously engrafted with MSLN ⁺ H1650 tumor cells to identify the best CAR candidate. To further improve the durability of T cell response, the MSLN CAR was expressed under the control of an engineered promoter (OP1) to dynamically regulate expression, and compared to CARs expressed constitutively using the MND promoter. CAR-T cells were repetitively stimulated (4x) by plate-bound antigen in combination with CD58 and ICAM1, and subsequently tested in a tumor spheroid assay for killing, cytokine production, CAR expression and T cell activation/ exhaustion markers (i.e., PD-1, TIGIT, CD39 and TIM3).

Results: Anti-MSLN VHH domains targeting region 1 or 3 (RI and RIII) of the MLSN extracellular domain were paired with a defined spacer library to define the optimal immune synapse. Contrary to the conclusions of prior studies, RI (membrane-distal) binders outperformed RIII (membrane-proximal) binders when paired with their optimal spacer. Based on tumor killing and cytokine production, RI targeted CARs preferred a spacer length of approximately 75 Å, consistent with our previously-reported optimal CAR synapse of ~200 Å. All spacer optimized MSLN-specific CARs resisted inhibition by soluble MSLN and MUC16. In vivo studies with two donors in NSG mice engrafted with MSLN ⁺ tumors showed that one CAR (B8S8) eliminated tumors at a dose of 2e6 and 4e6 CAR-T cells per animal, where other CAR constructs, including a clinically-used MSLN-specific CAR, failed to control tumors even at 8e6 cells per animal. Secondly, we assessed whether a novel “stim-on” promoter (OP1) could enhance CAR-T activity by preserving T cell function. Here, B8S8 CAR was expressed via OP1 or the constitutive MND promoter (n=3 donors). Following chronic antigen exposure by plate-bound antigen, OP1 regulated MSLN CAR T cells retained the capacity to eliminate MSLN ⁺ H226 tumor spheroids compared to MND-driven CAR T constructs. OP1-regulated CAR T cells also continued to produce high levels of cytokines (IFN-γ and IL-2) whereas MND CAR-T cells lost cytokine production ability. Further, phenotypic analyses showed that MND CAR-T cells expressed elevated levels of PD-1 and TIGIT compared to OP1 CAR-T cells.

Conclusions: Contrary to the conclusions of prior studies with MSLN-targeted CARs, membrane distal epitopes were superior to membrane proximal epitopes when binders were paired with the correct spacer, and shed MSLN and MUC16 were not inhibitory for the best-performing CARs, regardless of target epitope. Furthermore, regulated expression enhanced the potency and durability of response compared to constitutive expression. This approach may improve CAR activity in patients with MSLN ⁺ solid tumors or leukemia.