Off-Tumor Target Expression Levels Do Not Predict CAR-T Cell Killing: A Foundation For The Safety Of CD44v6-Targeted T Cells

Introduction Off-tumor expression of the target antigen raises justified safety concerns about newly designed chimeric antigen receptors (CARs). We have recently developed a CAR targeting the tumor-promoting antigen CD44v6 and demonstrated potent antitumor effects against acute myeloid leukemia (AML) and multiple myeloma (MM) both in vitro and in vivo. Despite promising activity against epithelial tumors, the administration of the CD44v6-specific mAb used for deriving our CAR (bivatuzumab) showed reversible myelosuppression and mucositis when conjugated with radioisotopes, and severe skin toxicity when conjugated with the potent cytotoxic drug mertansine. Preclinically evaluating the potential off-tumor toxicities of CD44v6-targeted T cells is therefore crucial before they can be safely translated to the clinic. Aim To profile the off-tumor expression of CD44v6 and to verify the susceptibility of expressing cells to CAR-T cell killing. Results Quantitative RT-PCR analysis on a wide panel of cDNA from normal tissues revealed restricted CD44v6 expression on flat stratified epithelia, like the skin, albeit at considerably lower levels compared with primary leukemic blasts. We therefore addressed the issue of keratinocyte recognition in co-culture experiments. Strikingly, at the E:T ratios allowing the potent antitumor effects of CD44v6-targeted T cells, keratinocytes were not killed and there was no cytokine production. Interestingly, comparative analysis of accessory molecules showed that, differently from leukemic blasts, keratinocytes expressed significant lower levels of adhesion/costimulatory molecules, including (ICAM-1, LFA-3 and B7.2), but higher levels of the critical checkpoint molecule PD-L1. Of the different cells of the hematopoietic system analyzed, only circulating CD14+ monocytes expressed CD44v6 and were killed by CD44v6-targeted T cells. Interestingly, by immunohistochemistry, we found no CD44v6 expression on bone-marrow monocytes, lymph-node macrophages, brain microglia, liver Kuppfer cells and dermal macrophages, suggesting a low risk for by-stander toxicity against these tissues. Moreover, CD44v6-targeted T cells did not interfere with the generation of virus-specific CTLs by antigen-specific stimulation in vitro. Importantly, both RT-qPCR and FACS demonstrated lack of CD44v6 expression on hematopoietic stem cells (HSCs) and progenitors. Accordingly, CD44v6-targeted T cells did not interfere with their clonogenic potential in vitro and, in co-culture experiments with whole bone marrow from MM patients, were able to selectively eliminate tumor cells, while sparing HSCs and progenitors. Finally, we tested the potential hematological toxicities of CD44v6-targeted T cells in NSG mice transgenic for human IL-3, SCF and GM-CSF (NSG-3GS). NSG-3GS mice transplanted with human CD34-selected cord blood cells showed enhanced myeloid reconstitution compared to NSG mice, including CD44v6+ monocytes. The infusion of CD44v6-targeted T cells in reconstituted NSG-3G mice resulted in the selective elimination of monocytes, but in the preservation of other cell subsets. Importantly, after in vivo exhaustion of CD44v6-targeted T cells, NSG-3G mice reconstituted monocytes de novo, indicating preservation of the HSC pool. For enabling rapid and conditional ablation of CD44v6-targeted T cells, we have finally co-expressed the CD44v6-CAR with TK or the inducible caspase-9 and validated the suicide gene approach in hyperacute xenogeneic GVHD surrogating maximal toxicity. Conclusions Our results indicate that off-tumor target expression levels do not automatically predict the susceptibility to CAR T-cell killing. Moreover they suggest that, differently from mAb-derived pharmaceuticals, therapeutic doses of suicidal CD44v6-targeted T cells might associate with acceptable and/or reversible toxicities. Disclosures: Bordignon: MolMed SpA: Employment.