CD70 CXCR2-Modified CAR T-Cells Against Acute Myeloid Leukemia

Introduction Chimeric antigen receptor (CAR) T cells have not proven as effective in acute myeloid leukemia (AML) as they have in B-cell malignancies. Reasons for therapeutic failure include lack of tumor-specific targets, antigen loss, tumor heterogeneity, suppressive leukemic microenvironment, and poor CAR T cell fitness. Our group previously developed a novel CD70 CAR T cell modified to constitutively express the IL-8 receptor, CXCR2 (8R-70CAR T cell), to treat glioblastoma (IND#23881, Huang). Objective As CD70 is recognized as a relevant target in AML and IL-8 is also overexpressed in AML, we tested whether 8R-70 CAR T cells may be active against AML. We also assessed whether pre-treatment with azacitidine, which has been shown to upregulate CD70 expression in AML, may enhance the effect of 8R-70 CAR T cells Methods We analyzed AML samples in The Cancer Genome Atlas (TCGA) by AML subtypes to see how CD70 expression may vary across AML subtypes. We tested 5 AML cell lines (K562, KG-1, MV4-11, Kasumi-1, and MOLM-13) for CD70 expression. We cocultured 105 cells with 8R-70CAR T cells overnight (generated as previously described). Supernatants were collected and IFN-γ was measured. These experiments were then repeated following azacitidine pre-treatment for 96 hours with 3 cell lines to determine impact on CD70 expression and IFN-γ secretion. We developed a 3D tumoroid model of AML using the previously-described type I collagen conjugated liquid-like solids (LLS) platform developed by Dr. Sawyer and imaged interaction of fluorescently labeled 8R-70CAR T cells with 3 cell lines by confocal microscopy over 118 hours. Results Analysis of TCGA revealed CD70 expression was found to vary among AML subtypes (highest in MLL rearranged, lowest in t(15;17) translocation). CD70 expression was near 0% for two cell lines (Kasumi-1 and K562), medium (40-70%) for two (KG-1 and MV4-11), and high (>90%) for MOLM-13. Azacitidine treatment resulted in increased CD70 expression in MV 4-11 (40% to 70%) and MOLM-13 (90% to near 100%) which also corresponded with increased IFN-γ secretion, but essentially no change in the negative cell line Kasumi-1 (<1% to 2%, no change in IFN-γ secretion). In our LLS 3D tumoroid model, no appreciable killing was observed in the CD70 negative line K562, but in the CD70+ lines KG-1 and MV4-11 8R-70CAR T cells cluster around AML cells and show evidence of cytotoxicity (Figure, arrows). Conclusion 8R-70CAR T cells are able to specifically recognize and kill CD70+ AML cells, and cytotoxicity may be augmented by enhancing CD70 expression with azacytidine pre-treatment. A LLS tumoroid model can aid in 3D visualization of 8R-70CAR T cells interactions AML cells. These findings combined with safety data obtained from our ongoing phase I clinical trial for adults with glioblastoma (NCT05353530) will support expansion of our existing IND to treat patients with AML using our novel 8R-70CAR.