TNFα Apoptotic Signaling By Antigen-Engaged CAR T-Cells Mediates Clearance of Target Antigen-Negative Acute Myeloid Leukemia

Background: The success of chimeric antigen receptor (CAR) T-cells against B-cell neoplasms has spurred investigation of such therapies in acute myeloid leukemia (AML). However, their clinical efficacy has been limited. These treatment failures are often attributed to the lack of an ideal target antigen that is universally expressed. Indeed, our preclinical models show that despite improved survival after CAR T-cell treatment, heterogeneous antigen expression predisposes to relapse with antigen-negative disease. Notably, the presence of a subset of tumor cells that lack the target antigen does not preclude complete response to CAR T-cells in other diseases such as B-cell acute lymphoblastic leukemia.

These observations prompted our investigation of factors that allow CAR T-cells to clear disease when target expression is not universal. In some tumors, malignant cells with insufficient target antigen can be killed by T-cells primed by antigen-positive cells. It is unknown if this occurs in AML, and we aimed to develop models to study these functions and potential resistance mechanisms.

Methods: CD33 and CLL-1 CAR T-cells were generated from donor T-cells by CD3/CD28 bead activation, lentiviral delivery of CAR transgenes, fluorescence-activated cell sorting (FACS) and IL-7/15 expansion. Where applicable, T-cell receptor (TCR) knockout (KO) was performed with ribonucleoprotein (RNP) CRISPR/Cas9 editing. CD33- and/or CLL-1-expressing human AML cell lines were selected for use, and target KO variants were created using RNP. Clonal lines were established by FACS, and we verified that target KO cells did not independently induce CAR T-cell activation. Each cell line was transduced to express a unique fluorescent protein and luciferase to allow differential quantitation of tumor burden using flow cytometry (FC) and bioluminescence imaging. For instance, in vitro killing assays were performed by coculture of labeled CD33_KO AML blasts with unlabeled CD33 wild type (WT) AML, and CD33 CAR T-cells or media control. Cytotoxicity was calculated by FC enumeration of viable fluorescent protein-expressing events in CAR T-cell conditions relative to controls.

We performed CRISPR KO screening using lentiviral delivery of individually cloned sgRNAs to Cas9-expressing CD33_KO AML cells. Transduced cells were FACS enriched, pooled and cultured in quadruplicate with unmodified CD33_WT AML cells, and CD33 CAR T-cells or media control. Integrated sgRNAs were PCR amplified and sequenced, and fold change was computed by count comparison with control conditions. Enrichment was determined relative to non-targeting sgRNAs using two-sided Student’s t test.

Results: We discovered that CD33_KO AML blasts are lysed in vitro by CD33 CAR T-cells only after activation by engagement with CD33_WT AML (PRIMECAR killing). This mechanism is independent of the TCR and is not unique to CD33 CARs— we observed similar findings using a CLL-1 CAR T-cell system. Importantly, hematopoietic stem cells appear resistant to these effector mechanisms.

To determine if PRIMECAR killing contributes to control of AML burden in vivo, we developed a murine model in which CD33_KO and CD33_WT AML xenografts could be individually monitored over time. Here, NOD.SCID.IL2rg^-/- (NSG) mice were injected with individually labeled CD33_WT and CD33_KO AML cell lines and then treated with CD33 or negative control CAR T-cells. CD33 CAR T-cells not only improved control of CD33_WT AML compared to control groups, but also significantly reduced CD33_KO burden.

Using transwell assays, we showed that although PRIMECAR killing of AML occurs via soluble factors, the killing is greatly enhanced by proximity. As death receptor (DR) apoptotic signaling is a T-cell mechanism enhanced by cell-cell contact, we performed pooled CRISPR KO screening to study DR pathways essential for PRIMECAR killing in AML. The TNFRSF1a KO phenotype was resistant to PRIMECAR killing (p=8.1e-05) indicating that TNFα apoptotic signaling is critical. As validation, anti-TNFα monoclonal antibodies significantly reduced PRIMECAR killing in vitro and in vivo.

Conclusion: Our findings indicate that cognate antigen-engaged CAR T-cells use DRs to mediate the clearance of proximal antigen-negative AML. Impaired apoptotic TNFα signaling is one resistance mechanism in AML, and further study of this pathway can inform strategies to improve the effectiveness of CAR T-cells in this disease.