Rational Chemical and Genetic Modifications Enhance Avidity and Function of CD70-Directed CAR-T-Cells for Myeloid Leukemia

Chimeric Antigen Receptor T cells (CAR-T) have changed the therapeutic landscape for lymphoid malignancies, but not yet in myeloid malignancies like acute myeloid leukemia (AML). The TNF-alpha family member CD70 has emerged as a promising surface target antigen in AML after high complete response rates (CR) were seen in a Phase 1 trial of the CD70 antibody cusatuzumab (Riether et al, Nature Medicine 2020). Disappointingly, phase II results with cusatuzumab found CR rates less one-half those seen in the phase 1 (Trudel ASCO 2020).

Because CAR-T cells may recognize lower antigen densities than monoclonal antibodies, we sought to develop a better CAR-T strategy for targeting the CD70 antigen in AML. A recent effort to improve on the 'first' generation natural-ligand based CD70 CAR (full length-CD27 fused to CD3zeta) compared a variety of CARs, including the single-chain variable-fragment as the binding moiety to CD70, found that the original 'first' generation CAR was superior (Shaffer Blood 2011, Sauer Blood 2021). Interestingly, a second generation natural-ligand based CAR that included the 4-1BB costimulatory domain was thought to be superior to first-generation (zeta-only) CARs with the same binder (Wang Clinical Cancer Research 2016).

We first confirmed that the ligand-based 4-1BB (Native) CAR, had activity against multiple AML targets in standard CAR-T assays including activation, cytolysis, and demonstrated activity in a NOD-SCID IL2R γ^null (NSG) Molm13 mouse model of AML. However, these models were not curative, even in combination with azacitidine, which we confirmed mediated increase CD70 expression on the AML cells. We hypothesized that surface cleavage of CD27, which is the natural ligand of CD70, attenuated the function of the Native CARs; we confirmed this hypothesis by measuring soluble CD27 in CAR T cells co-cultured with AML targets. To abrogate surface cleavage of ligand-based CARs, we generated and tested a panel of rationally designed, novel hinge CAR variants ('truncated', 'deleted', 'flexible', and 'CD8hinge&TM', Figure 1A). We found that the CD8hinge&TM variant had improved cytolysis against AML targets in vitro as well as higher binding avidity as measured by acoustic force microscopy (Figure 1B). Furthermore, there was no detectable soluble CD27 after co-culture with AML targets, suggesting successful abrogation of hinge cleavage.

When all the CAR variants were compared in vivo to the Native CAR, CD8hinge&TM CARs mediated improved tumor control, had higher CAR expansion in blood and bone marrow, a persistent central memory phenotype beyond 30 days, and mediated improved survival (Figure 1C). The effect of the CD8hinge&TM CARs was further enhanced in combination with azacitidine. We also found that tumor control (maximum flux) correlated most strongly with relative binding avidity of the hinge variants (R ² 0.906) compared to other measures of in vitro function such as IFNg production (R ² 0.548) or in vitro cytolysis (R ² 0.5982). Finally, we tested the Native and CD8hinge&TM CARs in patient-derived xenograft (PDX) models and confirmed that the Native CAR T cells did not control AML tumor and resulted in uniform lethality, whereas the CD8hinge&TM CAR T cells had superior in vivo expansion and were able to mediate AML eradication in mice. Our findings demonstrate that natural-ligand binding domains of CARs targeting CD70 in AML can be effective but require mechanisms to overcome surface cleavage. CD70-targeted CARs comprised of a fusion of truncated CD27 to a CD8 hinge and transmembrane domain have promise in patients with AML, with and without combination with azacitidine.

Figure 1

View largeDownload slide

Figure 1

View largeDownload slide

Close modal