Development of a Dual-Antigen Targeting CAR-T Therapy Directed Against CD70 and Active Integrin β2 for Acute Myeloid Leukemia

Introduction:

Acute myeloid leukemia (AML) remains a devastating disease with a great need for new therapies. While CAR-T therapy has achieved remarkable success in B-cell leukemias, lymphomas, and multiple myeloma, successful cellular therapy for AML remains elusive. A major contributing factor is thought to be antigen expression heterogeneity in AML, preventing complete tumor clearance with single antigen-targeting. Additionally, most leading AML targets (e.g. CD33, CD123) are also expressed on normal hematopoietic stem and progenitor cells (HSPCs), resulting in a limited therapeutic window and clinically deleterious hematologic toxicity.

Our group previously identified the active conformation of integrin β2 (aITGB2) as an AML-specific target and generated a CAR-T therapy that eliminated AML tumors while sparing normal HSPCs (Mandal et al., Nat Cancer 2023). More recently, our group developed a potent natural ligand-based CAR-T therapy targeting CD70 (Kasap et al., ASH 2023), another antigen expressed on many AML cells but absent on normal HSPCs. Both aITGB2 and CD70 have heterogeneous expression on AML blasts, however, preventing complete tumor clearance. Here, we combine the potent efficacy and low toxicity of these two single-antigen CAR-Ts into a dual-antigen targeting approach capable of eliminating a larger proportion of AML cells, thereby driving deeper remissions while achieving an acceptable therapeutic index.

Methods:

Primary AML patient samples were co-cultured with the HS-5 stromal cell line in RPMI media with 20% FBS and 25 ng/ml IL-3, SCF, and FLT3L. Primary AML cells were assessed for expression of CD70 and aITGB2 by flow cytometry. CAR constructs were lentivirally transduced into CD3-selected T-cells. For In vitro cytotoxicity assays, luciferase-expressing target cells were co-cultured with CAR-Ts for 24 hours, with leukemia cell killing measured by endpoint luminescence.

Results:

To first verify the AML-specificity of CD70, we analyzed the TARGET AML RNA-seq dataset, observing significantly higher expression in AML compared to normal CD34+ and bone marrow cells (p<0.0001). We then profiled 10 primary AML samples for co-expression of CD70 and aITGB2 using flow cytometry. To better simulate the tumor microenvironment, we co-cultured these AML cells with a bone marrow stromal cell line and exogenous cytokines for 7 days (see Methods). This approach resulted in a notable increase in the percentage of AML blasts displaying CD70 and aITGB2, possibly by reestablishing cell surface homeostasis after shedding of surface proteins with a freeze-thaw cycle. We showed that a dual-targeting, OR-gated approach could eliminate a mean 87±11% of blasts, compared to 72±28% targeting aITGB2 alone or 44±26% for CD70 alone (p= 0.015, 1-way ANOVA).

Utilizing our previously validated aITGB2 and CD70 binders, we designed 8 dual-targeting CAR-T constructs, differing in their costimulatory domains and relative CAR positions on the plasmid. Despite similar overall designs, we found surprisingly large differences in transduction and anti-tumor efficacy. We showed that an anti-aITGB2 CAR with a 4-1BB costimulatory domain, followed by an anti-CD70 CAR with CD28 costimulatory domain, demonstrated the best expression in T-cells. To prevent fratricide in CAR-Ts, which express both aITGB2 and CD70 when activated, we optimized a CRISPR-Cas9 protocol using a library of guide RNAs to knock out both targets in T-cells, with no defect in CAR transduction or killing ability observed.

Our best dual-targeting construct showed dose-dependent tumor killing (p<0.0001, 2-way ANOVA) in vitro against the AML cell lines NOMO-1 and THP-1, outperforming the other dual-targeting CAR designs and maintaining similar efficacy to single-targeting CARs. We then generated CD70 and ITGB2 knockouts for each cell line and observed that our dual-targeting CARs preserved tumor killing while single antigen-targeting CARs could no longer target their respective knockout cell lines (P<0.0001, 2-way ANOVA).

Conclusion:

Overall, using an ex vivo co-culture system for target validation and strategies for optimizing a bicistronic CAR design, we demonstrate that a dual-antigen targeted CAR-T therapy directed against both aITGB2 and CD70 is a highly promising approach for AML. We are continuing preclinical development of this strategy, including in vivo mouse model studies, to further assess efficacy and safety.