Targeting the Membrane-Proximal C2-Set Domain of CD33 for Improved CD33-Directed CAR T Cell Therapy

Background: There is increasing interest in targeting CD33 in malignant and non-malignant disorders, but available drugs are ineffective in many patients. As one limitation, therapeutic CD33 antibodies typically recognize the membrane-distal V-set domain. Likewise, currently tested CD33-directed chimeric antigen receptor (CAR) T cells likewise target the V-set domain and have thus far shown limited clinical activity. We have recently demonstrated that binding closer to the cell membrane enhances the effector functions of CD33 antibodies. We therefore raised antibodies against the membrane-proximal C2-set domain of CD33 and identified antibodies that bound CD33 regardless of the presence/absence of the V-set domain (“CD33^PAN antibodies”). Here, we tested their properties as targeting moiety in CD33^PAN CAR T cell constructs, using a clinically validated lentiviral backbone.

Methods: To generate CAR T cells, negatively selected CD8⁺ T cells were transduced with an epHIV7 lentivirus encoding the scFv from a CD33^PAN antibody (clone 1H7 or 9G2) linked to either a short (IgG₄ hinge only), intermediate (hinge plus IgG₄ CH3 domain), or long (hinge plus IgG₄ CH3 domain plus IgG₄ CH2 domain) spacer, the CD28-transmembrane domain, CD3zeta and 4-1BB intracellular signaling domains, and non-functional truncated CD19 (tCD19) as transduction marker. Similar constructs using scFvs from 2 different V-set domain-targeting CD33 antibodies, including hP67.6 (My96; used in gemtuzumab ozogamicin), were generated for comparison. CAR-T cells were sorted, expanded in IL-7 and IL-15, and used in vitro or in vivo against human AML cell lines endogenously expressing CD33 and cell lines engineered to lack CD33 (via CRISPR/Cas9) with/or without forced expression of different CD33 variants.

Results: CD33^V-set-directed CAR T cells exerted significantly more cytolytic activity against AML cells expressing an artificial CD33 variant lacking the C2-set domain (CD33^ΔE3-4) than cells expressing full-length CD33 at similar or higher levels, consistent with the notion that CD33 CAR T cell efficacy is enhanced when targeting an epitope that is located closer to the cell membrane. CD33^PAN CAR T cells were highly potent against human AML cells in a strictly CD33-dependent fashion, with constructs containing the short and intermediate-length spacer demonstrating robust cytokine secretion, cell proliferation, and in vitro cytolytic activity, as determined by ⁵¹Cr release cytotoxicity assays. When compared to optimized CD33^V-set CAR T cells, optimized CD33^PAN CAR T cells were significantly more potent in cytotoxicity, proliferation, and cytokine production without appreciably increased acquisition of exhaustion markers. In vivo, CD33^PAN CAR T cells extended survival in immunodeficient NOD.SCID.IL2rg^-/- (NSG) mice bearing significant leukemic burdens from various cell line-derived xenografts (HL-60, KG1α and MOLM14) with efficient tumor clearance demonstrated in a dose-dependent fashion.

Conclusion: Targeting the membrane proximal domain of CD33 enhances the anti-leukemic potency of CAR T cells. Our data provide the rationale for the further development of CD33^PAN CAR T cells toward clinical testing.