Autologous CD33CART for Children and Adolescents/Young Adults with Relapsed/Refractory AML: Phase 1/2 Clinical Trial Correlative Biology Analyses Demonstrate Potent in Vitro, In Vivo, and Ex Vivo Anti-Leukemia Activity

Background: Children and adolescents/young adults (AYAs) with multiply-relapsed/refractory acute myeloid leukemia (rrAML) have poor prognoses and rarely achieve complete remission (CR) with conventional chemotherapy. Given the success of CD19 and CD22 chimeric antigen receptor (CAR) T cell immunotherapies for B-ALL, we developed and preclinically validated CD33-redirected CAR T cells (CD33CART; Qin JITC 2021) and translated a best-in-class CD33CART-CD28/CD3z product to the clinic. In our multi-site CD33CART phase 1/2 trial (NCT03971799), 26 children and AYAs were treated in the autologous arm across four dose levels (DLs) with CR achieved in two of ten evaluable patients treated at the recommended phase 2 dose (DL4; 1e7 CD33CAR+ cells/kg) (Shah ASH 2023 #771). We now report correlative biology analyses of patients’ rrAML bone marrow specimens, excess clinical CD33CART products, and peripheral blood samples performed to elucidate biomarkers of immunotherapeutic response and failure.

Methods: Assays included flow cytometric assessment of CD33 antigen expression on primary rrAML cells pre- and post-CD33CART therapy, in vitro CAR T cell-mediated cytokine production and cell viability assays after co-incubation with CD33+ MOLM14 cells, creation of patient-derived xenograft (PDX) models for evaluation of in vivo anti-AML activity of patient CD33CARTs and positive control healthy donor (HD) CD33CARTs, and measurement of soluble CD33 (sCD33) and cytokine levels in serial blood plasma samples from patients following CD33CART infusion at days 1-14, 21, and 28.

Results: We observed equivalently potent in vitro cytotoxicity of autologous patient CD33CART products (n=30) against luciferase+ MOLM14 cells across all four DLs compared to that of HD CD33CARTs (n=2) in CellTiter-Glo viability assays. Interestingly and importantly, we did not detect differential killing potential in products from the two patients with CD33CART-induced CR. Similar in vitro IFN-g and IL-2 production was also detected across all cell DLs, as measured by ELISA at 1:1, 1:5, or 1:15 CD33CART effector to MOLM14 target (E:T) ratios. Comprehensive single-cell secretome analysis (Isoplexis) of patient CD33CARTs co-incubated with MOLM14 cells revealed comparable cytokine patterns to those of HD CD33CARTs. Amongst 32 assessed cytokines, most patient CD33CARTs produced moderate to high levels of GM-CSF, granzyme B, IFN-y, IL-5, MIP-1α, MIP-1β, perforin, TGF-β1, TNF-α, and TNF-β, while minimal production occurred with untransduced T cells (UTD). Indicators of CAR T cell quality (polyfunctionality) were somewhat variable across CD33CARTs with greater production of GM-CSF and other cytokines frequently observed in CD4+ versus CD8+ T cell subsets. Moderate to high sCD33 plasma levels were detected at baseline in 23 of 24 subjects with variable increases across serially-assessed post-CD33CART timepoints without clear correlation with cell DL or clinical response. Data correlation with AML CD33 antigen expression and genotype and ex vivo cytokine analyses is ongoing.

PDX modeling demonstrated successful engraftment of 7 of 14 (50%) rrAML patient specimens in busulfan-conditioned immunocompromised mice. Treatment of AML PDX mice (n=5/cohort) with saline or 5e6 UTD, patient CD33CARTs (n=5), or HD CD33CART as described (Qin JITC 2021, Niswander Haematologica 2023) demonstrated statistically similar in vivo CAR T cell expansion and inhibition of human leukemia proliferation in murine peripheral blood, spleen, and/or bone marrow between patient and HD CD33CART cohorts, suggesting effective anti-AML activity of autologous CD33CARTs across all cell DLs in preclinical model systems despite clinical responses seen only in DL4-treated subjects.

Conclusions: Autologous clinical CD33CART products generated from children and AYAs with rrAML demonstrate comparable in vitro and in vivo anti-AML functionality at all tested DLs to that of known-curative HD CD33CARTs in these bedside-back-to-bench studies. However, the limited CR rate observed in trial patients suggests that persistent barriers must be overcome to optimize in-human anti-leukemia functionality. Additional laboratory studies to elucidate potentially deleterious T cell- or AML-intrinsic and microenvironmental factors are in process. The trial is now accruing in an allogeneic arm using hematopoietic stem cell transplant donor-derived T cells.