Base edited "universal" donor CAR T-cell strategies for acute myeloid leukaemia

Abstract

Acute myeloid leukaemia (AML) is often aggressive and life-threatening with limited curative options. Immunotherapies including chimeric antigen receptor (CAR) T-cell approaches are under investigation, but high levels of disease heterogeneity remain a major hurdle to achieving durable responses. Targeting of multiple antigens may ensure complete immunological coverage of leukaemic blast populations, but such antigens are often also present on healthy haematopoietic populations. To address likely aplasia, strategies can be designed to bridge CAR T-cell therapies to allogeneic stem-cell transplantation (allo-SCT), as demonstrated in recent anti-CD7 CAR T-cell studies. Here we report that monotherapy using base edited "universal" donor CAR T cells against CD33, CLL-1, or CD7 delivered inhibition of AML in immunodeficient mice when antigen expression was homogenous, but combined use of BE-CAR33 and BE-CARCLL-1 T cells was required to address heterogenous CLL-1^-/+CD33^-/+ disease. We also demonstrate that removal of shared CD7 antigens enabled compatibility of BE-CAR33 and BE-CARCLL-1 with BE-CAR7 T cells, including in a patient-derived xenograft (PDX) model of AML. Therapeutic strategies envisage 'pick and mix' applications of base edited "universal" CAR T cells in combination determined by patient-specific antigen profiles. Such approaches also offer the possibility of deep, cell-based, de-bulking and conditioning ahead of allo-SCT and subsequent donor-derived reconstitution.

Fig. 1. BE-CAR7, BE-CAR33 and BE-CARCLL-1 T cells to target AML.

A Strategies employed to cause protein knockout using a cytidine base editor, through the C-to-T mediated conversion of arginine (arg), glutamine (gln) or tryptophan (trp) amino acid residues to premature stop codons (TGG, TGA, TAG), or the disruption of consensus sequences required for proper mRNA splicing. Created with BioRender.com. B Schematics of BE-CAR manufacture. For BE-CAR7 T cells, MNCs were electroporated with codon optimised mRNA encoding the cytidine base editor BE3 and sgRNAs against TRBC (introducing a premature stop codon), CD7 (introducing a premature stop codon) and CD52 (disrupting a splice donor site). Cells were then transduced with a pCCL-CAR7 and expanded prior to TCRɑβ magnetic column depletion. For BE-CAR33 and BE-CARCLL-1 T cells, MNCs were first transduced with pTTB lentiviral vector encoding anti-CD33 or anti-CLL-1 CAR and a sgRNA expression cassette against TRBC. coBE3 mRNA and sgRNAs were delivered by electroporation and residual TCRɑβ⁺ cells were removed through magnetic bead depletion. Created with BioRender.com. C Example of end-of-manufacture CD45⁺ BE-CAR7, D BE-CAR33 and BE-CARCLL-1 T cells with CAR⁺TCRαβ^- and CD7^-CD52^- populations in red. E Summary of multiple donor BE-CAR7 (n = 4), BE-CAR33 (n = 10) and BE-CARCLL-1 (n = 11) T cells with CAR⁺TCRαβ^- phenotype and residual TCRαβ, F CD7 and CD52 expression. G Example of EditR molecular analysis of on-target base conversion at TRBC, CD7 and CD52 editing sites with H percentage of on-target conversions from multiple healthy donor BE-CAR7 (n = 4), BE-CAR33 (n = 3) and BE-CARCLL-1 (n = 5) T cells. I Co-culture of BE-CAR7 T cells and BE-CAR33 T cells (with and without CD7 expression) indicating evasion of BE-CAR7 fractricidal effects after CD7 base editing. ** p < 0.01 (unpaired t test).

Fig. 2. BE-CAR T cells exhibited antigen specific effects against AML with homogenous antigen expression.

A (Top) ⁵¹Cr release from Kasumi-3 (left) or HL-60 (right) targets after co-culture with untransduced T cells or BE-CAR7, BE-CAR33 or BE-CARCLL1 effectors across E:T ratios. (Bottom) Cytokines released from untransduced, BE-CAR7, BE-CARCLL-1 or BE-CAR33 T cells (above 50 pg/mL as indicted by the dotted line) when co-cultured with Kasumi-3 (left) or HL-60 (right) targets at an E:T of 1:1. * p < 0.05, *** p < 0.001, **** p < 0.0001. (unpaired T test for BE-CAR7 comparison, one-way ANOVA with Tukey multiple comparison post-hoc for BE-CAR33 and BE-CARCLL-1 comparisons). B BE-CAR7 anti-leukaemic effects were assessed in vivo aginst a humanised murine model of AML using GFP⁺LUC⁺ Kasumi-3 cells. C Bioluminescent images and average radiance values (p/s/cm²/sr) D of mice that received 1×10⁶ Kasumi-3 targets on day 0 followed by 10×10⁶ untransduced T cells (n = 5) or 5 × 10⁶ BE-CAR7⁺ effectors (from 10 × 10⁶ total MNCs) (n = 5) on day 4. ** p < 0.01 (one-way ANOVA of AUC with Tukey multiple comparison post-hoc between D3 and D31). E BE-CAR33 and BE-CARCLL-1 anti-leukaemic effects were assessed in vivo against a humanised murine model of AML using GFP⁺LUC⁺ HL-60 targets. F Bioluminescent images and average radiance values (p/s/cm²/sr) G of mice that received 1 × 10⁶ HL-60 targets on day 0 followed by 10 × 10⁶ untransduced cells (n = 5), 10 × 10⁶ BE-CAR33⁺ (n = 5) or 10 × 10⁶ BE-CARCLL-1⁺ (n = 5) effectors (from 10×10⁶ total MNCs) on day 4. ** p < 0.01 (one-way ANOVA of AUC with Tukey multiple comparison post-hoc between D3 and D25). H CD45⁺GFP⁺ AML and CD45⁺CD2⁺/CD2⁺CAR⁺TCRαβ- effectors (highlighted in red) detected in bone marrow of mice from each group. I Normalised numbers of CD45⁺GFP⁺ AML cells and CD2⁺/CD2⁺CAR⁺TCRαβ- cells detected in bone marrow of mice (n = 5) from each treatment group. Limit of quantification is shown at 100 events per 100,000 bone marrow cells. ** p < 0.01 (one-way ANOVA with Tukey multiple comparison post-hoc).

Fig. 3. BE-CAR33 and BE-CARCLL-1 effectors combined effects against heterogenous AML in vivo.

A Humanised xenograft model of heterogenous AML with GFP⁺LUC⁺ variants used to assess a combined dosing strategy with BE-CAR33 and BE-CARCLL-1 effectors. B Bioluminescent images and average radiance values (p/s/cm²/sr) C of mice that received 1 × 10⁶ CD33^-/+CLL-1^+/- HL-60 targets on day 0 and 10 × 10⁶ untransduced T cells (n = 4), 4 × 10⁶ monotherapy BE-CAR33⁺ (n = 4) or 4 × 10⁶ BE-CARCLL-1 (n = 4) effectors, or 2 × 10⁶ BE-CAR33⁺ and 2 × 10⁶ BE-CARCLL-1⁺ combined effectors (n = 4) on day 4 (from a maximum of 10 × 10⁶ MNCs). * p < 0.05, ** p < 0.01, *** p < 0.001 (one-way ANOVA with Tukey multiple comparison post-hoc). D CD45⁺GFP⁺ AML targets and CD45⁺CD2⁺ effectors detected in bone marrow of mice from each group. E Normalised total CD45⁺GFP⁺ AML cell counts and F AML cell counts displaying CD33⁺CLL-1^-, CD33^-CLL-1⁺ or CD33^-CLL-1^- phenotypes detected in bone marrow of mice (n = 4) from each treatment group. Limit of quantification is shown at 100 events per 100,000 bone marrow cells. * p < 0.05, ** p < 0.01 (3E: one-way ANOVA with Tukey multiple comparison post-hoc, 3F: Mann-Whitney U test, * p < 0.01, *** p < 0.001).

Fig. 4. Removal of CD7 enables compatibility of BE-CAR7 and BE-CAR33 in a PDX model of AML.

A 1×10⁶ CD33^highCD7^low primary AML cells were injected into sub-lethally mice and monitored for disease engraftment through peripheral sampling for 3 weeks. Control CAR19 effectors, an antigen not expressed by AML or T cells, were injected alone or in combination with BE-CAR7 or BE-CAR33 to negate dose-dependent BE-CAR7 or BE-CAR33 effects when co-injected versus delivered in isolation. B Example flow cytometry and subsequent tracking C of peripheral CD33⁺ (left) and CD7⁺ (right) disease after treatment with 10 × 10⁶ CAR19⁺ T cells (n = 10) (from 10 × 10⁶ MNCs) or 10 × 10⁶ CAR19⁺ T cells and 6 × 10⁶ BE-CAR7⁺ T cells (n = 10), 10 × 10⁶ CAR19⁺ and 10 × 10⁶ BE-CAR33⁺ T cells (n = 10) or 10 × 10⁶ BE-CAR33 and 6 × 10⁶ BE-CAR7 T cells (n = 10) (from 20 × 10⁶ MNCs). # indicates animal death. ** p < 0.01, **** p < 0.0001 (one-way ANOVA of AUC with Tukey multiple comparison post-hoc between D-1 and D28). D Survival of mice treated with CAR19 T cells (n = 10), CAR19 and BE-CAR7 T cells (n = 10), CAR19 and BE-CAR33 T cells (n = 10), or BE-CAR33 and BE-CAR8 T cells (n = 10). *** p < 0.001, **** p < 0.0001 (Kaplan-Meir with Mantel-Cox test). E Example flow cytometry showing bone marrow CD33^+/-CD7^+/- disease following BE-CAR treatment.