Background: Approximately 50% of pediatric B-ALL patients treated with clinically approved CD19-targeting CAR-T cells do not remain in remission one year after therapy. CD22-targeting CAR-T cells appear to be curative in only a small fraction of CD19-refractory patients and this therapeutic strategy is primarily used as a bridge to stem cell transplant. Additional immunotherapeutic targets thus remain urgently needed. Our laboratory recently used cell surface proteomics to identify CD72 as a B-cell specific marker especially upregulated on poor prognosis, KMT2A/MLL-rearranged B-ALL (Nix et al., Cancer Discovery (2021)). In this published work, we used a best-in-class nanobody library displayed on yeast to develop binders to CD72. We demonstrated for the first time that fully synthetic nanobodies can generate CAR-T cells that are highly potent in vitro and in vivo. While we previously focused on these "nanoCARs" in KMT2A/MLLr B-ALL, in this follow-up study we aimed to 1) further expand our nanoCAR indications to other CD72-expressing B-cell malignancies; 2) biophysically characterize our synthetic nanobodies; 3) evaluate the potential for further humanization of the nanobody binder amino acid sequence while retaining anti-tumor efficacy; and 4) characterize the potency and T-cell immunophenotypes in the context of our lead nanobody binder ("NbD4") placed on different CAR backbones.
Methods: Flow cytometry of primary patient samples for CD72 was performed in a CLIA-certified laboratory. NbD4 nanobody was recombinantly expressed in E. coli and biolayer interferometry was used to determine the binding affinity to recombinantly-expressed CD72 extracellular domain. CAR-T cells were generated from peripheral blood donor CD4+ and CD8+ cells (1:1) ratio via lentiviral transduction. In vitro cytotoxicity assays were performed at a range of effector:tumor ratios. In vivo studies were performed in human cell line orthotopic xenografts in NSG mice. 1e6 luciferase-labeled Jeko cells were implanted at Day 0 followed by administration of 4e6 CAR-T cells at Day 6. Tumor burden was assessed by bioluminescence.
Results: Flow cytometry on primary non-Hodgkin B-cell lymphoma obtained from fine needle aspiration biopsy (n = 5) confirmed CD72 surface expression (not shown), consistent with RNA-seq across larger cohorts. Biolayer interferometry demonstrated that NbD4 bound with surprisingly low affinity to recombinant CD72 (K D ~800 nM) (Fig. 1A), with both slow on rate (k on 8.38e4 M -1s -1) and rapid off rate (k off 6.82e-2 s -1). This affinity stands in contrast to that reported for FMC63 single chain variable fragment (scFv) used in clinically approved CD19-targeting CAR-T cells (K D 0.3-5 nM), despite similar in vitro and in vivo efficacy of both products. Our NbD4 framework region shows ~82% homology to a human IgG variable heavy domain, significantly higher than FMC63 (~59% homology). We made additional substitutions in the framework domain to increase human homology up to 89%. In vitro cytotoxicity assays with SEM B-ALL cells showed several humanized variants with similar efficacy to NbD4 (Fig. 1B). We further evaluated the impact of placing NbD4 on different CAR backbones, including combinations of CD28 or 4-1BB costimulatory domains and CD8 or IgG4-based transmembrane and hinge regions (Fig. 1C). In vivo, CD72 nanoCARs with Backbone 3 showed significantly increased potency (Fig. 1D). Indeed, tumors treated with Backbone 3 CAR-Ts showed complete tumor clearance and did not develop new tumors even after re-challenge with 1e6 Jeko cells at Day 52 (Fig. 1D). Preliminary characterization of effector and memory CAR-T cell phenotypes before exposure to tumor suggested that Backbone 3 had an increased number of naïve T cells compared to empty CAR and CD19 CAR-T cells (data not shown).
Conclusions: Our results demonstrate that our fully synthetic CD72 nanoCARs can effectively eliminate CD72-expressing B-cell malignancy models despite low nanobody binding affinity. Humanized NbD4 variants may serve as clinical candidates with even further reduction in possible immunogenicity of the llama amino acid framework. Alterations to the CAR backbone have a major impact on anti-tumor efficacy and phenotypes of our synthetic nanobodies. CD72-targeting therapies may be effective therapeutics not only KMT2A/MLLr B-ALL but also across a broader spectrum of refractory B-cell malignancies.
No relevant conflicts of interest to declare.