Selection of the optimal single-chain variable fragment (scFv) of anti-CD33 by CAR-T cells

Previous work has identified a series of high-affinity antibody‒drug conjugates (Mylotarg) with binding specificity to distinct extracellular domains of CD33. We initially sought to develop a highly functional CAR against CD33 by cloning scFvs derived from the heavy- and light-chain pairs of these antibodies into a retroviral CAR vector with a Flag-tag for easy detection. This vector contained the 4-1BB stimulation domain. We tested scFvs derived from Mylotarg and novel sequences 1 and 2, and isolated T cells from healthy donors were transduced with each CAR (Fig. 2A, Supplement 2). The lysis potential of CAR-T cells was determined with a 20-hour luciferase killing assay with CD33+ Molm13 cells. Untransduced T cells were used as a negative control. We found substantial variation in the tumor cell lysis capacity of the CARs, with better performances of 1 and 2 than the control. We then cocultured CAR-T cells with Molm13 cells to assess CD33 CAR-T-cell activation and the proliferative ability. The levels of IL-2, IFN-γ, and TNF-α in the supernatant were measured 24 h after coculture, and CAR-T-cell proliferation was quantified after 7 days. Compared with the control, both the 1 CAR and the 2 CAR showed better efficacy in vitro. Then, we injected Molm13-luc cells into NSG mice to establish an AML model, and the mice were treated intravenously with 1.5 × 106 CAR-T cells on Day 7 after tumor development. Compared with the negative control and the conventional CD33 CAR structure, T cells expressing either the 1 or 2 CAR-binding domains elicited an antitumor response in the AML model (Supplement 3). However, seq-1 showed significantly greater toxicity to HSCs, resulting in a significantly decreased number of burst-forming unit-erythroid, colony-forming unit-granulocyte, colony-forming unit-granulocyte, erythrocyte, monocyte and megakaryocyte cells in the hematopoietic toxicity assay (Supplement 4).

Fig. 2

CAR structure and antileukemic efficacy of CAR-NK cells in vitro and in vivo. A CAR structure of CAR-T and CAR-NK cells; TM: transmembrane domain. B Cytotoxicity assay of CAR-NK cells with Molm-13 cells to show the killing efficacy of CAR-NK cells. C Concentrations of soluble IL-15 (E: T 1:1) in the control NK and CD33 CAR-NK cells. D Antitumor efficacy of CD33 CAR-NK cells in vivo

Umbilical blood cell-derived CAR-NK cell generation

To provide an off-the-shelf and safe immune cell treatment, based on efficacy and safety experiments in an animal model, we transduced sequence 2 into umbilical-derived NK cells to increase antitumor efficacy; a soluble IL-15 and P2A self-cleaving peptides were added to CD33 CAR-NK cells to increase antitumor efficacy in a coculture model (Fig. 2B), and increased levels of IL-15 were observed in vitro (Fig. 2C). In the Molm13 model, tumors were eliminated at 12 days after CAR-NK-mediated treatment (Fig. 2D). In general, CD33 CAR-NK-cell therapy is safe and effective for treating AML in vivo and in vitro.

Patient characteristics

Between December 2021 and June 2022, 12 patients were screened, 2 of whom were ineligible for NK cell infusion due to fatal infection and disease progression. Ten patients received one or more rounds of CAR-NK cell infusion. The median age of the patients was 42.5 (range, 18–65) years. All ten patients suffered from confirmed R/R AML after a median of 5 (range, 3–8) lines of treatment, and three patients received allogeneic HSCT (allo-HSCT), one patient received Mylotarg, and two patients had secondary AML (1 with myelodysplasia-related changes and 1 with confirmed lung cancer). Six patients had mutations associated with a poor prognosis. The patients’ baseline data are listed in Table 1. The efficacy of the final CAR-NK transduction for the infused product was 49.0% (range, 22.7–66.5%).

Table 1 Patient characteristicsSafety

After the infusion of CD33 CAR-NK cells, no immune effector cell–associated neurotoxicity syndrome (ICANS) or graft-versus-host disease (GVHD) was observed. Seven (70%) patients developed fever within two days after the first round of CAR-NK cell infusion; in 6 of the 7 patients, the fever was alleviated after symptomatic treatment. After the infusion of the second dose, a single patient developed recurrent fever, which was alleviated after one dose of 5 mg dexamethasone administered intravenously on the sixth day. No AEs above grade 3 were observed beyond hematological toxicity. None of the patients needed to be transferred to the intensive care unit (ICU). Severe bone marrow depression was self-limiting in the responding patients after a median of 27 days (range, 24–32 days) of supportive treatment (Table 2).

Treatment response

The median length of follow-up was 125 days (range, 45–258 days), and six of ten (60%) patients achieved minimal residual disease–negative complete response (MRD-CR), as indicated by imaging (imaging complete response, or iCR), at the day 28 assessment [13]. No significant difference was observed between the three dose groups (P = 0.93). Thus, most responses appeared within one month. Only one patient who bridged to allo-HSCT achieved long-term remission (258 days); the other patients had recurrent fatal disease that did not respond to further treatment. The median progression-free survival (PFS) was 71.5 days (range, 44–258 days), and the overall survival (OS) was 137 days (range, 50–258 days) for patients with complete remission (Fig. 3).

Fig. 3

Data for survival after CD33 CAR-NK cell treatment for patients with R/R AML enrolled in the clinical study

CAR-NK dynamics

CD33 CAR-NK cells could be observed by qPCR within 7 days. The peak concentration occurred 6 h after each infusion, followed by a second peak within two days of the initial peak (Fig. 4).

Fig. 4

The persistence of CD33 CAR-NK cells in the peripheral blood of patients enrolled in the clinical study was evaluated using qPCR