Nanobody-Engineered CLL-1 CAR T Cells: Optimizing Tumor-Specific Cytotoxicity and Minimizing Off-Tumor Toxicity

Abstract

Acute myeloid leukemia (AML) is an aggressive hematologic malignancy characterized by the rapid expansion of undifferentiated myeloid progenitors, leading to impaired hematopoiesis and poor patient prognosis. Although chimeric antigen receptor (CAR) T-cell therapy using single-chain variable fragments has revolutionized immunotherapy, clinical application in AML remains limited by on-target, off-tumor toxicities, largely due to shared antigen expression on normal hematopoietic stem and progenitor cells. To address this challenge, we developed a nanobody-based CAR T-cell platform directed against C-type lectin-like molecule-1 (CLL-1), a myeloid-restricted surface antigen minimally expressed on healthy hematopoietic stem cells but consistently enriched on AML blasts and leukemic stem cells. Leveraging the high specificity, solubility, and reduced immunogenicity of llama-derived single-domain variable heavy-chain antibodies, we engineered both CLL-1 and CD33 nanobody CAR constructs and systematically compared their functional activity. Functional validation included real-time cytotoxicity monitoring using IncuCyte imaging of mKate2-labeled AML cells, serial tumor rechallenge assays to assess sustained killing, and NOD/SCID/IL2Rγ^null xenograft models to evaluate in vivo efficacy under conditions of high leukemic burden. CLL-1 and CD33 CAR T cells demonstrated rapid and durable cytotoxicity, with significant killing efficiency at low effector-to-target ratios (0.33:1). Unlike CD33 CAR T cells, CLL-1–directed CARs spared normal hematopoietic progenitors, preserving colony-forming capacity. Importantly, CLL-1 CAR T cells retained a favorable memory phenotype with stable proliferation and viability, whereas cytokine release assays confirmed effective yet antigen-specific immune activation. In vivo, treatment with CLL-1 CAR T cells resulted in profound and sustained tumor regression in AML xenografts, accompanied by the persistence of functional CAR T cells. Together, these findings establish CLL-1–targeted nanobody-based CAR T cells as a precision-engineered immunotherapy with potent antileukemic activity, reduced off-target toxicity, and enhanced translational potential. This platform provides a promising therapeutic avenue to overcome current barriers in AML CAR T-cell development and improve patient outcomes.

Significance:

Nanobody-based CLL-1 CAR T-cell therapy balances potent antitumor activity with hematopoietic preservation, highlighting the potential of our CLL-1 CAR T-cell platform as a next-generation, safer, and clinically superior strategy for effective AML treatment.

Introduction

Acute myeloid leukemia (AML) is a highly aggressive hematologic malignancy, often associated with poor outcomes despite advancements in novel therapeutic approaches (1). AML represents 15% to 20% of acute leukemias in the pediatric population and is the most common form of acute leukemia in adults (2). The goal of AML therapy is disease eradication, when possible. This is typically accomplished by inducing complete remission (CR) with initial therapy, followed by consolidation and/or maintenance therapy to maximize treatment response. Treatment response and overall prognosis vary based on several patient- and tumor-specific factors, including age, performance status, and cytogenetic profile (3). Despite intensive chemotherapy, 10% to 40% of patients with newly diagnosed AML fail to achieve CR. Although substantial progress has been made in understanding AML biology, conventional treatments such as intensive chemotherapy remain the standard of care for eligible patients and are associated with significant toxicity and high rates of treatment failure (4).

Studies suggest that leukemic stem cells (LSC) give rise to leukemia-regenerating cells (LRC), which are resistant to chemotherapy and contribute to disease relapse. As a result, combination strategies targeting LSCs alongside cytoreductive chemotherapy have been proposed as potential therapeutic approaches for AML. However, this concept has been challenged by recent research demonstrating that chemotherapy-resistant LRCs in minimal residual disease are not necessarily derived from naïve LSCs (5–16).

Allogeneic hematopoietic stem cell (HSC) transplantation (allo-HSCT) has undergone significant advancements and remains a potentially curative treatment for AML by leveraging the graft-versus-leukemia effect. Despite its therapeutic potential, allo-HSCT carries substantial risks, including hematopoietic reconstitution failure, delayed immune recovery, graft-versus-host disease (GVHD), and posttransplant relapse, which continue to limit its broader application (17–19).

More recently, chimeric antigen receptor (CAR) T-cell therapy has revolutionized the treatment landscape for hematologic malignancies but has shown limited success in AML due to the lack of an ideal target antigen (20). Several AML-associated targets have been identified, including CD33, CD70, CD123, and FLT3 (21, 22). However, targeting AML with CAR T cells is particularly challenging because many myeloid antigens are expressed on both leukemic and healthy myeloid cells. This overlap increases the risk of damaging normal hematopoietic tissues, leading to potentially severe on-target, off-tumor toxicity (23, 24). For example, CD33, a transmembrane receptor of the sialic acid–binding immunoglobulin-like lectin (SIGLEC) family, is highly expressed on AML blasts but is also present on healthy hematopoietic stem and progenitor cells, resulting in hematopoietic toxicity when targeted by CD33 CAR T cells (25–27).

To overcome these challenges, we engineered a nanobody-based CAR targeting C-type lectin-like molecule-1 (CLL-1). CLL-1, a member of group V of the C-type lectin-like receptor family, is predominantly expressed on myeloid lineage cells. CLL-1 is present on AML blasts, with reported expression frequencies ranging from 85% to 92%. Importantly, HSCs typically do not express CLL-1 although some studies have suggested that a minor subset of these cells may exhibit low-level expression (5, 28). This expression pattern enables selective targeting of AML cells while largely sparing normal HSCs.

Multiple clinical trials are currently underway evaluating CAR T-cell therapies for AML that target myeloid-associated antigens using either single- or dual-targeting strategies. Most CLL-1–directed clinical trials focus on patients with relapsed or refractory AML, for whom therapeutic options remain limited. A primary objective across these trials is safety evaluation, which continues to be a critical concern in CAR T cell–based therapies.

Several early-phase clinical studies (phase I/II) evaluating CLL-1–targeted CAR T cells have reported promising efficacy with manageable safety profiles in adult patients with relapsed or refractory AML (28, 29). In addition, a CRISPR-edited allogeneic anti–CLL-1 CAR T-cell trial has completed phase I evaluation (30), and a separate phase I trial investigating CLL-1–directed CAR NK cells has also been initiated (31).

Furthermore, dual-targeting CAR constructs against CLL-1 and CD33 are under investigation, as well as multitargeted combinations involving CLL-1 with CD33 or CD123, tailored to individual patient antigen expression profiles. Early phase I results have demonstrated encouraging safety and preliminary efficacy of CD33/CLL-1 dual CAR T cells in patients with relapsed or refractory AML (32, 33).

Similarly, CD33-directed CAR T-cell clinical trials have progressed through phase I and II evaluations, including single-arm studies designed to assess the safety and tolerability of functionally enhanced CD33 CAR T cells in relapsed or refractory AML (34, 35). Although complete responses were frequently accompanied by incomplete hematologic recovery and aplasia (34–36), several additional CD33 CAR T-cell trials remain active or are recruiting patients for phase I evaluation (37, 38).

Collectively, findings from completed and ongoing phase I/II trials underscore the therapeutic potential of CAR T-cell therapy as a future strategy for AML treatment. Most of these studies employ conventional single-chain variable fragment (scFv)–based CARs, and to our knowledge, no nanobody-based CARs are currently in clinical trials for AML.

Nanobodies, also known as single-domain antibodies, are antibody fragments derived from heavy-chain–only antibodies found in camelids such as llamas and alpacas. These nanobodies consist of a single variable heavy-chain (VHH) domain and lack light chains, resulting in a much smaller molecular weight (12–15 kDa) compared with conventional scFvs (25–30 kDa) derived from human immunoglobulin G (IgG) antibodies (∼150 kDa; refs. 39–42). This compact size, combined with excellent solubility, stability, and high antigen-binding affinity, makes nanobodies highly versatile for therapeutic and diagnostic applications and has generated significant interest in antitumor therapies and imaging diagnostics (43, 44).

In recent years, nanobodies have demonstrated promising potential in cancer medicine, including CAR T-cell therapy (45). Compared with scFv-based CAR T cells, nanobody-based CAR T cells have been reported to exhibit reduced immunogenicity and enhanced cytokine release, which may improve therapeutic outcomes (46, 47). Notably, the FDA-approved CAR T-cell therapy ciltacabtagene autoleucel (trade name Carvykti) for multiple myeloma employs a nanobody-based design, underscoring the translational potential of this approach (48, 49).

In the present study, we developed a novel nanobody-based CAR T-cell therapy targeting CLL-1 to enhance the efficacy and specificity of CAR T-cell therapy for AML. Our in vitro studies demonstrated selective tumor cell killing by CLL-1 CAR T cells and sustained cytotoxic activity following repeated tumor rechallenge. In a xenograft model using U937 cell line–derived tumors in NOD/SCID/IL2Rγ^null (NSG) mice, bioluminescence imaging (BLI) revealed complete tumor elimination in mice treated with CLL-1 CAR T cells. Although CLL-1 CAR T cells performed comparably with CD33 CAR T cells both in vitro and in vivo, they did not exhibit the same myelotoxicity in colony formation assays. Collectively, these findings support the potential of nanobody-based CLL-1 CAR T-cell therapy to reduce relapse risk and improve overall survival, particularly in patients who are not eligible for conventional treatment modalities.

Materials and Methods

Antibody generation and construct design

To generate CLL-1 and CD33 nanobodies, two llamas (Lama glama) were immunized with recombinant target proteins from Diaclone. Four injections were administered on days 0, 14, 28, and 35 using a protein cocktail containing 100 μg of recombinant human CD33 or CLL-1 (ACROBiosystems). Immunized libraries were constructed, and phage display panning was performed by QVQ, which provided the sequences of the lead CLL-1 and CD33 VHH panels.

Using these VHH sequences, CLL-1 and CD33 CAR constructs were engineered with a CD8α hinge and transmembrane domain. Intracellular signaling domains consisted of CD3ζ in combination with the 4-1BB (CD137) costimulatory domain to enhance T-cell persistence and functionality. An N-terminal FLAG tag was included to confirm CAR expression levels by flow cytometry. All CAR constructs were cloned into the third-generation lentiviral transfer plasmid pCDH-EF1α (RRID: Addgene_170445) under the control of the EF1α promoter using standard molecular biology techniques.

To generate AML target cell lines for in vitro and in vivo experiments, the far-red fluorescent protein mKate2, containing a nuclear localization signal, and firefly luciferase (fLuc) were cloned into the pCDH-EF1α plasmid in a similar manner, separated by a P2A sequence.

Lentiviral vector production

HEK 293T cells (RRID: CVCL_0063) were adapted to suspension culture in serum-free LV-MAX production medium (Gibco) and designated as 293TS cells.

Lentiviral particles were generated using pCDH-derived transfer plasmids together with a third-generation packaging system comprising pMDLg/pRRE (RRID: Addgene_12251), pRSV-Rev (RRID: Addgene_12253), and pMD2.G (RRID: Addgene_12259; ref. 50). Plasmids were cotransfected into 293TS cells using the Gibco LV-MAX transfection kit. After 55 hours of incubation following transfection, viral supernatants were harvested by centrifugation at 1,300 × g for 15 minutes using a swinging-bucket rotor to remove cells and debris. Supernatants were subsequently filtered through a 0.45 μm polyethersulfone filter, aliquoted, and stored at −80°C until use.

T-cell isolation and T-cell transduction

CD3⁺ T cells were isolated from leukopaks (AllCells) using magnetic‑activated cell sorting technology (Miltenyi). Cells were activated in vitro for 24 hours in the presence of CD3/CD28 Dynabeads (Human T-Expander; Gibco, cat. #11141D) and recombinant human interleukin 7 (IL7) and IL15 (10 ng/mL each; PeproTech).

For transduction, culture plates were coated with retronectin (20 μg/mL; Takara) overnight at 4°C or for 2 hours at room temperature. Lentiviral vectors were added to the retronectin-coated plates and centrifuged at 2,000 × g for 2 hours at 32°C. Activated T cells were then resuspended and transferred onto virus-coated plates, followed by a second centrifugation at 1,000 × g for 20 minutes at 32°C to enhance transduction efficiency. CAR expression was assessed 96 hours after transduction by flow cytometry.

To support T-cell expansion and maintenance, cells were cultured in complete medium (CM) consisting of CTS OpTmizer T-Cell Expansion SFM (Gibco), supplemented with 2.5% OpTmizer T-Cell Expansion Supplement (Gibco), 2% human AB serum (BioIVT), 4 mmol/L L-glutamine (Gibco), 1× penicillin–streptomycin (Gibco), and recombinant IL7 and IL15 (10 ng/mL each; PeproTech). Cell viability and CAR expression were monitored throughout expansion.

Cell lines and primary AML samples

U937 (RRID: CVCL_0007), THP-1 (RRID: CVCL_0006), Jurkat (RRID: CVCL_0065), HL-60 (RRID: CVCL_0002), MOLM-13 (RRID: CVCL_2119), EOL-1 (RRID: CVCL_0258), HEL 92.1.7 (RRID: CVCL_2481), Raji (RRID: CVCL_0511), and HEK293T cells were obtained from the American Type Culture Collection (ATCC) and cultured under standardized conditions. Cell authentication was not repeated, as experiments were conducted using early-passage cells directly derived from the original ATCC master stock. Human leukemia and myeloid cell lines (HL-60, MOLM-13, U937, EOL-1, and THP-1) were cultured in RPMI 1640 medium supplemented with 2 mmol/L L-glutamine, 10% to 20% FBS (Gibco, cat. #A5256701), and 1% penicillin–streptomycin. All cell lines were maintained at 37°C in a humidified incubator with 5% CO₂. Original ATCC master stocks were stored in liquid nitrogen, and cultures were routinely monitored for contamination. Mycoplasma testing was performed every 3 months using the MycoAlert Mycoplasma Detection Kit (Lonza, cat. #LT07-318). To minimize genetic drift and phenotypic variability, cells were not maintained beyond 14 passages following thawing from master stocks.

293T cells were cultured in Gibco LV-MAX medium at 37°C in a humidified incubator with 8% CO₂. AML cell lines were transduced with lentiviral vectors encoding mKate2 and luciferase to generate fluorescent and bioluminescent target cells. Jurkat cells were retrovirally transduced with a nuclear factor of activated T cells (NFAT)–inducible green fluorescent protein (GFP) reporter construct and single-cell cloned by limiting dilution to generate Jurkat NFAT-GFP (JNG) reporter cells.

Primary human AML peripheral blood samples were obtained from Discovery Life Sciences. Expression of CLL-1 and CD33 was evaluated by flow cytometry using assays conducted by the contract research organization BRIDGENT.

Flow cytometry

Fluorochrome-conjugated isotype controls and monoclonal antibodies specific for human CD95 (Fas), CD8, CD3, CD45RA, CD45RO, CD62L, CCR7, CD137 (4-1BB), PD-L1, CD25, CD69, LAG-3, CD33, CLL-1 (FITC), and FLAG (allophycocyanin) were obtained from BioLegend (full catalog and RRID information as provided in the original article). Recombinant extracellular domain (ECD) proteins, including Alexa Fluor 647–conjugated CD33 and phycoerythrin (PE)-conjugated CLL-1, were obtained from ACROBiosystems.

Cells were washed with phosphate-buffered saline (PBS) and stained with Fixable Viability Dye eFluor 780 (Thermo Fisher Scientific) for 30 minutes at 4°C. Following viability staining, cells were washed and incubated with pretitrated antibody cocktails in FACS buffer (PBS supplemented with 0.2% BSA and 0.1% sodium azide) for 30 minutes at 4°C. Cells were then washed to remove unbound antibodies prior to acquisition.

Single-color compensation controls were prepared using UltraComp eBeads Plus (Thermo Fisher Scientific) according to the manufacturer’s instructions. Samples were acquired on an Attune NxT flow cytometer (Thermo Fisher Scientific), and data were analyzed using FlowJo software (version 10; BD Biosciences; RRID: SCR_008520). Gating strategies excluded doublets and dead cells prior to analysis.

Cytokine release assays

CAR T cells (1 × 10⁵) were cultured alone or cocultured with 1 × 10⁵ target cells in 200 μL of CM per well in 96-well plates. After 24 hours of incubation at 37°C with 5% CO₂, supernatants were collected and either analyzed immediately or stored at −80°C. Interferon gamma (IFNγ) levels were quantified using a Human IFNγ DuoSet ELISA kit (R&D Systems) according to the manufacturer’s instructions. All conditions were tested in triplicate.

Cytotoxicity assay

We evaluated CAR T-cell cytotoxicity against target cells in vitro using live-cell imaging. CLL-1⁺ and CD33⁺ AML cell lines (U937, HL-60, and EOL-1) expressing mKate2 and luciferase, as well as antigen-negative Raji control cells, were seeded in poly-D-lysine–coated 96-well plates (50 μg/mL; Gibco) at 10,000 cells per well. CAR T cells were added at effector-to-target (E:T) ratios of 10:1, 3.3:1, 1.1:1, and 0.33:1. Cytotoxicity was monitored using the IncuCyte S3 Live-Cell Analysis System (Sartorius), with images acquired every 2 to 3 hours for up to 48 hours. mKate2 fluorescence intensity was used to quantify viable tumor cells and was normalized to baseline values (51). Experiments were performed in triplicate.

CAR T-cell serial killing assays

To assess the target-killing ability and persistence of CAR T cells after multiple rounds of antigen challenge, we used IncuCyte live-cell imaging. Briefly, CAR T cells and U937-mKate2-Luc cells were cocultured at E:T ratios of 10:1, 3.3:1, 1.1:1, and 0.33:1 for 48 hours. Following the initial coculture, CAR T cells were rechallenged with fresh U937-mKate2-Luc cells at the same E:T ratios for an additional 48 hours (52). This rechallenge was repeated three times. After each round, CAR T cells were transferred to fresh poly-D-lysine–coated plates with target cells (10,000 cells/well). Total integrated mKate2 red fluorescence intensity per well was quantified to measure the number of viable mKate2⁺ tumor cells. Fluorescence values were normalized to controls containing tumor cells only. Each experiment was performed in triplicate to ensure reproducibility.

Coculture assay

CAR-transduced or untransduced T cells and JNG reporter cells (1 × 10⁵ cells/well) were cocultured with tumor cell lines (2 × 10⁵ cells/well) in 96-well plates at an E:T ratio of 1:2. Cultures were maintained in the absence of exogenous cytokines to assess the intrinsic activation of T cells and reporter cells. After 24 hours of coculture, cells were harvested and stained with Brilliant Violet 510–conjugated anti-CD3 (BioLegend) to identify T cells and PE-conjugated anti-CD69 (BioLegend) to assess T-cell activation.

For JNG reporter cell assays, HL-60-mKate2, U937-mKate2, and Raji-mKate2 target cells were used at an E:T ratio of 1:2 for 24 hours. Following incubation, cells were harvested and stained with Fixable Viability Dye eFluor 780 (Thermo Fisher Scientific) to exclude dead cells. Samples were then washed with FACS buffer and analyzed for GFP expression by flow cytometry, which was used as a readout of NFAT activation in Jurkat cells. This analysis was performed to evaluate the activation status of both T cells and reporter cells in response to target cells with or without antigen expression.

Colony formation assay

CLL-1 and CD33 CAR T cells, as well as untransduced T cells, were generated from peripheral blood mononuclear cells, and CD34⁺ cells enriched from bone marrow (BM) mononuclear cells were procured from matched donors (AllCells). A total of 15,000 CAR T cells or untransduced T cells per sample (E:T ratio of 10:1) were mixed with methylcellulose-based medium (MethoCult H4434 Classic; STEMCELL Technologies) and vortexed to ensure even distribution. Subsequently, 1,500 CD34⁺ cells were added to each test condition, followed by another round of vortexing for uniform mixing. Triplicate cultures were set up in 35 mm dishes and incubated at 37°C with 5% CO₂ for 14 days. After incubation, colony-forming units (CFU) were quantified, including burst-forming units, erythroid (BFU-E); CFUs, granulocyte–macrophage (CFU-GM); and CFUs, multipotent granulocyte–erythroid–macrophage–megakaryocyte (CFU-GEMM).

Xenograft model of AML and BLI

Female NSG mice (6–8 weeks old) were procured from The Jackson Laboratory (RRID: BCBC_4142) and allowed to acclimatize for 1 week at the Charles River facility in Thousand Oaks, CA, prior to experimental use. Following acclimatization, mice were intravenously injected with 1 × 10⁶ U937-mKate2-Luc cells to establish a xenograft model of AML. After tumor establishment, mice were randomized and allocated into experimental groups. Tumor burden was monitored longitudinally using BLI to quantify photon emission (photons/second/cm²/steradian) using the Revvity in vivo imaging system (IVIS; PerkinElmer).

Seven days after tumor cell injection, treatment groups were established, and 1 × 10⁷ CLL-1 CAR T cells, CD33 CAR T cells, untransduced control T cells, or PBS were administered intravenously. All animal procedures adhered to Institutional Animal Care and Use Committee (IACUC) guidelines at Charles River Laboratories (Thousand Oaks, CA, USA), ensuring compliance with ethical standards for animal welfare. Tumor progression was assessed weekly by BLI. On day 45, blood samples were collected from the remaining mice for flow cytometry analysis to quantify circulating CAR T cells. Mice in the PBS and untransduced T-cell groups were euthanized due to excessive tumor burden and signs of morbidity, in accordance with institutional guidelines for humane euthanasia.

To minimize bias, treatment group assignments were concealed from investigators during data collection and analysis. Animal health was closely monitored throughout the study, and mice showing signs of distress or disease progression beyond predefined ethical thresholds were humanely euthanized according to the Charles River Animal Care Unit protocol.

Statistical analysis

GraphPad Prism 5 software (GraphPad Software, RRID: SCR_002798) was used for statistical analyses, and data are presented as means ± standard error of the mean (SEM). Comparisons between two groups were performed using a two-tailed Student t test. For mouse experiments, survival was analyzed from the time of T-cell injection by constructing Kaplan–Meier curves and applying log-rank (Mantel–Cox) tests.

Results

CLL-1 and CD33 expression in AML cell lines and primary AML samples

To validate CLL-1 and CD33 as target antigens for CAR T-cell therapies against AML, we performed a comprehensive expression analysis of these antigens across a diverse panel of AML cell lines using flow cytometry. The human B lymphocyte cell line Raji, which inherently lacks CLL-1 and CD33 expression, served as a negative control to ensure assay specificity. Additionally, to enhance the utility of these cell lines in functional in vitro and in vivo studies, we engineered all cell lines to stably express the mKate2 far-red fluorescent protein and firefly luciferase (Luc). This dual-labeling strategy enables both real-time monitoring of mKate2⁺ cells in vitro and noninvasive imaging of tumor burden in vivo, providing a robust platform for evaluating CAR T-cell efficacy.

Our flow cytometry–based analysis revealed differential expression patterns of CLL-1 and CD33 across the AML cell lines, underscoring the heterogeneity of antigen presentation within this malignancy. High levels of CLL-1 expression were detected in U937, HL-60, and THP-1 cells, whereas EOL-1, MOLM-13, and HEL-92 cells exhibited comparatively lower expression levels (Fig. 1A). In contrast, CD33 expression was found to be consistently high across a broader range of AML cell lines, including EOL-1, HL-60, HEL-92, THP-1, MOLM-13, and U937 (Fig. 1B). CLL-1 and CD33 expression were analyzed in three primary AML peripheral blood samples to validate the relevance of cell line data to human samples (Fig. 1C and D).

Figure 1.

CLL-1 and CD33 expression in AML cell lines and patient blood samples. A, Surface expression of CLL-1 was analyzed in AML cell lines (EOL-1, HEL-92, HL-60, MOLM-13, THP-1, U937) and the CLL-1⁻CD33⁻ lymphoma cell line (Raji) using flow cytometry. Cells were stained with a FITC-conjugated anti–CLL-1 antibody (green) or left unstained (gray) to assess specific binding. B, Surface expression of CD33 was evaluated in the same AML and lymphoma cell lines using flow cytometry. Cells were stained with a PE-conjugated CD33-specific antibody (green), and fluorescence intensity was compared with the unstained control (gray). C, Gating strategy of blast population side scatter (SSC) low and CD45 low. D, Expression of CD33 and CLL-1 in the primary AML peripheral blood samples. PE- conjugated CD33-specific antibody and PE-conjugated CLL-1–specific antibody (green) and unstained control (gray). Shown are overlays of histograms from one representative experiment, including the percentage of stained cells with a higher fluorescence intensity than the unstained control. Primary AML samples were used; n = 3. Cell line data were generated from three independent experiments.

Robust characterization of CLL-1 and CD33 expression across AML cell lines and primary AML samples has significant implications for the development of CAR T-cell therapies. Selecting appropriate in vitro and in vivo models based on these expression profiles is critical to accurately assess the therapeutic potential, specificity, and cytotoxic efficiency of CAR T cells targeting these antigens. Moreover, the observed heterogeneity in CLL-1 expression highlights potential challenges associated with antigen escape. This comprehensive approach not only strengthens the translational relevance of our findings but also lays a solid foundation for advancing effective immunotherapies tailored to the antigenic diversity of AML.

Development and screening of CLL-1 and CD33 CAR constructs

We successfully developed novel CAR constructs targeting the AML-associated antigens CLL-1 and CD33 by utilizing nanobody (VHH) sequences derived from llama immunization with recombinant target antigens. VHH phagemid libraries were constructed using high-quality metrics, including optimal size and insert ratios. After two rounds of panning, the enriched libraries were analyzed for antigen-binding potential using enzyme-linked immunosorbent assay (ELISA) with crude periplasmic extracts and screened by sequencing. Clones were categorized based on CDRH3 sequence clustering and ELISA results, allowing for the identification of the most promising antigen-binding candidates.

CAR constructs were engineered to incorporate a VHH specific to either CLL-1 or CD33, fused to a CD8α hinge and transmembrane domain. The intracellular signaling domains included CD3ζ in combination with the 4-1BB (CD137) costimulatory domain. This modular design was aimed at enhancing T-cell activation, persistence, and antitumor efficacy while mitigating exhaustion.

In this study, CAR functionality was validated using two systems: JNG reporter cells and primary human T cells. Functional assessments included evaluation of CAR activation, tonic signaling (ligand-independent self-activation), cytotoxic potential, cytokine secretion, and memory phenotype. These parameters were tested in response to AML cell lines expressing differential levels of CLL-1 and CD33 antigens. Untransduced T cells were used as a negative control to ensure assay robustness and eliminate confounding variables.

To optimize construct performance, CAR T cells were screened for their ability to specifically activate and kill AML target cells while minimizing off-target activity and tonic signaling. Constructs exhibiting high specificity, potent cytotoxicity, and minimal self-activation were selected for further characterization (Fig. 2A). The CLL-1 and CD33 candidates were subsequently advanced for detailed in vitro and in vivo analyses to confirm their efficacy and safety profiles.

Figure 2.

CAR-transduced JNG reporter cells exhibit robust in vitro activation in response to CLL-1⁺ and CD33⁺ AML cell lines. A, Schematic representation of CLL-1 and CD33 CAR constructs. B and C, Flow cytometry analysis confirming CLL-1 and CD33 CAR expression in JNG reporter cells, compared with untransduced (Unt.) JNG cells. D, Flow cytometry-based assessment of tonic signaling in CAR-transduced JNG reporter cells. E and F, CAR-transduced JNG cells exhibited significant activation upon overnight coculture (1:1 E:T ratio) with the target antigen-positive AML cell lines HL-60 and U937. G, No activation of CAR-transduced JNG cells was observed when cocultured with the antigen-negative Raji cell line (1:1 E:T ratio, overnight). Stimulation of Unt. JNG cells with anti-CD3/CD28 antibodies was included as a positive control. Statistical significance was tested with data from three independent experiments using the Tukey multiple (TM) comparisons test. ****, P < 0.0001. AF647, Alexa Fluor 647.

The experimental workflow and corresponding results, including CAR construct design, systematic screening strategies, and functional validation, are detailed in the Supplementary Figures. Supplementary Figure S1A and S1B depict the design and surface expression of CLL-1–targeting CAR constructs in primary human T cells following lentiviral transduction. Basal signaling activity and tonic signaling evaluation in the absence of antigen engagement are shown in Supplementary Fig. S1C. Antigen-dependent activation of CLL-1 CAR T cells upon target recognition is shown in Supplementary Fig. S1D–S1G. Posttransduction phenotypic characterization of CLL-1 CAR T cells is provided in Supplementary Fig. S2A–S2F. Cytotoxic potential against AML cell lines expressing high, low, or undetectable levels of CLL-1 is evaluated in Supplementary Fig. S3A–S3C.

An analogous screening strategy was applied to CD33-directed CAR constructs. Supplementary Figure S4A and S4B illustrate the design and expression of CD33-targeting CARs in primary human T cells, with tonic signaling shown in Supplementary Fig. S4C. Antigen-specific activation following CD33 engagement is detailed in Supplementary Fig. S4D–S4F. Phenotypic profiling of CD33 CAR T cells after transduction is presented in Supplementary Fig. S5A–S5E, whereas cytotoxic efficacy is demonstrated in Supplementary Fig. S6A and S6B. Based on these integrated assessments, the CLL-1–specific VHH sequence X1 and the CD33-specific VHH sequence W2 were selected for downstream validation.

Our iterative engineering and validation processes culminated in the development of highly specific and functional CAR constructs capable of targeting AML cells with precision while minimizing adverse effects.

CLL-1 and CD33 CAR T cells demonstrate strong in vitro activation upon engagement with CLL-1⁺ and CD33⁺ target cells

To evaluate the activation and function of CAR T cells in response to target cells, we first expressed CLL-1 and CD33 CAR constructs in JNG reporter cells using lentiviral transduction. To confirm CAR expression, we assessed transduction efficiency and expression levels in JNG reporter cells using flow cytometry (Fig. 2B and C). Flow cytometry analysis confirmed that the CARs were successfully expressed on the surface of Jurkat cells, allowing us to proceed with subsequent activation assays. Flow cytometry analysis also confirmed that the CARs did not exhibit tonic signaling (Fig. 2D).

Next, we tested CAR activation by coculturing transduced JNG reporter cells with target cells. These target cells included U937 and HL-60 (Fig. 1A and B), both of which express CLL-1 and CD33 antigens. This was compared with JNG reporter cells cocultured with the Raji cell line (Fig. 2G), which does not express either target antigen. The E:T ratio was set at 1:1 to ensure a balanced number of T cells and target cells for optimal cell contact during the assay.

As controls, untransduced JNG reporter cells were included as a negative control. In addition, untransduced JNG reporter cells were stimulated with an anti-CD3/CD28 activation cocktail to serve as a positive control, which activates the TCR signaling pathway and induces NFAT-mediated transcriptional activation, resulting in GFP expression in JNG reporter cells (53).

The experimental setup was based on the principle that JNG reporter cells are engineered to express GFP under the control of the NFAT promoter, which is induced by TCR and CAR signaling.

After a 24-hour coculture period, GFP expression was analyzed by flow cytometry to assess the level of T-cell activation. In Fig. 2E–G, GFP expression data are shown for JNG reporter cells cocultured with AML cell lines U937 (Fig. 2E) and HL-60 (Fig. 2F). Both CLL-1 and CD33 CAR constructs demonstrated activation levels nearly equivalent to the positive control (anti-CD3/CD28 activation cocktail), indicating that these CAR T cells were successfully activated by target cells expressing the respective antigens (CLL-1 and CD33). Robust GFP expression under these experimental conditions suggests that CAR-transduced cells efficiently recognize and respond to target cells.

In contrast, when JNG reporter cells were cocultured with Raji cells, which do not express the target antigens, no GFP expression was observed (Fig. 2G). This lack of activation confirms that CAR-mediated T-cell activation requires the presence of the specific target antigens (CLL-1 and CD33) and that the absence of these antigens on Raji cells prevents nonspecific T-cell activation.

In summary, the results shown in Fig. 2E–G demonstrate that CARs targeting CLL-1 and CD33 are functionally active and capable of inducing T-cell activation in response to target cells expressing the corresponding antigens. Furthermore, the absence of activation in response to Raji cells highlights the antigen specificity of these CAR constructs, as activation was observed only in the presence of the appropriate target antigen. These results underscore the potential of these CAR constructs for selective targeting of AML cells while minimizing off-target effects.

CLL-1 and CD33 CARs show no significant impact on T-cell viability, expansion, or memory phenotype

Based on the results of the JNG reporter assay, we further characterized human T cells following transduction with CLL-1 and CD33 CAR constructs. Flow cytometric analysis confirmed successful expression of CLL-1 and CD33 CARs in transduced T cells, as demonstrated by staining with the respective ECD proteins derived from the target antigens (Fig. 3A and B). Comparative analyses were conducted with untransduced T-cell controls to ensure assay accuracy. We also examined the CD4/CD8 T-cell ratio after transduction; however, no significant difference was observed (Fig. 3C).

Figure 3.

Primary human CAR T cells maintain viability, expansion capacity, and phenotype after transduction. A and B, Flow cytometry analysis confirming the surface expression of CLL-1 and CD33 CARs in human T cells. CAR expression was detected using Alexa Fluor 647 (AF647)–conjugated CD33 and PE-conjugated CLL-1 ECD proteins, demonstrating successful transduction. C, The proportion of CD4⁺ and CD8⁺ T cells was assessed after transduction to evaluate potential shifts in the T-cell subset composition. D, CAR T cells were evaluated for tonic signaling by detecting CD137 (4-1BB) expression in CAR⁺ cells compared with CD134 expression in all T cells for the untransduced (Unt.) control. No significant spontaneous activation was observed, indicating an absence of constitutive signaling. E and F, CAR T-cell expansion and viability were assessed after transduction using an automated cell counter, comparing CAR T cells to Unt. T cells. No significant differences in expansion rates or viability were observed, suggesting that CAR expression did not impair T-cell growth or survival. G, Memory phenotype analysis was performed to characterize CAR T cells after transduction, comparing them with Unt. T cells to determine whether the transduction process influenced differentiation into stem cell–like memory (T_SCM), central memory (T_CM), effector memory (T_EM), or terminally differentiated effector T cells (T_EMRA). T_SCM were identified as CD95⁺CD45RA⁺CD45RO⁻CCR7⁺ cells, T_CM as CD45RA⁻CD45RO⁺CCR7⁺ cells, T_EM as CD45RA⁻CD45RO⁺CCR7⁻ cells, and T_EMRA as CD45RA⁺CD45RO⁻CCR7⁻ cells. Data were generated from three independent experiments. A two-way ANOVA was performed to determine the statistical significance between groups based on two or more independent variables. The Tukey multiple comparisons test was used. ****, P < 0.0001; ns = not significant.

To assess basal signaling activity, we evaluated the expression of the activation marker CD137 (4-1BB) in the absence of target antigen engagement. The results demonstrated no evidence of tonic signaling or spontaneous CAR T-cell activation (Fig. 3D). Furthermore, CAR transduction was evaluated for its impact on T-cell growth and viability, with no significant differences observed between CAR-transduced and untransduced T cells (Fig. 3E and F).

The memory phenotype of CAR-transduced T cells was also examined and compared with that of untransduced T cells. Analysis revealed no significant alterations in the distribution of memory subsets, suggesting that the transduction process preserves intrinsic T-cell memory characteristics.

Overall, these findings indicate that CLL-1 and CD33 CAR transduction is well tolerated by primary human T cells, maintaining normal viability, growth, and memory phenotype without inducing tonic signaling or other adverse effects.

CLL-1 CAR induces an antigen-specific cytokine response upon target recognition

To evaluate the cytokine response of CLL-1 and CD33 CAR T cells, as well as untransduced T cells, we quantified IFNγ release upon coculture with U937 and HL-60 target cells expressing the antigens or antigen-negative Raji cells. The coculture was conducted at an E:T ratio of 1:1 for 24 hours, with culture medium used as a background control. Supernatants were collected after incubation to measure IFNγ secretion by ELISA, serving as an indicator of CAR T-cell function. Both CLL-1 and CD33 CAR T cells demonstrated robust IFNγ production when cocultured with antigen-expressing target cells, whereas minimal cytokine production was observed with antigen-negative target cells and untransduced T-cell controls (Fig. 4D).

Figure 4.

Nanobody-based CLL-1 and CD33 CAR T cells effectively kill antigen-expressing tumor cells. A, CLL-1 CAR T cells, CD33 CAR T cells, or untransduced (Unt.) T cells were cocultured with U937-mKate2 cells, a CLL-1 and CD33-expressing AML cell line, at varying E:T ratios ranging from 10:1 to 0.33:1. Cytotoxicity was monitored in real time over 42 hours using the IncuCyte imaging system, with target cells labeled with mKate2 far-red fluorescent protein. B and C, CLL-1 CAR T cells, CD33 CAR T cells, or Unt. T cells were cocultured with HL-60-mKate2 cells, another CLL-1 and CD33-expressing AML cell line, at E:T ratios ranging from 10:1 to 0.33:1. Cytotoxicity was again monitored in real time over 42 hours using the IncuCyte imaging system, with target cells labeled with mKate2 far-red fluorescent protein. Raji cells were included as a negative control to assess the specificity of CAR T-cell activity. Unt. T cells were used to evaluate nonspecific killing. Shown are the mean values of four images acquired by the IncuCyte at each time point and condition, from one representative experiment out of three independent repeats with three different donors. D, IFNγ cytokine release of CLL-1 and CD33 CAR T cells or Unt. T cells in response to a 24-hour coculture with antigen-positive U937 and HL-60 target cells. Raji cells were used as a negative control, and plain culture medium was used as a background control. Data represent mean values of triplicate wells ± standard deviation. N = 3 T-cell donors were used. Cytokine release data were generated from three independent experiments. A two-way ANOVA was performed to determine the statistical significance between groups based on two or more independent variables. Šídák’s multiple comparisons test was used. *, P < 0.0168; ****, P < 0.0001; ns, not significant.

CLL-1 CAR T cells secreted slightly higher levels of IFNγ compared with CD33 CAR T cells, indicating a stronger activation and effector response upon antigen engagement. Although IFNγ contributes to the antitumor activity of CAR T cells (54, 55), excessive and uncontrolled release of IFNγ can result in serious adverse effects, such as cytokine release syndrome (CRS; refs. 56, 57). As CRS is a complex phenomenon involving multiple additional cell types, such as myeloid immune cells, the risk of CRS cannot be adequately assessed in this in vitro experiment and requires evaluation in suitable animal models.

CLL-1 CAR T cells exhibit robust cytolytic activity against AML cells expressing target antigens

To evaluate the cytotoxic potential of CLL-1 CAR T cells, we utilized two AML cell lines: U937 (high CLL-1 expression) and EOL-1 (low CLL-1 expression), along with Raji cells as a negative control (no CLL-1 expression). CAR T cells were cocultured with these cell lines at varying E:T ratios ranging from 10:1 to 0.33:1. U937 cells exhibited robust cytotoxicity, even at the lowest E:T ratio of 0.33:1, highlighting the high efficacy of CLL-1 CAR T cells (Supplementary Fig. S3A). In contrast, significant cytotoxicity against EOL-1 cells was observed only at higher E:T ratios (10:1; Supplementary Fig. S3B). Importantly, no cytotoxic activity was observed against Raji cells, confirming the specificity of CLL-1 CAR T cells (Supplementary Fig. S3C). Based on these results, the X1 CLL-1 CAR construct was selected for further analysis.

Similarly, CD33 CAR T cells were screened using U937 and HL-60 cell lines, both of which express high levels of CD33. Four distinct CD33 VHH nanobody constructs (A1, R1, U2, and W2) were tested. Based on cytotoxicity and activation data, the W2 CD33 CAR construct demonstrated superior performance and was selected for downstream analysis (Supplementary Figs. S4–S6).

To evaluate the combined functionality of the selected constructs, X1 CLL-1 and W2 CD33 CAR T cells were cocultured with U937 and HL-60 cells, which coexpress both CLL-1 and CD33 antigens. Raji cells served as a negative control. Cocultures were performed across multiple E:T ratios (10:1, 3.3:1, 1.1:1, and 0.33:1) to assess cytotoxicity under varying conditions. Cytotoxicity was monitored in real time over 42 hours using the IncuCyte imaging system, with target cells labeled with mKate2 far-red fluorescent protein. Fluorescence intensity was normalized to cocultures with untransduced T cells as negative controls. Both CAR T cells displayed significant cytotoxicity against U937 and HL-60 cells across all E:T ratios while exhibiting no detectable cytotoxicity against Raji cells, underscoring the antigen specificity of the CAR T cells (Fig. 4A–C).

These results provided strong evidence for the efficacy and specificity of the CAR T cells, supporting further investigations into their persistence. We evaluated the functional activity of nanobody-based CLL-1 CAR T cells in comparison with their scFv-based counterparts. Both CAR formats demonstrated comparable cytolytic potency against the CLL-1–high U937 cell line, indicating equivalent target-specific efficacy (Supplementary Fig. S8E). Notably, when assessed against the CLL-1–low EOL-1 cell line, nanobody-based CAR T cells exhibited enhanced cytotoxic activity, particularly at lower E:T ratios, reflecting improved antigen sensitivity and superior performance under conditions of reduced antigen density (Supplementary Fig. S8A–S8G). As no scFv-based CLL-1 CAR T-cell therapy has yet reached clinical approval, these findings highlight nanobody-based CLL-1 CAR T cells as a promising and compact option with potential therapeutic benefit for targeting AML.

A serial killing assay was conducted to assess the persistence and serial killing capacity of CAR T cells. Both CD33 and CLL-1 CAR T cells were cocultured with U937 cells expressing both CLL-1 and CD33 antigens. Following an initial 48-hour incubation, CAR T cells were rechallenged with fresh target cells for three consecutive cycles. Robust cytotoxicity was observed across all cycles, even at the lowest E:T ratio of 0.33:1 (Fig. 5), indicating sustained antitumor activity. CAR T cells proliferate upon initial antigen encounter, increasing effector numbers and enhancing cytotoxicity during subsequent rechallenges, particularly at low E:T ratios. At high E:T ratios, rapid target elimination saturates cytotoxic capacity, limiting further impact upon rechallenge. This serial killing and expansion dynamic has been described previously, highlighting the role of CAR T-cell proliferation and persistence in enhancing functional activity (58, 59).

Figure 5.

Nanobody-based CLL-1 and CD33 CAR T cells exhibit sustained cytotoxicity against antigen-expressing tumor cells. CLL-1 CAR T cells, CD33 CAR T cells, and untransduced (Unt.) T cells were cocultured with U937-mKate2 cells, an AML cell line with high CLL-1 and CD33 expression, at varying E:T ratios ranging from 10:1 to 0.33:1. Cytotoxicity was continuously monitored for 42 hours using the IncuCyte live-cell imaging system, with target cells labeled with mKate2 far-red fluorescent protein. To assess serial killing capacity, CAR T cells were rechallenged with fresh U937-mKate2 cells three consecutive times. CLL-1 and CD33 CAR T cells demonstrated comparable sustained killing efficiency even at low E:T ratios. Unt. T cells were included as a control to evaluate nonspecific killing. Shown are the mean values from four images acquired by the IncuCyte at each time point and condition, from one representative experiment out of three independent repeats with three different T-cell donors.

These findings highlight the persistent cytotoxic capability of CLL-1 and CD33 CAR T cells and provide a compelling rationale for their further evaluation in preclinical mouse models.

Nanobody-based CAR T-cell therapy exhibits strong antileukemic activity and prolonged survival in animal models

To evaluate the antileukemic activity of CLL-1 and CD33 CAR T cells in vivo, we developed a xenogeneic systemic AML model using the U937 leukemic cell line, which was genetically modified to stably express the luciferase reporter gene (U937-Luc) for noninvasive tumor monitoring. NSG mice were intravenously injected with 1 × 10⁶ U937-Luc cells on day 1 to establish systemic leukemia. Following tumor engraftment, mice were randomized into four treatment groups: CLL-1 CAR T cells, CD33 CAR T cells, untransduced T cells (mock), and PBS. On day 7 after tumor inoculation, mice received an intravenous injection of 1 × 10⁷ cells from their assigned treatment group (Fig. 6A). Tumor progression was monitored weekly using BLI, and body weight changes were recorded as a measure of systemic health (Fig. 6E).

Figure 6.

Nanobody-based CLL-1 and CD33 CAR T cells effectively eliminate human AML in vivo in a xenograft model. A, Schematic representation of the U937 xenograft model. To establish leukemia in vivo, NSG mice were intravenously injected via the tail vein with 1 × 10⁶ U937-mKate2-Luc AML cells expressing firefly luciferase on day 0. Disease engraftment was confirmed on day 7 using BLI, and mice were subsequently randomized into four treatment groups. On day 7, mice received i.v. injections of either CLL-1 CAR T cells (1 × 10⁷), CD33 CAR T cells (1 × 10⁷), untransduced (Unt.) T cells (1 × 10⁷), or PBS as a control. AML burden was longitudinally monitored via serial BLI, with bioluminescence radiance serving as a surrogate marker for tumor progression. B, Tumor burden assessment through BLI. Representative BLI images of mice from each treatment group were taken at key time points: before treatment (day 7), early response (day 14), and throughout the study period until day 63 after tumor injection. Changes in tumor burden were visualized and quantified to assess therapeutic efficacy. C, Kaplan–Meier survival analysis. Survival curves were generated to compare overall survival between treatment groups. Statistical analyses were performed using the log-rank (Mantel–Cox) test to determine significant differences in survival outcomes. D, Quantitative analysis of bioluminescence signal over time. Serial BLI measurements were used to track leukemia progression and response to therapy. The bioluminescent signal was quantified and plotted for each treatment group, represented as mean radiance (photons/second/cm²/sr) ± standard deviation (SD). E, Body weight monitoring was initiated on day 14, 1 week following CAR T-cell administration, as a measure of overall treatment tolerance and potential toxicity. Mouse body weight was recorded throughout the study to assess potential treatment-related toxicity. No significant weight loss was observed, indicating that CLL-1 and CD33 CAR T-cell therapy was well tolerated. F, The graph showed the CD8⁺ CAR expression in mouse blood at day 45. Tumor burden data used two-way ANOVA to determine the statistical significance between groups based on two or more independent variables. A t test was performed for CD8⁺ CAR detection to determine statistical significance. Each treatment group consisted of n = 5 mice. *, P < 0.0197; ****, P < 0.0001.

Control mice treated with mock T cells or PBS displayed rapid disease progression, with the majority succumbing to leukemia by day 20 (Fig. 6B). In contrast, mice treated with CLL-1 CAR T cells and CD33 CAR T cells showed a marked reduction in systemic leukemic burden, as evidenced by a sustained decrease in bioluminescent signal intensity over time (Fig. 6B and D). Remarkably, complete tumor eradication was achieved in CAR T cell–treated groups within 1 week of infusion, as corroborated by a significant reduction in bioluminescent signal intensity (Fig. 6D). This rapid and efficient tumor clearance highlights the potent antileukemic activity of both CLL-1 and CD33 nanobody-based CAR T cells.

Despite these promising outcomes, one mouse in the CLL-1 CAR T cell–treated group developed GVHD symptoms after 5 weeks of treatment. Consistent with Charles River IACUC ethical guidelines, this mouse was euthanized upon confirmation of GVHD symptoms. GVHD is a well-documented complication in xenograft models employing human CAR T cells, arising from mismatches between human T cells and the immunocompromised host.

Survival analysis revealed a significant extension of lifespan in mice treated with CLL-1 CAR T cells and CD33 CAR T cells compared with mock T-cell or PBS controls (P < 0.001; Fig. 6C), demonstrating the therapeutic efficacy of these CAR T cells against systemic AML. We also monitored mouse body weight as an indicator of overall health status. Weight measurements began on day 14, following IVIS imaging used to confirm tumor engraftment. The untreated and PBS control groups were euthanized on day 20 due to significant disease burden, as verified by IVIS imaging; therefore, body weight data for these groups were not collected beyond this time point. In contrast, no significant changes in body weight were observed in any of the CAR T cell–treated groups (Fig. 6E).

To further investigate the persistence and activation status of CAR T cells, we analyzed peripheral blood from sacrificed mice. No tumor cells were detected in the blood of CLL-1 and CD33 CAR T cell–treated mice. Flow cytometric detection of mKate2-expressing tumor cells enabled monitoring of their presence in the peripheral blood of CAR T cell–treated mice (Supplementary Fig. S7A), whereas CAR T cells were observed at elevated frequencies in the peripheral blood of the CLL-1 CAR T cell–treated group compared with the CD33 CAR T-cell group (Fig. 6F). The activation marker CD69 was expressed at low levels in both CLL-1 and CD33 circulating CAR T cells, indicating that these cells may exist in a quiescent state in the bloodstream and remain capable of reactivation upon antigen encounter (Supplementary Fig. S7B and S7C). This suggests that CLL-1 (and, to a lesser extent, CD33) CAR T cells exhibit both long-term persistence and a poised activation state, enabling sustained antileukemic activity without continuous activation.

These findings provide strong preclinical evidence supporting the application of CLL-1 and CD33 CAR T-cell therapy for hematologic malignancies expressing CLL-1. The combination of efficient tumor clearance, long-term CAR T-cell persistence, and antigen-driven activation potential underscores the therapeutic promise and safety of CLL-1 CAR T-cell strategies for AML treatment.

Nanobody-based CLL-1 CAR T cells demonstrate selective targeting without affecting hematopoietic stem cells

After gathering in vivo efficacy data, we further evaluated the on-target, off-tumor toxicity of our CAR T-cell constructs using a colony formation assay. The CFU assay is a well-established and highly sensitive method for evaluating the effects of therapeutic agents on the differentiation and proliferative capacity of early hematopoietic progenitors (60, 61). In a CD70 CAR T-cell study, in vitro colony formation assays were utilized to assess off-tumor toxicity as an alternative to in vivo models (62). Moreover, previous research has demonstrated a strong correlation between in vitro CFU assay results and in vivo hematopoietic toxicity outcomes (63, 64). Given their human relevance, cost-effectiveness, and ethical considerations, CFU assays serve as a valuable alternative to in vivo studies.

On-target, off-tumor toxicity remains one of the primary challenges in CAR T cell–based therapy, particularly for targets expressed on normal tissues. Previous studies have demonstrated that even minimal expression of target antigens on nonmalignant cells can lead to severe, unintended toxicity (26, 65, 66). In the context of hematologic malignancies, CAR T cells targeting antigens shared with HSCs may disrupt normal hematopoiesis, leading to significant hematologic toxicity. Therefore, assessing the impact of our nanobody-engineered CLL-1 and CD33 CAR T cells on hematopoietic progenitor function was critical for determining their therapeutic safety profile.

To evaluate potential hematopoietic toxicity, we conducted a colony formation assay using CD34-enriched normal BM samples. CLL-1 and CD33 CAR T cells were generated from healthy donors and cocultured with autologous CD34⁺ HSCs for 6 hours at an E:T ratio of 10:1. Following coculture, cells were plated in methylcellulose-based media and incubated for 14 days to allow colony formation. The number and types of colonies formed were then quantified, including BFU-E, CFU-GM, and CFU-GEMM, which are indicative of erythroid, myeloid, and multipotent hematopoietic progenitors, respectively.

Our results revealed significant differences in BFU-E counts between groups. Specifically, CD33 CAR T cell–treated samples exhibited a marked reduction in BFU-E formation compared with both nontransduced T-cell controls and CLL-1 CAR T cell–treated samples, indicating potential erythropoietic suppression. As initially hypothesized, our findings support that CLL-1 CAR T cells did not significantly affect BFU-E formation, suggesting that CLL-1 targeting does not impair erythroid progenitor function (Fig. 7A and C). Furthermore, CFU-GM and CFU-GEMM colony counts remained comparable across all experimental conditions, indicating that neither CD33 nor CLL-1 CAR T cells substantially affected myeloid progenitor viability or differentiation capacity (Fig. 7B and D). Moreover, analysis of total CFU counts confirmed that overall progenitor activity was preserved in the presence of CLL-1 CAR T cells, whereas CD33 CAR T-cell treatment selectively reduced erythroid lineage output (Fig. 7E).

Figure 7.

CD33 CAR T cells, but not CLL-1 CAR T cells, exhibit myelotoxicity in vitro. CLL-1 and CD33 CAR T cells, as well as untransduced (Unt.) T cells, were cocultured with donor-matched BM-derived CD34⁺ hematopoietic progenitor cells at an E:T ratio of 10:1 under hematopoiesis-inducing conditions for 14 days. A, Representative image of BFU-E formation after coculture of CD34⁺ cells with CAR T cells. B, Representative image of CFU-GM formation. C, Quantification of BFU-E formation. D, Quantification of CFU-GM formation. E, Quantification of total CFU formation. Data represent mean values of triplicates ± standard deviation (SD). **, P < 0.0025; ns, not significant.

These findings provide critical insights into the safety profile of our nanobody-engineered CLL-1 CAR T cells. Unlike CD33 CAR T cells, which may pose a risk of hematopoietic toxicity due to their impact on erythroid progenitors, CLL-1 CAR T cells seem to spare normal HSC-derived colony formation. This suggests that compared with CD33 CAR T cells, CLL-1 CAR T-cell therapy could represent a viable and safer immunotherapeutic strategy for AML, minimizing the risk of myelosuppression and off-target hematologic toxicity.

Discussion

We have developed a novel nanobody-based CLL-1 CAR T-cell therapy, specifically designed to target AML blasts and LSCs while minimizing the risk of normal HSC depletion. CLL-1, a glycosylphosphatidylinositol-anchored protein, is a highly expressed antigen on the surface of AML cells, particularly on LSCs, and is absent in most normal hematopoietic progenitor cells, making it an ideal target for selective immunotherapy (1, 2). The key challenge in developing targeted therapies for AML lies in avoiding damage to normal hematopoiesis. Based on a comprehensive analysis of completed and ongoing clinical trials, most early-phase CAR T-cell studies have primarily focused on evaluating safety profiles. Initial clinical data indicate that CLL-1 represents a safer and more effective target for CAR T-cell therapy in AML (29–33). This is largely due to its restricted expression on malignant myeloid cells and minimal expression on normal hematopoietic lineages, which addresses off-tumor toxicity as one of the major safety concerns of CAR T-cell therapy. Consequently, CLL-1–directed CAR T cells have demonstrated promising efficacy while potentially minimizing collateral damage to healthy tissues, supporting their continued development in both relapsed/refractory and newly diagnosed AML settings. Our data demonstrate that CLL-1 CAR T cells effectively target AML cells without impairing normal HSC function. Colony formation assays performed on human HSCs cocultured with CLL-1 CAR T cells showed no significant decrease in HSC viability or differentiation potential, supporting the hypothesis that the CLL-1 antigen is a viable therapeutic target with minimal off-target toxicity compared with CD33 CAR T cells.

To evaluate the functional performance of CLL-1 and CD33 CAR T cells, we focused on assessing antigen-dependent signaling and effector function while carefully monitoring basal activity in the absence of target engagement. Tonic signaling, which arises from antigen-independent CAR activation, remains a significant challenge in CAR T-cell therapy, as it drives inappropriate cytokine release, accelerates T-cell exhaustion, and compromises long-term persistence. Therefore, ensuring minimal basal signaling is critical for maintaining CAR T-cell fitness and therapeutic durability.

Although traditional CAR T-cell strategies rely on bulkier scFvs from monoclonal antibodies for antigen recognition (48, 66–68), the smaller size of nanobody-based CARs may allow for more efficient antigen binding with fewer steric restraints and may also reduce the risks of immunogenicity and nonspecific binding. In line with this, our nanobody-based CLL-1 CAR T cells demonstrated highly specific activation and cytokine responses when cocultured with CLL-1⁺ AML cell lines, but not with antigen-negative controls. In cytotoxicity assays, CLL-1 CAR T cells exhibited potent and dose-dependent tumor cell killing, even at an E:T ratio as low as 0.33:1. This indicates that the CAR T cells can efficiently eliminate AML blasts at low cell concentrations, which is a critical feature for effective therapeutic application. Additionally, serial killing assays, in which CAR T cells were cocultured with AML target cells over multiple rounds, demonstrated sustained cytotoxicity, further supporting the long-lasting efficacy of CLL-1 CAR T cells in targeting and eliminating leukemia cells.

To validate the preclinical efficacy of CLL-1 CAR T cells in vivo, we employed a xenograft mouse model using U937 mKate2-Luc cells, a bioluminescent human AML cell line. Upon engraftment, robust tumor growth was observed in control mice, whereas CLL-1 and CD33 CAR T cell–treated mice exhibited rapid tumor regression, as evidenced by bioluminescent imaging. CLL-1 CAR T cells efficiently targeted and eliminated AML cells, resulting in a significantly reduced tumor burden. However, residual tumor growth was detected for several weeks in some animals, which may be attributable to the persistence of antigen-negative or low antigen–expressing leukemic cells within the tumor. This finding suggests that future refinements to CAR T-cell constructs or combination therapies may be necessary to further enhance efficacy and overcome antigen escape. Importantly, CLL-1 CAR T cells did not significantly affect overall mouse health, with minimal signs of off-target toxicity or damage to normal tissues, highlighting the specificity of this approach.

Despite these promising results, we observed GVHD in one of the CLL-1 CAR T cell–treated mice. GVHD is a well-documented complication in allogeneic settings, particularly when human T cells are transferred into immunocompromised host mice. GVHD results from the recognition of host tissues as foreign by infused T cells, leading to tissue damage and systemic inflammation. To mitigate graft rejection, we utilized immunocompromised NSG mice, which are deficient in both adaptive and innate immune responses, thereby minimizing alloreactivity by host immune cells. However, the risk of GVHD remains a challenge in xenograft models, which may exaggerate human immune cell activation due to the lack of human-specific immune regulation.

Traditional xenograft models may also exaggerate GVHD pathology due to inflammatory conditions induced by pretransplant conditioning regimens, limiting their translational relevance (69). Emerging humanized mouse models, such as NSG–human leukocyte antigen mice engineered to express human MHC molecules (44, 70), provide a more accurate representation of human immune responses and may offer improved platforms for GVHD studies. Other mouse models, such as the NOD B6 SCID Gamma W41 (NBSGW) strain, also enable engraftment of human cells without radiation-based conditioning and may provide more accurate insights into GVHD dynamics (71–73). Additional approaches, including engineering CAR T cells with altered costimulatory domains or fine-tuning activation thresholds, have been explored to minimize alloreactivity (74, 75). However, although these considerations are important in true allogeneic and post-HSCT settings, GVHD is not generally a concern in autologous CAR T-cell applications (76). These models underscore the need for caution when extrapolating preclinical findings to clinical outcomes.

Our study highlights a novel nanobody-based CLL-1 CAR T-cell strategy that enhances antitumor efficacy through target-dependent activation while minimizing off-target toxicity. By leveraging nanobody technology, this approach offers superior specificity and a reduced risk of autoreactivity, as evidenced by robust preclinical data. The use of nanobody-based CARs targeting the CLL-1 antigen demonstrates strong potential to improve therapeutic outcomes in AML, particularly by preserving normal hematopoiesis and avoiding excessive inflammation. These findings lay the groundwork for future clinical translation, with the ultimate goal of developing safer and more effective CAR T-cell therapies for patients with AML.

Conclusion

This study presents a nanobody-based CLL-1 CAR T-cell strategy that targets CLL-1 in AML. By harnessing the specificity of CLL-1 antigen recognition, this CAR T-cell approach not only drives potent and selective T-cell activation but also enhances cytotoxicity while minimizing off-target effects. These promising results underscore the therapeutic potential of this strategy, paving the way for clinical translation and offering a transformative solution to improve outcomes for patients with AML.

Supplementary Material

Supplementary Figure S1

Figure S1. CLL-1 nanobodies clones screening in Jurkat NFAT GFP reporter cell line

Supplementary Figure S2

Figure S2. CLL-1 nanobodies clones screening in Human T cells

Supplementary Figure S3

Figure S3. Screen the cytotoxic potential of CLL-1 nanobodies clones.

Supplementary Figure S4

Figure S4. CD33 nanobodies clones screening in Jurkat NFAT GFP reporter cell line

Supplementary Figure S5

Figure S5. CD33 nanobodies clones screening in Human T cells

Supplementary Figure S6

Figure S6. Screen the cytotoxic potential of CD33 nanobodies clones.

Supplementary Figure S7

Figure S7. CAR expression and activation in mice blood

Supplementary Figure S8

Figure S8. Nanobody-based CAR T cells represent an effective and compact alternative to scFv-based CAR constructs

Graphical Abstract

Graphical Abstract

Data Availability

Data and materials generated in this study are available from the corresponding authors upon request. Supplementary data are available online.

Authors’ Disclosures

C. Tripathi reports a patent to USPTA Patent file: 19/259,537 pending. J. Nguyen reports a patent to USPTA Patent file: 19/259,537 pending. V. Natarajan reports other from Theragent Inc. and the Sino-American Cancer Foundation (SACF) during the conduct of the study as well as other from Theragent Inc. and SACF outside the submitted work; a patent to 19/259,537 pending; and employment with Theragent, Inc. (October 2020–March 2024) when the work described in this article was conceived, initiated, and substantially developed, and during that time, the CAR T cells and nanobody programs were established, and early data were generated. V. Natarajan, along with team members, subsequently transitioned to SACF in April 2024, where aspects of the work were continued; accordingly, both Theragent and SACF were involved at different stages of the research timeline. Y. Yen reports a patent to 19/259,537 pending. No disclosures were reported by the other authors.

Authors’ Contributions

C. Tripathi: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. S. Zolov: Data curation, formal analysis, investigation, methodology, writing–review and editing. J. Nguyen: Data curation, investigation, methodology, writing–review and editing. F. Luh: Formal analysis, writing–review and editing. V. Natarajan: Conceptualization, supervision, investigation, project administration, writing–review and editing. Y. Yen: Conceptualization, supervision, funding acquisition, writing–review and editing.

Ethics Approval and Consent to Participate

All animal experiments were conducted in accordance with protocols approved by the IACUC at the Charles River facility (Thousand Oaks, CA, USA; Protocol No. 2024-2122). Human peripheral blood samples were obtained from Discovery Life Sciences under Ethics Committee approval for medical research (Protocol No. SB-LB01_v.1.1). Sample collection followed established standard operating procedures, and written informed consent was obtained from all donors prior to participation.