Article Text

Original research
Asymmetric anti-CLL-1×CD3 bispecific antibody, ABL602 2+1, with attenuated CD3 affinity endows potent antitumor activity but limited cytokine release
  1. Eunhee Lee1,
  2. Shinai Lee1,
  3. Sumyeong Park1,
  4. Yong-Gyu Son1,
  5. Jiseon Yoo1,
  6. Youngil Koh2,
  7. Dong-Yeop Shin2,
  8. Yangmi Lim1 and
  9. Jonghwa Won1
  1. 1ABL Bio Inc, Seongnam, Korea (the Republic of)
  2. 2Department of Internal Medicine, Seoul National University Hospital, Jongno-gu, Korea (the Republic of)
  1. Correspondence to Dr Jonghwa Won; jonghwa.won{at}; Dr Yangmi Lim; yangmi.lim{at}


Background Acute myeloid leukemia (AML) is a type of leukemia in adults with a high mortality rate and poor prognosis. Although targeted therapeutics, chemotherapy, and hematopoietic stem cell transplantation can improve the prognosis, the recurrence rate is still high, with a 5-year survival rate of approximately 40%. This study aimed to develop an IgG-based asymmetric bispecific antibody that targets CLL-1 and CD3 for treating AML.

Methods ABL602 candidates were compared in terms of binding activity, T-cell activation, and tumor-killing activities. ABL602-mediated T-cell activation and tumor-killing activities were determined by measuring the expression of activation markers, cytokines, cytolytic proteins, and the proportion of dead cells. We evaluated in vivo tumor growth inhibitory activity in two mouse models bearing subcutaneously and orthotopically engrafted human AML. Direct tumor-killing activity and T-cell activation in patient-derived AML blasts were also evaluated.

Results ABL602 2+1 showed a limited CD3 binding in the absence of CLL-1, suggesting that steric hindrance on the CD3 binding arm could reduce CLL-1 expression-independent CD3 binding. Although the CD3 binding activity was attenuated compared with that of 1+1, ABL602 2+1 exhibited much stronger T-cell activation and potent tumor-killing activities in AML cell lines. ABL602 2+1 efficiently inhibited tumor progression in subcutaneously and orthotopically engrafted AML mouse models. In the orthotopic mouse model, tumor growth inhibition was observed by gross measurement of luciferase activity, as well as a reduced proportion of AML blasts in the bone marrow, as determined by flow cytometry and immunohistochemistry (IHC) staining. ABL602 2+1 efficiently activated T cells and induced the lysis of AML blasts, even at very low effector:target (E:T) ratios (eg, 1:50). Compared with the reference 1+1 antibody, ABL602 did not induce the release of cytokines including interleukin-6 and tumor necrosis factor-α in the healthy donor-derived peripheral blood mononuclear cell.

Conclusions With its potent tumor-killing activity and reduced cytokine release, ABL602 2+1 is a promising candidate for treating patients with AML and warrants further study.

  • cytotoxicity, immunologic
  • immunotherapy
  • T-lymphocytes
  • hematologic neoplasms

Data availability statement

Data are available on reasonable request.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Acute myeloid leukemia (AML) is a major therapeutic challenge with a high unmet need in hematologic oncology, however, there have been major huddles such as severe cytokine syndrome or pancytopenia in developing T cell redirecting antibodies and chimeric antigen receptor-T cells (CAR-T).


  • This study presents that CD3 binding activity can be attenuated by the adjacent Fab arm-driven steric hindrance and CD3 binding can be potentiated in the presence of CLL-1 expressing cells.

  • With reduced CD3 binding activity, ABL602 2+1 shows lack of cytokine release, especially tumor necrosis factor-α and interleukin 6, without sacrificing tumor inhibitory efficacy.

  • Much stronger tumor cytolytic activity was observed by 2+1 format compared to 1+1 format potentially due to more stabilized and proximal association of T cells and tumor cells.


  • We expect that two main characteristics, including lowered cytokine release and potentially limited pancytopenia by targeting CLL-1, would place ABL602 2+1 as a new therapeutic option in treating AML.


Acute myeloid leukemia (AML) is one of the most common and fatal hematological malignancies in adults, with a majority of them having a poor prognosis. Despite major progress in the field of immunotherapy over the last four decades, AML is a major therapeutic challenge with a high unmet need in hematologic oncology.1 2

Leukemia surface proteins, including CD33, CD123 (the alpha chain of the interleukin 3 receptor), FLT-3 (FMS-like tyrosine kinase 3), and CLL-1, also known as CLEC12A (C-type lectin domain family 12 member A), have been extensively developed as antibody-drug conjugates, T-cell engagers (TCE), and chimeric antigen receptor-T cells (CAR-T).3 4 The CD33 antibody-drug conjugate, gemtuzumab ozogamicin, was the first antibody-drug conjugate (ADC) approved for treating patients with AML, validating CD33 as a therapeutic target.5–9 However, other immunotherapies such as CAR-T and TCE are still in early clinical stages and many of them have been discontinued. This is partly due to the absence of AML-specific targets and the other is one of the technical issues associated with TCE, for example, using CD3 binding arm with high affinity, leading to poor biodistribution and severe cytokine release syndrome.9–11 Currently, there are no AML/leukemic stem cell-specific antigens, such as CD19 in B-cell acute lymphoblastic leukemia, and most tumor-associated antigens (TAA) are expressed in normal myeloid lineage cells, such as neutrophils, monocytes, and thrombocytes. Therefore, a certain degree of functional loss of the protective immune response against infection or impairment of blood clotting is unavoidable. For example, gemtuzumab ozogamicin was associated with prolonged pancytopenia and CD123 targeting TCE, flotetuzumab, induced treatment-emergent grade 3 thrombocytopenia in the phase 1/expansion study.7 12 Acute pulmonary edema (XmAb14045 (CD123×CD3)), leukopenia (AMG330 (CD33×CD3)), and death have also been observed with several TCE.10 12 Expression of CD123 was also reported in non-hematopoietic tissues including several vital organs such as the lung, gut, liver, and kidney.13 In terms of expression, CLL-1 is in the better position in that CLL-1 is overexpressed in leukemia stem cells and AML blasts but not in normal hematopoietic stem cells (HSCs) and megakaryocyte-erythroid lineage cells.14 15

The pharmacological activities of CD3 bispecific antibodies (bsAbs) are influenced by multiple factors, such as the binding epitope and affinity of the individual targeting arms.16 17 Targeting membrane-proximal epitopes led to stronger tumor cell killing potentially due to the formation of tight cytolytic synapses between tumor and T cells.18 A review of first-generation CD3 bsAbs showed a high affinity for CD3 arms with an equillibrium dissociation constant (KD) of approximately 1 nM. CD3 bsAbs with high affinity for CD3 showed potent tumor cell killing but elicited severe cytokine storms.16 19 CD3 TCE with high affinity also displays poor biodistribution, and it has been demonstrated that lowering the CD3 affinity improves the pharmacokinetics and biodistribution of bsAbs.16 To widen the therapeutic window, the next generation of CD3 bsAbs has applied various strategies, including lowering CD3 binding affinity,20–22 leveraging steric hindrance on the CD3 binding arm to allow conditional CD3 exposure,23 and screening anti-CD3 antibodies eliciting low cytokine levels.24 The thresholds for T-cell activation and cytolytic activity were reported to be different, and it is possible to find an anti-CD3 antibody with potent cytolytic activity but without cytokine release syndrome (CRS).22

To develop more potent and safe treatment modalities for patients with AML, we compared two CLL-1×CD3 bsAbs with different formats that is, 2+1 and 1+1 with respect to CD3 binding affinity, efficacy, and basal cytokine release. Positioning the TAA arm on top of the CD3 arm was assumed to present steric hindrance for CD3 binding, which has never been proved empirically. This study is the first to prove the steric hindrance of the CD3 binding arm by the adjacent arm in 2+1 format and provides a comprehensive preclinical characterization of ABL602 2+1, serving as a foundation for an upcoming clinical trial in patients with relapsed or refractory AML.


Cell lines and cell culture

AML cell lines OCI-AML2 and PL-21 were purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen (Braunschweig, Germany). Jurkat, U937, HL60, HL60-luc, and Raji cells were purchased from ATCC (Manassas, Virginia, USA). All cell lines were maintained at 37°C in 5% CO2.

Induced T-cell binding

To determine if ABL602 binds to T cells in a CLL-1 expression-dependent manner, an increasing amount of recombinant His-tagged CLL-1 protein (0–5 µg) (Sino Biological) was spiked onto anti-His6 antibody-coated dynabeads (Invitrogen, 10103D) and incubated with 50 nM of ABL602 1+1 or 2+1 in 96-well plate for 1 hour at room temperature. Subsequently, T cells (5×105) isolated from human peripheral blood mononuclear cells (PBMCs) were added to the plates and incubated for 40 min at room temperature. After washing the cells, T cells were stained with PE-conjugated anti-human IgG (Jackson) for 30 min at 4°C, then assessed using flow cytometry. All data were analyzed using the FlowJo software (V.10.6.1).

Binding to Jurkat T cells and AML cells by flow cytometry

The binding of ABL602 to AML and Jurkat cells was evaluated by flow cytometry. Cells (1×105 cells/well) were added to a 96-well plate. Fourfold or fivefold serially diluted ABL602 was added and incubated for 1 hour at 4°C. The cells were washed with flow cytometry buffer (1% bovine serum albumin in phosphate-buffered saline), stained with a PE-labeled goat anti-human IgG secondary antibody (Jackson) for 1 hour at 4°C, and analyzed with flow cytometry.

In vitro and ex vivo cytotoxicity and T cell activation

Human T cells were isolated from the PBMC of healthy donors using MACS beads (Miltenyi Biotec). AML cells (U937, HL60, OCI-AML2, or PL-21) (2×104 cells) were seeded in a 96-well plate with purified human T cells (1×105 cells/well) and incubated with 150 ng of Fc block (BD Biosciences). The bsAbs with 10-fold serial dilutions from 50 nM were added to the wells and incubated for 48 hours in a 5% CO2 incubator at 37°C. The cells were stained with anti-CD45, anti-CD8, anti-CD25, anti-CD69 (BD Biosciences), and anti-CD4 antibodies (BioLegend) to determine T-cell activation and with anti-CD33 antibody (BD Horizon) to detect AML cell lines and blasts. A fixable viability kit (BioLegend) was used to assess the viability of the AML cells. The cells were washed twice and analyzed using LSRfortessa (BD Biosciences). The percentage of cytotoxicity was calculated using the following formula: [1- (live cell numbers of total CD33+ AML cells in the treated group/ live cell numbers of total CD33+ AML cells in the media control group)]×100.

For the ex vivo cytotoxicity assay, 2×105 PBMCs from primary AML whole blood were dispensed into 96-well U bottom plates and treated with serially diluted bsAbs from 50 to 0.00005 nM. After 72 hours of incubation, cytotoxicity toward CLL1+ AML blasts was determined by gating the CLL1+CD45dimSSClow population. Activated T-cells were defined as CD25 positivity. This study was conducted under institutional review board (IRB)-approved protocol (IRB number H-2105-138-1220). All data were analyzed using FlowJo V.10.6.1 and the GraphPad Prism software.

Cytokine analysis

Human PBMC (2×105) from eight healthy donors were cultured with ABL602 2+1 or MCLA-117 analogs at the maximum killing dose (approximately EC90) for each antibody in 96-well U bottom plates. After 48 hours, the supernatant was harvested to measure the human cytokine levels. The tumor necrosis factor (TNF)-α, interleukin (IL)-2, interferon (IFN)-γ, and IL-6 concentrations in cell culture supernatants were determined by ELISA (R&D systems) kit per the manufacturer’s protocol.

In vivo mouse studies to evaluate the antitumor activity

NOD/Shi-scid, IL-2Rγ KOJic (CIEA NOG, 6–7 weeks old, female) mice were inoculated subcutaneously with 0.5×106 OCI-AML2 cells on day 0. On day 5, the mice were inoculated intraperitoneally with ex vivo activated/expanded 1×107 pan-T cells. When the mean tumor volume reached approximately 150 mm3, all mice were randomly divided into four groups (eight mice per group). Vehicle or ABL602 was administered intraperitoneally twice weekly. The tumor volume and body weight were monitored for 29 days.

To generate the HL60-luc orthotopic mouse model, NOD.Cg-PrkdcscidIl2rgtm1Sug/JicTac (CIEA NOG, 8 weeks old, female) mice were pretreated with 150 mg/kg cyclophosphamide (Sigma) and intravenously inoculated with 1×107 HL60-luc cells (marked as Day 0). Five days later 1.5×107 human T cells were transplanted into the peritoneal cavity of mice (denoted as Day 5). Mice were randomized into groups (7 mice per group) and ABL602 was administered twice a week at 0.5, 0.05, and 0.005 mg/kg, starting on day 7. Bioluminescence imaging was performed four times (days 7, 14, 21, and 28) using an In Vivo Imaging System (IVIS) Spectrum in vivo imaging system (PerkinElmer).

In all studies, the human IgG Isotype control (Bio X Cell) was injected intraperitoneally at 30 mg/kg 24 hours before ABL602 dosing to block Fc receptor binding.

Immunohistochemical staining of tumor tissues

Tumor tissues were collected on day 13 in the OCI-AML2 model and whole femur tissues were collected on day 21 in the HL60-luc orthotopic model. Whole femur tissues were subjected to decalcification. All tissues were fixed with formalin solution and embedded in paraffin blocks, which were cut into 4 µm thick sections. Sections were blocked with 10% normal goat serum and stained with anti-CD8 (Abcam, ab237709) or anti-CD33 (Abcam, ab199432) using VECTASTAIN Elite ABC kits (Vector Laboratories, PK-6101). Tissues were subsequently developed using a 3, 3’-diaminobenzidine (DAB) peroxidase substrate kit (Vector Laboratories, SK-4100) and counterstained with Mayer’s hematoxylin (Dako, S330930). Slides were mounted and visualized using a ZEISS Axio Observer.Z1 microscope as images.


Statistical analyses were performed using GraphPad Prism Software V.9. Data from the in vitro studies were analyzed using Dunnett’s multiple comparison test. For in vivo studies, a two-way or one-way analysis of variance with Tukey’s or Bonferroni’s multiple comparison tests was used.


ABL602 2+1 induces more potent T-cell activation and cytolytic activity than 1+1, while exhibiting lower binding activity to CD3

The CLL-1×CD3 bsAb, ABL602, comprises an anti-human CD3 monoclonal antibody (clone W3311-2.306.4-z1) and an anti-huCLL-1 (clone 33C2) monoclonal antibody, where each heavy chain combines the knobs-into-holes technology. ABL602 1+1 (1+1 antigen binding valency) had anti-CD3 Fab and anti-CLL-1 Fab in each arm, whereas ABL602 2+1 (2+1 antigen binding valency) had one more anti-CLL-1 Fab connected to the N-terminal end of the anti-CD3 heavy chain by (G4S)3 linker (figure 1A). Both candidates have a single amino acid substitution (N297A) in the Fc to eliminate FcγR binding activity. Constant domains of anti-CD3 Fab are derived from the α and β chains of the T-cell receptor (TCR) to permit proper assembly of heavy and light chains of the anti-CD3 antibody and avoid mispairing with the anti-CLL-1 antibody.

Figure 1

In contrast to ABL602 1+1, ABL602 2+1 shows induced binding to CD3 in a CLL-1 expression-dependent manner and potent tumor cytolytic activity. (A) Schematic structure of ABL602 1+1 and ABL602 2+1. In ABL602 2+1, anti-CLL-1 Fab and anti-CD3 Fab are linked by (G4S)3. The schematic was made using (B, C) Binding activities of two different formats to T cells were measured in the presence of beads conjugated with increasing amounts of CLL-1 recombinant proteins. (B) Relative CD3 binding proportion of ABL602 2+1 compared with that of ABL602 1+1 is indicated as %. (C) T-cell binding of ABL602 2+1 and ABL602 1+1 was determined by flow cytometry. (D) Cell binding activities were determined by incubating Jurkat T cells or U937 cells with serially diluted antibodies, followed by flow cytometry. (E, F) T cells isolated from PBMC of healthy donors were incubated with U937 cells with an E:T ratio of 5:1 for T-cell activation (E) and cytotoxicity (F), respectively, in the presence of serially diluted antibodies for 48 hours. E:T, effector:target; MFI, mean fluorescence intensity; PBMC, peripheral blood mononuclear cell.

We hypothesized that (1) the binding activity of ABL602 2+1 to CD3 would be lower than that of ABL602 1+1 because of steric hindrance by the extra anti-CLL-1 Fab positioned on top of the anti-CD3 Fab and (2) this steric hindrance could be overcome when anti-CLL-1 Fab binds to CLL-1. To test this hypothesis, ABL602 candidates and purified T cells from human PBMC were incubated with beads coated with increasing CLL-1 concentrations (online supplemental figure 1A). The relative binding % of ABL602 2+1 is displayed compared with ABL602 1+1 by standardizing the binding level of ABL602 1+1 to 100%. In the absence of CLL-1, the CD3 binding proportion of ABL602 2+1 was negligible compared with that of ABL602 1+1. The binding activity of ABL602 increased in a CLL-1-dependent manner, reaching a level comparable to that of ABL602 1+1 in the presence of beads loaded with high amounts of CLL-1 (figure 1B,C). This suggests that the binding of ABL602 2+1 to CD3+ T cells is minimal in the absence of CLL-1+ cells but is potentiated in the presence of CLL-1+ cells, potentially in the tumor microenvironment (TME). Although a fairly high amount of ABL602 1+1 bound to T cells in the absence of CLL-1 as we expected, its binding to CD3 also augmented as CLL-1 increased (figure 1C).

Supplemental material

Supplemental material

Surface plasmon resonance revealed that the binding activity of ABL602 2+1 to huCD3 was approximately 1.64-fold lower than that of ABL602 1+1 (KD of 138 nM compared with 84 nM). Cell binding analysis with Jurkat T cells expressing CD3 also showed a lower binding activity of ABL602 2+1 compared with ABL602 1+1, whereas binding activities to CLL-1-expressing U937 cells were comparable (figure 1D). T-cell activation and cytotoxic activity of ABL602 2+1 compared with ABL602 1+1 were investigated using the U937 cell line. T cells were incubated with U937 cells at the Effector:Target (E:T) ratio of 5:1. The T-cell activation and cytotoxicity of ABL602 1+1 and ABL602 2+1 showed a potent efficacy compared with the control bsAb. T cells were greatly activated by ABL602 2+1, approximately 45-fold higher than that by ABL602 1+1, as indicated by the increased proportion of CD69+CD8+ T cells (figure 1E). Approximately 18.5-fold stronger cytolytic activity was observed with ABL602 2+1 than with ABL602 1+1 (figure 1F). Claudin 18.2x CD3 2+1 was used as a control bsAb (mock/CD3 2+1), in which the same CD3 antibody to ABL602 was used. Claudin18.2 was not expressed in AML cell lines as well as PBMC and did not induce either T-cell activation or cytolytic activity.

ABL602 2+1 shows strong CLL-1 binding activity and potent cytotoxicity against CLL-1-expressing AML cells

Based on in vitro binding and efficacy data shown in figure 1, ABL602 2+1 was selected as the final format. ABL602 2+1 bound to various AML cell lines including OCI-AML2, HL60, and PL21 with half maximal effective concentration (EC50) values of 0.15 nM (figure 2A). In accordance with the previously reported expression profile for CLL-1,25 ABL602 2+1 was also bound to granulocytes, monocytes, myeloid dendritic cells, and neutrophils known to express CLL-1, but not to natural killer cells or B cells (online supplemental figure 2). Flow cytometry analysis of normal PBMC also showed similar binding of ABL602 to both CD3+CD4+ and CD3+CD8+ T cells (online supplemental figure 2).

Supplemental material

Figure 2

ABL602 2+1 binds to CLL-1 expressing acute myeloid leukemia (AML) cell lines and induces T-cell activation and cytotoxicity in a CLL-1-dependent manner. (A) Acute myeloid leukemia cells (HL60, OCI-AML2, PL-21) were incubated with ABL602 2+1 and determined for CLL-1 binding activity using flow cytometry. (B, C) Purified human T cells were incubated with various AML cell lines at an E:T ratio of 5:1 in the presence of indicated concentrations of ABL602 2+1 for 48 hours. CD69 expression was used as a readout for CD8+T-cell activation (B) and target cell lysis was quantified by flow cytometry relative to the PBS control (C). EC50, half maximal effective concentraion; E:T, effector:target; MFI, mean fluorescence intensity; PBS, phosphate-buffered saline.

ABL602 2+1 induced T-cell activation, as indicated by the upregulation of CD69, an early activation marker, and CD25, an intermediate activation marker (data not shown), in the presence of CLL-1 positive cell lines, but not in response to CLL-1 negative cells such as Raji cells (figure 2B). Comparable T cell-activating efficacy was also observed in CD4+ T cells (online supplemental figure 3A). Granzyme B and perforin expression were also observed in CD8+ T cells in the presence of ABL602 2+1 (online supplemental figure 3B).

Supplemental material

The tumor-killing activity of ABL602 2+1 was investigated in AML cell lines U937, HL60, OCI-AML2, and PL-21. Purified human T cells were co-cultured with cancer cells at an E:T ratio of 5:1 in the presence or absence of ABL602 2+1 followed by flow cytometry. In all cases, potent target cell lysis was detected with EC50 ranging from 0.75 to 6.45 pM in over 85% of cells in all four cell lines with all donors (figure 2C). No cytotoxicity was observed in the CLL-1 negative cell line or with control bsAbs (figure 2C).

ABL602 2+1 exhibits potent in vivo antitumor activity both in subcutaneously and orthotopically engrafted AML model

We evaluated in vivo efficacy of ABL602 2+1 in two established xenograft models of NOG mice engrafted with human T cells. In the established subcutaneous OCI-AML2 tumor model, treatment with ABL602 2+1 at 0.1, 0.3, and 1 mg/kg elicited tumor growth inhibition of 45%, 77%, and 78%, respectively, compared with PBS-treated control animals on day 16 (p<0.001 for 0.3 mg/kg and 1 mg/kg) (figure 3A). ABL602 2+1 at 1 mg/kg resulted in the complete regression of tumors in seven out of eight mice on day 29. IHC analysis of the harvested tumors showed CD8+ T-cell infiltration in the tumors of ABL602 2+1-treated mice in a dose-dependent manner, whereas PBS-treated mice had almost no detectable CD8+ T cells (figure 3B,C).

Figure 3

ABL602 exhibits potent tumor inhibitory activity in a T cell-humanized subcutaneously tumor-engrafted mouse model. (A) T cell-humanized NOG mice were inoculated with OCI-AML2 subcutaneously (0.5×106 cells). After 7 days, mice were intraperitoneally administered with ABL602 2+1 at 0.1, 0.3, and 1 mg/kg. Tumor volume was measured twice weekly, and the results are presented as the mean tumor volume with SE of the mean for each group. Statistical significance is denoted by asterisks as * and *** for p<0.05 and p<0.001, respectively. (B) On day 13 after two doses of ABL602, tumors were harvested and stained with anti-human CD8 antibody to evaluate T-cell infiltration in the tumor. (C) For the quantification of CD8 T cells, three high-powered fields (HPF, ×200) were randomly selected from each mouse tumor tissue, three mice per group. Using the ImageJ Software, the DAP-stained CD8 positive areas were converted as a percentage of each total area. The CD8 positive area (%) was evaluated as the average of three regions. Statistical significance was denoted as **p<0.01 and ***p<0.001 compared with PBS, respectively. DAP, 2,2’-Diaminobenzidine; PBS, phosphate-buffered saline.

In the established disseminated HL-60-luc model, ABL602 2+1 was administered after confirming the homing of AML cells to the bone marrow using bioluminescence. ABL602 2+1 significantly inhibited tumor growth in a dose-dependent manner as assessed by bioluminescence (31%, 60%, and 86%, respectively) compared with PBS-treated mice (figure 4A). The tumor burden in the bone marrow, spine, and hind limbs was significantly reduced, as observed using bioluminescence (figure 4B). Tumor burden was also determined by the proportion of CD33+ AML blasts in live bone marrow cells using flow cytometry. On day 29, AML cells were barely detectable in mice treated with 0.5 mg/kg of ABL602 2+1 (figure 4C, p<0.01). Bone marrow was fixed in formalin-fixed paraffin-embedded tissue and prepared for IHC staining. The vehicle control group showed highly infiltrated CD33+ AML cells with barely infiltrated CD8+ T cells. In contrast, there was a dose-dependent decrease in AML cells and an increased infiltration of T cells in the ABL602 2+1 treated groups on day 22 (figure 4D,E). These data show that ABL602 2+1 inhibited tumor growth in the two AML tumor models by recruiting T cells to the tumor site and inducing T-cell activation and killing.

Figure 4

ABL602 inhibits the progression of systemically engrafted HL60-luc tumor cells in T cell-humanized mice. T cell-humanized NOG mice were inoculated with HL60-luc (1×107 cells) intravenously. Seven days later, mice were intraperitoneally administered with ABL602 2+1 at 0.005, 0.05, and 0.5 mg/kg twice a week. (A, B) Bioluminescence was measured once weekly, and photon intensity (A) and representative images of live animal imaging of bioluminescence (B) on days 7, 14, 21, and 28 are shown. (C) AML blast in the bone marrow was measured by staining with an anti-CD33 antibody, followed by flow cytometry. The proportion of CD33+ AML blasts in bone marrow live cells is indicated as %. (D) Bone marrows were harvested, fixed in FFPE, and stained for AML blasts and CD8 T cells using anti-CD33 antibody and anti-CD8 antibody, respectively. (E) For the quantification of CD8 T cells, seven regions were randomly chosen from one bone marrow slide of each group. The number of CD8 T cells was counted and presented as the average of seven regions. Statistical significance was shown as *p<0.05 and ***p<0.001 compared with PBS, respectively. AML, acute myeloid leukemia; FFPE, formalin-fixed paraffin-embedded; HPF, high powered fields; PBS, phosphate-buffered saline.

ABL602 2+1 induces T-cell activation, resulting in lysis of AML blasts in AML patient-derived PBMC

To determine the CLL-1 expression level in patients with AML, PBMC was isolated from 11 patients with AML and analyzed using flow cytometry (table 1). CLL-1 expression in AML blasts was comparable to that of CD33 in terms of both the percentage of positive cells and MFI. CD45dim SSClow immune cells were defined as AML blasts (online supplemental figure 4).26 AML blasts were observed in the blood of most patients, except for two (patient numbers 6 and 13), which were not included in the cytotoxicity assay. Patient 9 had a large proportion of AML blasts, however, did not have CLL-1+ cells (0.26%) and was excluded from the cytotoxicity study. Most patients expressed CLL-1 and CD33 with a few exceptions. CLL-1, as well as CD33 expression level, was quite variable with relative MFI of 0.48–64.83 and 1.37–72.70, respectively (figure 5A and table 1).

Supplemental material

Table 1

CLL-1 and CD33 expression and cytolytic activity of ABL602 2+1 on AML blasts

Figure 5

Most of the AML blasts express CLL-1 and are lysed by ABL602 2+1 in primary AML samples. (A) Flow cytometry analysis for CLL-1 and CD33 expression in 11 patients with AML samples. The proportion of CLL-1 or CD33 expressing cells within AML blasts as denoted by CD45dimSSClow is indicated as a percentage. MFI of CLL-1 or CD33 stained cells are indicated as relative MFI compared with IgG control. (B) PBMC from patients with AML were incubated with increasing ABL602 2+1 doses for 72 hours followed by assessment of target cell lysis and CD8+ T-cell activation by flow cytometry. Representative data are presented for CLL-1 expression (left panel), lysis of CLL-1+ AML blast (middle panel), and T-cell activation (right panel). AML, acute myeloid leukemia; MFI, mean fluorescence intensity; PBMC, peripheral blood mononuclear cell.

The ability of ABL602 2+1 to induce cytotoxicity in a clinically relevant context was assessed by an ex vivo cytotoxicity assay using PBMC from patients with AML. This system relies on the presence of autologous T cells in the blood of patients to kill AML cells. The E:T ratio of the PBMC used for the cytotoxicity test was relatively low in most patients (1:4.35–1:62). ABL602 2+1 exhibited strong cytotoxicity on CLL-1+ AML blasts (EC50, 70–300 pM) and T-cell activation as indicated by the upregulation of CD25 as an intermediate activation marker (EC50, 0.3–40 pM) (figure 5B) in four patient samples. These data indicate that ABL602 2+1 was effective in killing CLL-1+ AML cells, even at a very low E:T ratio in a more physiological ex vivo setting.

ABL602 2+1 shows about 1,000-fold stronger cytotoxicity compared with MCLA-117 analog while inducing limited TNF-α or IL-6 expression

CD3 is consistently expressed in all T cells, including resting T cells, making it a difficult target for the development of agonistic antibodies. Consequently, CD3 targeting TCE have been associated with poor pharmacokinetics and CRS.20 21 To restrict T-cell activation in the TME, a TCE with a much lower binding affinity for CD3 has been developed and has shown some promising outcomes.20 27–29 By using the 2+1 format, we diminished the CD3-binding affinity to 138 nM from 84 nM of the parent CD3 antibody and proved that CLL-1 binding led to induced and potentiated binding to CD3 T cells (figure 1B, online supplemental table 1 and figure 1).

MCLA-117 is a bsAb with a 1+1 format, and its affinity for CD3 is 177 nM.25 Compared with the MCLA-117 analog, ABL602 2+1 showed potent cytolytic activity and T-cell activation by approximately 1,000-fold against HL-60 and U937 cells at an E:T ratio of 5:1 (figure 6A and B). Around EC90 of ABL602 2+1 and MCLA-117 analog in the cytolytic assay (0.005 and 5 nM, respectively) was incubated with healthy donor-derived PBMC. As indicated in online supplemental figure 2, normal PBMC contains myeloid populations expressing CLL-1; therefore, T cells can be activated even in the absence of AML cancer cells. While MCLA-117-induced pro-inflammatory cytokines such as TNF-α and IL-6, ABL602 2+1 did not, suggesting ABL602 2+1 would have limited CRS in humans potentially (figure 6C). Comparable results were obtained when normal PBMC was incubated with an even higher dose (50 nM) (data not shown).

Figure 6

ABL602 2+1 shows about 1,000-fold potent cytolytic activity and T-cell stimulating activity against AML cancer cell lines compared with the MCLA-117 analog. (A, B) Serially diluted ABL602 2+1 (red), mock/CD3 2+1 (gray), or MCLA-117 analog (blue), and human purified T cells were co-cultured with AML cell lines at an E:T ratio of 5:1. After 48 hours, the proportion of activated T cells (CD69+CD8+) (A) and lysed AML cells (B) were determined using flow cytometry. (C) To determine potential cytokine release in the absence of AML cells at the efficacious dose, healthy donor-derived PBMCs were cultured with corresponding EC90 doses of ABL602 2+1 or MCLA-117 analog for 48 hours. EC90 values from cytotoxicity assay (B) were used for cytokine release assay. The supernatant was analyzed for the indicated cytokines by ELISA. AML, acute myeloid leukemia; EC90, 90% maximal effective concentration; E:T, effector:target; IFN, interferon; IL, interleukin; PBMC, peripheral blood mononuclear cell; TNF, tumor necrosis factor.

To compare the cytokine release in more controlled conditions, purified T cells were incubated with ABL602 2+1 and MCLA-117 analog in the presence or absence of U937 (online supplemental figure 5). Both ABL602 and MCLA-117 analog induced cytokines in the presence of U937 but not in the absence of U937. This suggests that augmented pro-inflammatory cytokines elicited by MCLA117 in PBMC are from non-T cells, potentially monocytes, stimulated by MCLA-117 either directly or indirectly.

Supplemental material


Activation of T cells with monoclonal antibodies targeting TCRs or co-stimulatory receptors, such as CD137, has been associated with many difficulties, such as poor pharmacokinetics, systemic cytokine release, liver toxicity, and even death. One major obstacle in targeting CD3 is its consistent expression on T cells, which results in TAA-independent T-cell activation. To avoid on-target/off-tumor toxicities, we used a CD3 arm with low binding activity and designed the antibody in a 2+1 format. We showed that ABL602 2+1 bound weakly to CD3 in the absence of CLL-1, but its binding was significantly potentiated in the presence of CLL-1 as shown in a binding assay using CLL-1-spiked beads. In agreement with binding analysis using beads, ABL602 2+1 alone is weakly bound to Jurkat T cells; however, its T cell-activating potency dramatically increased in the presence of CLL-1-expressing AML cells. It is even more encouraging that ABL602 2+1 demonstrated potent cytolytic activity under conditions where T-cell proportions in AML blasts were very low, which is frequently observed in patients with AML. Furthermore, ABL602 2+1 strongly inhibited tumor progression in two AML models, including subcutaneously and orthotopically engrafted AML tumor cells. These results warrant further development of ABL602 2+1.

Compared with ABL602 1+1, ABL602 2+1 showed negligible binding to T cells when incubated with empty beads. The binding activity of ABL602 2+1 dramatically increased as the amount of loaded CLL-1 increased, reaching a level like that of ABL602 1+1. Because the anti-CD3 arm of ABL602 1+1 was consistently exposed, we expected it to bind to T cells independent of the amount of CLL-1 coated on the beads. However, the binding of ABL602 1+1 to T cells also tended to increase as the amount of loaded CLL-1 increased. ABL602 1+1 bound to CLL-1-loaded beads likely increased its avidity toward CD3, leading to augmented binding to CD3.

The head-to-tail 2+1 format has been reported to have many advantages over the 1+1 format. By having two arms for the tumor, target avidity-driven CD3 binding can be achieved, and this is applied for TAA overexpressed in tumors but also expressed in normal tissues.30–32 Two TAA binding arms, as well as their position near the T cell-binding arm, create the appropriate distance for solidifying the immunologic synapse between the tumor and T cells, leading to maximal T-cell activation and cytolytic activity.30 33 34 Steric hindrance on CD3 binding in the absence of TAA was also suggested as a way of limiting non-specific T-cell activation but has not been proved before.35 36 This is the first to prove that TAA binding leads to enhanced binding of antibodies to T cells.

CLL-1 is expressed in most myeloid lineage cells, and myeloid cell depletion-related adverse effects are unavoidable. Nevertheless, these adverse effects are expected to be transient, and patients might recover after cessation of treatment, considering their limited expression in myeloid lineage cells, excluding HSCs and multipotent progenitors. Compared with CLL-1, CD33, and CD123 are widely expressed in myeloid cells and HSC, which might be one of the reasons why most CD3 bsAbs targeting these TAAs failed to move forward to Phase 2 clinical trials.11 37

MCLA-117 is a CLL-1×CD3 bispecific antibody with a binding affinity of 3 nM and 177 nM for CLL-1 and CD3, respectively. ABL602 2+1 has similar binding affinities to those of MCLA-117 with binding affinities of 3.75 nM and 138 nM for CLL-1 and CD3, respectively. However, ABL602 2+1 exhibited a 1000-fold superior activity in T-cell activation and target cell lysis compared with MCLA-117 and ABL602 1+1. The two arms binding to CLL-1 might stabilize antibody binding to target cells. The formation of immunological synapses and increased interaction of tumor/T cells by 2+1 TCE have been reported previously.31 34 38 Moreover, 2+1 might make tumor cells more vulnerable to T cell-mediated killing by keeping those two cells in proximity. Bacac et al demonstrated that 2+1 has a significantly superior tumor-killing activity to 1+1 in vitro. It has been suggested that the asymmetric head-to-tail 2+1 format may promote interactions between T cells and tumor cells.30 33

Although ABL602 2+1 showed superior T-cell activation and target cell lysis, it barely elicited pro-inflammatory cytokines such as TNF-α and IL-6 in the absence of AML cells under EC90 dose (0.005 nM) of cytotoxicity assay (figure 6C) and even up to 50 nM (data not shown). In contrast, MCLA-117 induced expression of TNF-α and IL-6 at an EC90 of 5 nM. Because CLL-1 is normally expressed in resting myeloid cells, T cells can be activated to a certain degree even in the absence of AML cells and ABL602 2+1 activated T cells, as indicated by CD69+CD3+ T cells and IFN-γ secretion. It is unclear why ABL602 2+1 did not induce the expression of TNF-α and IL-6 but did induce IFN-γ.

Functional decoupling between T-cell activation, which leads to cytokine production, and cytotoxicity has been previously reported.18 24 39 40 Faroudi et al proved that T-cell activation had higher thresholds than that of eliciting tumor-killing activity by enumerating TCR- pMHC (peptide-Major Compatibility Complex) complexes required for each activity.39 This concept was further solidified by T cell agonistic bsAbs that drove potent tumor killing but showed very low levels of cytokine release.24 TNF-α is a central mediator of monocyte activation and systemic cytokine release; however, it is dispensable for T-cell cytotoxicity.18 It is plausible that lack of cytokine release, especially TNF- α and IL-6, may place ABL602 2+1 in a good position in terms of safety.

Data availability statement

Data are available on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

All included patients provided written informed consent and this study was approved by the Institutional Review Board of Seoul National University Hospital. Mouse studies have been approved by the IACUC of ABL Bio and the approval number is IACUC-21-10.


Supplementary materials


  • Contributors Conception and design: EL, SL, SP, YL and JW. Antibody production: JY. In vitro study: EL, SL, SP and Y-GS. Animal experiment: EL, SL and SP. Analysis and interpretation of data: EL, SL, SP, Y-GS, YL and JW. Provided patient samples and served as scientific advisors in terms of patient data interpretation: YK and D-YS. Writing and review of the manuscript: EL, SL, SP, Y-GS, JY, YK, D-YS, YL and JW, Responsible for the overall content as the guaranter: YL and JW.

  • Funding This work was supported by a grant from ABL Bio Inc., Republic of Korea.

  • Competing interests EL, SL, SP, Y-GS, JY, YL and JW are employees of ABL Bio, Inc. The other authors declare no potential conflicts of interest.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.