A Chemo-enzymatically Linked Bispecific Antibody Retargets T Cells to a Sialylated Epitope on CD43 in Acute Myeloid Leukemia

Acute myeloid leukemia (AML) is a high-risk hematologic malignancy, with high mortality rates especially in elderly patients (1, 2). With a median age of 68 years at diagnosis, AML is predominantly a disease of the elderly population (3). For most AML subtypes, current standard of care has not improved over the past three decades and still relies on chemotherapy and allogeneic hematopoietic stem cell transplantation (HSCT; ref. 4). The severe side effects and comorbidities of these treatments limit their applicability especially in older patients (>60 years) and/or patients with poor physical fitness (5, 6). This poses an unmet medical need and highlights the need for better tolerated, broadly applicable AML treatments. To develop new therapeutic options for patients with AML, the identification of novel AML-specific targets is essential.

Earlier, we have examined whether patients with AML successfully treated with allogeneic HSCT generate AML-specific antibodies (7–9). Allogeneic HSCT can evoke an immunotherapeutic graft-versus-leukemia reaction, which includes a B-cell response and the generation of tumor-specific antibodies (10). From a number of allogeneic HSCT–treated patients with AML in durable remission we immortalized B cells, to generate clonal lines of donor-derived B cells, which produce antibodies reacting with autologous and allogeneic AML tumor samples (11, 12).

One of these antibodies is AT1413, which recognizes a unique sialylated epitope on CD43 (CD43s) and binds to all AML and myelodysplastic syndrome (MDS) blast samples that we tested (n = 80) including the patient's own blasts (8). These samples represent all World Health Organization 2008 types of AML and MDS. AT1413 also binds to healthy granulocytes and monocytes, but to a lower extend compared with AML or MDS blasts from the same donor, and weakly binds endothelial cells. No binding to other normal cells was detected in an IHC screen containing samples of multiple tissues (8).

AT1413 induces antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC) in vitro and inhibited growth of an AML cell line in an in vivo mouse model of AML, suggesting that AT1413 may have impacted the favorable clinical outcome of the patient it was isolated from.

Because of the preferential reactivity of AT1413 with AML cells, we reasoned that a bispecific T-cell–engaging antibody (bTCE) based on AT1413 could have increased cytotoxic potency and therapeutic potential. Many bTCEs are in clinical development, and two bTCEs have reached market approval (13). One of these, blinatumomab, is used to treat patients with relapsed or refractory B-cell precursor acute lymphatic leukemia and is undergoing clinical testing for other types of leukemia. Hence, the bTCE concept is clinically validated for hematologic malignancies (14, 15).

It has been suggested that the target of a bTCE needs to be located sufficiently close to the cell membrane, to form a stable cytolytic synapse between the target tumor cell and the T cell (16). AT1413 binds a sialylated epitope in the extracellular domain of CD43 (8). Being highly glycosylated, this extracellular domain is described to protrude away from the plasma membrane in a rod-like manner with a length of about 45 nm (17). While other CD43-targeting antibodies bind to membrane-distal CD43 regions, the CD43s epitope was mapped within a membrane-proximal part of the extracellular domain (8). This further encouraged us to generate a bTCE against CD43s as this favorable location of the CD43s epitope will likely support the bTCE's efficacy and potency (16).

To prepare bTCEs, antibodies or antibody fragments have successfully been equipped with single-chain variable fragments (scFv) derived from established CD3ϵ-binding antibodies (18–21). To convert AT1413 into a bTCE, we chose to chemo-enzymatically link a CD3ϵ-binding scFv, derived from the antibody UCHT1 (22), to the C-termini of the AT1413 IgG heavy chains in a C-to-C orientation. C-to-C fusion of the IgG and UCHT1 scFv is achieved by combining sortase transpeptidation and click chemistry (23–25). To abolish interaction with other non-T-cell immune effector cells, we removed Fc-gamma receptor (FcγR) interaction sites by introducing two point mutations G236R and L328R into the AT1413 heavy chains (AT1413-Fc0), as described previously (26). The resulting conjugation product is a bispecific antibody, in which two CD3ϵ-binding scFv are coupled to one IgG molecule. Here, we describe the generation of an AT1413 bTCE and demonstrate that this bTCE efficiently induces T-cell–mediated lysis of CD43s-expressing AML cells in vitro and in vivo.

S. aureus sortase A (residues 82–249) with an N-terminal His₆-tag was expressed and purified as described previously (27). The E. coli strain expressing S. aureus sortase A was a gift from Hidde Ploegh (Harvard Medical School, Boston, MA).

C-terminally LPETGGH₆-tagged (tag sequence: GGGGSLPETGGHHHHHH) antibodies were cloned, expressed, and purified as described previously (24).

The UCHT1 scFv sequence was derived from the antibody UCHT1 (22, 28), by fusing the UCHT1 heavy and light chain variable regions, with the heavy chain variable domain oriented N-terminal to the light chain variable domain and a (G₄S)₃-linker sequence in between both domains, and adding a C-terminal LPETGGH₆-tag. The construct was cloned into a pXC19 vector (Lonza), using synthetic codon-optimized open reading frames (GeneArt) encoding the human VK3 leader sequence. LPETGGH₆-tagged UCHT1 scFv was stably expressed in CHO-GS cells and purified by immobilized metal affinity chromatography (IMAC) and size exclusion chromatography (SEC). IMAC was conducted on HisTrap excel and SEC on HiLoad Superdex 200 16/600 (GE Healthcare) columns using an ÄKTA Explorer 10S System (GE Healthcare).

AT1413 bTCE and AT1002 bTCE molecules were generated and purified as described previously (25).

bTCE stability was assessed by incubation of 0.1 μmol/L bTCE in PBS or IMDM + 8% FBS at 37°C for up to 21 days followed by SDS-PAGE and flow cytometric analysis.

Study protocols were approved by the institutional Medical Ethical Committee of Amsterdam UMC, location AMC (Amsterdam, the Netherlands). Human peripheral blood mononuclear cells (PBMC) were obtained from buffy coats of healthy blood donors (Sanquin) and AML blasts from peripheral blood (PB) or bone marrow (BM) of patients with AML by ficoll density gradient centrifugation, after obtaining written informed consent, and cryopreserved or used directly.

For cytotoxicity assays, PBMCs and AML blasts were thawed and rested in IMDM Medium (Gibco) supplemented with 8% FBS (Gibco) and 100 IU/mL penicillin and 100 μg/mL streptomycin (PS; Roche) overnight.

CD8-positive T cells for T-cell cultures were isolated from PBMCs by FACS on a FACSAria System (BD Biosciences). T cells were expanded in the presence of 1 eq irradiated (25 Gy) JY cells and 10 eq irradiated (100 Gy) PBMCs with 0.5 μmol/L phytohaemagglutinin (BioTrading) in Yssel medium (29) containing 1% human serum (ATCC), 0.1 nmol/L recombinant human IL2 (Sigma), and PS. Medium was renewed every 3–4 days and fresh IL2 was added.

Cell lines were obtained from DSMZ (SH-2, Molm13, BV173, and Kasumi3) or ATCC (THP-1, HL-60, Jurkat, and U266) and endothelial cells from Lonza (HAEC and HUVEC) and were maintained as recommended by the supplier. Jurkat TCRα^−/− cells were a gift from Ton Schumacher (Netherlands Cancer Institute, Amsterdam, the Netherlands). Cell lines and endothelial cells were tested monthly for Mycoplasma by PCR. Cell lines were authenticated using short tandem repeat analysis.

Flow cytometry was conducted on a FACS Canto II (BD Biosciences) or a FACS LSR Fortessa X-20 (BD Biosciences). Data were processed with FlowJo 10.1 (FlowJo, LLC) and Prism 8.0.1 (GraphPad Software, Inc.).

FACS binding assays were carried out in 384-well plates in 1% BSA (Roche) in PBS (Gibco) at 4°C. Cells were harvested, washed, and incubated with antibody, scFv, or bTCE at 12,000 cells per well for 30 minutes. After washing, surface bound antibody and bTCE were detected with AF647 goat anti-human IgG H+L (Thermo Fisher Scientific) and UCHT1 scFv with AF647 anti-PentaHis (Qiagen) by incubating for 30 minutes in the dark. Cells were washed and resuspended in 10 nmol/L 4′,6-diamidino-2-fenylindole (DAPI; Sigma) directly before cytometric analysis.

Cytotoxicity was assessed by a modified Calcein AM retention assay adjusted for T-cell engagement (30). PBMCs or T-cell–depleted PBMCs were added at an effector-to-target (E:T) ratio of 10:1 (unless otherwise indicated) to target cells before incubation at 37°C and 5 % CO₂ for 6–44 hours.

Whole-blood assays were performed likewise, but by adding PB from healthy volunteers instead of PBMCs to target cells before incubating for 5 hours at 37°C and 5 % CO₂. Leukocytes were stained with anti-human CD66b PE and CD14 APC-Cy7 (BioLegend) and live cells with an eFluor660 Viability Marker (eBioscience). Red blood cells were lysed by adding BD FACS Lysing Solution (BD Biosciences), incubating for 15 minutes at 22°C, centrifuging, discarding supernatant, and repeating this step once. Accudrop beads were added and samples analyzed as above.

AML target cells were allowed to opsonize AT1413 bTCE or control as during cytotoxicity assays and were incubated with PBMCs (E:T ratio 10:1 or 2:1) or T cells (E:T ratio of 1:1), for 24–48 hours at 37°C and 5% CO₂. T cells were stained with anti-human CD4 PE-Cy7, CD8a APC-Cy7, CD25 APC, CD45RO FITC (all BioLegend), CD69 PE/APC (BioLegend/BD Biosciences), CD107a APC (BD Biosciences), and/or CD197 (CCR7) PE (BD Biosciences) for 30 minutes in the dark or during incubation (CD107a). Cells were washed, resuspended in 10 nmol/L DAPI, and analyzed by flow cytometry. T cells were selected as DAPI-negative, gated for CD4- or CD8-positive, where applicable for CD45RO- and CD197-positive or -negative, and for CD69-, CD25-, or CD107a-positive cells. Statistical testing was performed with Prism 8.0.1.

AML target cells were incubated with AT1413 bTCE or control and T cells for 48 hours as for T-cell activation assays. Cell-free supernatants were collected and diluted 4–10× with ELISA/ELISPOT Diluent (Thermo Fisher Scientific). ELISAs were performed with a Human IFN gamma ELISA Ready-SET-Go! Kit (Thermo Fisher Scientific) in a 384-well format. Absorbance at 450 nm was monitored with an EnVision 2104 Multilabel Reader (PerkinElmer).

Cultured T cells were washed with PBS, stained with 5 μmol/L Celltrace CFSE in PBS at 37°C and 5% CO₂ for 20 minutes and washed with IMDM medium + 8% FBS. In 96-well U-bottom plates, 25,000 target cells per well were preincubated with AT1413 bTCE or control in IMDM medium + 8% FBS for 15 minutes at 22°C. T cells were added (E:T ratio 1:1) and samples were incubated for 5 days at 37°C and 5% CO₂; then stained with APC-Cy7 anti-human CD8a as described for T-cell–activation assays, washed, resuspended in 10 nmol/L DAPI, and analyzed by flow cytometry.

All animal protocols were carried out in accordance with Dutch and European laws. Experiments were approved by the AMC Animal Experiments Commission (Dierexperimentencommissie) and conducted according to AMC's institutional guidelines.

Human immune system (HIS) mice were generated using NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG; The Jackson laboratory) mice. Briefly, NSG newborns before 5 days of age were irradiated with 1 Gy using a Cs¹³⁷ source and injected within 24 hours with 5 × 10⁴ Lin⁻CD34⁺CD38⁻ human stem cells obtained from fetal liver.

At 7–8 weeks, 12 female NSG mice (19.9 ± 1.5 g) were injected intravenously in the tail vein with 10 × 10⁶ luciferase/ZsGreen-labeled SH-2 cells in 100 μL of PBS (day 0). Starting after 15 days, tumor growth was assessed weekly by bioluminescence imaging (BLI) using a Photon Imager (Biospace Lab). Mice were injected intraperitoneally with VivoGlo Luciferin, In Vivo Grade (3.75 mg, Promega) 15 minutes before image acquisition. During the acquisition, the mice were anesthetized using Isoflurane and maintained on a thermally controlled pad. Mice were weighted twice per week. Mice losing more than 15 % of their maximum weight were humanely sacrificed.

At day 19, mice were randomized and allocated to groups of 6 mice each. Mice were injected intravenously with 5 × 10⁶ PBMC and directly given the first dose of 40 μg bTCE in 100 μL PBS i.p. Subsequent bTCE doses were biweekly injected intravenously from day 22 to day 33. Blood samples were collected from the submandibular vein. Mice were sacrificed on day 36 or day 40 by cervical dislocation after deep isoflurane anesthesia. The drug injection and measurements were performed by technicians blinded to the group allocation. Tumor growth inhibition was calculated as the median of tumor size of the AT1413 bTCE–treated group divided by the one from the control group at the end of the experiment.

Studies with HIS-NSG mice were conducted in the same way, but without PBMC injection and with the following variations. At 12–13 weeks of age, 7 female HIS-NSG mice (20.3 ± 1.4 g) were injected intravenously with 8 × 10⁶ luciferase/ZsGreen-labeled SH-2 cells. Weekly BLI was started on day 12. Mice were treated biweekly with 40 μg of bTCE (4 mice AT1413 bTCE group, 3 mice control group) from day 23 to 40 after tumor engraftment. Blood samples were collected on day 30, day 40, and at sacrifice (day 43). Statistical testing was performed in Prism 8.0.1.

At 4–7 weeks, 12 females NOD.Cg-Prkdc^scid Il2rg^tm1Sug Tg(SV40/HTLV-IL3,CSF2)10-7Jic/JicTac (hGM-CSF/hIL-3 NOG; Taconic; 16.2 g ± 0.9 g) were injected intravenously with AML blasts isolated from the PB of a patient with AML (BL059); T cells were depleted using a CD3 MicroBeads MACS Kit (Miltenyi Biotech) and 1.2 × 10⁶ cells/mouse were injected intravenously. The engraftment of the AML cells was followed through time by blood sampling and FACS analysis for CD45⁺CD33⁺ cells. When AML blasts could be reliably detected in the blood (≥1% of live cells), mice were injected with PBMC and treated as described above.

C-to-C fusion of AT1413-Fc0 and UCHT1 scFv into a bTCE-format was accomplished by the combination of sortase transpeptidation and click chemistry (Fig. 1A; refs. 23–25). Methyltetrazine (MeTz) and trans-Cyclooctene (TCO) were used as complementary click chemistry handles (Fig. 1B; ref. 31). To prepare AT1413 bTCE, AT1413-Fc0-MeTz was incubated with UCHT1-TCO; the reaction product was purified by SEC (Supplementary Fig. S1A).

Coupling of AT1413-Fc0-MeTz and UCHT1-TCO scFv by TCO-tetrazine cycloaddition proceeded nearly quantitatively as indicated by reducing and nonreducing SDS-PAGE (Supplementary Fig. S1B). The integrities of AT1413 bTCE, as well as its components, AT1413-Fc0 and UCHT1 scFv, were maintained for up to 21 days when incubated in PBS at 37°C (Supplementary Fig. S2).

Analogously, a control bTCE was prepared from the antibody AT1002 that recognizes an irrelevant target (influenza group 2 HA-protein; ref. 24).

To confirm that AT1413 bTCE retained dual binding capacity to CD43s- and CD3ϵ-expressing cells, binding to the SH-2 AML and Jurkat T-cell lines was compared with that of bTCE components, AT1413-Fc0 and UCHT1 scFv, by flow cytometry. Comparison of median fluorescence intensity signals for AT1413-Fc0 and AT1413 bTCE on SH-2 cells indicated only a slight signal reduction for the latter (Supplementary Fig. S3). Both, AT1413 bTCE and the control bTCE, bound Jurkat cells with similar half-maximal concentrations as unmodified UCHT1 scFv. TCRα^−/− Jurkat cells, devoid of CD3ϵ surface expression, were used to confirm binding specificity toward CD3ϵ as part of the TCR complex. Complete lack of bTCE and UCHT1 scFv binding to TCRα^−/− Jurkat cells was observed.

bTCE binding to SH-2 and Jurkat cells was maintained for up to 21 days of incubation at 37°C in PBS or IMDM medium (Supplementary Fig. S4).

The binding of AT1413 bTCE on SH-2 and Jurkat cells indicates that the bTCE retained the binding capacities of both components, AT1413-Fc0 IgG and UCHT1 scFv.

AT1413 bTCE was tested for inducing T-cell–mediated lysis of CD43s-expressing SH-2 cells in an in vitro Calcein AM retention assay (30). Using PBMCs as effector cells at an E:T ratio of 10:1, concentration-dependent target cell lysis was observed with EC₅₀ values of 30–60 pmol/L (Fig. 2A; Table 1). Neither the control bTCE, nor UCHT1 scFv or AT1413-Fc0 induced target cell death. Incubation of TCRα^−/− Jurkat cells with AT1413 bTCE and PBMC did not induce target cell lysis, assuring target cell specificity (Fig. 2A). At decreased E:T ratios of as low as 0.5:1, AT1413 bTCE–induced SH-2 target cell lysis was retained, albeit with gradually reduced efficacies (Supplementary Fig. S5).

AT1413 bTCE was also tested on other cell lines and primary AML blast cells. AT1413 binds the AML cell lines THP-1, Molm13, HL-60, and Kasumi3, as well as the B-cell precursor leukemia cell line BV173, although at reduced levels compared with SH-2 cells (8). Notably, AT1413 bTCE demonstrated in vitro cytotoxicity also toward these cell lines, with EC₅₀ values in the range of 40–250 pmol/L (Fig. 2B; Table 1). Activity of AT1413 bTCE toward primary AML blast cells was tested on 14 samples of PB or BM from patients with various AML subtypes (Fig. 2B; Supplementary Table S1). AT1413 bTCE induced lysis in all 14 tested AML blast samples; blasts were lysed with a wide range in efficacy (maximal lysis: 9%–92%). No correlations between the extend of maximal target cell lysis and AML subtype or AT1413 binding were observed in samples, where corresponding data were available (Supplementary Fig. S6A). Consistently, no significant differences in AT1413 binding were found in a set of 52 blast samples of various AML subtypes (Supplementary Fig. S6B).

To confirm that AT1413 bTCE–induced lysis is mediated by T-cell cytotoxicity, Calcein AM retention assay was performed with T-cell–depleted PBMCs as effector cells, which abolished target cell lysis (Fig. 2C).

Previously, AT1413 was shown to bind normal endothelial cells, but no triggering of ADCC nor of CDC against endothelial cells was observed (8). Because AT1413 bTCE has a higher cytotoxic potency than AT1413, we investigated whether AT1413 bTCE triggers T cells to lyse endothelial cells isolated from human aorta (HAEC) and human umbilical vein (HUVEC). Although AT1413 binds to these endothelial cells at IgG concentrations of ≥6.7 nmol/L, neither HAEC nor HUVEC were lysed by PBMCs in the presence of AT1413 bTCE (Fig. 2D).

To confirm that AT1413 bTCE activates T cells only in the presence of CD43s-expressing target cells, upregulation of the T-cell activation markers CD69 and CD25 was monitored. In the presence of SH-2 cells, CD69 expression was observed only on T cells incubated with AT1413 bTCE, but not with the control AT1002 bTCE, AT1413-Fc0, or UCHT1 scFv alone (Fig. 3A). In the absence of target cells, AT1413 bTCE did not induce CD69 expression. CD69 expression on CD4- and CD8-positive T cells from PBMCs of four different donors was compared (Supplementary Fig. S7A). While PBMC donors showed variations in the maximal percentage of CD69- CD4-positive and CD8-positive T cells, potency was similar for all donors.

CD25 expression was also detected on both CD4- and CD8-positive T cells within PBMCs, but percentages of CD25-positive T cells were lower than those of CD69-positive T cells (Fig. 3B). In addition, the EC₅₀ values for CD25 expression (140–170 nmol/L for CD4-positive T cells and 170–200 pmol/L for CD8-positive T cells) were higher than those for CD69 expression. Importantly, no activation markers were induced in PBMC in the absence of target cells, which implies that monocytes despite expression of low levels of CD43s do not activate T cells in the presence of AT1413 bTCE.

Using previously stimulated, resting CD8-positive T cells, higher CD25-expression levels were observed with AT1413 bTCE, whereas cells treated with the control bTCE remained CD25-negative. In addition to SH-2, other AML cell lines, such as THP-1 and HL-60, induced CD25 expression when incubated with AT1413 bTCE in this setting (Supplementary Fig. S7B).

Expression of CD69 and CD25 was compared for naïve, effector (E), central memory (CM), and effector memory (EM) T-cell subsets (Fig. 3C). Both activation markers were significantly upregulated on CD4 and CD8 T cells across most subsets in AT1413 bTCE compared with AT1002 bTCE–treated samples. However, activation was more pronounced in some subsets compared with others. Of CD8 T cells, primarily those of CM or EM phenotype expressed CD69 and CD25. On CD4 T cells, CD25 expression showed a similar pattern, although less pronounced, whereas CD69 expression was also high on naïve CD4 T cells.

Cell surface expression of CD107a throughout the incubation with bTCE and target cells was also monitored for different T-cell subsets (Fig. 3D). Consistent with their primary cytotoxic function, mainly CD8 T cells stained CD107a positive. CD107a-positive CD8 T cells spread across the CM, E, and EM phenotypes, whereas naïve CD8 T cells, with low levels of preformed cytotoxic granules remained widely CD107a negative as expected.

Next, we studied the ability of AT1413 bTCE to induce cytokine (IFNγ) production, and to stimulate T-cell proliferation. IFNγ production was only observed when previously stimulated T cells were used as effector cells and was dependent on AT1413 bTCE and the presence of CD43s-expressing target cells (Fig. 4A). With the control bTCE or in the absence of target cells, no IFNγ was detected. With SH-2 as target cells, IFNγ production was observed at lower AT1413 bTCE concentrations than for THP-1 and HL-60 cells, consistent with AT1413 binding (Fig. 4B). When coincubated with SH-2 cells and AT1413 bTCE over 5 days, T cells proliferated in a concentration-dependent manner whereas control bTCE–treated T cells remained unaffected (Fig. 4C).

In conclusion, our analyses of T-cell marker expression, cytokine production, and T-cell proliferation induced by AT1413 bTCE indicate that, in the presence of CD43s-positive target cells, AT1413 bTCE is capable of activating T cells and engaging them to kill CD43s-expressing AML cells.

To test the in vivo efficacy of AT1413 bTCE we used a xenograft NSG mouse model, inoculated with human luciferase/ZsGreen-labeled SH-2 cells and engrafted with human PBMC. Mice were treated biweekly with AT1413 bTCE or control (AT1002) bTCE with 2 mg/kg i.v. Treatment with AT1413 bTCE for 15 days resulted in tumor growth inhibition of 99% (P < 0.001) compared with AT1002 bTCE–treated mice, demonstrating that AT1413 bTCE is efficacious against SH-2 cells in vivo (Fig. 5A). During the course of the experiment, no obvious signs of discomfort or weight loss were observed.

Next, AT1413 bTCE was tested in a HIS-mouse xenograft model, to better mimic its effect on human immune cell subsets. Successfully engrafted female HIS-NSG mice were inoculated with luciferase/ZsGreen-labeled SH-2 cells. AT1413 bTCE or control bTCE treatment (biweekly, 2 mg/kg i.v.) was initiated 23 days after AML inoculation. In this model, after 20 days of treatment, BLI revealed a tumor growth inhibition of 89% (P < 0.01) in AT1413 bTCE–treated mice compared with the control bTCE group (Fig. 5B).

To detect whether AT1413 bTCE had an effect on the engrafted human hematopoietic cells in vivo, populations of human hematopoietic stem cells, granulocytes, and monocytes were analyzed at sacrifice (after 20 days of treatment with AT1413 bTCE). In BM, the percentages of neither human hematopoietic stem cells (CD34⁺), nor monocytes (CD14⁺), nor granulocytes (CD66b⁺) were significantly reduced by AT1413 bTCE (Fig. 5C). In addition, liver, PB, and spleen monocytes were present at comparable percentages within the human compartment compared with the control group. To further investigate a potentially cytotoxic effect of AT1413 bTCE toward human monocytes or granulocytes, whole blood from healthy donors was incubated with AT1413 bTCE in the presence or absence of SH-2 cells in an in vitro assay. While up to 29% of SH-2 cells were lysed in this setup at 1 nmol/L AT1413 bTCE, cytotoxicity toward monocytes and granulocytes remained at a background level and was hardly affected by SH-2 cell lysis (Supplementary Fig. S8).

Taken together, AT1413 bTCE revealed strong tumor growth inhibition against SH-2 AML cells in vivo. In a HIS-NSG mouse model, AT1413 bTCE did not show a depletion of the engrafted human myeloid cells.

To assess its efficacy in vivo against primary AML blast cells, AT1413 bTCE was tested in a patient-derived xenograft (PDX) mouse model. hGM-CSF/hIL-3 NOG mice were engrafted with patient-derived (BL059), T-cell–depleted AML blast samples. Greater than or equal to 1% CD45⁺CD33⁺ cells were observed on day 49 after engraftment. Treatment was initiated 4 days later and CD45⁺CD33⁺ cells were monitored on day 59.

On day 59, CD45⁺CD33⁺ cells were reduced in AT1413 bTCE–treated mice by a median of 89% compared with day 49 (Fig. 6A). In the same period, CD45⁺CD33⁺ cells further increased by a median of 57% in the AT1002 bTCE–treated mice. At sacrifice, the percentages of human CD33⁺ cells within human CD45⁺ cells were assessed in PB, BM, liver, and spleen (Fig. 6B). In all three organs and PB, CD33⁺ cells were significantly reduced in AT1413 bTCE compared with AT1002 bTCE–treated mice.

With AT1413 bTCE we have generated a T-cell–engaging antibody against CD43s as a potential new drug for the treatment of AML. Previously we had demonstrated that AT1413 induced ADCC in vitro (5). By converting AT1413 into a bTCE, we expected to increase its cytotoxic potency. Indeed, AT1413 bTCE induced SH-2 target cell death with EC₅₀ values of 30–60 pmol/L, whereas naked AT1413 antibody induced ADCC with EC₅₀ values only in the low nanomolar range (8). In addition, we demonstrated AT1413 bTCE to induce lysis of other AML cell lines, as well as patient-derived primary AML blast cells. Consistently, we observed more efficient tumor growth inhibition in vivo using a lower dose of AT1413 bTCE (2 mg/kg) compared with the naked AT1413 antibody (15 mg/kg). Further, AT1413 bTCE also inhibited growth of patient-derived primary AML blasts in an in vivo PDX model. Despite the potency increase, AT1413 bTCE did not mediate lysis of monocytes or granulocytes, which expressed some CD43s, in an in vitro cytotoxicity assay in whole blood. Consistently, human monocytes were not depleted by AT1413 bTCE in the in vivo HIS-NSG mouse model. We conclude therefore that AT1413 bTCE has a very strong tumor reducing activity without significantly impacting the normal myeloid cells.

Although the final reason for this difference remains elusive, we expect that the lower level of CD43s expression protects normal myeloid cells from T-cell–mediated lysis or at least reduces their susceptibility. To directly compare lysis of primary AML blasts and normal myeloid cells from the same donor, whole-blood Calcein AM retention assays could be conducted on fresh AML patient samples, which contain similar amounts of AML blasts compared with the normal myeloid populations. In addition to the CD43s expression level, protection of normal myeloid cells against T-cell–mediated lysis may explain the difference. For instance, naïve blood monocytes have been described to be able to suppress T-cell function (32), which is partially dependent on nitric oxide synthase (NOS). If this is applicable, AT1413 bTCE–mediated T-cell cytotoxicity against this monocyte population should be enhanced by the addition of a NOS inhibitor. Finally, we speculate that the target, CD43s, may be exposed differently on AML cells compared with normal myeloid cells hampering efficient formation of a cytolytic synapse.

Compared with other bTCE molecules, features of our bTCE-format are, its relatively high molecular weight and its bivalent binding to the tumor antigen. Unlike smaller bTCE-formats such as the bispecific T-cell engager or the dual-affinity retargeting protein formats, our bTCE-format is expected to have a longer in vivo half-life, as it is not prone to renal clearance (33–35). Further advantages of our bTCE-format are the perpetuation of integrity and stability of the full-length parent IgG, as well as its bivalency for tumor cell binding. Particularly if the T-cell–engaging antibody binds a glycopeptide epitope, as in case of AT1413 bTCE, antibodies often have lower affinities as when targeting peptide epitopes (36). Thus, the avidity effect of bivalent target binding gives a valuable addition to its overall functional affinity.

Our bTCE-format contains two CD3ϵ-binding domains. Knowing that CD3ϵ-targeting antibodies such as UCHT1 and OKT3 are mitogenic toward T cells and observing that this effect is not fully abrogated by using an antibody with abolished FcγR interaction, one of our initial concerns was whether bivalent binding of AT1413 bTCE to CD3ϵ could cause CD43s-independent T-cell activation (37). However, AT1413 bTCE neither induced lysis of CD43s-negative TCRα^−/− Jurkat cells nor activation of T cells in the absence of CD43s-positive target cells. Likewise, the AT1002 control bTCE had no effect on either target cell death or T-cell activation. Hence, triggering of T-cell activation and cytotoxicity by this bTCE-format is dependent on target binding, which is consistent with previously described T-cell–engaging antibody formats with bivalent CD3ϵ binding (19, 38).

AT1413 bTCE is the first bTCE generated through chemo-enzymatic linkage of a CD3ϵ-targeting scFv to an IgG by combining sortase-catalyzed transpeptidation and click chemistry. This bTCE-format provides a platform for simple conversion of any recombinantly expressed antibody into a T-cell–engaging format by adding just the short six amino acid sortase-recognition motif (LPETGG) to its heavy chain C-termini. Thus, extensive antibody engineering, which is often required to prepare an antibody-fusion product, or other bispecific formats (e.g., knob-into-hole) may be bypassed. This allows for quick screening of panels of antibody candidates for use as a bTCE. Alternative to T-cell–engaging antibodies, bispecific antibodies with other modalities (e.g., cytokines) or antibody–drug conjugates are readily accessible, by coupling a functional group of interest to the antibody C-termini via the presented sortase + click procedure (39).

To summarize, with AT1413 bTCE we present a bTCE against CD43s, a novel target overexpressed in AML. Efficacy of AT1413 bTCE toward AML cell lines and toward primary AML blasts in vitro and in vivo validates both the antibody and the target for T-cell engagement, and confirms the therapeutic potential of AT1413 in a T-cell–engaging format.

M.A. Gillissen has ownership interest (including stock, patents, etc.) in AIMM Therapeutics. A.Q. Bakker is a senior scientist at AIMM Therapeutics. H. Spits reports receiving commercial research grant from, has ownership interest (including stock, patents, etc.) in AIMM Therapeutics, and is a consultant/advisory board member for GSK. K. Wagner has ownership interest (including stock, patents, etc.) in AIMM Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G. de Jong, A.Q. Bakker, M.D. Hazenberg, H. Spits, K. Wagner

Development of methodology: L. Bartels, M.A. Gillissen, E. Yasuda, C. Bru, P.M. van Helden, J. Villaudy, K. Wagner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Bartels, G. de Jong, E. Yasuda, V. Kattler, C. Bru, C. Fatmawati, S.E. van Hal-van Veen, M.G. Cercel, J. Villaudy, M.D. Hazenberg, K. Wagner

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Bartels, E. Yasuda, C. Fatmawati, J. Villaudy, K. Wagner

Writing, review, and/or revision of the manuscript: L. Bartels, G. de Jong, G. Moiset, P.M. van Helden, J. Villaudy, M.D. Hazenberg, H. Spits, K. Wagner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Kattler, C. Fatmawati, G. Moiset, A.Q. Bakker, K. Wagner

Study supervision: A.Q. Bakker, P.M. van Helden, M.D. Hazenberg, H. Spits, K. Wagner

Other (in vivo and in vitro experiments): C. Bru

The authors thank the patients for participating in this study and the Trial Office of the Amsterdam UMC, location AMC, Department of Hematology for the provision of the AML samples. We would also like to thank Andrea Solar and Wouter Pos for their initial contribution to the work on bispecific antibodies at AIMM Therapeutics and Sophie Levie for her assistance in testing AT1413 binding to AML blast samples. This study was financially supported by the Netherland's Organization for Scientific Research (NWO), the Dutch Cancer Society (KWF), and Amsterdam UMC, location AMC. Zwaartekracht grant by the Netherland's Organization for Scientific Research (NWO) to H. Spits (ICI00004 to L. Bartels). Intramural AMC PhD Scholarship (to M. A. Gillissen). KWF Kankerbestrijding to H. Spits and M.D. Hazenberg (CA301006 to G. de Jong and G. Moiset). NWO ZonMW VIDI to M.D. Hazenberg (grant no. 91715362).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.