The human neuroblastoma cells lines, NB69, SK-N-FI and IMR-5/75, were purchased from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (Braunschweig, Germany) and cultured in Roswell Park Memorial Institute (RPMI) 1640 media (GIBCO) supplemented with 10% fetal calf serum. SK-N-BE(2) cells were purchased from the American Type Culture Collection (Wesel, Germany), and cultured in Dulbecco’s modified Eagle’s medium (DMEM, GIBCO) supplemented with 10% fetal calf serum. The NXS2 murine neuroblastoma cell line19 was kindly provided by Holger N. Lode (University Medicine Greifswald, Germany) and cultured in DMEM supplemented with 10% fetal calf serum, 100 U/mL penicillin and 100 µg/mL streptomycin. All cell lines were cultured at 37°C in 5% CO₂. Cell line identity was assured by short tandem repeat DNA genotyping. Cultures were routinely tested for mycoplasma using the PlasmoTest Kit (Invitrogen).

SUREK, EKTOMUN, TRBs011, TRBs012 and Me361 antibodies were kindly provided by Trion Research or Lindis Biotech (Martinsried, Germany), where they were generated by quadroma technology and purified by affinity and ion exchange chromatography as previously described in detail.20

QuantiBRITE PE calibration beads (BD Biosciences) were used to determine GD2 antigen density on neuroblastoma cells according to the manufacturer’s instructions as previously described.21 To assess binding of EKTOMUN and SUREK to GD2+ neuroblastoma cell lines, we incubated 200 000 SK-N-BE(2) cells with EKTOMUN and 200 000 NXS2 cells with SUREK for 1 hour at 4°C. Subsequently, cells were stained with a PE-labeled secondary antibody directed against mouse IgG2a (cat#407108, Biolegend) for 30 min at 4°C and flow cytometric analysis was performed. To visualize trifunctional effector and target cell binding, we coincubated 2×105 NB69 cells expressing GFP_ffluc with 2×106 PBMCs with or without 1 µg/mL EKTOMUN for 5 min at 37°C. After vortexing, the mixture was fixed immediately using BD cytofix/cytoperm according to the manufacturer’s instructions. Subsequently flow cytometric analysis was performed using antibodies against CD11b (cat#101236, Biolegend), CD4 (cat#357407, Biolegend) and CD8 (cat# 300912, Biolegend). Data were analyzed using FlowJo software (BD).

In vitro antibody-mediated tumor cytotoxicity was quantified as previously described22 using a biophotonic luciferase assay. In brief, GFP_ffluc-expressing NB69, SK-N-FI, IMR-5/75 or SK-N-BE(2) neuroblastoma cells were cocultured with freshly thawed peripheral blood mononuclear cells (PBMCs) added 3 hour after neuroblastoma cell seeding in the presence of test antibodies. FcγR blocking was performed using the FcR blocking reagent human (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s instructions. Bioluminescence was measured 72 hours after adding 28.6 µg of Xenolight D-Luciferin (PerkinElmer, Rodgau, Germany) per 150 µl media, with an exposure time of 180 s. Instrumental background was subtracted. Antibody-mediated tumor cell lysis was calculated using following formula: % specific lysis = [(RLU_{tumor cells +PBMCs})−(RLU_{tumor cells + PBMCs + antibody})]/[(RLU_{tumor cells + PBMCs})−(RLU_PBMCs)]×100%. To obtain individual cell populations as required to assess bispecific trAb trifunctionality, T cells were isolated from PBMCs by magnetic-activated cell separation using the Pan T-Cell Isolation Kit (Miltenyi Biotec). Non-T cells retained in the column were eluted and used as accessory immune cells (AICs). IL2 and interferon-γ (IFNG) release by T cells was quantified in media conditioned (24 and 48 hours) by coculture of neuroblastoma cell lines (seeded at 2×10⁵ cells/well in 24-well plates 24 hours before adding PBMCs at an effector ratio of 10:1) using OptEIA enzyme‐linked immunosorbent assays (BD, Heidelberg, Germany) according to the manufacturer’s instructions. T-cell activation and IFNG producing effector cells were flow cytometrically quantified (as previously described22 48 hours after neuroblastoma-PBMC coculture (identical set up to cytokine release assays) using LIVE/DEAD cell stain kit (Invitrogen, Carlsbad, CA, USA) to gate out dead cells and antibodies against CD3 (cat#300328, Biolegend, San Diego, CA, USA), CD4 (cat#300546, Biolegend), CD8a (cat#344742, Biolegend), CD25 (cat#302612, Biolegend), CD56 (cat#318306, Biolegend), CD69 (cat#310904, Biolegend) and IFNG (cat#502532, Biolegend). IFNG was stained intracellularly using cytofix/cytoperm (cat#554722, BD) according to the manufacturer’s instructions. Murine cell surface activation markers were detected using antibodies against Cd3 (cat#100218, Biolegend), Cd4 (cat#100559, Biolegend), Cd8a (cat#100744, Biolegend), Cd25 (cat#101916, Biolegend) and Cd69 (cat#104506, Biolegend). Data were analyzed using FlowJo software (BD). All assays were conducted in triplicates.

The ADCC Reporter Bioassay (Promega, Walldorf, Germany) was used in 96-well format to detect cytotoxicity in NB69 cells (25 000 cells/well, effector to target ratio of 3:1) exposed to antibody and engineered Jurkat effector cells (stably expressing the relevant Fc gamma receptor and a luciferase reporter gene under control of the NFAT response element). CDC in NB69 cells (stably expressing luciferase, 30 000 cells/well) exposed to test antibodies was performed as previously described23 in 96-well format. After 24 hours, background bioluminescence was measured in culture media before adding 28.6 µg of XenoLight D-luciferin (PerkinElmer) per 150 µl media for either assay (ADCC and CDC), and bioluminescence was quantified. Relative light units (RLUs) were reported directly as surrogates for ADCC, and CDC was calculated by [1−(RLU_{tumor cells + serum + antibody})/(RLU_{tumor cells + serum})]×100%.

Male and female A/J mice (The Jackson Laboratory, Bar Harbor, ME, USA) were bred in-house and group-housed under pathogen-free conditions on a 12-hour day/night cycle with free access to food and water, according to institutional guidelines and compliant with national and EU regulations for animal use in research. Tumor-infiltrating lymphocytes (TILs) were analyzed in A/J mice (8–12-week-old males, 7 per treatment group) intravenously injected with 1×10⁶ NXS2 cells on day 0, then intraperitoneally injected with either 10 µg SUREK in 100 µL phosphate-buffered saline (PBS) or only PBS on day 17. Mice were euthanized on day 21 and macroscopic metastases were dissected from the liver. Metastases were mechanically disaggregated and then enzymatically digested with dispase II (1 mg/mL, Sigma, St. Louis, MO, USA), papain (0.37 mg/mL, Sigma) DNAse I (0.1 mg/mL, AppliChem, Darmstadt, Germany) disolved in calcium-free and magnesium-free Hank’s balanced salt solution (GIBCO, Grand Island, NY, USA) for 20 min at room temperature and filtered through 70 µm nylon cell sieves (Falcon, BD) to obtain a single-cell suspension. Erythrocytes were lysed before flow cytometric analysis of TILs. Spleen tissue was minced, passaged through a 70 µm filter and erythrocytes were lysed before further analysis. Liver metastases were quantified in A/J mice (8–12-week-old females) intravenously injected with 5×10⁵ NXS2 cells on day 0, then intraperitoneally injected with 5 µg of either SUREK, TRBs012 or ch14.18 in 100 µl PBS (or PBS alone) on days 1 and 5, followed by 1 µg of the same agent on day 9. Mice were euthanized on day 21, and macroscopic liver metastases were counted by two independent researchers (blinded to testing group), and results were averaged.

Differences between testing groups in in vitro assays were analyzed using unpaired t tests. Differences between testing groups in in vivo assays were compared using the Mann-Whitney test. P values<0.05 were considered statistically significant. Statistical analyses were performed using GraphPad Prism V.8.3 software (GraphPad Software, San Diego, CA, USA).

To assess EKTOMUN response against neuroblastoma cells presenting a clinically relevant range of target molecule density, we selected four human neuroblastoma cell lines expressing different GD2 levels from intermediate-risk or high-risk genomic backgrounds, either lacking or harboring MYCN amplifications. Quantification of GD2 molecules on the cell surface in these four human neuroblastoma cell lines and the murine NXS2 neuroblastoma cell line, which is used in our syngeneic mouse model, showed that GD2 molecule density was highest in the MYCN-amplified IMR-5/75 cell line (599 163±5095 molecules/cell) and lowest in the SK-N-FI cell line (4357±327 molecules/cell; figure 1A). The murine NXS2 cell line expressed 19 645±1266 molecules/cell, ranking in the lower half of the tested cell lines. We investigated binding of the bispecific trifunctional antibodies, EKTOMUN (binding of human T cells) and SUREK (binding of murine T cells), and control antibodies (binding specificities and structures in figure 1B) to appropriate target-expressing cells in our cell line panel. Both EKTOMUN and SUREK were able to bind the neuroblastoma cell lines SK-N-BE(2) and NXS2 (figure 1C). The bivalent GD2-directed monoclonal Me361 antibody, the parental antibody to the GD2-directed antibody half of EKTOMUN and SUREK, bound at lower concentrations compared with monovalent EKTOMUN and SUREK. The more effective binding of EKTOMUN to SK-N-BE(2) compared with SUREK binding to NXS2, representing the higher GD2 density on SK-N-BE(2) cells. Both TRBs011 and TRBs012 did not bind to neuroblastoma cells and are used as tumor-blind negative controls. To show trAb-mediated effector and target cell interaction, we coincubated PBMCs together with a GFP+ neuroblastoma cell line with or without EKTOMUN. We stained the mixture with fluorochrome-labeled antibodies against CD4 and CD8 to detect T cells, and against CD11b to detect FcγR+ monocytic AICs. We detected 8.4% triple+effector-target cell clusters (CD11b+/CD4+ or CD8+/GFP+) in the presence of EKTOMUN and only 0.82% without EKTOMUN via flow cytometry (figure 1D). In conclusion, we show differential GD2 expression on neuroblastoma cell lines and provide evidence that both SUREK and EKTOMUN bind to neuroblastoma cells in a concentration-dependent manner. We also show simultaneous interaction of EKTOMUN with all effector and target cell types.

In order to evaluate EKTOMUN-dependent neuroblastoma cytotoxicity in vitro, we used human PBMCs from healthy donors in coculture with neuroblastoma cell lines. We first assessed the ‘trifunctionality’ of EKTOMUN by testing, whether all three EKTOMUN specificities (GD2-expressing tumor cells, CD3-expressing T cells and FcγR-expressing AICs) are able and necessary to induce neuroblastoma cell death. The PBMCs were analyzed for their proportion of T cells and AICs by flow cytometry to assess T or AIC contributions to tumor cytolysis. The PBMCs used, consisted of 70% T cells and 30% AICs as determined by flow cytometry (online supplemental figure 1). To differentially investigate each of the three antibody functions, NB69 cells were incubated with (i) T cells at an effector to target (E:T) ratio of 7:1, (ii) AICs (E:T ratio of 3:1) or (iii) or both (E:T ratio of 10:1). PBMC or T cells were not activated prior to incubation. To block trAb–effector cell interaction, we treated one condition with SUREK, which binds murine instead of human CD3 but has the same Fc fragment and GD2 binding arm as EKTOMUN. To prevent FcγR binding, we incubated PBMCs with a FcγR-blocking reagent before adding them to the coculture. The use of SUREK or only AIC completely abolished the CD3-mediated cytotoxic effect and the use of FcγR blocking reagent or only T cells massively decreased tumor cell lysis. EKTOMUN concentration-dependent neuroblastoma cell lysis was only fully induced when both T cells and AICs were present and all EKTOMUN functions were operative (figure 2A), demonstrating EKTOMUN trifunctionality to recognize target cells (via GD2-binding), activate T cells (via CD3-binding) and activate AICs (via FcγR-binding).

Next, we compared EKTOMUN-mediated lysis of the human NB69 neuroblastoma cell line using PBMCs from three different donors to exclude donor-dependent factors. Very low EKTOMUN concentrations (0.1 ng/mL) already induced tumor cell lysis, which reached nearly 100% using 1 ng/mL at an E:T ratio of 10:1 independent of the PBMC donor (online supplemental figure 2). The half maximal effective concentration (EC₅₀) ranged between 0.04 and 0.15 ng/mL. All following experiments were conducted with PBMCs from donor #2, as they displayed a mid-range cytotoxic potential.

We next compared EKTOMUN efficacy against four different GD2+ neuroblastoma cell lines cocultured with PBMCs to the standard-of-care monoclonal ch14.18 antibody and tumor-blind trAb TRBs011 negative control. Treatment with EKTOMUN in the presence of PBMCs lysed 73%±4.0 and 77%±17.4 of SK-N-FI and NB69 tumor cells, respectively, independent of target antigen expression. TRBs011 mediated only limited NB69 cytotoxicity, and ch14.18 had almost no cytotoxic effect in this experimental setting (figure 2B). The two highly aggressive MYCN-amplified cell lines, IMR-5/75 and SK-N-BE(2), proved more resistant to EKTOMUN, despite very high GD2 surface densities. Resistance was overcome by increasing either the EKTOMUN concentration or E:T ratio (figure 2C), which however, also increased non-specific effects, as seen by the increased cytotoxicity mediated by the tumor-blind trAb, TRBs011. NB69 cytotoxicity in the presence of PBMCs and the indicated antibody was also viewed microscopically. Tumor cells detached when EKTOMUN was added, indicating their imminent death, while NB69 cells remained adherent when exposed to ch14.18, TRBs011 or no antibodies (figure 2D). In a similar experiment green-fluorescent SK-N-FI and NB69 tumor cells were treated with EKTOMUN or the control TRBs011 trAb in the presence of PBMCs. EKTOMUN treatment reduced GFP+ tumor cells, indicating GD2-specific tumor cell death (online supplemental figure 3). Here we show that the GD2-directed trAb, EKTOMUN, mediates a GD2-dependent cytotoxic effect on neuroblastoma cells harboring or lacking MYCN amplifications, and that activity requires the interplay between tumor, T and AICs.

The identical Fc regions in EKTOMUN and SUREK each contain one mouse and one rat heavy chain, in contrast to the fully human-derived Fc region of ch14.18. It is unknown, how effectively the murine-derived Fc regions induce ADCC and CDC. Thus, we evaluated the biological activity of the trAb Fc fragment in comparison to ch14.18. NK cell-mediated ADCC is considered to be the main method through which ch14.18 delivers tumor cell death.24 We used a commercially available ADCC reporter assay in which Jurkat cells stably express the human FcγR IIIa (CD16), which can be bound by human and murine Fc regions with similar affinity.25 FcγR IIIa binding on the engineered Jurkat cells mediates a quantifiable luciferase activation as a surrogate for ADCC activity. SUREK was used to evaluate the Fc region instead of EKTOMUN because it does not recognize human CD3, thus avoiding trAb-CD3 interaction on the surface of CD3+ Jurkat cells. As expected, ch14.18 produced a strong concentration-dependent bioluminescent signal. SUREK also induced Jurkat activation, but to a lesser extent (21% relative to ch14.18 at 25 µg/mL, figure 3A), demonstrating the noticeably weaker ability of the murine Fc region to induce ADCC.

CDC is thought to be the main mechanism of action for neuropathic pain, a common side effect of ch14.18. Whether the trAb Fc region induces CDC is also currently unknown, but is interesting in regard to its potential ability to induce neuropathic pain. We performed a CDC assay where the neuroblastoma cell line NB69 was treated with ch14.18 or EKTOMUN in culture medium with 12.5% complement factor-containing human serum. While ch14.18 induced a strong complement-dependent cytotoxic effect after 24 hours (specific lysis: 74.9%±5.4), EKTOMUN induced significantly less CDC (specific lysis: 24.3%±16.0, figure 3B). Our data demonstrate that the cytotoxic effect mediated by the trAb Fc fragment is lower compared with ch14.18, possibly avoiding ch14.18-dependent side effects.

In order to decipher which immune cell subtype is activated on EKTOMUN treatment, we flow cytometrically characterized PBMCs after 48 hours coculture with NB69 neuroblastoma cells and EKTOMUN or control antibodies. CD4+ and CD8+ T cells, NK cells (CD56+CD3−) and NKT (CD56+CD3+) cells were effector cells of this activity. CD3+ cells strongly upregulated activation markers (CD25 and CD69) in the presence of EKTOMUN and, to a lesser extent, in the presence of the unspecific TRBs011 trAb (figure 4A). Similarly, we detected CD25 and CD69 upregulation in CD3−CD56+ cells, indicating that NK cells were engaged and activated by the Fc region (figure 4A). CD4+ T cells expressed the highest CD25 levels (33.6%±6.4 positive cells), while NKT cells expressed the highest CD69 levels (89.0%±2.2 positive cells). After coculturing PBMCs with neuroblastoma cells and EKTOMUN, T-cell effector function was assessed by quantifying IFNG and IL2 cytokine release (figure 4B). EKTOMUN induced PBMCs to release IFNG and IL2 release more strongly than treatment with TRBs011, while ch14.18 induced no cytokine release. Intracellular IFNG staining in effector cells demonstrated that CD4+ and CD8+ T cells, NK (CD56+CD3−) and NKT (CD56+CD3+) cells were IFNG+. While most NKT cells were identified as IFNG-producing (more than 40%), the numbers of CD4+ T cells made them the largest IFN-producing population (figure 4C). However, other immune cell types than those we have analyzed, such as M1 macrophages, could also contribute to IFNG production. Our data demonstrate that treatment with EKTOMUN, strongly activates T, NK and NKT cells as effectors within human PBMCs cocultured with neuroblastoma cells in vitro. Furthermore, T cells exert their effector function by releasing IL2 and IFNG on exposure to EKTOMUN.

The presence of activated TIL has been associated with improved outcomes for neuroblastoma patients.26 We investigated whether SUREK has an impact on TILs and their activation status in a highly aggressive murine model mimicking neuroblastoma minimal residual disease. We used the established murine neuroblastoma cell line, NXS2, which is a hybrid of the murine C1300 neuroblastoma cell line (A/J background) and murine dorsal root ganglia cells (C57BL/6 background) in order to achieve GD2 expression.19 We transduced NXS2 cells with GFP to support flow cytometric identification after tumor cell harvest. After intravenous NXS2 cell injection, immunocompetent A/J mice predominantly developed liver metastases.24 Hence, animals were treated as shown in figure 5A, spleens and liver metastases were harvested on day 21 and Cd3+ cells were measured via flow cytometry (gating strategy to identify Cd3+ TILs is shown in online supplemental figure 4). No difference was seen in the frequency of TILs between mice receiving SUREK or the negative control, PBS (online supplemental figure 5A). Interestingly, the composition of T-cell subsets within the metastases changed dramatically on SUREK treatment. The proportion of Cd4+ T cells significantly decreased and there was a trend toward an increase of the Cd8+ T-cell proportion in SUREK-treated animals (online supplemental figure 5B,C). Consequently, the Cd4/Cd8 ratio of TILs was different between mice treated with SUREK (0.59±0.28) and control mice receiving PBS (2.2±0.13), thus, shifting the ratio toward cytotoxic Cd8+ T cells. This effect was also much more pronounced in T cells harvested from tumors, compared with spleens, from SUREK-treated animals (figure 5B). To assess TIL activation, we flow cytometrically determined Cd25 and Cd69 expression on Cd4+ and Cd8+ TILs compared with background activation observed in splenic T cells. Both Cd4+ and CD8+ TILs specifically upregulated the Cd69 activation marker compared with TILs from the PBS control cohort or to splenic T cells from SUREK-treated mice (figure 5C). The Cd25 activation marker was significantly upregulated in Cd4+ TILs from SUREK-treated animals compared with TILs from PBS-treated controls and splenic T cells from SUREK-treated mice (figure 5D). Though Cd25 expression on Cd8+ TILs from SUREK-treated mice was higher than in splenic T cells from the same animals, expression did not significantly differ from CD8+ TILs from PBS-treated mice. Here, we show that SUREK treatment causes a shift in the Cd4-expressing and Cd8-expressing TIL population toward a higher proportion of cytotoxic Cd8+ cells, while also inducing TIL-specific activation of Cd4+ and Cd8+ T cells.

We used our murine model mimicking neuroblastoma minimal residual disease to directly compare the antitumor potency of SUREK to that of the GD2-directed monoclonal ch14.18 antibody, the current standard-of-care. We intravenously injected NSX2 tumor cells into mice, before treating cohorts every 4 days for a total of three cycles with either SUREK, the unspecific TRBs012 trAb, ch14.18 or PBS starting on day +1. This design mimicks the clinical setting, in which ch14.18 is used as postconsolidation therapy for minimal residual disease (figure 6A). On day +21 post-tumor cell injection, mice were sacrificed to quantify macroscopically visible metastases on the liver surface (examples shown in figure 6B). No superficial liver metastases were detected in four of six mice (67%) belonging to the SUREK treatment group. Single metastasis was detected in each of the other two mice in this group. In contrast, only two of six mice (33%) receiving ch14.18 were free of liver metastases, while the majority developed multiple liver metastases (up to 23) on the liver surface (figure 6C). The tumor-blind trAb, TRBs012, also reduced the number of liver foci compared with the PBS-treated control group, but to a lesser extent compared with SUREK. In conclusion, we provide preclinical evidence that bispecific trifunctional antibodies directed against GD2 show superior efficacy to prevent the outgrowth of liver metastases compared with the standard-of-care monoclonal ch14.18 antibody.