The study protocol was approved by the Ethics Committee of the University of Science and Technology of China (Approval No. USTCACUC212301032).

Male or female NSG mice were purchased from SHANGHAI MODEL ORGANISMS (Shanghai, China). Mice were kept in specific pathogen-free conditions according to the National Guidelines for Animal Usage in Research (set by the Chinese government) at the University of Science and Technology of China. Mice between 6 weeks and 8 weeks of age were used.

Cell lines (MM.1S-luc, RPMI 8226) were purchased from the Cell Bank of the Type Culture Collection of Chinese Academy of Sciences (Shanghai, China). The U266 cell line was purchased from Procell (Wuhan, China). The NCI-H929-luc cell line was purchased from SHANGHAI MODEL ORGANISMS (Shanghai, China). NK92 cells were purchased from the American Type Culture Collection (Manassas, USA). 293T cells were purchased from the Cell Bank of the Type Culture Collection of Chinese Academy of Sciences (Shanghai, China). ExpiCHO was purchased from Thermo (Waltham, USA). The cell lines (MM.1S-luc, RPMI 8226, U266, NCI-H929-luc) were cultured at 37°C in an atmosphere of 5% CO₂ in RPMI 1640 medium (VivaCell, China) supplemented with 10% fetal bovine serum, penicillin (100 U/mL) and streptomycin (100 U/mL). NK92 cells were cultured at 37°C in an atmosphere of 5% CO₂ in α-MEM (HyClone, Logan, USA) supplemented with 12.5% fetal bovine serum, 12.5% horse serum, 100 units/mL human recombinant IL-2, 10 mM inositol, 0.1 mM 2-mercaptoethanol, 0.02 mM folic acid, penicillin (100 U/mL) and streptomycin (100 U/mL). 293T cells were cultured at 37°C in an atmosphere of 5% CO₂ in DMEM (VivaCell, Shanghai, China) supplemented with 10% fetal bovine serum, penicillin (100 U/mL) and streptomycin (100 U/mL). ExpiCHO cells were cultured at 37°C in an atmosphere of 8% CO₂ in ExpiCHO™ Expression Medium (Thermo, Massachusetts, USA).

BT1 and BK1 were constructed by recombinant DNA technology and purified from the supernatants of transfected ExpiFectamine™ CHO cells using the ExpiFectamine™ CHO Transfection Kit (Thermo, Massachusetts, USA). The supernatants were then affinity purified with protein A. The molecular weight and purity of BT1 and BK1 were determined by SDS-PAGE.

Blood samples were collected from adults at the Blood Center of Anhui Province (Hefei, China). PBMCs were isolated by centrifugation through a density gradient according to the manufacturer’s instructions (GE Healthcare, Chicago, USA). Negative selection was conducted to purify NK cells and T cells using a MACS kit according to the manufacturer’s instructions (Miltenyi Biotec, Bergisch Gladbach, Germany).

For the binding assay, tumor cells or effector cells were incubated with various concentrations of BT1 or BK1 for 30 min at 4°C. After washing, tumor cells or effector cells were stained with anti-IgG Fc antibodies for 30 min at 4°C. Then, the positive proportion was detected using flow cytometry.

For binding of a bispecific antibody to antigen, antigens (1 μg/mL) were incubated in each well overnight at room temperature. The wells were blocked for 2 h with blocking buffer and incubated for 2 h with bispecific antibodies (IgG, BT1, or BK1) at serial dilutions. The wells were washed with PBST and incubated for 1 h with HRP-conjugated AffiniPure Mouse Anti-human IgG (BOSTER, Wuhan, China). Then, the cells were incubated with 1-Step™ Ultra TMB-ELISA Substrate Solution (Thermo, Massachusetts, USA) after the wells were washed, followed by stop Solution (Multi Sciences, Hangzhou, China). The absorbance at 450 nm (OD450) was read.

Cytokine concentrations were analyzed from supernatants of cytotoxicity assays using the Human Th1/Th2/Th17 CBA Kit (human) Kit (BD, New Jersey, USA) and LEGENDplex™ Human CD8/NK Panel (13-plex) (Biolegend, San Diego, USA) following the manufacturer’s instructions.

RPMI 8226, NCI-H929-luc and MM.1S-luc cells were labeled with CellTrace™ Violet (CTV) (Thermo, Massachusetts, USA) and incubated with PBMCs, purified human NK cells, or purified human T cells at the indicated E:T ratios in the presence of IgG, BT1, BK1 or BK1 combined with BT1 for 6 h or 24 h. For the spontaneous death control, CTV-labeled target cells were cultured alone under identical conditions. After 6 h or 24 h, the wells were harvested, and 7-aminoactinomycin D (7-AAD) (BD, New Jersey, USA) was added before analyses. Samples were mixed thoroughly and analyzed by flow cytometry.

For flow cytometry, CellTrace™ Violet (CTV) was obtained from Invitrogen (C34557, Thermo, Massachusetts, USA). 7-AAD (559925), anti-CD107a (563869), anti-CD45 (557833), anti-CD69 (560711), anti-CD14 (563079) and anti-TNF (559321) were obtained from BD Biosciences (USA); anti-CD56 (362538), anti-CD3 (300328), anti-IFN-γ (506504), anti-TCR α/β (306705), anti-FOXP3 (320113), anti-CD16 (980112) and anti-IgG-Fc (410712) were obtained from BioLegend (San Diego, USA).

For staining of the surface markers, PBMCs were cocultured with tumor cells in the presence of BT1, BK1 or BK1 combined with BT1 (1 μg/mL) for 6 h or 24 h. After incubation, PBMCs were stained for CD25, CD69 or CD107a.

For staining of intracellular cytokines, PBMCs were cocultured with tumor cells in the presence of BT1, BK1 or BK1 combined with BT1 (1 μg/mL) for 6 h or 24 h. After incubation, PBMCs were stained for surface markers, fixed, and permeabilized with fixation buffer according to the manufacturer’s instructions (eBioscience, Massachusetts, USA). The fixed cells were then stained with antibodies against IFN-γ and TNF. All samples were acquired on an LSR-II or Celesta flow cytometer (BD, New Jersey, USA) and were analyzed using FlowJo (TreeStar, Ashland, USA).

For cell sorting experiments, human CD56⁺ CD16^- NK cells were sorted from human CD56⁺ NK cells using BD Fusion (BD Biosciences, USA).

NSG mice were injected intravenously (i.v.) with luciferase-expressing MM.1S (MM.1S-luc) tumor cells (1 × 10⁶) on day 0 followed by i.v. of CD16-overexpressing NK92 (NK92-CD16) cells (2 × 10⁶) every 6 days for a total of 4 doses beginning on day 4. Tumor-bearing mice were injected i.v. with BK1 (50 μg) or PBS every 3 days for a total of 8 doses beginning on day 4. In addition, tumor-bearing mice were injected i.p. with IL-2 (50,000 units; Jiangsu Kingsley Pharmaceuticals, Nanjing, China) every 2 days for a total of 11 doses beginning on day 4.

NSG mice were injected intravenously (i.v.) with luciferase-expressing NCI-H929 (NCI-H929-luc) tumor cells (2 × 10⁶) on day 0 followed by i.v. of PBMCs (5 × 10⁶ or 1 × 10⁷) or a 1:1 mixture of purified NK and T cells (5 × 10⁵) on day 9. Tumor-bearing mice were injected i.v. with IgG, BK1, and BT1 (50 μg) every 3 days for a total of 3 doses beginning on day 9 combined with IL-2 (10,000 units; Jiangsu Kingsley Pharmaceuticals, Nanjing, China) every 3 days for a total of 3 doses beginning on day 9.

NSG mice were injected intravenously (i.v.) with luciferase-expressing NCI-H929 (NCI-H929-luc) tumor cells (5 × 10⁶) on day 0 followed by i.v. of PBMCs (2 × 10⁷) on day 11. Tumor-bearing mice were injected i.v. with IgG, BT1, BK1, and BT1+BK1 (50 μg/mouse) on day 11 with IL-2 (10000 units; Jiangsu Kingsley Pharmaceuticals, Nanjing, China). Serum cytokines were detected 6 hours after treatment with a V-PLEX Proinflammatory Panel 1 (human) Kit (LX-K15049D-1, LabEx, Shanghai, China).

Mice were injected i.p. with d-luciferin (15 mg/mL; Gold Biotechnology, St. Louis, USA) at 150 mg/kg bodyweight. Mice were placed into an in vivo imaging system (Caliper Life Sciences, Waltham, USA) when fully anesthetized by isoflurane. Luciferase expression was imaged using Spectral In Vivo (PerkinElmer, Waltham, USA) and calculated using Living Image (PerkinElmer, Waltham, USA).

Statement: For animal studies, Tumor-bearing mice were randomized before administration.

Comparisons between 2 groups were analyzed using the unpaired one/two-tailed Student’s t test. Comparisons between ≥ 3 groups were analyzed using 2-way analysis of variance. Kaplan-Meier analyses and Mantel-Cox tests were used to analyze mouse survival. The results are expressed as the mean ± SEM. P < 0.05 was assumed to be significant in all analyses.

To compare the function of T-cell engaging bispecific antibody (T-cell engager) and NK-cell engaging bispecific antibody (NK-cell engager), we constructed BT1 (BCMA×CD3 T-cell engager) and BK1 (BCMA×CD16A NK-cell engager), respectively. BT1 is a two-arm IgG1-based human antibody consisting of two single chain variable fragments (scFvs) and Fc ( Figure S1A ). The purity of BT1 was > 90%, as determined by SDS-PAGE ( Figure S1B ). To ascertain the binding ability of BT1, we detected the binding ability of BT1 to the BCMA extracellular segment and CD3 protein through ELISA. The results indicated that the binding ability of BT1 improved with increasing concentrations ( Figures S1C, D ). Furthermore, BCMA⁺ MM.1S-luc cells and CD3⁺ T cells were incubated with Alexa Fluor 647-labeled BT1, and the proportion of Alexa Fluor 647-positive cells increased with BT1 concentrations, indicating that BT1 binds to these BCMA⁺ target cells and CD3⁺ T cells successfully ( Figures S1E, F ).

In addition, we constructed an NK-cell engager (BK1) with the same structure as BT1, consisting of two scFvs and Fc ( Figure 1A ). The purity of BK1 was > 90%, as determined by SDS-PAGE ( Figure 1B ). To ascertain the binding ability of BK1, we detected the binding ability of BK1 to the BCMA extracellular segment and CD16 protein through ELISA. The results indicated that the binding ability of BK1 improves with increased concentrations ( Figures 1C, D ), and it binds to CD16 mainly through its Fab segment ( Figure S2 ). Furthermore, BCMA⁺ MM.1S-luc cells and purified CD16⁺ NK cells were incubated with Alexa Fluor 647-labeled BK1, and the proportion of Alexa Fluor 647-positive cells increased with BK1 concentrations, indicating that BK1 binds to these BCMA⁺ target cells and CD16⁺ NK cells successfully ( Figures 1E, F ). Altogether, these data indicated that BT1 and BK1 have similar structures and sizes and are capable of binding to tumor cells and effector cells.

To compare the effect of BT1 and BK1 on the activation and function of corresponding effector cells, we examined the cytotoxic effect of BT1 and BK1, respectively. PBMCs (culture with 100 IU of IL-2/mL 1640 complete medium) were incubated with MM.1S-luc cells or NCI H929-luc cells in the presence of BK1 for 6 hours, and the results showed that BK1 significantly improved the specific lytic ability of PBMCs against BCMA⁺ tumor cells in a concentration-dependent manner ( Figures 2A, B , S3 ). Furthermore, the incubation of CD16-overexpressing NK92 (NK92-CD16) cells and MM.1S-luc cells also showed significantly improved lytic ability of NK92-CD16 cells against tumor cells in the presence of BK1 ( Figure 2C ). The lytic ability of effector cells was also increased with an elevated E:T ratio ( Figures 2D , S3D–J ). Meanwhile, comparison between the cytotoxicity of CD16⁺ NK cells vs. CD16^- NK cells under the effect of BK1 proved that BK1 could only act through CD16⁺ NK cells to exert its anti-tumor function ( Figures S4A-C ). In addition, BK1 significantly enhanced the expression of CD69, CD107a, IFN-γ and TNF-α on CD3^-CD56⁺ NK cells compared with IgG after coculturing PBMCs with MM.1S-luc cells ( Figures 2E–H , S5A-H ). Furthermore, BK1 also significantly enhanced the secretion of IFN-γ, TNF-α, granzyme A, granzyme B, perforin and sFasL ( Figures 2I, J , S5I-L , S6 ). On the other hand, BT1 significantly improved the lytic ability of PBMCs or purified T cells against BCMA⁺ tumor cells ( Figures S7A-F ). Furthermore, BT1 significantly enhanced the expression of CD25, CD69, IFN-γ and TNF-α on CD56^-CD3⁺ T cells ( Figure S7G ) and the secretion of IFN-γ, IL-2 and TNF-α in the supernatant ( Figures S7H-J ). These data indicated that BK1 was able to activate CD16⁺ NK cells, while BT1 was able to activate CD3⁺ T cells, and both specifically killed BCMA⁺ tumor cells.

Antitumor cytotoxicity is one of the most important functions of bispecific antibodies. We aimed to assess whether BT1 or BK1 improves antitumor function in vivo. Here, we established xenogenic models using human NK92-CD16 cells, a purified NK and T-cell mixture, or PBMCs from NSG mice. First, we established xenogeneic models using NK92-CD16 cells to assess the antitumor effect of BK1. These mice were loaded with MM.1S-luc cells and treated with BK1 or PBS ( Figure 3A ). Tumor growth was significantly inhibited by treatment with BK1 ( Figure 3B ). Quantification of luminescence also confirmed significantly delayed tumor growth on days 15, 22 and 29 after BK1 treatment ( Figures 3C–F ). Consistent with the results of tumor growth, xenogeneic mice treated with BK1 also showed significantly improved survival ( Figure 3G ).

Since the quantity of T cells (55-65%) is far more numerous than that of NK cells (5-20%) in the peripheral blood, to compare the antitumor effect of BK1 vs. BT1, we used a 1:1 mixture of purified NK cells and T cells to reconstitute the xenogenic models. These mice were then loaded with NCI-H929-luc cells and treated with IgG, BT1 or BK1 ( Figures 4A–C ). The results indicated that BK1 (BCMA×CD16A) significantly inhibited tumor growth and elicited a stronger antitumor effect than BT1 when the numbers of NK cells and T cells were equal. We also established xenogenic models using human PBMCs, and these mice were then loaded with NCI-H929-luc cells and treated with IgG, BT1 or BK1 ( Figure 5A ). The results showed that tumor growth was significantly inhibited by treatment with BT1 or BK1, and BT1 and BK1 showed comparable antitumor effects in these xenogeneic models reconstituted with human PBMCs ( Figures 5B, C ). Altogether, these data indicated that both BT1 and BK1 are effective against tumors, among which BK1 is stronger than BT1 when the numbers of NK cells and T cells are equal. Comparable antitumor effects were achieved when the number of T cells was approximately 3 times greater than the number of NK cells, suggesting that BK1 (BCMA×CD16A) is a stronger antitumor bispecific antibody than BT1 (BCMA ×CD3).

BT1 and BK1 both showed strong antitumor effects in tumor-bearing mice. Next, we wanted to determine whether the combined treatment could elicit stronger cytotoxicity. We used PBMCs to coculture with MM.1S-luc cells in the presence of BK1, BT1 or IgG in vitro, and the results indicated that both BT1 and BK1 significantly improved the cytotoxicity after 6 hours of coculturing, in which BK1 (BCMA×CD16A) showed stronger cytotoxicity against target cells than BT1 (BCMA ×CD3) ( Figures 6A, B ). We further assessed the cytotoxicity of BK1, BT1 or BK1 combined with BT1 in vitro, and the results indicated that combined treatment significantly improved the cytotoxicity of PBMCs against MM.1S-luc cells after 6 hours of coculturing compared to using either treatment alone ( Figure 6B ). To rule out the influence of other immune cells, we purified NK cells and T cells and cocultured a 1:1 mixture of purified NK and T cells with MM.1S-luc target cells for 6 hours. The results also showed that combined treatment significantly improved the cytotoxicity of the mixed effector cells compared to using either treatment alone ( Figure 6C ).

In addition to the in vitro analysis, we also established xenogenic models using human PBMCs with NSG mice to assess the antitumor effect of the combined treatment in vivo ( Figure 6D ). Bioluminescence imaging of mice 15 days after tumor challenge showed that BT1 or BK1 alone could reduce tumor growth, while combined treatment with BT1 and BK1 further delayed tumor growth with the strongest antitumor effect ( Figures 6E, F ). Altogether, these data indicated that combining BT1 and BK1 improved the cytotoxicity of effector cells in vitro and restrained tumor growth in vivo, suggesting that the maximum antitumor effect could be achieved by using combinatorial treatment.

Cytokine storms and neurotoxicity are obstructive factors for immunotherapy. To evaluate the potential risks of BT1 and BK1, the cytokine levels in the supernatant were determined after coculture of PBMCs and MM.1S-luc cells for 24 hours in the presence of IgG, BK1, BT1 or BK1 combined with BT1 in vitro. The results indicated that BT1 induced significantly higher levels of cytokine secretion, including IFN-γ, IL-2, TNF, IL-10 and IL-6, than BK1 ( Figures 7A–E , S8 ), in which IL-6, TNF and IL-10 are strong inducers of cytokine storms. Interestingly, significantly lower levels of cytokines, including IL-2, IL-10 and IL-6, were observed in the supernatant of the combined treatment compared with that of BT1, suggesting that the combined usage of BT1 and BK1 reduces cytokine secretion by BT1. To explore the underlying mechanism, we evaluated the PBMC composition after 24 hours of co-culture. The percentage of regulatory T cells and monocytes remained the same after treatment with BT1, BK1 or BT1 combined with BK1 ( Figures S9A, B ). Interestingly, the percentage of T cells decreased after BT1 treatment but recovered after BK1 or BT1 combined with BK1 treatment, suggesting that BK1 could enhance the proliferation of T cells ( Figure S9C ). We further used CTV-tagged PBMCs co-cultured with tumor cells in the presence of BT1 or BT1 combined with BK1 for 3 days to evaluate the proliferation of T cells. As shown in Figures S9D-F , combinatorial treatment resulted in higher T-cell proliferation than BT1 treatment alone.

To mimic the in vivo microenvironment, we constructed a mouse model of cytokine release syndrome to assess cytokine secretion in the blood serum of tumor-bearing mice. NSG mice were injected intravenously with NCI-H929-luc tumor cells on day 0 followed by intravenous injection of PBMCs on day 11. Tumor-bearing mice were then injected intravenously with IgG, BT1, BK1 or combined treatment on the same day followed by cytokine detection 6 hours later ( Figure 7F ). The results showed that treatment with BK1 induced fewer proinflammatory cytokines, including TNF, IL-6, IL-10, IL-8, IL-12p70, IL-4 and IL-13, than treatment with BT1 ( Figures 7G–M ). These cytokines are closely related to cytokine release syndrome from immunotherapy (24, 25). Surprisingly, combined treatment with BT1 and BK1 in vivo induced fewer cytokines than BT1 ( Figures 7G–M ), suggesting that BK1 somehow reduces the cytokine secretion induced by BT1. Altogether, these data indicated that NK-cell engager was safer than T-cell engager with less proinflammatory cytokine production. In addition, BK1-activated NK cells somehow reduced excessive cytokine production by BT1-activated T cells, suggesting that not only NK-cell engager results in less proinflammatory cytokine production but also helps to reduce the proinflammatory cytokines secreted by T cells.