The Raji, SP2/0, Jurkat, Reh, and Daudi cell lines were acquired from the Cell Bank of the Chinese Academy of Sciences. The NCI-H929 and NK92MI cell lines were obtained from the American Type Culture Collection (ATCC). L-1236 and MOLM-13 cell lines were purchased from Shanghai Qiansi Biotechnology Co., Ltd., and Nanjing Cobioer Biosciences Co., Ltd., respectively. The modification of CHO-CD38 and FcgRIIIA (158 V) target-stimulated NK (FcR-TANK™) cells were carried out in our laboratory. Cells in all cell lines were harvested after they reached the logarithmic growth phase. Jurkat, SP2/0, L-1236, Daudi, MOLM-13, Reh, Raji, and NCI-H929 cells were incubated at 37°C in 5% CO₂. RPMI-1640 media (Gibco, Cat#11875093) was utilized for the cultivation of all the cells. The media were supplemented with 1% penicillin–streptomycin (Gibco, Cat# 15140122) and 10% fetal bovine serum (Gibco, Cat# 10091148). CHO-CD38 cells were cultured in EX-CELL^® 302 serum-free medium (Sigma, Cat# 24326C) supplemented with 1% penicillin–streptomycin. Serum-negative TANK media (Immuneonco, Cat# CT001–1) was utilized for cultivating FcR-TANK cells.

After BALB/c mice were immunized with CHO cells overexpressing hCD38 and the human CD38 full-length extracellular domain (43–300) fused to mIgG1-Fc, the conventional hybridoma method was used to screen for anti-human CD38 antibodies. Positive fusions were assessed for specific interactions with CHO-CD38, but not with CHO, by fusing immunized animal splenocytes with Sp2/0 myeloma cells. Based on the epitope and sequence homology, we selected nine anti-CD38 antibodies for detailed investigation and development of CD38/CD47 BsAbs.

TransFx-C CHO transient transfection media (HyClone, Cat# SH30942.02) was utilized for the cultivation of CHO-S cells (Gibco, Cat# A29127). By transient transfection, the expression vectors expressing the heavy and light chains of antibodies were cotransfected using a polyethylenimine transfection reagent (Polysciences, Cat# 24765). The supernatant of the cotransfected cells was collected after 8–10 days and added to Protein A Sepharose columns (Bestchrom, Cat# AA0273). Subsequently, wash buffer [NaCl (140 mM) + phosphate buffer (PB; 20 mM); pH = 7.4 ± 0.1] was added to the columns, and elution buffer [NaAc (25 mM) + NaCl (100 mM); pH = 3.5 ± 0.1] was used to elute the antibodies. The pH of the collected fraction was adjusted to 5.2 ± 0.2 with 2 M Tris. Finally, size-exclusion high-performance liquid chromatography (SEC-HPLC) was carried out to determine the purity of the acquired antibodies.

The antibodies binding to CD38+ tumor cells (including Raji, Daudi, NCI-H929, L-1236, MOLM-13, and Reh), RBCs, and platelets were assessed via flow cytometry. hIgG1-Fc (in-house) was used as an isotype reference. The samples were incubated at 4°C in tumor cells at multiple serial dilutions for 45 min. Then, PBS + 0.5% BSA (Sangon Biotech; Cat#A500023–0100) was added to remove the free antibodies by centrifugation. Subsequently, the samples were treated with FITC-linked anti-human IgG Fc secondary antibodies (500-fold dilution; Sigma, Cat# F9512) at 4°C in the dark for 45 min. After washing, the cells’ FITC fluorescence signals were assessed by flow cytometry analysis (Luminex, Guava^® easyCyte™ 8HT Base System). The acquired data were analyzed using guavaSoft_33_x64 software.

Similar to the above binding assay, after Jurkat or Reh cells were pretreated with various concentrations of the CD38/CD47 BsAb for 30 min, in this mixture, 1 µg/mL SIRPα-mFc (in-house) was added at a 1:1 (v/v) ratio for 45 min. The cells were then washed with PBS supplemented with 0.5% BSA, the secondary antibody PE-conjugated anti-mouse IgG (500-fold dilution; Biolegend, Cat# 405307) was added, and the cell PE fluorescence signal was detected via flow cytometry analysis.

Using PBS, the antibodies were diluted to 10 µg/mL, and CD38⁺ tumor cells were resuspended to a density of 1 × 10⁶ cells/mL. The antibodies and cells were mixed at a 1:1 ratio (by volume) and added to a 96 well flat-bottom plate (BeyoGold, Cat#FCP968) containing 100 µL of the mixture in each well. For 15 min, the plates were pre-incubated in the dark at ambient temperature. The nicotinamide guanine dinucleotide (NGD) substrate solution (Sigma Aldrich, Cat# N5131) was diluted to 120 µM, and 50 µL of this solution was added to each well containing the mixture. Then, for 2 h, the plates were preincubated in the dark at ambient temperature. The production of cGDPR was quantified by reading fluorescence signals at an excitation wavelength of 300 nm (EX300) and an emission wavelength of 410 nm (EM400) using a multifunctional enzyme-labeling instrument (Molecular Devices, SpectraMax M3). The percent inhibition was calculated as follows: inhibition % = [(1-(Antibodies-blank)/(positive-blank))×100]%. The antibody group contained 50 µL of cell + 50 µL of antibody + 50 µL of NGD, the blank group contained 50 µL of cell + 100 µL of PBS, and the positive group contained 50 µL of cell + 50 µL of PBS + 50 µL of NGD.

The incubation of the target cells was carried out at 37°C in various concentrations of antibodies in 5% CO₂ for 4 h before staining with propidium iodide (PI) solution (Sigma, Cat# P4170). The cells were collected by flow cytometry, and PI-positive cells were subsequently assessed.

The in vivo efficacy of CD38/CD47 BsAbs and the anti-CD38 antibody combined with the SIRPα Fc fusion protein was evaluated in NCI-H929 xenograft models. Further, 200 µL of 5 × 10⁶ cells were resuspended in cold PBS/Matrigel (v/v = 1:1) and administered on the right side of the CB17-SCID mouse back near the axilla. After the tumors reached an average volume of 120 mm³, the mice were randomly categorized into treatment groups based on tumor size and mouse weight. Drugs were injected intraperitoneally in all the experiments. The body weights and tumor volumes of the mice were determined twice per week, and when the tumor size reached 3,000 mm³, the mice were euthanized.

GraphPad Prism 8.0 (GraphPad Software, Inc.) was used for all the statistical analyses. The differences between three or more groups were elucidated by one-way ANOVA with Holm–Sidak correction, whereas the intergroup differences were elucidated by Student’s t-tests. p ≤ 0.05 indicated statistically significant differences among and between groups. In the figures, asterisks denote statistical significance (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

Using mouse hybridoma screening technology, we obtained nine ideal anti-CD38 antibodies that specifically target CD38. We analyzed epitopes using the BLI-based in-tandem orientation method, sequenced cDNA using the SMARTer^® RACE 5’/3’ Kit (Taraka, Cat#634858), and assessed sequence homology using DNAMAN software. The epitope binning technique is used to cluster different antibodies based on their recognized epitopes [epitopes are the specific regions on the antigen (Ag) recognized by the antibody (Ab)]. The results showed that the epitopes of the nine anti-CD38 antibodies were clustered into four categories: 35G5, 26B4, and 25B5 were in one category; 65A7 and 82G10 were in the same category; 12C10 and 32D6 were in one category; and 4A8 and 16C2 were in one category ( Supplementary Figure 1A ). The results of the homology analysis of the heavy and light chain amino acid sequences of the antibodies showed that the heavy and light chain sequence homology was relatively low, especially that of the heavy chain, which indicated a high diversity of anti-CD38 antibodies ( Supplementary Figures 1B–D ).

The affinity of anti-CD38 antibodies for human CD38 was assessed using BLI technology. The affinity and binding kinetics were determined to calculate the dissociation constants and determine the KD. The binding of CD38 antibodies to various hematologic tumor cell lines (Raji, Daudi, L-1236, NCI-H929, Reh, and MOLM-13) in vitro was assessed via flow cytometry. The CD38 expression profile of the tumor cell lines was analyzed using the anti-CD38 mouse IgG1 antibody 26B4 ( Supplementary Figure 2 ). The EC50 was defined as the antibody concentration that resulted in an approximately 50% response when the antibodies were specifically bound to CD38-expressing target cells in these experiments.

The calculated KD values for all anti-CD38 antibodies (except 65A7) that bound to purified human CD38 were lower than the corresponding KD values for daratumumab and isatuximab ( Supplementary Figure 3 ). The results of flow cytometry assays showed that all anti-CD38 antibodies bound to CHO-CD38 cells in a concentration-dependent manner, but they did not bind to CHO cells ( Supplementary Figures 4A, B ). However, the anti-CD38 antibodies that bound to monkey CD38 included only 65A7, 12C10, 32D6, 82G10, 4A8, and 16C2 ( Supplementary Figure 4C ). The results of the flow cytometry assays showed that all anti-CD38 antibodies bound to cells in various CD38⁺ tumor cell lines in a concentration-dependent manner; 35G5 and 25B5 exhibited a stronger binding ability ( Supplementary Figures 5A–F ).

In the development of therapeutic antibodies targeting anti-CD38, our focus was on prioritizing IgG1. This choice was driven by its high affinity for binding and activating FcγRs, leading to the potent induction of ADCP and ADCC against CD38⁺ tumor cells. IgG1 formats were selected for anti-CD38 antibodies, with the substitution of E333A, S298A, and K334A in the Fc region to enhance Fc-mediated effector functions. In vitro, anti-CD38 antibodies induced the lysis of CD38⁺ tumor cells through CDC, ADCP, and ADCC. Anti-CD38 antibodies exhibited strong ADCC activity against cells associated with hematological malignancies (Daudi, L-1236, Raji, and NCI-H929 cell lines) in a concentration-dependent manner via FcR-TANK (effector). The activity of all the anti-CD38 antibodies was similar and was significantly greater than that of daratumumab ( Figures 1A–D ). By conducting flow cytometry assays, phagocytosis was evaluated in the monocytic cell line THP-1 (effector) against the target cell lines Raji, L-1236, and NCI-H929 labeled with CFSE by measuring the green fluorescence of THP-1 cells following 2 h of incubation with anti-CD38 antibodies. The anti-CD38 antibodies also exhibited strong concentration-dependent ADCP activity against the Raji, L-1236, and NCI-H929 cell lines ( Figures 2A–C ). Antibody CDC induction was related to the density of the target protein and the inhibitory complement regulatory proteins (CD46, CD55, and CD59). Cell lysis was used as a marker for CDC, as determined by the uptake of propidium iodide (PI) via flow cytometry analysis. Hematologic cancer cells (Daudi, L-1236, Raji, and NCI-H929 cell lines) were incubated with anti-CD38 antibodies in the presence of complement-containing human serum. The CDC activity of different antibodies was different, and the CDC activity of the same antibody in different cell lines was also different. None of the anti-CD38 antibodies showed CDC activity against NCI-H929 cells (data not presented). Additionally, the 12C10, 35G5, 82G10, and 26B4 antibodies showed strong CDC activity against Raji, L-1236, and L-1236 cells ( Figures 2D–F ).

Daudi, L-1236, Raji, and NCI-H929 cells were cultured with anti-CD38 antibodies for 24 h, and apoptosis was measured by PI staining. The apoptotic activity of different antibodies differed, and the activity of the same antibody in different cell lines also differed. None of the antibodies showed apoptotic activity against NCI-H929 cells. The 12C10 and 82G10 antibodies showed the strongest apoptosis-inducing activity against Daudi, L-1236, and L-1236 cells ( Figures 3A–D ).

The effect of anti-CD38 antibodies on the ADP-ribosyl cyclase activity of CD38 was examined by incubation with CD38-expressing cells in assays using nicotinamide guanine dinucleotide (NGD) as a substrate. In assays where NGD was used as the substrate, 26B4, 35G5, and 82G10 exhibited the strongest ability to inhibit the enzymatic cyclase activity of CD38-expressing Daudi, Raji, NCJ-H929, and L-1236 cells. In contrast, the 4A8, 12C10, 65A7, and 32D6 antibodies did not influence the enzymatic cyclase activity ( Figures 3E–H ).

Peripheral blood was collected from humans and incubated with anti-CD38 antibodies in vitro. RBCs were identified based on their light-scattering properties using a flow cytometer. Unlike the CD47 antibody Hu5F9, even at a concentration of 100 nM, the anti-CD38 antibodies did not bind to RBCs and did not induce RBC agglutination ( Supplementary Figures 6A, B ). Human blood samples incubated with anti-CD38 antibodies were counterstained with anti-CD61 antibodies to identify platelets in whole blood. Similarly, anti-CD38 antibodies did not bind to platelets at a concentration of 100 nM ( Supplementary Figure 6C ).

Based on in vitro activity data, we selected five anti-CD38 antibodies (4A8, 12C10, 26B4, 35G5, and 65A7) for the development of CD38/CD47 BsAbs using a 2 + 2 “mAb-trap” platform ( Table 1 ). The CD38/CD47 BsAbs consisted of the anti-CD38 antibody and the SIRPα IgV domain (SIRPα domain 1, SD1), connected to the N-terminus of the heavy chain (IMM5605) or the light chain (IMM5606) of the antibody ( Figure 4A ). We also used daratumumab and isatuximab antibody frameworks to design CD38/CD47 BsAbs as controls. IMM5605-ISA and IMM5605-DARA were connected via SD1 to the N-terminus of the heavy chain of isatuximab and daratumumab, respectively. The purity of the CD38/CD47 BsAbs was evaluated by HPLC-SEC and SDS–PAGE, and the results showed that the purity was greater than 90% ( Supplementary Figures 7A, B ). We also used BLI technology to evaluate the affinity of CD38/CD47 BsAbs for CD47 and CD38 targets. The findings revealed that the affinity of the BsAbs for the CD38 target was significantly greater (except for 65A7) than that of the controls IMM5605-ISA and IMM5605-DARA. The affinity of the CD38/CD47 BsAbs for the CD38 target was significantly greater than that for the CD47 target, except for 65A7 ( Supplementary Figures 8 , 9 ). These results indicated that CD38/CD47 BsAbs preferentially bound to CD38⁺/CD47⁺ tumors and avoided binding to CD47⁺ normal tissues, thus reducing the potential for off-target effects. The results of the flow cytometry analysis showed that compared to IMM01 (SIRPα Fc fusion protein), all CD38/CD47 BsAbs had a greater ability to bind to CD38⁺ tumor cell lines in a concentration-dependent manner, and the binding ability of 35G5 and 12C10 was greater than that of the other BsAbs ( Figures 4B–F ). We also analyzed the binding ability of CD38/CD47 BsAbs to human RBCs and found that even at a concentration of 100 nM, BsAbs had very little binding to RBCs, but Hu5F9 strongly bound to RBCs ( Figure 5A ). All CD38/CD47 BsAbs could bind to platelets in a dose-dependent manner but were weaker than Hu5F9 ( Figure 5B ). More importantly, none of the CD38/CD47 BsAbs induced RBC agglutination ( Figure 5C ).

We used BLI technology to evaluate the ability of CD38/CD47 BsAbs to simultaneously bind purified CD47 and CD38. The results indicated that, irrespective of whether BsAbs first bound to CD47 or CD38, they could bind to the other antigen ( Figures 6A, B ). We also performed a flow cytometry assay to evaluate the ability of CD38/CD47 BsAbs to simultaneously bind CD47⁺ cells and CD38⁺ cells. The results showed that BsAbs could bind to cellular CD47 and CD38 simultaneously. 35G5 and 12C10 exhibited the strongest ability to simultaneously bind to the CD47 and CD38 targets ( Figures 6C–E ).

As CD47 is a ubiquitously expressed molecule, we designed a panel of CD38/CD47 BsAbs, which consisted of a CD47-binding arm with lower affinity than the CD38-binding arm. To assess the specificity of the CD38/CD47 BsAbs, we obtained the binding profile of CD38⁺/CD47⁺ Reh cells mixed with CD38^-/CD47⁺ NK92MI cells at a 1:1 ratio by performing flow cytometry assays. CD38/CD47 BsAbs preferentially bound to CD38⁺/CD47⁺ Reh cells, indicating that compared to CD47 therapeutic Abs, BsAbs have lower potential for on-target, off-tumor effects in vivo ( Figure 7 ).

The ability of CD38/CD47 BsAbs to block the interaction between CD47 and SIRPα was tested in a CD47⁺/CD38^low Jurkat cell line and a CD47⁺/CD38⁺ Reh cell line. The results indicated that the ability of BsAbs to block the interaction between CD47 and SIRPα was lower than that of the IMM01 in the Jurkat cell line ( Figure 8A ). In contrast, the blocking activity of the CD38/CD47 BsAbs was 200–500-fold greater than that of IMM01 in the Reh cell line ( Figure 8B ). Our findings showed that the CD38/CD47 BsAbs strongly blocked the CD47/SIRPα interaction in CD38/CD47 double-positive tumor cells, indicating that the inhibition of CD47/SIRPα depended on the engagement of CD38. We evaluated the inhibitory effect of CD38/CD47 BsAbs on the ADP ribosyl-cyclase enzymatic activity of cellular CD38 in vitro and found that 26B4 had the strongest inhibitory effect on cellular CD38 enzymatic function, followed by 35G5; however, the other BsAbs had no inhibitory effect ( Figures 8C, D ).

We also selected IgG1 for designing CD38/CD47 BsAbs, as well as an engineered Fc domain (S298A, E333A, and K334A) for enhancing Fc-dependent effector functions. We used an engineered NK cell line and primary PBMC as effector cells to evaluate ADCC activity of CD38/CD47 BsAbs. The CD38/CD47 BsAbs exhibited potent ADCC activity against CD47⁺/CD38⁺ Raji and NCI-H929 cells in a concentration-dependent manner. Except for 26B4, which had a lower activity than the other BsAbs, the difference in activity between the other BsAbs was 1–3 times ( Figures 9A, B ; Supplementary Figure 10 ). Additionally, the BsAbs did not exhibit an ADCC effect on HL-60 (CD47⁺/CD38^–) cells, which indicated that the CD38 arm was the major contributor to ADCC ( Figure 9C ). No major ADCC was detected in normal cells expressing CD47, suggesting that the BsAbs exhibited less on-target, off-tumor toxicity. The evaluation of phagocytosis was carried out in the monocytic cell line THP-1 (effector) against the target cell lines Raji and NCI-H929 labeled with CFSE by measuring the THP-1 green fluorescence after 2 hours of incubation with CD38/CD47 BsAbs. The BsAbs also exhibited strong ADCP activity against the Raji and NCI-H929 cell lines in a concentration-dependent manner and showed ADCP activity similar to that exhibited by the IMM5605-ISA and IMM5605-DARA controls ( Figure 9D ). We assessed the CDC effects of CD38/CD47 BsAbs against Daudi cells and found that 12C10, 35G5, and 26B4 exhibited relatively weak CDC effects ( Figure 9E ). The ability of CD38/CD47 BsAbs to directly induce apoptosis in the Daudi cell line was evaluated by flow cytometry assays, and the results showed that 12C10 exhibited the strongest ability to induce apoptosis ( Figure 9F ).

We next assessed the in vivo antitumor killing efficacy of BsAbs targeting CD47 and CD38 in a mouse xenograft model using CB17-SCID mice implanted with the NCI-H929 multiple myeloma cell line. IMM5605-ISA was significantly more efficacious in controlling tumor growth, with a D21 tumor growth inhibition (TGI) of 94.07%, than was the control. IMM01, isatuximab, and their combination had partial effects on tumor growth, with final TGIs of 72.12%, 86.52%, and 86.88%, respectively ( Figures 10A, B ). Next, we evaluated the therapeutic effects of IMM01 (SIRPαD1 Fc fusion protein, IgG1) or IMM0120 (SIRPαD1 Fc fusion protein, IgG4) combined with the anti-CD38 antibody 26B4 in an NCI-H929 xenograft model. Tumor growth was monitored beyond the end of the treatment period, and the results indicated that the combination treatment decreased tumor growth for a longer period than any single-drug treatment ( Figures 10C, D ). These results confirmed that co-targeting CD47 and CD38 greatly enhanced antitumor efficacy. Based on these findings, we designed BsAbs targeting CD47 and CD38 using the 2 + 2 “mAb trap” platform and evaluated the effectiveness of the BsAbs through in vitro and in vivo experiments. The NCI-H929 tumor xenograft model was used to evaluate the efficacy of the CD38/CD47 BsAbs, including IMM5605-ISA, IMM5605-DARA, IMM5605–4A8, 12C10, 26B4, 35G5, and 65A7. All CD38/CD47 BsAbs (1 mg/kg, D0/D3, IP) showed high antitumor efficacy, among which IMM5605–12C10 exhibited the highest complete response rate and eradicated all tumors ( Figures 11A–C ).