Targeting BCMA in multiple myeloma with a trifunctional NK cell engager

TL;DR

A new trifunctional NK cell engager called SAR’514 shows strong anti-tumor effects against multiple myeloma by targeting BCMA with minimal side effects.

Contribution

SAR’514 is a novel trifunctional NK cell engager optimized for BCMA-targeted therapy in multiple myeloma with superior efficacy and low cytokine release.

Findings

SAR’514 demonstrates potent and selective anti-tumor activity in vitro and in vivo.

It induces minimal cytokine release compared to T cell engagers targeting the same antigen.

SAR’514 efficiently activates NK cells from multiple myeloma patients and kills resistant malignant cells ex vivo.

Abstract

Multiple myeloma (MM), the second most common hematologic malignancy, remains incurable, highlighting the need for durable therapies. Natural killer (NK) cell engagers (NKCEs) represent a promising alternative to T cell therapies, offering potent anti-tumor activity with limited cytokine release. SAR445514 (SAR’514) is a trifunctional NKCE that co-engages NKp46 and FcγRIIIa to activate NK cells while targeting B cell maturation antigen (BCMA) on MM cells. After exploring several molecular formats with varied BCMA and NKp46 valency, we selected SAR’514, a monovalent format with enhanced antibody-dependent cellular cytotoxicity (ADCC). SAR’514 has potent and selective anti-tumor activity in vitro and in vivo, outperforming other FcγRIIIa-immune cell engagers, while inducing minimal cytokine release compared to T cell engagers targeting the same antigen. Ex vivo, SAR’514 efficiently activates NK cells from MM patients and induces cytotoxicity against autologous malignant cells, including those resistant to standard therapies. These findings support further development of SAR’514.

•SAR’514 targets BCMA in myeloma and CD16/NKp46 on NK cells via trifunctional design•Various formats and valencies have been evaluated to optimize anti-tumor activity•Monovalency for each binder shows superior efficacy against BCMA+ myeloma cells•SAR’514 outperforms standard-of-care therapies in patient samples ex vivo

SAR’514 targets BCMA in myeloma and CD16/NKp46 on NK cells via trifunctional design

Various formats and valencies have been evaluated to optimize anti-tumor activity

Monovalency for each binder shows superior efficacy against BCMA+ myeloma cells

SAR’514 outperforms standard-of-care therapies in patient samples ex vivo

Tang et al. describe SAR’514, an NK cell engager targeting BCMA in multiple myeloma. This trifunctional molecule demonstrates potent anti-tumor activity in vitro and in vivo with minimal cytokine release and efficiently activates patient NK cells against myeloma cells, including those not responding to daratumumab.

Figures

ADCC-optimized BCMA-NKp46-FcγRIIIa NKCE drives anti-tumor activity *in vitro* and *in vivo*(A) Comparison of cytotoxicity of BCMA-NKp46-FcγRIIIa NKCE molecules engineered for enhanced ADCC (-ADE) vs. non-engineered (wild-type Fc) counterparts. RPMI 8226 MM cells were used as target, and purified resting NK cells served as effector. Data from one representative NK cell donor is shown (*n* = 13).(B) EC_50_ and maximum cytotoxic activity of F25 (*N* = 21), F25-ADE (*N* = 13), CODV-1:1 (*N* = 29), CODV-1:1-ADE (*n* = 14), F33 (*n* = 15), F33-ADE (*n* = 13), CODV-2:2 (*n* = 23), and CODV-2:2-ADE (*

SAR’514 outperforms FcγRIIIa-engager *in vitro* and mediates dose-dependent anti-MM activity *in vivo*(A) Comparison of cytotoxicity of BCMA-NKp46-FcγRIIIa NKCE (CODV-1:1-ADE; red) and FcγRIIIa-based NK cell engager molecule targeting BCMA (FcγRIIIa-engager-tool; blue). RPMI 8226 and MM.1R cells were used as targets, with purified resting NK cells as effectors. Data from two representative NK donors out of *n* = 10 (RPMI 8226) and *n* = 6 (MM.1R) are shown.(B) EC_50_ and maximum cytotoxicity activity of BCMA-NKp46-FcγRIIIa NKCE (CODV-1:1-ADE; red) and FcγRIIIa-engager (blue) against RPMI 8226

BCMA-NKp46-FcγRIIIa NKCE promotes NK cell activation and MM cell killing *ex vivo*(A) Cytotoxicity of MM patient-derived PBMCs (*n* = *13*) against Karpas 620 MM cells (100:1 PBMC-to-target ratio). Cells were treated with dose range of BCMA-NKp46-FcγRIIIa NKCE (0.0005–80 nM, white circles) or IC-NKp46-FcγRIIIa NKCE (80 nM, black squares).(B) Karpas 620 cell death and MM patient NK cell activation in response to BCMA-NKp46-FcγRIIIa NKCE treatment (80 nM). Resting PBMCs from MM patients were co-cultured with Karpas 620 cells and treated with BCMA-NKp46-FcγRIIIa NKCE, IC-NKp46-FcγRIIIa NKCE, dara

SAR’514 promotes autologous NK cell activation and MM cell killing *ex vivo*(A) NK cell activation and myeloma cell killing in BM samples from 13 MM patients at the time of diagnosis (*n* = 3) or relapse/progression (*n* = 10). BM mononuclear cells were treated with SAR’514 (80 nM) and analyzed by flow cytometry after 18 h. Untreated samples served as controls. (Left) Percent myeloma cell death. (Right) Percent CD69-positive cells among CD3^−^ CD56^+^ NK cells. (Paired *t* test, two-tailed; ∗∗*p* < 0.01, ∗∗∗*p* < 0.0005).(B and C) NK cell activity against autologous MM cells of PCL patients (*

Full text

Introduction

Multiple myeloma (MM) is the second most common hematologic cancer in Western countries, accounting for 10% of all blood cancers. It is defined by the uncontrolled proliferation of abnormal plasma cells in the bone marrow (BM) and the secretion of monoclonal M proteins, leading to bone damage, kidney dysfunction, anemia, hypercalcemia, and infections.1^,^2 The global incidence is 1.78 per 100,000, with a higher prevalence among older men, especially in high-income countries.2^,^3 In the United States, the 2024 incidence rate was 7.3 per 100,000, with 36,110 new cases and 12,030 expected deaths, and a 5-year survival rate of 62.4%.4 MM remains incurable5^,^6 despite significant therapeutic advances. Modern treatment regimens combining proteasome inhibitors (PIs), immunomodulatory drugs (IMiDs),6 and anti-CD38 monoclonal antibodies have markedly improved survival.7 However, most patients eventually relapse, and outcomes remain poor for those who become refractory to these three classes of agents. This has driven the rapid development of immunotherapies targeting novel plasma cell-restricted antigens, among which B cell maturation antigen (BCMA) has become a cornerstone of current myeloma therapy.

BCMA, a member of the tumor necrosis factor receptor superfamily expressed almost exclusively on plasma cells, plays a critical role in their survival through interaction with a proliferation-inducing ligand (APRIL) and B cell-activating factor (BAFF).8 Its restricted expression and essential biological function have made it an ideal therapeutic target. As of 2025, six BCMA-directed agents—including three bispecific antibodies (teclistamab, elranatamab, and linvoseltamab), two CAR-T products (idecabtagene vicleucel and ciltacabtagene autoleucel), and one antibody-drug conjugate (belantamab mafodotin)—are approved for the treatment of MM, with several moving into earlier lines of therapy. These agents have established BCMA-directed immunotherapy as a core treatment component alongside PIs, IMiDs, and anti-CD38 antibodies. The key challenge ahead is to refine and expand BCMA-based strategies to enhance efficacy, durability, and accessibility while minimizing toxicity. Despite their efficacy, current T-cell-redirecting BCMA therapies are limited by cytokine release syndrome (CRS), neurotoxicity, prolonged cytopenias, and, for CAR-T cells, complex manufacturing and limited persistence. Moreover, T cell exhaustion and dysfunction represent a significant barrier to sustained efficacy, leading to poor response or disease progression with T cell-engaging therapies.9^,^10 These constraints highlight the need for alternative immune effector platforms capable of targeting BCMA with high potency but improved safety and scalability.

Natural killer (NK) cells represent a promising effector population in this context. They mediate antibody-dependent cellular cytotoxicity (ADCC) and contribute to the activity of several standard MM therapies.11^,^12^,^13^,^14 Unlike T cells, NK cells do not require antigen priming and can mediate potent cytotoxicity while maintaining a favorable safety profile.15 Harnessing NK cells through multispecific antibodies—NK cell engagers (NKCEs)—offers an innovative approach to redirect innate immunity toward tumor cells without the risk of severe cytokine storms.15 Recent advances in NKCE engineering have enabled the design of trifunctional molecules that co-engage NK cell-activating receptors such as NKp46 and FcγRIIIa while targeting a tumor antigen.16^,^17 These constructs have demonstrated potent antitumor activity and a favorable safety profile in preclinical and early clinical studies. Given the central role of BCMA in myeloma biology and the proven therapeutic value of its targeting, BCMA-directed NKCEs represent a logical next step in expanding immune engagement beyond T cells.

Here, we describe SAR445514 (SAR’514), a trifunctional NK cell engager that co-targets NKp46 and FcγRIIIa on NK cells and BCMA on myeloma cells. SAR’514 was designed to optimize both NK cell activation and ADCC, providing potent antitumor activity with minimal cytokine release. Through systematic exploration of molecular valency and Fc engineering, we identified a monovalent 1:1 configuration with an optimized Fc domain as the most effective format. Preclinical evaluation of SAR’514 demonstrates robust efficacy across in vitro, in vivo, and ex vivo models.

Results

Anti-BCMA antibody CA10_V7 cross-reacts on cynomolgus monkey and blocks APRIL

To develop an optimal BCMA binder for potential use in an NKCE format targeting MM, humanized Trianni mouse platform were immunized with the recombinant extracellular domain of human BCMA. The resulting anti-human BCMA antibodies were screened for physicochemical properties and cross-reactivity to cynomolgus monkey BCMA. An antibody, named CA10, emerged as the lead candidate based on its in vitro characteristics and was humanized via CDR grafting (Figure S1). The top humanized variant, CA10_V7, exhibited high-affinity binding to both human and cynomolgus monkey BCMA, with sub-nanomolar monovalent KD values of 0.2 ± 0.1 nM and 0.3 ± 0.1 nM, respectively (Figure S1A). Given the important role of BCMA in MM survival and growth via activation by its cognate ligands APRIL and BAFF,18^,^19^,^20 the ability of CA10_V7 to block the binding of APRIL (high-affinity ligand) was evaluated by SPR competition assay (Figure S1B). The results showed that CA10_V7 efficiently inhibited APRIL binding to BCMA, suggesting that a BCMA-NKp46-FcγRIIIa NKCE molecule incorporating CA10_V7 could not only kill myeloma cells but also disrupt BCMA-driven tumor survival.

Further structural characterization of CA10_V7 provided insight into its mechanism of action. X-ray crystallographic analysis of the CA10_V7 Fab-BCMA complex at 3.18 Å resolution (Figures S1C and S1D) revealed that CA10_V7’s epitope overlaps with the APRIL-binding site in the BCMA-APRIL complex (PDB: 1XU2). This structural evidence demonstrated the ability to compete with APRIL for BCMA binding. These findings reinforced the potential of the antibody as a dual-function therapeutic antibody capable of enhancing NK cell-mediated cytotoxicity against BCMA-positive tumor cells, while blocking APRIL-BCMA binding, which may interfere with downstream signaling.

1:1 BCMA-NKCE format promotes anti-tumor activity in vitro and in vivo

The NKCE technology we have developed and previously described17 is designed to recruit NK cells by co-engaging NKp46 and FcγRIIIa. This approach relies on molecular formats that engage the different target entities, including activating receptors and the tumor-associated antigen (TAA), in a monovalent manner (Figure 1A). To explore alternative molecular configurations and valencies for BCMA targeting in MM treatment, we developed and evaluated various NKCE formats. Bivalent formats targeting BCMA and/or the NKp46 receptor were designed based on two distinct molecular architectures: Antibody-based NK cell engager therapeutics (ANKET)17^,^21 and cross-over dual variable Ig-like format (CODV).22^,^23 Molecules such as CODV-2:1, which binds bivalently to the tumor-associated antigen BCMA and monovalently to NKp46, and CODV-2:2 and F33, which bind bivalently to both BCMA and NKp46, were generated (Figure S1E). These bivalent formats were compared to monovalent trifunctional NKCE molecules CODV-1:1 and F25, which engage both BCMA and NKp46 in a monovalent manner. All five BCMA-NKp46-FcγRIIIa NKCE formats incorporated a functional human IgG1 Fc domain, enabling co-engagement of NKp46 and FcγRIIIa on NK cells. These constructs were based on the anti-NKp46 antibody 3D9, previously validated for the clinical development of NKCE technology.24^,^25^,^26 In the context of CODV-1:1 architecture, 3D9 antibody binds to human and cynomolgus monkey NKp46 with similar KD values of 12.5 ± 0.7 nM and 16.9 ± 0.8 nM, respectively (Figure S1A). The cytotoxic activity of these molecules was first evaluated in vitro using a ^51^Cr-release assay with purified NK cells from four healthy donors against two human myeloma cell lines (HMCLs), RPMI 8226 and MM1.R (Figure 1B; Figure S2A). In selected experiments, a regular human immunoglobulin G1 (IgG1) antibody with the same anti-BCMA binder was used as a positive control for ADCC. CODV-2:1, which binds bivalently to BCMA, did not exhibit superior cytotoxicity compared to the monovalent formats CODV-1:1 and F25, suggesting that increased tumor antigen avidity alone does not enhance NK cell-mediated killing. However, all BCMA-NKp46-FcγRIIIa NKCE formats demonstrated picomolar EC_50_ values that were 100- to 1,000-fold lower than that of the anti-BCMA IgG1 antibody (Figure 1B; Figure S2B), confirming the superior potency of NKCE constructs over conventional IgGs. Notably, bivalent constructs for both BCMA and NKp46 (CODV-2:2 and F33) showed similar potency compared to their 2:1 and 1:1 counterparts (Figure 1C; Figure S2C). This finding suggested that co-engagement of NKp46 and FcγRIIIa plays a more significant role in NKCE potency than increased antigen avidity alone. We further compared monovalent (1:1) and bivalent (2:2) binders, containing one or two arms targeting BCMA and NKp46, respectively, in a syngeneic recombination activating gene 1 (RAG^−/−^)-immunodeficient mouse model expressing human NKp46 on NK cells (huNKp46-Tg x Rag1^−/−^ mice)21^,^27 (Figures 1D and 1E). Murine T cell lymphoma RMA cells transduced to express human BCMA and a dsRED-tracking protein served as targets, while RMA cells expressing GFP were used as antigen-negative controls. BCMA-positive (red) and BCMA-negative (green) RMA cells were mixed in a 3:2 ratio and intravenously engrafted into huNKp46-Tg x Rag1^−/−^ mice (Figure 1D). Mice were treated with either monovalent 1:1 (F25 or CODV-1:1) or bivalent 2:2 constructs (F33 or CODV-2:2) at the time of RMA engraftment, and RMA cells in liver biopsies were analyzed by flow cytometry 48 h post-treatment (Figure 1E). All BCMA-targeting NKCE molecules selectively depleted BCMA-positive RMA cells with comparable efficacy at a low dose of 12.5 pmol/mouse, resulting in near-complete elimination in the liver and a more than two-log reduction in the red-to-green cell ratio (p < 0.0001). Although no statistically significant difference was observed between the 1:1 and 2:2 formats, the 1:1 molecules consistently exhibited a trend toward superior efficacy across all tested doses (Figure S3A). This trend cannot be explained simply by differences in the pharmacokinetic (PK) properties of the two molecules, since they showed a similar PK profile in huFcRn tg32 mice (Figure S3B).Figure 11:1 BCMA-NKp46-FcγRIIIa NKCE drives anti-tumor activity in vitro and in vivo(A) Schematic of BCMA-NKp46-FcγRIIIa NKCE mode of action. The molecule binds to BCMA on MM cells and recruits/activates NK cells by co-engaging NKp46 and FcγRIIIa, leading to MM cell killing and/or cytokine production.(B) (Left) Cytotoxicity comparison of F25, CODV-1:1, CODV-2:1, BCMA-IgG1, and IC-NKp46-FcγRIIIa NKCE control (IC-F25). RPMI 8226 and MM.1S MM cells were used as targets, with purified resting NK cells from 4 healthy donors as effectors. (Right) EC_50_ and maximum cytotoxic activity of BCMA-IgG1 and BCMA-NKp46-FcγRIIIa NKCE against RPMI 8226 and MM.1S MM cells. Delta maximum lysis (ΔMax lysis)—defined as percent maximum lysis minus background lysis of the isotype control (IC-F25) at corresponding concentration—and EC_50s_ were monitored from dose-response curves after 4-h incubation, plotted separately for all MM cell lines and NK donors (n = 4). ANOVA/Tukey’s multiple comparisons test ∗p < 0.05, ∗∗∗p < 0.0005.(C) Comparison of cytotoxicity between 1:1 (F25, CODV-1:1) and 2:2 (F33, CODV-2:2) BCMA-NKp46-FcγRIIIa NKCE formats with the same setting as in (B). Two representative NK donors are shown (n = 17).(D) (Left) Flow cytometry analysis of fluorescent RMA cells expressing human BCMA (dsRed) or lacking BCMA expression (eGFP), showing BCMA expression only on dsRed cells. (Middle) Experimental setup. A 2:3 mixture of eGFP (BCMA-negative) and dsRed (BCMA-positive) RMA cells was intravenously injected into huNKp46-Tg x Rag1^−/−^ transgenic mice. Tumor-bearing mice (n = 27–28 per group) received a single dose of 12.5 pmol/animal (0.078 mg/kg) of BCMA-NKp46-FcγRIIIa NKCEs or vehicle control. Liver biopsies were collected 48 h post-treatment, and infiltrating RMA cells quantified by flow cytometry. (Right) Flow cytometry analysis of BCMA expression on dsRed RMA cells pre-engraftment (in vitro, red histogram) and post-engraftment in liver biopsies (gray histogram).(E) (Left) Absolute count of liver-infiltrated RMA cells. (Right) dsRed/eGFP cell ratio in liver biopsies, analyzed by flow cytometry 48 h after treatment. Statistical analysis: Mann-Whitney test (ns, p > 0.05; ∗p < 0.01).(F) Kaplan-Meier survival curves for huNKp46-Tg x Rag1−/− mice bearing disseminated EL4-huBCMA tumors treated with BCMA-NKp46-FcγRIIIa NKCEs lacking Fc optimization for enhanced ADCC. Treatment groups (red) received 5, 0.5, or 0.05 mg/kg of BCMA-NKp46-FcγRIIIa NKCE, compared to the isotype control NKCE (gray, 5 mg/kg) and vehicle (black). Statistical analysis: log rank (Mantel-Cox) test (p < 0.05, ∗p = 0.0398, ∗∗p = 0.0082, ∗∗∗p = 0.0001, ∗∗∗∗p < 0.0001).See also Figures S1–S5.

We further confirmed the anti-tumor potential of the CODV-1:1 BCMA-NKp46-FcγRIIIa NKCE in a long-term disseminated EL4-huBCMA tumor model using again huNKp46-Tg x Rag1^−/−^ mice (Figure 1F). Treatment with CODV-1:1 molecule controlled the EL4-huBCMA tumor growth in a dose-dependent manner. The lowest dose of 0.05 mg/kg (8 pmol/mouse) resulted in a median survival of 44 days, compared to 19 days for the isotype control at 5 mg/kg. At 0.5 mg/kg (80 pmol/mouse), 70% of mice were rescued, defining the plateau activity in this model. We also compared the BCMA-NKp46-FcγRIIIa CODV-1:1 molecule to an anti-BCMA IgG1 antibody and a CODV-1:1 molecule in which binding to Nkp46 was disabled (BCMA-IC-FcγRIIIa) (Figure S3C). We confirmed that co-engaging NKp46 and FcγRIIIa on NK cells led to a better and significant control of the EL4-huBCMA tumor growth than the bivalent IgG1 targeting the same antigen. Moreover, the monovalent CODV-1:1 molecule BCMA-IC-FcγRIIIa, engaging only FcγRIII on mouse NK cells, tended to be less efficient than BCMA-NKp46-FcγRIIIa NKCE, confirming with the same format architecture the advantages of the cis co-engagement of NKp46 and FcγRIIIa on NK cells as already described.17^,^24

Altogether, these data provided strong rationale for the further development and optimization of BCMA-NKp46-FcγRIIIa NKCE molecules in the CODV-1:1 format.

BCMA-NKp46-FcγRIIIa NKCE optimized for ADCC promotes anti-tumor activity in vitro and in vivo

To enhance NK cell receptor engagement and maximize the cytotoxic potential of NKCEs, we engineered the Fc domain of monovalent CODV-1:1 construct using clinically validated strategies aimed at improving therapeutic antibody efficacy.28^,^29 Specifically, we optimized the affinity for FcγRIIIa through afucosylation or ADE or DE mutations, either alone or in combination with disulfide bond (DSB) engineering to optimize manufacturing (Figures S4A and S4B). All engineering strategies tested enhanced the killing potency of CODV-1:1, yielding comparable geometric mean EC_50_ values with 2.39 pM (95% confidence interval [CI]: 1.16–4.95 pM) for afucosylated molecules, 1.51 pM (95% CI: 0.88–2.59 pM) for DE mutants, and 2.28 pM (95% CI: 1.74–2.97 pM) for ADE mutants. These values represent a significant improvement over the non-engineered molecule, which had an EC_50_ of 24.06 pM (95% CI: 18.30–31.63 pM). The average ΔMax lysis levels observed for afucosylated (46.1 ± 8.0%), DE-mutated (45.6 ± 9.0%), and ADE-mutated (47.9 ± 8.4%) molecules were not significantly different from the non-engineered reference (39.2 ± 8.0%), suggesting that the engineering strategies primarily enhanced potency rather than overall efficacy. Building on these findings, we further investigated the ADE mutations in both the monovalent (1:1) and bivalent (2:2) NKCE formats (Figures 2A and 2B). As previously reported,17 Fc engineering significantly improved ADCC potency. CODV-1:1-ADE and F25-ADE exhibited lower geometric mean EC_50_ values of 1.3 pM (95% CI: 0.8–2.1 pM) and 1.1 pM (95% CI: 0.5–2.7 pM), respectively, compared to 15.9 pM (95% CI: 13.1–19.3 pM) and 14.5 pM (95% CI: 10.7–19.7 pM) for non-optimized CODV-1:1 and F25 formats in the 4-h cytotoxic assay (Figures 2A and 2B, upper). Monovalent CODV-1:1 and F25 formats, with regular or optimized Fc, demonstrated comparable efficacy in cytotoxicity assays. This suggested that NKp46-targeting NKCEs can accommodate various binding geometries while maintaining optimal NK cell activation potential (Figure 2A, upper, and 2B, lower). Surprisingly, the significant potency enhancement observed with Fc engineering in the 1:1 configuration was not observed in the bivalent 2:2 format, as reflected in similar EC_50_ values for CODV-2:2-ADE (1.8 pM, 95% CI: 1.1–3.1 pM), F33-ADE (1.9 pM, 95% CI: 0.9–3.7 pM), CODV-2:2 (2.9 pM, 95% CI: 2.1–4.1 pM), and F33 (2.9 pM, 95% CI: 1.7–5.1 pM) (Figure 2A, lower, and 2B, upper). Overall, the 1:1 formats (F25, F25-ADE, CODV-1:1, and CODV-1:1-ADE) exhibited similar maximum lysis efficacy in vitro, with average values of 41.8% ± 7.5%, 50.2% ± 9.4%, 39.7% ± 8.4%, and 44.7% ± 11.8%, respectively. In contrast, the 2:2 formats (F33 or CODV) showed a tendency for lower efficacy in 4-h cytotoxicity assay, as indicated by their maximum lysis. The introduction of ADE mutations in the 2:2 format led to a slight but significant increase in efficacy, although it never surpassed the efficacy observed in the 1:1-ADE format (Figures 2A and 2B, lower).Figure 2ADCC-optimized BCMA-NKp46-FcγRIIIa NKCE drives anti-tumor activity in vitro and in vivo(A) Comparison of cytotoxicity of BCMA-NKp46-FcγRIIIa NKCE molecules engineered for enhanced ADCC (-ADE) vs. non-engineered (wild-type Fc) counterparts. RPMI 8226 MM cells were used as target, and purified resting NK cells served as effector. Data from one representative NK cell donor is shown (n = 13).(B) EC_50_ and maximum cytotoxic activity of F25 (N = 21), F25-ADE (N = 13), CODV-1:1 (N = 29), CODV-1:1-ADE (n = 14), F33 (n = 15), F33-ADE (n = 13), CODV-2:2 (n = 23), and CODV-2:2-ADE (n = 13) molecules against RPMI 8226 MM cells. ΔMax lysis and EC_50s_ were determined from dose-response curves and plotted separately for each NK donor. Statistical analysis: ANOVA/Tukey’s multiple comparisons test (∗p < 0.005; ∗∗∗p < 0.0001).(C) Kaplan-Meier survival curves for huFcγR-Tg mice bearing disseminated EL4-huBCMA tumors and treated with surrogate BCMA-moNKp46-FcγRIIIa NKCE lacking ADCC optimization (wild-type Fc) (CODV-1:1, pink) or ADCC-optimized BCMA-moNKp46-FcγRIIIa NKCE (CODV-1:1-ADE, blue) at 0.5 mg/kg. Statistical analysis: log rank (Mantel-Cox) test (∗∗p < 0.001).See also Figures S2 and S4.

Based on these results, we conducted in vivo investigations to compare the efficacy of a mouse NKp46 surrogate CODV 1:1, with or without optimized Fc, in a disseminated EL4-huBCMA tumor model using transgenic mice expressing human FcγRs (huFcγR-Tg mice)30 (Figure 2C). The efficacy of CODV-1:1, with or without optimized Fc, was assessed at a dose of 0.5 mg/kg, which partially controlled the growth of EL4-huBCMA cells in the huNKp46-Tg x Rag1^−/−^ model (Figure 1F). CODV 1:1 (without optimized Fc) at 0.5 mg/kg slightly improved survival compared to the isotype control, with median survival of 60 days compared to 44 days. In contrast, treatment with optimized Fc-bearing molecule at the same dose significantly improved survival, with nearly all treated mice surviving (9 of 10).

This result highlighted the superior clinical potential of Fc engineering over the non-mutated format. Taken together, these findings demonstrate that bivalent binding to the target antigen does not further enhance the potency and efficacy of ADCC-optimized molecules. Instead, the mode of NK cell activation via co-engagement of FcγRIIIa and NKp46 plays a critical role in determining killing efficacy. The BCMA-NKp46-FcγRIIIa NKCE CODV-1:1-ADE-DSB format shows sustained killing activity over time, as assessed in cytotoxicity assays over periods of up to 21 h (Figure S4C), and binds to human FcγRIIIa with KD of 13 ± 3 nM and 23 ± 6 nM for the V158 and F158 alleles, respectively (Figure S4D). The BCMA-NKp46-FcγRIIIa NKCE CODV-1:1-ADE-DSB molecule was subsequently selected for further development and is hereafter referred to as SAR’514.

SAR’514 induces NK cell-mediated killing of BCMA-positive tumor cells in vitro and in vivo

We further investigated the cytotoxic activity of SAR’514 in relation to BCMA expression on MM cells (Figure S5). The median BCMA expression on patient samples is approximately 1,500 sites per MM cell.31 To evaluate the impact of BCMA expression on SAR’514 efficacy, we quantified BCMA surface expression across several HMCLs (Figure S5A) and assessed SAR’514-induced tumor cell killing using a representative panel of cell lines with varying BCMA expression levels (Figure S5B). This panel included RPMI 8226 and L-363, which exhibit mean BCMA densities comparable to the average observed in patient samples. These assays were conducted using resting NK cells purified from healthy donors (Figures S5C–S5E). Notably, with the exception of HUT78 cells—which lack BCMA expression (Figure S5B) and were consequently not targeted by SAR’514 (Figure S5D)—all other HMCLs were efficiently lysed by SAR’514-activated NK cells, regardless of BCMA expression levels (Figure S5C). HUT78 cells were effectively killed in the presence of an anti-KIR3DL2 antibody32 (Figure S5D), confirming that their resistance to SAR’514 was due to the absence of BCMA and not due to intrinsic insensitivity to NK cell-mediated lysis. Furthermore, quantitative cytotoxicity assays revealed that SAR’514’s killing potency and efficacy were not correlated with BCMA expression density on target cells (Figure S5E). These findings demonstrated that SAR’514 effectively enhances NK cell-mediated lysis of HMCLs, with potency and efficacy independent of BCMA density, at least within the examined expression range, further supporting its broad therapeutic potential.

We next compared the efficacy and potency of SAR’514 to an FcγRIIIa-engaging tool targeting BCMA, derived from a clinically evaluated molecule (Figures 3A and 3B).33 A ^51^Cr-release assay using purified NK cells from 10 and 6 healthy donors against RPMI 8226 and MM1.R cells, respectively, demonstrated that both molecules achieved comparable ΔMax lysis efficacy. However, SAR’514 exhibited superior potency over the FcγRIIIa engager, with a geometric mean EC_50_ of 4.5 pM (95% CI: 2.2–9.1 pM) for RPMI 8226 compared to 48.0 pM (95% CI: 25.1–91.7 pM) for the FcγRIIIa-engager and a geometric mean EC_50_ of 2.6 pM (95% CI: 1.3–5.4 pM) for MM1.R, compared to 37.8 pM (95% CI: 18.6–76.8 pM) for the FcγRIIIa-engager (Figures 3A and 3B).Figure 3SAR’514 outperforms FcγRIIIa-engager in vitro and mediates dose-dependent anti-MM activity in vivo(A) Comparison of cytotoxicity of BCMA-NKp46-FcγRIIIa NKCE (CODV-1:1-ADE; red) and FcγRIIIa-based NK cell engager molecule targeting BCMA (FcγRIIIa-engager-tool; blue). RPMI 8226 and MM.1R cells were used as targets, with purified resting NK cells as effectors. Data from two representative NK donors out of n = 10 (RPMI 8226) and n = 6 (MM.1R) are shown.(B) EC_50_ and maximum cytotoxicity activity of BCMA-NKp46-FcγRIIIa NKCE (CODV-1:1-ADE; red) and FcγRIIIa-engager (blue) against RPMI 8226 and MM.1R cells. Delta maximum lysis (Δ Max lysis) and EC_50s_ were determined from dose-response curves and plotted for each HMCL-NK donor pair (n = 10 for RPMI 8226, n = 6 for MM.1R). Paired t test, two-tailed; ∗∗p ≤ 0.01, ∗p ≤ 0.05.(C) (Upper) Experimental setup. Human NK cells were purified and amplified in vitro for 14 days in the presence of K562 cells engineered to express CD86 and 4-1BB ligand, IL-15 (50 U/mL), and IL-21 (100 U/mL). Expanded NK cells were adoptively transferred into irradiated NOG-IL-15-Tg mice (n = 10 per group) 7 days before MM1.R HMCL engraftment (day 0). Mice were treated once on day 1 with BCMA-NKp46-FcγRIIIa NKCE at doses of 0.05, 0.5, 2.5, 5, and 10 mg/kg, or with the IC-NKp46-FcγRIIIa NKCE control molecule at 5 mg/kg. (Lower) Kaplan-Meier survival curves of treated mice. Endpoint significance was calculated in a log rank (Mantle-Cox) test. n = 10/group. ∗p < 0.05, ∗∗∗∗p < 0.0001.See also Figures S5 and S6.

Importantly, SAR’514 induced minimal to no cytokine release at active concentrations in human whole blood from 11 healthy donors loaded with RPMI 8226-RFP cells as compared to a BCMA-CD3-TCE tool included in the study as a positive control (Figure S6).

To further assess the therapeutic efficacy of SAR’514, we evaluated its anti-tumor activity in immunodeficient NOG mice expressing human IL-15 (NOG-IL-15-Tg mice) engrafted with MM1.R cells and adoptively transferred with amplified human NK cells from healthy donors.34^,^35 A dose-dependent response was observed, with median survival of 57 and 111 days at SAR’514 doses of 0.05 and 0.5 mg/kg compared to 45 days for the isotype control. Near-complete rescue of mice was achieved at 2.5, 5, and 10 mg/kg, demonstrating the potent therapeutic efficacy of SAR’514 in this in vivo humanized model (Figure 3C).

SAR’514 outperforms daratumumab in NK cell activation and MM cell killing ex vivo

We performed immunophenotyping of NK cells to analyze their phenotype in peripheral blood and BM from 4 newly diagnosed MM patients, 15 relapsed MM patients, and one patient (MM#0282) with plasma cell leukemia (PCL), an aggressive form of PC disorder characterized by the presence of malignant cells in the periphery and poor prognosis.36 The relapsed patients were categorized into two groups based on prior treatments: those who had never received anti-CD38 mAbs (referred to as “anti-CD38 naive”) and those who had received anti-CD38 mAbs, daratumumab or isatuximab, in previous lines (referred to as “not anti-CD38 naive”) (Figure S7). Notably, NK cells from both peripheral blood and BM consistently expressed NKp46, one of the two activating receptors targeted by SAR’514. However, in 3 of 20 patients, the frequency of FcγRIIIa^+^ NK cells in the BM was reduced compared to peripheral blood levels. In contrast, no significant differences were observed between peripheral blood and BM for other activating receptors, including NKp30, NKG2D, DNAM-1, and 2B4 (Figure S7).

We next assessed the ability of SAR’514 to promote NK cell-mediated killing of Karpas 620 HMCL cells using NK cells derived from 13 MM patients. SAR’514 was compared to the standard-of-care (SoC) anti-CD38 mAb daratumumab, clinically approved for MM treatment. Due to the co-expression of BCMA and CD20 on Karpas 620 cells,37 obinutuzumab, an anti-CD20 mAb approved for non-Hodgkin’s lymphoma and chronic lymphocytic leukemia, was included as a positive control (Figure 4). In all conditions, SAR’514 induced higher maximum specific lysis on average compared to the approved molecules (Figure 4B) with 60% ± 23% for SAR’514, 43% ± 26% for obinutuzumab, and 50% ± 22% for daratumumab.Figure 4BCMA-NKp46-FcγRIIIa NKCE promotes NK cell activation and MM cell killing ex vivo(A) Cytotoxicity of MM patient-derived PBMCs (n = 13) against Karpas 620 MM cells (100:1 PBMC-to-target ratio). Cells were treated with dose range of BCMA-NKp46-FcγRIIIa NKCE (0.0005–80 nM, white circles) or IC-NKp46-FcγRIIIa NKCE (80 nM, black squares).(B) Karpas 620 cell death and MM patient NK cell activation in response to BCMA-NKp46-FcγRIIIa NKCE treatment (80 nM). Resting PBMCs from MM patients were co-cultured with Karpas 620 cells and treated with BCMA-NKp46-FcγRIIIa NKCE, IC-NKp46-FcγRIIIa NKCE, daratumumab, or obinutuzumab (all at 80 nM). NK cell activation was assessed by flow cytometry, measuring CD107a expression and intracellular production of IFN-γ and MIP-1β. Karpas 620 cell death was determined via flow cytometry based on CD38 and CD138 staining modulation. Data for all MM patients (n = 13) are shown, with individual patient values corresponding to same symbols as in (A). One-way ANOVA Tukey’s multiple comparison test; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Mean (horizontal bar) and error (SD) are indicated on graphs.See also Figures S7 and S8.

Moreover, SAR’514 enhanced NK cell activation, as evidenced by a higher proportion of IFN-γ-producing and cytotoxic CD107a^+^ NK cells compared to daratumumab. Specifically, 19% ± 13% of IFN-γ^+^ NK cells were observed with SAR’514 vs. 7% ± 6% with daratumumab (p = 0.0006; n = 13), and less than 3% for the controls (p = 0.0011). Similarly, 35% ± 10% of CD107a^+^ NK cells were detected following SAR’514 stimulation, compared to 22% ± 7% with daratumumab (p < 0.0001), and less than 9% for the controls (p < 0.0001) (Figure 4B). Additionally, SAR’514 induced higher frequencies of MIP1-β-producing NK cells, with 70% ± 18% of positive cells vs. 63% ± 22% for daratumumab (p = 0.0174) and 56% ± 23% for the isotype control (p = 0.0136).

Importantly, CD38 is highly expressed on MM cells but also present on NK cells, in contrast to BCMA (Figure S8A). Consequently, daratumumab infusion in patients can lead to NK cell depletion.38^,^39 In contrast, while CD38-targeting NKCEs induced fratricidal NK cell killing in vitro, BCMA-targeting NKCEs, including SAR’514 and other Fc-optimized molecules, did not (Figure S8B).

Taken together, these findings demonstrate that SAR’514, an Fc-optimized NKCE targeting BCMA, outperforms the SoC antibody daratumumab in an allogenic setting in vitro, while preserving NK cells. Baseline characteristics of all MM patients involved in the ex vivo study is summarized in Table 1.Table 1MM patient characteristicsSamplesGenderStatusCytogeneticCD38 treatmentBCMA treatmentMM#0286MMMDND––MM#0303MMMDno t(4;14), no t(11;14), no del1p, no 1q––MM#0305MMMDND––MM#0309FMMDno t(4;14), no t(11;14), no del1p, 1q, no del17p––MM#0022MMMRno del1p, 1q, no del17pexposed to daratumumabnaiveMM#0092FMMRNDexposed to daratumumabexposed to CAR-TMM#0139MMMRno t(11;14), no del17pexposed to daratumumabnaiveMM#0182MMMRNDexposed to daratumumabexposed to BCMAxCD3MM#0192FMMRNDexposed to daratumumabnaiveMM#0194MMMRNDexposed to daratumumabnaiveMM#0212MMMRNDexposed to daratumumabnaiveMM#0238MMMRNDexposed to daratumumabnaiveMM#0253MMMRNDexposed to daratumumabnaiveMM#0264FMMRno t(4;14), no t(14;16), no del1p, 1qexposed to daratumumabexposed to BCMA x CD3MM#0277MMMRNDnaivenaiveMM#0285FMMRt(4;14), 1q, del17pexposed to daratumumabnaiveMM#0287MMMRNDexposed to daratumumabnaiveMM#0289FMMRt(4;14), no t(11;14), no del1p, no 1q, del17pnaivenaiveMM#0298FMMRno t(4;14), no del17pnaivenaiveMM#0308FMMRdel1p, no 1q, no del17pexposed to daratumumabnaiveMM#0311FMMRdel17pnaivenaiveMM#0312FMMRno t(4;14), no del17pnaivenaiveMM#0313FMMRNDnaivenaiveMM#0326MMMRno t(4;14), no del17pexposed to isatuximabnaiveMM#0329FMMRno t(4;14), no t(11;14), no t(14;16), no del1p, 1q, no del17pexposed to daratumumabnaiveMM#0330FMMRt(4;14)exposed to daratumumabnaiveMM#0353FMMRno del17pnaivenaiveMM#0359FMMRt(4;14), del17pnaivenaiveMM#0361MMMRno t(14;16), no del1p, no 1q, no del17pnaivenaiveMM#0289FMMRRno t(11;14), no del1p, del17pnaivenaiveMM#0288FMMRRno t(11;14), no del1pexposed to daratumumab, relapse under IsatuximabnaiveMM#0221FPCLRRt(11;14), del17pexposed to daratumumab, progression under CD38xCD3exposed to BCMA-ADCMM#0282MPCLRRNDexposed to daratumumabnaiveADC, antibody drug conjugate; BCMA, B-cell maturation antigen; CART, chimeric antigen receptor T-cell; CD3, cluster of differentiation 3; D, diagnosis; F, female; M, male; MM, multiple myeloma; PCL, plasma cell leukemia; R, relapse; RR, relapse refractory. Related to Figures 4 and 5.

SAR’514 promotes autologous NK cell activation and MM cell killing ex vivo

We further assessed the in situ efficacy of SAR’514 in BM and PB samples from patients diagnosed with MM (n = 13) or PCL (n = 3). SAR’514 consistently induced the activation of NK cells and mediated efficient lysis of MM cells across the majority of patient samples (Figure 5). This included autologous NK cell-mediated cytotoxicity against MM cells derived from BM aspirates of both newly diagnosed (n = 3) and relapsed (n = 10) patients, independent of prior treatment regimens, including therapies targeting CD38 or BCMA (p = 0.0002, n = 13). SAR’514-induced cytolysis was accompanied by a significant upregulation of CD69 expression on NK cells (p = 0.0046, n = 13), indicative of cellular activation (Figure 5A). Furthermore, in PB samples from three patients with PCL (MM#0221, MM#0282, and MM#0289), SAR’514 induced substantial depletion of CD138^+^ MM cells, with elimination rates of 26.2%, 95.4%, and 73.3%, respectively (Figures 5B and 5C). Patient MM#0282, who appeared resistant to daratumumab, exhibited a marked response to SAR’514 treatment. Importantly, SAR’514 elicited robust cytotoxic responses in samples harboring a hallmark of high-risk disease (del17p, 1q and/or t(4;14)), achieving an average of 41% MM cell death across six such cases. The efficacy of SAR’514 in promoting NK cell-mediated cytotoxicity was primarily dependent on the number of effector NK cells in the BM samples, highlighting the crucial role of NK cell abundance in its therapeutic potential (Figure 5D). These findings demonstrated the potent ability of SAR’514 to mobilize NK cells for MM tumor cell killing in patients, regardless of prior treatment with BCMA- or CD38-targeting agents.Figure 5SAR’514 promotes autologous NK cell activation and MM cell killing ex vivo(A) NK cell activation and myeloma cell killing in BM samples from 13 MM patients at the time of diagnosis (n = 3) or relapse/progression (n = 10). BM mononuclear cells were treated with SAR’514 (80 nM) and analyzed by flow cytometry after 18 h. Untreated samples served as controls. (Left) Percent myeloma cell death. (Right) Percent CD69-positive cells among CD3^−^ CD56^+^ NK cells. (Paired t test, two-tailed; ∗∗p < 0.01, ∗∗∗p < 0.0005).(B and C) NK cell activity against autologous MM cells of PCL patients (n = 3). PBMCs from patients with PCL were treated with SAR’514 (80 nM, black histogram), daratumumab (140 nM, gray histogram), or isotype control NKCE (80 nM, white histogram), and death of MM cells was assessed by flow cytometry after 18 h. Myeloma cells were identified as CD3^−^ CD138^+^ cells. (B) CD138 and CD3 staining of PBMCs and gating on CD3^−^ CD138^+^ myeloma cells. Percentages of MM cell population among PBMCs are indicated on graphs in (C). Presence of myeloma cells in samples treated with SAR’514 or daratumumab, expressed as percent relative to control condition (IC-NKp46-FcγRIIIa NKCE).(D) Correlation between SAR’514-mediated myeloma cell killing and effector-to-target (NK:MM) ratio in all patient samples (n = 14). The NK:MM ratio was calculated based on absolute counts of CD138^+^ myeloma cells and NK cells (CD3^−^ CD56^+^) by flow cytometry. Spearman r correlation, p < 0.0001, r = 0.9149. Patients with cytogenetic hallmarks associated with high risk (samples with 17p deletion, del17p, and/or 1q gain, 1q, and/or t(4;14) translocation) are indicated by a star.See also Figure S7.

Discussion

Despite advances in MM treatment, the disease remains incurable, necessitating the continuous development of novel therapies. High-risk MM patients often relapse within 18 months, highlighting the urgency for new strategies. While CAR-T and bispecific antibodies are expanding treatment options, significant gaps remain, particularly in addressing high-risk and refractory MM. CAR-T therapies have shown remarkable efficacy in RRMM, but their use is limited by a personalized, centralized manufacturing model that delays treatment. Recent evaluations of BCMA-CD3 bispecific TCEs (e.g., teclistamab, elranatamab, and linvoseltamab) are associated with significant toxicities, including CRS, neurotoxicity, and infections. There is a critical need for alternative strategies that provide effective and accessible treatment options, particularly for MM. One major challenge is T cell exhaustion, which can arise from the disease itself or as an adverse effect of certain therapies, compromising the efficacy of T-cell-based immunotherapies.40 NK cells play a crucial role in cancer immunotherapy due to their ability to recognize and eliminate malignant cells without prior sensitization. Thus, NK cell-based therapies represent a promising, safe, and versatile alternative to T cell therapies. NKCEs, by selectively targeting NK cell-activating receptors (NKp46 and FcγRIIIa) rather than CD3, hold promise for improved safety profiles and lower manufacturing costs, making them an attractive option for broader clinical implementation. We generated and evaluated SAR’514 as a BCMA-targeting NKp46-FcγRIIIa NKCE. SAR’514 demonstrated potent anti-tumor activity in vitro and in vivo, outperforming other immune cell engagers in multiple aspects. It showed superior cytotoxic potency compared to FcγRIIIa-immune cell engagers, effectively eliminated tumors at moderate antigen expression levels, and exhibited strong in vivo efficacy across multiple models, including humanized NK cell mouse models. Importantly, unlike a TCE targeting the same antigen, SAR’514 induced minimal to no cytokine release in ex vivo assays, suggesting a lower risk of toxicity compared to TCEs and CAR-T therapies.

The development of immune cell engagers has emerged as a promising strategy to mobilize effector cells against tumors. These molecules exist in various molecular formats, offering multiple options for targeting specific antigens and activating immune cell receptors with different valencies. Natural cytotoxic antibodies, such as IgG1, bind antigens in a bivalent manner, ensuring strong avidity for the target while interacting monovalently with Fcγ receptors (FcγRs) on immune effector cells, typically with low affinities (K_D_ in the μM range), except for FcγRI (K_D_ in the nM range).41 The 2:1 engagement format represents a natural model of immune activation. Among innate immune cell engagers, those targeting FcγRIIIa, such as TandAbs (Tandem diabodies), are designed with a bivalent 2:2 molecular arrangement, engaging both FcγRIIIa on NK cells and tumor antigens.42 In contrast, most TCEs in clinical development, including those already approved, use a 1:1 monovalent format, binding CD3 on T cells and tumor antigens.43 NKCEs were initially designed in a 1:1 format, similar to TCEs and have demonstrated potent anti-tumor activity in preclinical models. For example, the CD123-targeting NKp46-FcγRIIIa NKCE (SAR'579) developed for AML follows the same 1:1 structure.24 These NKCEs have demonstrated strong efficacy in preclinical studies, excellent tolerability in patients, and promising signs of clinical activity.44 However, since the cytotoxic activity of immune cell engagers depends on their spatial configuration and valency and considering that each tumor antigen has unique structural characteristics and epitope positioning, optimizing the molecular design of NKCEs for each target remains a critical and complex challenge. Literature suggests that converting TCEs from a 1:1 to a 2:1 format may enhance their cytotoxic potential.16^,^31 In this study, we aimed to optimize the cytotoxic potential of NKCEs targeting BCMA by systematically investigating how the spatial organization and valency of the BCMA and NKp46 binders influence efficacy in vitro and in vivo. To achieve this, we generated and evaluated multiple molecular formats, featuring two spatial arrangements for the BCMA and NKp46 binders relative to the Fc domain (cis vs. trans) and three valency options (1:1, 2:1, and 2:2). The CODV format (cis arrangement) and the ANKET format (trans arrangement) allowed us to systematically compare the impact of valency and spatial orientation on NKCE-mediated tumor killing. Our in vitro studies revealed that antigen-binding valency did not significantly affect molecular efficacy. Molecules in a 2:1 format, which mimic the natural IgG1 configuration, did not show superior activity compared to the 1:1 format in cytotoxicity assays. This suggests that NK cell activation, rather than target avidity, is the primary driver of NKCE activity. Furthermore, the 2:1 NKCE format, which co-engages NKp46 and FcγRIIIa, was significantly more effective than conventional IgG1 antibodies, highlighting the importance of dual receptor engagement for NK cell activation. Additionally, no significant efficacy differences were observed between cis- and trans-arranged formats, indicating that NKCE molecules co-engaging NKp46 and FcγRIIIa are functionally tolerant to different spatial organizations. We also explored the impact of bivalent NKp46 engagement and Fc domain optimization. Our in vitro data showed that while bivalent NKp46 engagement with FcγRIIIa co-engagement enhanced activity, the effect was moderate. However, in vivo comparisons between 1:1 and 2:2 molecules revealed that 1:1 formats tended to be more effective, even at higher doses, suggesting that enhanced in vitro cytotoxicity does not always translate into improved in vivo tumor control. Based on PK in huFcRn tg32 mice, PK in efficacy model is expected to be similar considering similar target-binding affinity for the two constructs. This rather suggests that while dual NK activation through NKp46 and FcγRIIIa enhances engager activity, monovalent target antigen binding may be more favorable. This could result in more engagers binding monovalently to BCMA on tumor cells, increasing the formation of bridges between effector and tumor cells.

Conversely, Fc domain optimization significantly enhanced NKCE potency, consistent with previous findings.17 Notably, affinity optimization for FcγRIIIa had a much lower impact on 2:2 molecules compared to 1:1 molecules. However, Fc optimization led to a dramatic improvement in in vivo activity, particularly in the huFcγR-Tg mouse model, where the optimized format achieved near-complete tumor control at clinically relevant doses.

Our findings led to the selection of the 1:1 configuration with an Fc domain optimized for ADCC as the optimal format for SAR’514 as a BCMA-targeting NKCE for MM treatment. SAR’514 demonstrated potent anti-tumor activity in vitro and in vivo and effectively activated and mobilized NK cells from MM patients against BCMA-expressing tumor cells in both autologous and allogeneic settings. It also showed efficacy in BM-resident NK cells, which is particularly relevant given the BM is the primary site of MM. Notably, SAR’514 activated NK cells even in heavily pretreated MM patients, including those resistant to anti-CD38 therapies. These findings highlight the potential of SAR’514 as an effective NK cell engager for MM treatment. Recent advances in the treatment of autoimmune diseases such as rheumatoid arthritis and systemic lupus erythematosus have shown potential in depleting autoreactive B cells.45^,^46 Given the expression of BCMA on plasma cells, therapies targeting BCMA may be explored in autoimmune diseases.

Continued research and clinical trials are crucial to fully realize its therapeutic benefits and address the unmet needs of patients.

Limitations of the study

Direct head-to-head comparisons between SAR’514 and TCEs present significant technical challenges that affect both in vitro and in vivo evaluations. In vitro comparisons would require PBMC-based experimental systems, which inherently favor T cell responses due to their higher abundance in peripheral blood compared to NK cells. Furthermore, standard in vitro assays can neither adequately capture the complex dynamics of immune cell recruitment from circulation, a key aspect of in vivo efficacy, nor fully reflect the differential toxicity profiles between NKCEs and TCEs (Figure S6). In vivo comparative studies face additional complexities, requiring mouse models simultaneously compatible with both human NK and T cell populations and suitable for evaluating engineered Fc of SAR’514.

Ex vivo studies with MM patient samples are most translatable but have significant limitations due to sample scarcity and fragility, restricting the number of experimental conditions that could be tested and sometimes necessitating selective testing approaches.

Resource availability

Lead contact

Further information and requests for reagents and resources should be directed to and will be fulfilled by the lead contact, Dr. Marielle Chiron (marielle.chiron@sanofi.com).

Materials availability

The molecular organization of the ANKET used in the present study can be found in patent PCT/EP2022/057824. Requests for new materials generated in this paper are to be directed to and will be fulfilled (pending MTA and associated restrictions) by the lead contact (marielle.chiron@sanofi.com).

Data and code availability

• •Crystallography data have been deposited to the Protein Data Bank (PDB) repository and are publicly available as of the date of publication. Accession numbers and link (https://www.rcsb.org/structure/9MQO) are listed in the key resources table (PDB: 9MQO).
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

Acknowledgments

We thank V. Boisrobert D’heilly and E. Perlat (Sanofi) for performing cellular studies. We thank F. Windenberger for statistical analysis; members of Sanofi Immuno-Oncology Research, particularly A. Rohaut and V. Croizé; and members of Translational Medicine Unit, particularly F. Gallen and C. Kamperschroer, Biomarker and Clinical Bioanalysis, Large Molecule Research and Chemistry, Manufacturing, and Control teams who contributed to this project. We thank A. Joulin-Giet and all members of Innate Pharma research laboratory who contributed. We thank C. Kervoelen and E. Menoret (Nantes University) for performing the ex vivo assays. The study was performed and supported by Sanofi and Innate Pharma. Medical writing support was provided by Sarabjeet Kaur and Himanshi Bhatia (Sanofi). E.V.’s lab at CIML is funded by the Institut National Du Cancer (PLBIO23-061 to E.N.-M), Canceropole PACA (emergence grant 2022 to E.N.-M.), European Research Council (ERC) under the European Union’s Horizon (TREATLIVMETS, grant agreement no. 101118936), MSD Avenir, Innate Pharma, and institutional grants to CIML (Inserm, CNRS, and Aix-Marseille University). E.V. is “Natural Killer Cells” Chair, Fondation Gustave Roussy. The authors acknowledge the Cytocell-Flow Cytometry and FACS core facility (SFR Bonamy, BioCore, Inserm UMS 016, CNRS UAR 3556, Nantes, France) for technical expertise, member of Scientific Interest Group (GIS) Biogenouest, and the Labex IGO program supported by French National Research Agency (n°ANR-11-LABX-0016-01).

Author contributions

A.T., L.G., A.V.-O., M.C., V.F., Y.M., and E.V., conceived the project. A.T., L.G., C.A., A.B., D.B., B.R., N.G., A.M., C.N., and C.P.-D. designed experiments, supervised studies and analyzed data. F.G., M.G., E.Z., P.R., V.C, N.Z., L.M., Y.Q., and E.M. performed experiments. C.T., C.P.-D. contributed to MM patient sample collection and ex vivo studies. Y.Q. and J. Batchelor solved crystallography structures. L.G. and E.N.-M. wrote the manuscript with help from A.T., A.V.-O., M.C., E.V., and other co-authors. All authors reviewed and approved the final manuscript.

Declaration of interests

L.G., B.R., A.B.-A., F.G., M.G., N.G., A.M., Y.M., and E.V. are Innate Pharma employees. A.T., E.Z., V.C., N.Z., J. Beninga, C.A., A.B., D.B., C.N., P.R., L.M., J. Batchelor, V.F., M.C., and A.V.-O are Sanofi employees. Y.Q. is employee of AstraZeneca. L.G., B.R., A.M., A.T., M.C., A.V-O., N.G., and J. Beninga hold patents related to multifunctional antibodies engaging NK cells.

STAR★Methods

Key resources table

REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesAnti-human CD3-BUV496BD Biosciences****Cat# 612940; RRID:AB_2870222Anti-human CD3-PBBD BiosciencesCat# 558117; RRID:AB_397038Anti-human CD56-BV786BD BiosciencesCat# 564058; RRID:AB_2738569Anti-human CD56-BUV737BD BiosciencesCat# 612767; RRID:AB_2860005Anti-human CD56-APCMiltenyi BiotecCat# 130-113-310; RRID:AB_2726088Anti-human CD16-BUV395BD BiosciencesCat# 563784; RRID:AB_2744293Anti-human CD16-BUV615BD BiosciencesCat# 751572; RRID:AB_2875567Anti-human CD16-PEBD BiosciencesCat# 556619; RRID:AB_396491Anti-human CD32-PEBeckman CoulterCat# IM1935; RRID:AB_131295Anti-human CD269 (BCMA)-PEBioLegendCat# 357504; RRID:AB_2561926Purified anti-human CD269 (BCMA)BioLegendCat# 357502; RRID:AB_2561924Purified Mouse IgG2a, κ Isotype Ctrl AntibodyBioLegendCat# 400202; RRID:AB_2927399Anti-human CD335 (NKp46)-PEBD BiosciencesCat# 557991; RRID:AB_396974mIgG1-PEBD BiosciencesCat# 555749; RRID:AB_396091mIgG2a-PEBioLegendCat# 400213; RRID:AB_2800438Mouse IgG1, k isotype control BUV395BD BiosciencesCat# 563547; RRID:AB_2869503Anti-human NKp46-PEBD BiosciencesCat# 557991; RRID:AB_396974Anti-human CD226-BB515 (DNAM-1)BD BiosciencesCat# 565152; RRID:AB_2739081Mouse IgG1, k isotype control BB515BD BiosciencesCat# 564416; RRID:AB_2721017Anti-human CD244-PE-Cy7BioLegendCat# 329520; RRID:AB_2572017Mouse IgG1, k isotype control PECY7BioLegendCat# 401402; RRID:AB_2801451Anti-human NKG2D-BV421BioLegendCat# 320822; RRID:AB_2566511Mouse IgG1, k isotype control BV421BioLegendCat# 400158; RRID:AB_11150232Anti-human TIGIT-BV711BD BiosciencesCat# 747839; RRID:AB_2872302Mouse IgG2B, k isotype control BV711BD BiosciencesCat# 563125; RRID:AB_2869460Anti-human NKp30-APCMiltenyi BiotecCat# 130-121-995; RRID:AB_2784148Mouse IgG1, k isotype control APCMiltenyi BiotecCat# 130-113-196; RRID:AB_2733440Anti-human CD107a(LAMP-1)-APCMiltenyi BiotecCat# 130-119-869; RRID:AB_2751898Anti-human CD69-BV650BioLegendCat# 310934; RRID:AB_2563158Anti-human CD69-FITCMiltenyi BiotecCat# 130-113-523; RRID:AB_2733656Anti-human CD138-PerCP-eF710ThermoFisher ScientificCat# 46-1388-41; RRID:AB_2815145Anti-human CD38-BUV395BD BiosciencesCat# 563811; RRID:AB_2744372Anti-human MIP1β-PEBD BiosciencesCat# 550078; RRID:AB_393549Anti-human IFNγ-AF488BioLegendCat# 502515; RRID:AB_493029BCMA-PE-Dazzle594BioLegendCat# 357511; RRID:AB_2566530mIgG2aPE-Dazzle594BioLegendCat# 400275; RRID:AB_3097694ObinutuzumabRocheGazyva/GazyvaroDaratumumabGenmab/Jensen biotechDARZALEXHuman NK activation panel CD107a/b-APC cocktail (CD56-PE-Vio770/CD107a-APC/CD107b-APC/CD3-VioBlue)Miltenyi Biotec130-095-212 (custom)Anti-human CD69-FITCMiltenyi BiotecCat# 130-113-523; RRID:AB_2733656Anti-human MIP1β-PEBD BiosciencesCat# 550078; RRID:AB_393549Anti-human TNFα-BUV395BD BiosciencesCat# 563996; RRID:AB_2738533Anti-human IFNγ-BV605BioLegendCat# 502536; RRID:AB_2563881NK cell engager/F25Innate PharmaIn this paperNK cell engager/IC-F25Innate PharmaIn this paperNK cell engager/F33Innate PharmaIn this paperNK cell engager/IC-F33Innate PharmaIn this paperNK cell engager/CODV-1:1SanofiIn this paperNK cell engager/IC-CODV-1:1SanofiIn this paperNK cell engager/CODV-2:1SanofiIn this paperNK cell engager/CODV-2:2SanofiIn this paperNK cell engager/IC-CODV-2:2SanofiIn this paperNK cell engager/SAR’514SanofiIn this paperNK cell engager/F25-DEInnate PharmaIn this paperNK cell engager/IC-F25-DEInnate PharmaIn this paperNK cell engager/F33-DEInnate PharmaIn this paperNK cell engager/IC-F33-DEInnate PharmaIn this paperNK cell engager/CODV-1:1-DESanofiIn this paperNK cell engager/CODV-2:2-DESanofiIn this paperNK cell engager/IC-CODV-1:1-SanofiIn this paperNK cell engager/SAR’514SanofiIn this paperNK cell engager/CODV-1:1-DE-DSBSanofiIn this paperNK cell engager/CODV-1:1-AFEvitriaIn this paperControl antibody/BCMA-IgG1Innate PharmaIn this paperIsotype Control antibody/IC-IgG1Innate PharmaIn this paperIsotype Control antibody/IC-IgG1-DEInnate PharmaIn this paperIsotype Control antibody/IC-IgG1-ADEInnate PharmaIn this paperNK cell engager/Surrogate CODV-1:1SanofiIn this paperNK cell engager/Surrogate CODV-1:1-ADESanofiIn this paperNK cell engager/Surrogate IC-CODV-1:1SanofiIn this paperT cell engager/BCMA-CD3SanofiIn this paperAlexa Fluor® 647 AffiniPure® Goat Anti-Human IgG, Fcγ fragment specificJacksonCat# 109-605-098; RRID:AB_2337889Biological samplesHuman buffy coats (consented healthy Donors)EFS MarseilleIn this paperHuman buffy coats (consented healthy Donors)EFS (Runigs, France)In this paperHuman whole bloodFramingham, MA, USAIn this paperMM patient samplesNCT03807128In this paper (Table 1)Chemicals, peptides, and recombinant proteinsHis-tagged Human FcγRIIIaV (CD16aV); GenBank AAH36723; V158Innate PharmaGauthier et al.17https://doi.org/10.1016/j.cell.2019.04.041His-tagged Human FcγRIIIaF (CD16aF); GenBank AAH36723; F158Innate PharmaGauthier et al.17https://doi.org/10.1016/j.cell.2019.04.041His-tagged Human FcgRIIIb (CD16b); GenBank AAI28563Innate PharmaGauthier et al.17https://doi.org/10.1016/j.cell.2019.04.041His-tagged Human FcgRIIa (CD32a); GenBank AAH20823Innate PharmaGauthier et al.17https://doi.org/10.1016/j.cell.2019.04.041His-tagged Human FcgRIIa (CD32b); GenBank NP_003992Innate PharmaGauthier et al.17https://doi.org/10.1016/j.cell.2019.04.041His-tagged Human FcgRI (CD64); GenBank P12314R&D Systems1257-FC-050His-tagged human NKp46; GenBank NM_004829.5Innate PharmaGauthier et al.17https://doi.org/10.1016/j.cell.2019.04.041M2-tagged cynomolgus NKp46; GenBank NP_001271509.1Innate PharmaDemaria et al.21https://doi.org/10.1016/j.xcrm.2022.100783His-tagged human BCMA; GenBank BAB60895.1SanofiIn this paperHis-tagged cynomolgus BCMA; GenBank NP_001183.2SanofiIn this paperImmunocult Human CD3/CD28 T cell ActivatorStemcell TechnologiesCat#10971High yield lysis bufferINVITROGENHYL250GentamycinGIBCO15750-060B2-MercaptoethanolGIBCO31350-010huIL-15BioLegend570308huIL-21BioLegend571208Fetal bovine serumGIBCO10500-056Critical commercial assaysNK Cell Isolation Kit, humanMiltenyi Biotec130-092-657CELLQUANT calibrator kitBioCytex7208MSD custom U-plex cytokine assayMesoscale DiscoveryCat# K15067L-2Lot# 404471MACSxpress® Whole Blood NK Cell Isolation Kit humanMiltenyi Biotec130-127-695Deposited dataCrystal structure of BCMA in complex with CA10v2 FabThis paperPDB: 9MQOhttps://www.rcsb.org/structure/9MQOExperimental models: Cell linesHUT78ATCCTIB-161; CVCL_0337MM.1SATCCCRL-2974; CVCL_8792MM.1RATCCCRL-2975; CVCL_8794RPMI 8226ATCCCCL-155; CVCL_0014NCI-H929ATCCCRL-9068; CVCL_1600JJN3DSMZACC 541; CVCL_2078KMS12BMDSMZACC 551; CVCL_1334L-363DSMZACC 49; CVCL_1357LP1DSMZACC 41; CVCL_0012MOLP8DSMZACC 569; CVCL_2124OPM2DSMZACC 50; CVCL_1625U266DSMZACC 9; CVCL_0566KMS11JCRBJCRB1179; CVCL_2989RMAPr. Eric VivierN/AEL4ATCCTIB-39; CVCL_0255huBCMA-expressing EL4 cells (wild type EL4 cell line obtained from ATCC, TIB-39)SanofiThis paperK562-CD86+/CD137+ (wild type K562 cell line obtained from ATCC, CLL-243)SanofiThis paperExperimental models: Organisms/strainshuNKp46-TgxRag1^−/−^ mice (Tg hNKp46 Low x B6;129S7-Rag1tm1Mom/J)Sanofi Internal breeding (Montpellier, France)Walzer et al.27https://doi.org/10.1073/pnas.0609692104huFcγR-Tg mice created and obtained by Jeffrey V. Ravetch laboratory from licensing from Rockefeller University, USASanofi Internal breeding (Montpellier, France)Smith et al.30https://doi.org/10.1073/pnas.1203954109NOD.Cg-Prkdc^scid^Il2rg^tm1Sug^ Tg(CMV-IL2/IL15)1-1Jic/JicTacTaconic Biosciences (Ejby, Denmark)Model No: 13683-F; RRID:IMSR_TAC:13683B6.Cg-Fcgrttm1Dcr Tg(FCGRT)32Dcr/DcrJThe Jackson LaboratoryStrain #:014565; RRID:MGI:4944214Recombinant DNAHuman BCMA extracellular domainUniprotQ02223, amino acid L2 to A54Cynomolgus BCMA extracellular domainUniprotA0A2K5UD97, amino acid L2 to A53Diphtheria toxin fragment AUniprotQ5PY51, amino acid G2 to R194Software and algorithmsFlowJo v.10.5.2BD BiosciencesFlowJo 11 - Download | FlowJo, LLCGraphPad Prism, version 8.1.1GraphPad Softwarehttps://www.graphpad.com/scientific-software/prism/MSD Discovery WorkbenchMesoscale Discoveryhttps://www.mesoscale.com/en/products_and_services/softwareSAS® 9.4 Software for Windows 10SAS Institute IncSAS 9.4 Software Overview for the CustomerOtherCulture-treated 96-well plateBD Falcon353077LumaPlate-96Perkin Elmer6006633Pancoll tubePan BiotechP04-60125Separating LS ColumnsMiltenyi30-042-401Pre-Separation Filter, 70 μMMiltenyi130-095-823Corning® Costar® Ultra-Low Attachment Multiple 96-Well PlateThermoFisher Scientific7007MultiRad 160 X-IrradiatorPrecision X-ray IncMR160MACSmix Tub RotatorMiltenyi BiotecCat#130-090-753MACSxpress SeparatorMiltenyi BiotecCat#130-098-308Roswell Park Memorial Institute (RPMI) 1640 + L-GlutamineGIBCO21875-034Fetal bovine serum (FBS)SigmaF7524Sodium pyruvateGIBCO11360-039Non-Essential Amino Acids (NEAA) 100×GIBCO11140-035L-glutamineGIBCO25030-024Ionomycin (Iono)SIGMAI0634Phorbol 12-myristate 13-acetate (PMA)SIGMAP8139BD GolgiStop, Protein Transport InhibitorBD Biosciences554724BD Cytofix/CytopermBD Biosciences554722Perm/Wash BufferBD Biosciences554723BD CytofixBD Biosciences554655BD Pharmingen Stain Buffer BSABD Biosciences554657Chromium-51 (^51^Cr) Radionucleide, 74 MBqPerkin ElmerNEZ030002MCTergitolSIGMA15S9-100 MLPhosphate buffer solution (PBS)GIBCO14190-094Bovine serum albumin (BSA)SIGMAA9418-100GEthylenediaminetetra-acetic acid (EDTA)Invitrogen15575-038Sodium azideSIGMA71290-100GBD CellFixBecton Dickinson340181calcein-AMThermoFisher ScientificC3100MPProbenecidThermoFisher ScientificP36400Triton X-Merck9002-93-1

Experimental protocol and study participant details

Cell lines

NCI-H929 (ATCC CRL-9068), MM.1S (ATCC CRL-2974), MM.1R (ATCC CRL-2975) and RPMI 8226 (ATCC CCL-155) human myeloma cell lines (HMCLs), and HUT78 cutaneous T cell lymphoma cell line (ATCC TIB-161), were purchased at American Type Culture Collection (ATCC, USA). JJN3 (DSMZ ACC 541), KMS12BM (DSMZ ACC 551), L-363 (DSMZ ACC 49), LP1 (DSMZ ACC 41), MOLP8 (DSMZ ACC 569), OPM2 (DSMZ ACC 50) and U266 (DSMZ ACC 9) HMCLs were purchased at DSMZ (Germany). KMS11 (JCRB1179) HMCL was purchased at JCRB (Japan). Cells were cultured in complete RPMI medium (RPMI-1640 containing 10% FBS, 2 mM L-glutamine). For the in vitro cytokine release assay, RPMI 8226-RFP cells were derived from RPMI 8226 cells obtained from ATCC and transfected with red fluorescent protein (RFP) by Sanofi. The modified RPMI 8226-RFP cells were cultured in RPMI 1640 (Thermo Fisher Scientific, Inc., Waltham, MA, USA) supplemented with 10% FBS (Thermo Fisher Scientific, Inc.) and 1% Penicillin-Streptomycin (10,000 U/mL, Thermo Fisher Scientific, Inc.) at 37°C in a humidified atmosphere of 5% CO_2_. For in vivo mouse study, RMA (kindly provided by Pr Eric Vivier) murine T lymphoblasts were first engineered to express either eGFP or dsRed fluorescent proteins. RMA-eGFP and RMA-dsRed cells were subcloned and selected for stable expression of the fluorescent proteins. The EL4 murine T lymphoblasts (ATCC TIB-39) and RMA-dsRed cells were engineered to express human BCMA by transduction with a lentivirus encoding human BCMA and selected based on BCMA expression level.

Biological samples

Whole blood from 11 consented healthy donors on-site at Sanofi (Framingham, MA, USA) was collected by venipuncture into 10 mL vacutainer tubes containing sodium-heparin anticoagulant. Samples were gently mixed and kept under ambient conditions until study initiation. Healthy human buffy coats were provided by the Etablissement Français du Sang (EFS, the French blood service, Marseille or Ile de France-Agreement N° 12/EFS/131) with the written consent of each volunteer blood donor. Peripheral mononuclear cells (PBMC) were isolated from buffy coats by Ficoll density gradient centrifugation. Human NK cells were purified from PBMC with a bead-based negative selection kit from STEMCELL Technologies or Miltenyi Biotec. After informed consent, blood and/or bone marrow samples from multiple myeloma patients were collected at the University Hospital of Nantes, Department of Hematology (MYRACLE cohort; ClinicalTrials.gov registration: NCT03807128).47 Bone marrow mononuclear cells (BMMC) and PBMC were isolated by gradient density centrifugation using Ficoll-Hypaque.

Mice

huNKp46-TgxRag1^−/−^ mice (Tg hNKp46 Low x B6;129S7-Rag1tm1Mom/J)28^,^33 are deficient for the expression of recombinant activating gene 1 (RAG1) and transgenic for the expression of human NKp46 protein. huFcγR-Tg (C57BL/6 FcgR human) mice30 are FcγR humanized mice in which all murine FcγRs were deleted and human FcγRs encoded as transgenes were inserted into the mouse genome. Mice were bred in specific and opportunistic pathogen-free conditions animal facility of “Center de Recherche et Développement de Montpellier” (Sanofi). Female immunodeficient NOG mice expressing human IL-15 cytokine (NOD.Cg-Prkdc^scid^ Il2rg^tm1Sug^ Tg(CMV-IL2/IL15)1-1Jic/JicTac, also called NOG-IL-15-Tg mice) were purchased from Taconic Biosciences (4623 Lille Skensved, Ejby, Denmark).

Animal care

All animal procedures were approved by the Sanofi Animal Care and Use Committee, followed the French and European regulations on care and protection of the Laboratory Animals, and in accordance with the standards of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).

Method details

Recombinant human and cynomolgus BCMA - Cloning, production, and purification

For the generation of recombinant human and cyno BCMA (TNFRSF17) a synthetic DNA fragment coding for the extracellular domain of human (Uniprot Q02223, amino acid L2 to A54) or cynomolgus (Uniprot A0A2K5UD97, amino acid L2 to A53) BCMA fused to diphtheria toxin fragment A (DTA, Uniprot Q5PY51, amino acid G2 to R194) and a His6 purification tag was ordered from Atum (Newark, CA, USA). The synthetic DNA fragment was cloned into a pTT5 mammalian expression vector using Gateway Cloning technology (Life Technologies - ThermoFisher Scientific, Carlsbad, CA, USA) and expressed in Expi293 F cells (ThermoFisher). Protein was purified from cell supernatant after 6 days of growth in conditioned media (CM) using NiNTA agarose affinity chromatography (Qiagen), followed by preparative SEC preformed using a Superdex 200 (16/60) column (Cytiva Life Sciences). The final sample was concentrated in PBS to a concentration of >5 mg/mL.

Generation of anti-BCMA antibody CA10

The anti-BCMA antibody CA10 was generated by immunizing humanized Trianni mouse® platform using recombinant human BCMA proteins. Because the selected anti-BCMA CA10 antibody contained a murine lambda chain, it was further humanized by standard CDR-grafting methods using human Vk and Ck backbone.

Structure of BCMA/CA10 fab complex by X-ray crystallography

The mixture of BCMA residues 1–54 and CA10 Fab was purified by size exclusion chromatography after incubation at 4°C for 1 h. The complex-containing fractions were pooled and concentrated to 15 mg/mL in 20 mM HEPES pH 7.0, 100 mM NaCl buffer. The CD40-CP-870,893 Fab complex was crystallized in sitting drops composed of 0.18 M Sodium Chloride, 0.09 M Phosphate/citrate pH 4.2, 20% w/v PEG 8000, 0.1 M Sodium Malonate and 25% of glycerol was used as cryoprotectant for crystal freezing. Data were collected at ESRF ID-30 b, and the structure was solved by molecular replacement, revealing six BCMA/CA10 Fab complexes in the crystallographic asymmetric unit.

NKCE expression and purification

F25- and F33-based NKCEs were generated, produced and purified as described previously.24 Briefly, the expression vectors encoding the different polypeptide chains were used to co-transfect EXPI-293 F or HEK-293 FS cells (Life Technologies). The cells were used to seed culture flasks at a density of 10^6^ cells per mL and the supernatant was harvested after six to seven days and passed through a Stericup filter with 0.22 μm pores. The NKCE molecules were purified from the supernatant with rProtein A Sepharose Fast Flow resin (GE Healthcare, ref. 17-1279-03), followed by cation ion-exchange chromatography (CIEX) after dialysis against 50 mM Na_2_HPO_4_/KH_2_PO_4_ phosphate buffer pH 6.2. For generation of CODV-based NKCEs, the expression plasmids encoding the different chains of the constructs were propagated in E.coli DH5a, and plasmids used for transfection were prepared with EndoFree Plasmid Mega kits (Qiagen). HEK 293-FS cells growing in F17 serum free suspension culture (Invitrogen) were transfected with indicated plasmids using Polyethyleneimine transfection reagent. The proteins were captured from supernatant after 6 days of cultivation at 37°C with 8% CO_2_ on MabSelect SuRe (Cytiva), eluted with 0.1 M Citrate buffer pH 3.0 and neutralized with 1 M Tris pH9. After polishing the proteins by size exclusion chromatography (SEC) using a Superdex200 26/60 (Cytiva) and 0.22 p.m. filtration and UV280 concentration determination the proteins were used for further characterization. All NKCE molecules reached a monomer purity higher than 95% (in SE-HPLC) and an endotoxin level lower to 0.5 EU/mg for in vivo use. All BCMA-NKp46-FcγRIIIa NKCE molecules were built using the humanized anti-NKp46 antibody 3D917 associated to the humanized anti-BCMA antibody CA10_V7. Isotype control molecules were built by replacing the BCMA-binder with anti-Lysozyme Ab D1.3.48

Quantification of BCMA molecules on HMCLs

Quantification of surface BCMA expression level on HMCLs (ABC; Antibody binding capacity) was performed by flow cytometry on a MACSQuant Analyzer 16 Flow Cytometer or MACSQuant X Flow Cytometer with the CELLQUANT calibrator kit (BioCytex, ref#7208) according to the protocol recommended by the supplier. Anti-human BCMA antibody clone 19F2 (Biolegend, ref. 357502) and the isotype control clone MOPC-173 (Biolegend, ref. 400201) were used at saturating concentration (10 μg/mL) for the quantification. Labeled cells and calibration beads were analyzed on a flow cytometer, and a standard regression line between fluorescence intensity and Ag density expressed as ABC (molecules per cell) was calculated. Specific ABC (sABC) was determined by subtracting the ABC background of the isotype control.

NK cell-based cytotoxic assay

MM.1R, MM.1S, L-363, HUT78 and RPMI 8226 cells were loaded with ^51^Cr. Depending on experiments, dilution ranges from 100 to 50 to 1–0.5∗10^−5^ nM (1/10e serial dilution) were performed for both test and control items. The molecules, labeled target cells (3,000 cells) and human purified NK cells (30,000 cells) from healthy donors were successively added to each well of round-bottomed 96-well plates to obtain an effector to target (E:T) ratio of 10:1. After 4 h of co-incubation, the supernatant was transferred to a Lumaplate. ^51^Cr released from dead target cells was determined with a TopCount NXT™ (Microplate Scintillation and Luminescence Counter; PerkinElmer). Radioactivity was measured by counting γ-emission for 60 s for each well. The results are expressed in counts per minute (cpm). The percent specific lysis was calculated with the following formula:

$$[eqn]$$

where ER = experimental release (cpm or RFU), SR = spontaneous release (cpm or RFU) and MR = maximal release (cpm or RFU).

Alternatively, RPMI 8226 target cells were loaded with calcein-AM (TermoFisher Scientific, C3100MP, 50 μg) previously reconstituted in 25 μL of DMSO. Labeled cells were incubated at 37°C in the presence of 5% CO_2_ for 30 min. Antibody serial dilutions were prepared and added to a U-bottom 96-well plate (100 μL/well) (Corning® Costar® Ultra-Low Attachment Multiple 96-Well Plate; Thermo Scientific, Denmark, ref #7007). A dilution range from 100 nM to 0.01 pM (1/10^th^ serial dilution) was performed for both test and control items.

Target cells were washed 3 times with 5 mL of complete medium (every wash was followed by a centrifugation of 300 g for 5 min and the supernatant was discard). Final wash was performed in complete medium containing Probenecid (4× concentrated) (ThermoFisher Scientific, P36400). Target cells were counted and seeded at 5000 cells/50 μL per well. Finally, NK cells at E:T ratio of 10:1 were added to the suspensions of target cells and antibodies. As controls, target cells alone and target cells with NK, but without antibodies, were added. Moreover, to reach maximal target cell death, 2% of Triton X-(CAS 9002-93-1; T9284) was considered. Each condition was performed in duplicate. Culture plates were incubated at 37°C in the presence of 5% CO_2_ for 2 h at the end of which 100 μL of supernatant was collected, transferred to a black 96-well microplates flat-bottom med binding, (Greiner ref#655076) to assess the calcein release from dead target cells with TECAN 1000Pro machine.

For the killing analysis, the percent specific lysis was calculated using the following formula:

$$[eqn]$$

• ER = experimental release (target cell + NK + [Antibody])
• SR = spontaneous release (target cell alone)
• MR = maximal release (target cells +2% Triton X-)

The EC_50_ was determined for each antibody by drawing an appropriate non-linear regression curve (choice of “log(agonist) vs. response – variable slope (four parameters)” model) with Graphpad Prism Software.

Cytokine release assay in human whole blood

Whole blood from consented healthy donors (11 donors) was collected on-site at Sanofi (Framingham, MA, USA) by venipuncture into 10 mL vacutainer tubes containing sodium-heparin anticoagulant. Samples were gently mixed and kept under ambient conditions until study initiation. After the addition of RPMI-8226-RFP cells (at final density of 20,000 cells per well) and controls or SAR’514, fresh whole blood samples (200 μL) were then added to the wells and the plates were incubated at 37°C in a humidified atmosphere of 5% CO_2_ for 24 h. Experiments were performed in triplicate. Negative controls were media alone (untreated condition) and isotype control (IC-NKp46-FcγRIIIa NKCE, identified as RA16936297, Sanofi) at a final concentration of 2.4 μM (300 μg/mL). Positive controls included Immunocult™ Human CD3/CD28 T cell Activator at 25 μL/mL (Stemcell Technologies, Vancouver, BC, Canada) and BCMA-CD3 TCE at 0.7 μM (100 μg/mL, identified as RA17045303, Sanofi). SAR’514 was tested at concentrations of 0.008, 0.08, 0.8, and 2.4 μM (1, 10, 100, and 300 μg/mL).

At the end of the incubation period, plates were centrifuged at 500 × g for 10 min. Plasma was collected from these plates and then transferred directly to a new 96-well cell culture plate and then centrifuged at 2000 × g for 10 min to remove remaining cellular debris. Plasma samples were assessed for levels of IFN-γ; (quantification range of 9.45–27,000 pg/mL), TNF-α (quantification range of 0.66–2,940 pg/mL), macrophage inflammatory protein 1 alpha (MIP-1α; quantification range of 14.5–5,580 pg/mL), granulocyte-macrophage colony-stimulating factor (GM-CSF; quantification range of 0.33–10,200 pg/mL), interleukin-2 (IL-2; quantification range of 1.24–1990 pg/mL), interleukin-4 (IL-4; quantification range of.

0.16–1,780 pg/mL), interleukin-10 (IL-10; quantification range of 0.35–3,770 pg/mL), interleukin-6 (IL-6; quantification range of 0.89–2,050 pg/mL), interleukin-8 (IL-8; quantification range of 0.35–2,180 pg/mL), and interleukin-1 beta (IL-1β; quantification range of 0.35–4430 pg/mL) using an MSD U-PLEX assay (Catalog #K15067L-2; Lot #404471) according to manufacturer instructions (Mesoscale Discovery, Rockville, Maryland, USA). Cytokine levels were defined using calibration curves for each cytokine by fitting the signals from the calibration standards to a 4-parameter logistic or sigmoidal dose-response model with a 1/Y^2^ weighting. The calculations to determine cytokine concentrations were carried out using the MSD DISCOVERY WORKBENCH analysis version 4.0 (Mesoscale Discovery, Rockville, Maryland, USA) and concentrations were expressed in units of pg/mL. The data was entered in GraphPad Prism software, version 9.1.2 (GraphPad Software, San Diego, CA, USA) and the mean value for each set of replicates was calculated. The mean for each donor/condition was plotted on a log scale and the median for each condition calculated and plotted. Samples where the mean fell at or below the lower limit of quantification (LLOQ) were reported as the LLOQ and this value was used in all calculations and in the graphs generated. Samples where the mean fell above the upper limit of quantification (ULOQ) were reported as the extrapolated value obtained, and this value was used in all calculations and in the generated graphs.

Surface plasmon resonance studies

A Biacore T200 instrument (Cytiva, Uppsala, Catalog No. 28975001) was used with a Series S CM5 sensor chip (Cytiva, Uppsala, Catalog No. 29149603). HBS-EP+ (Cytiva, Uppsala, Catalog No. BR1006-69) served as running buffer in all experiments. Affinity capture of the NKCE sample was achieved with the human antibody capture kit (Cytiva, Uppsala, Catalog No. BR1008-39). The anti-Fc capture antibody was coupled to the CM5 chip by standard amine coupling with the amine coupling kit (Cytiva, Uppsala, Catalog No. BR-100-50) to yield approximately 8000 response units (RU). Serial ½ dilutions of either human or cynomolgus recombinant BCMA (6.25–200 nM), or human and cynomolgus NKp46 (15.625–1,000 nM), or recombinant human CD16aV (158 V allele), CD16aF (158 F allele), CD32a, CD32b, CD16b (20.6–5000 nM) and CD64 (18.65–300 nM) were injected over the NKCE–immobilized proteins and allowed to dissociate for 10 min were injected for 240 s at a flow rate of 40 μL/min, followed by a dissociation phase of 600 s, onto 30 to 60 RU of anti-Fc captured NKCE. 50 nM analyte concentration was run in duplicate, together with multiple buffer blanks for double referencing. The capture surface was regenerated with regeneration solution (3 mol/L MgCl_2_) at a flow rate of 40 μL/min for 60 s. The data for binding kinetics were evaluated with Biacore T200 Evaluation Software version 3.0 (Cytiva, Uppsala) using either Langmuir 1:1 binding model with mass transport limitation or steady state model.

For competition study with APRIL, 60–70 RU of BCMA-DTA-6xHis recombinant proteins were captured onto an anti-His antibody (Qiagen, 34670, batch 172044782) immobilized on a CM5 chip as described above. SAR’514 (at a concentration of 10 μg/mL) was injected twice over the captured BCMA protein to ensure full saturation of the antigen, and recombinant APRIL protein (R&D systems, 5860-AP/CF, batch TCF1522032; at a concentration of 5 μg/mL) was injected over the SAR’514/BCMA complex. For the reference control of APRIL binding to captured BCMA, the same sequence of injections was performed but with running buffer instead of SAR’514. Sensorgrams were normalized at 100 RU for BCMA capture and further aligned on the x and y axis at the start of APRIL injection.

Efficacy studies in short term disseminated mouse models

The activity of the BCMA-NKp46-FcγRIIIa NKCE molecule was evaluated in a disseminated tumor model using mouse RMA leukemia cells, transduced RMA-dsRed-huBCMA Cl.E6 (BCMA-positive) or not RMA-eGFP Cl.5A6 (BCMA-negative) for the expression of human BCMA and mixed at 1:1 ratio for injection in the tails vein on day 0 of huNKp46-TgxRag1^−/−^ mice (N = 14). Mice were divided in two groups, treated at day 0 with the vehicle (N = 7) or with BCMA-NKp46-FcγRIIIa molecule (N = 7) at a flat dose of 12.3 picomoles per mice. Mice were sacrificed 48 h after treatment and disseminated RMA cells extracted from liver by crushing with OctoMacs® and Percoll gradient isolation. RMA cells infiltrating the liver were analyzed and counted by flow cytometry.

Efficacy studies in long term disseminated syngenic mouse models

Disseminated tumor model was established in huNKp46-TgxRag1^−/−^ mice by injecting 5 × 10^5^ huBCMA-expresssing EL4 cells into the caudal vein on Day 0. Mice were treated at Day 1 with BCMA targeting or isotype control molecules as following: huNKp46-TgxRag1^−/−^ mice (10 per group) were assigned on Day 1 to following treatment groups – control group without treatment (vehicle only), IC-NKp46-FcγRIIIa NKCE isotype control, at 5 mg/kg, or BCMA-NKp46-FcγRIIIa NKCE (CODV-1:1) administered IP at three doses of 5, 0.5, and 0.05 mg/kg, or BCMA-IC-FcγRIIIa NKCE control, at 5 mg/kg, or anti-BCMA-IgG1 antibody, at 10 mg/kg. huFcγR-Tg mice (10 per group) were assigned on Day 1 to following treatment groups – Isotype control IC-NKp46-FcγRIIIa NKCE (non-engineered Fc), surrogate BCMA-moNKp46-FcγRIIIa NKCE (non-engineered Fc) and surrogate BCMA-NKp46-FcγRIIIa NKCE-ADE (CODV-1:1 Fc engineered with the ADE mutation). Molecules were administered intraperitoneally (IP) at 0.5 mg/kg. Surrogate NKCE molecules were built using the anti-moNKp46 antibody 29A1.4.27 Mice were monitored daily, or at least twice weekly, for adverse clinical reactions, individual body weight and survival until day 60 (huNKp46-TgxRag−/−experiments) or day 90 (huFcγR-Tg experiment). Mice were euthanized when moribund, according to predefined criteria, to avoid suffering which included critical clinical signs such as limb paralysis, ascites, palpable internal tumor mass, morbidity, or weight loss of 20% or more. The individual days of death (if any) of each mouse was recorded. Median survival time (MST) was determined for each group. Data were summarized using Kaplan-Meier curves. survival differences between control or treated groups was evaluated with the log-rank test followed by Bonferroni-Holm adjustments. Statistical analyses were performed using SAS® version 9.4 for Windows 10, with a 5% significance level.

Efficacy studies in NK-humanized NOG-IL-15-Tg mice bearing disseminated MM.1R cells

Female NOG-IL-15-Tg mice were irradiated at 0.9 Gy. After a 4-day resting period in a quiet environment, irradiated NOG-IL-15-Tg mice were humanized by injecting IV 10 × 10^6^ amplified NK cells per mouse in the tail vein. Eight days after adoptive transfer, MM1R cells were implanted in the tail vein of NK-humanized mice (5 × 10^6^ cells per mouse in PBS suspension) at day 0. Control groups were left untreated. SAR’514 was administered at 10, 5, 2.5, 0.5, and 0.05 mg/kg by intraperitoneal injections on days 1, 4, and 7 after tumor implantation. Survival studies was performed until day 120.

Phenotypic study of MM patient NK cells by flow cytometry

The phenotype of NK cells from BM and peripheral blood of MM patients was investigated by multicolor flow cytometry. Patients’ mononuclear cells were isolated by density gradient centrifugation and stained with an antibody panel containing LIVE/DEAD (ThermoFisher, L34976), anti-CD3 (BD Biosciences, 612940), anti-CD56 (BD Biosciences, 564058), anti-CD16 (BD Biosciences, 563785), anti-CD226 (DNAM-1) (BD Biosciences, 565152), anti-NKp46 (BD Biosciences, 557991), anti-CD244 (2B4) (Biolegend, 329520), anti-NKG2D (Biolegend, 320822), anti-TIGIT (BD Biosciences, 747839) and anti-NKp30 (Miltenyi, 130-121-995), or an FMO (full minus one) antibody panel for CD16, CD226 (DNAM-1), NKp46, CD244 (2B4), NKG2D, TIGIT and NKp30 markers. Data were acquired on a BD FACSymphony A5 with BD FACSDiva Software (Cytocell-Flow Cytometry and FACS core facility, SFR Bonamy, BioCore, Inserm UMS 016, CNRS UAR 3556, Nantes, France). The analyses were done with FlowJo software.

Ex vivo functional activity of BCMA-NKCE on MM patient samples-allogeneic settings

For allogeneic study with MM patient PBMCs, Karpas 620 cells used as target were validated at each thawing by flow cytometry as described.49 Cells were cultured in RPMI1640 medium in the presence of 5% fetal calf serum (FCS), and labeled with calcein-AM (ThermoFisher, C3099) the day of the assay. The day before the assay, PBMC from MM patients were stained with anti-CD3 and anti-CD56 to determine the percent of CD3^-^/CD56^+^ cells and kept in RPMI1640 culture medium containing 10% FCS at 37°C and 5% CO_2_ overnight. The day of the assay, resting PBMC were co-cultured with calcein-AM-loaded Karpas 620 cells (3,000 cells/well) at E:T ratio of 100:1 in presence of tested and control molecules for 3 h. MM cell lysis was assessed by Calcein release assay as described in the NK-cell based cytotoxicity assay section. Alternatively, resting PBMC and Karpas 620 cells (30,000 cells/well) were co-cultured in presence of tested and control molecules for 4 h at E:T ratio of 10:1. Expression of activation markers (CD69 and CD107a), intracellular production of cytokines (IFN-g) and chemokines (MIP-1β) on NK cells, and Karpas 620 cell death were monitored by flow cytometry with anti-CD69 (Biolegend, 310934), anti-CD107a (Miltenyi, 130-119-869), anti-IFN-g (Biolegend, 502515), anti-MIP-1b (BD Biosciences, 550078), anti-CD38 (BD Biosciences, 563811) and anti-CD138 (ThermoFisher, 46-1388-41) mAbs. Kapas 620 cell killing was assessed by the disappearance of CD138 staining.50 Samples data were acquired on a BD FACSymphony A5 with BD FACSDiva Software. The analyses were done with FlowJo software.

Ex vivo functional activity of BCMA-NKp46-FcγRIIIa NKCE on MM patient samples-autologous settings

The day of the assay, Ficoll Hypaque-isolated mononuclear cells from BM or PB of patients with MM were incubated (400,000 cells/well) in presence of BCMA-NKCE, Isotype control NKCE, or Daratumumab for 18 h. NK cell activation and death of myeloma cells were analyzed by flow cytometry as described above. For negative control, BMMC or PBMC cells were incubated without antibody. At the time of the assay, some BMMC or PBMC cells were reserved to assess the expression of BCMA on CD138^+^ MM cells by staining with anti-BCMA (Biolegend, 357511) antibody and the expression of FcγRIIIa and NKp46 on NK cells (defined as CD3^−^CD56^+^ cells). Acquisition was performed on a BD FACSymphony A5 with BD FACSDiva Software (Cytocell-Flow Cytometry and FACS core facility). The analyses were done with FlowJo software. Data were analyzed using GraphPad Prism 9 (GraphPad Software Inc.).

Quantification and statistical analysis

Statistical analyses were performed using GraphPad Prism and SAS. Flow cytometry data were analyzed using FlowJo software, EC_50_ values were determined from dose-response curves. Survival data were analyzed using Kaplan-Meier survival curves. P-values were calculated using: the log rank (Mantel-Cox), One-way ANOVA Tukey’s multiple comparison test, paired two-tailed t test, as indicated in each Figure legend.