ScFv fragments fused to a 6×His-FLAG tag (scFv-HF) were expressed in E. coli HB2151 using the pComb3X vector. Colonies were cultured to OD600 = 0.6–0.8, induced with 0.5 mM IPTG at 16°C for 8 h, and lysed with polymyxin B (50 μg/mL). Supernatants were purified via Ni-NTA affinity chromatography, eluted with a 250 mM imidazole gradient, and buffer-exchanged into PBS.

mCD16/CD16A-His (5 μg/mL) was coated onto plates, blocked with 2% BSA-PBS, and incubated with serially diluted scFv-HF (0.1–300 nM). IL6-HisFlag was used as a labelling control antibody. Bound antibodies were detected using HRP-conjugated anti-FLAG antibody, followed by TMB substrate. EC₅₀ values were calculated via four-parameter logistic regression (GraphPad Prism 5).

For protein analysis, 10 ng of recombinant CD16A-His protein was mixed with 5× reducing loading buffer and denatured at 95°C for 10 min in a final volume of 20 μL. Samples and pre-stained protein markers were resolved on SDS-PAGE gels through electrophoresis, terminated when the dye front approached the gel bottom. Proteins were transferred onto PVDF membranes at 250 mA for 60 min using wet transfer methodology. Membranes were blocked with 5% non-fat milk in TBST (Tris-buffered saline with 0.1% Tween-20) for 1 h at room temperature to minimize non-specific binding. Primary antibody incubation was performed overnight at 4°C with anti-Flag antibody (1 μg/mL in blocking buffer). Following three 5-min TBST washes, membranes were probed with HRP-conjugated mouse anti-Flag secondary antibody (1:5,000 dilution) for 1 h at room temperature. Four additional TBST washes (5 min each) preceded signal development using enhanced chemiluminescent substrate (ECL), with image acquisition performed via a digital gel documentation system.

Cells were harvested with trypsin-EDTA, washed, and resuspended in 5% BSA-PBS. Surface CD16A expression was assessed using 10 μg/mL scFv-HF, detected with APC-conjugated anti-His antibody (BioLegend). For PBMC analysis, FITC-conjugated goat anti-rabbit secondary antibodies (Abclonal) were utilized.

A tandem scFv format was employed, linking GPC3-scFv-hFc(N297A) to CD16A-scFv (N297A mutation in Fc to eliminate glycosylation and ADCC activity) (32). This architecture follows recent biophysical insights showing that G4S linkers improve conformational flexibility and antigen engagement. Constructs were cloned into CHO-K1 (ATCC CCL-61) and HEK 293F cells via Hind III digestion and selected with puromycin. Following PEI-mediated transfection, genetically modified cells were subjected to puromycin selection (30 μg/mL) in 96-well plates 72 h post-transfection. Surviving monoclonal colonies were expanded and screened for antibody secretion using quantitative sandwich ELISA. High-producing clones exhibiting superior antibody titers were prioritized for scale-up production. Selected stable cell pools were propagated in suspension culture using Erlenmeyer shake flasks under standardized conditions (37°C, 5% CO₂, 120 rpm). Culture supernatants were harvested at peak productivity phases, clarified by centrifugation (4,000 × g, 20 min), and subjected to Protein A affinity chromatography for IgG purification. Target antibodies were eluted using low-pH buffer (0.1 M glycine-HCl, pH 2.7) followed by immediate neutralization and buffer exchange into PBS via tangential flow filtration.

Microplates were immobilized with 50 μL of recombinant GPC3/CD16A(ECD)-His protein (5 μg/mL in PBS, pH 7.4) through 30-min incubation at 37°C. Following three PBST (PBS containing 0.05% Tween-20) washes, non-specific binding sites were blocked using 2% BSA in PBST for 30 min at 37°C. Serial three-fold dilutions of bispecific antibodies (100 nM starting concentration in PBS) were applied to the plates (50 μL/well) and incubated for 30 min at 37°C. After removing unbound antibodies and performing two PBST washes, HRP-conjugated rabbit anti-human IgG (1:8,000 dilution; Boiron, Shanghai) was added (50 μL/well) for 30 min at 37°C. Following four PBST wash cycles, enzymatic activity was quantified by adding 50 μL TMB substrate (KPL) for 5 min protected from light. The reaction was terminated with 50 μL of 0.5 M H₂SO₄, with absorbance measured at 450 nm using a multimode microplate reader (Molecular Devices).

Adherent cells were enzymatically dissociated using 0.25% trypsin-EDTA (Thermo Fisher Scientific, Waltham, MA), pelleted by centrifugation (300 × g, 5 min), and resuspended in ice-cold PBS supplemented with 5% BSA at a density of 1 × 10⁶ cells/mL. Cells were incubated with primary antibody (10 μg/mL in blocking buffer) for 90 min on ice with gentle agitation. Following two washes with cold PBS, secondary staining was performed using Cy5-conjugated goat anti-human IgG (1:500 dilution; Jackson ImmunoResearch, West Grove, PA) for 45 min under light-protected conditions. Cell suspensions were analyzed on a BD FACS Calibur flow cytometer (BD Biosciences, Franklin Lakes, NJ), with fluorescence signals quantified using CellQuest Pro software.

Tumor cells (LG; A431, G1, Huh7) at a density of 5,000 per well were seeded into 24-well plates and subsequently cultured at 37°C with 5% CO₂ for a period of 24 hours. Thereafter, CY5-labeled CD16AM5 and MA4-hFc-CD16AM5 antibodies (10 µg/ml) were added to the PBMCs corresponding to the tumor cells at a ratio of PBMC:tumor = 5:1. The incubation period was continued for a duration of 24 hours, during which direct microscopic observation was performed. The bright field channel is utilized for the observation of the morphology and distribution of PBMCs. The CY5 channel is employed for the detection of the fluorescence signal of the double antibody on the surface of tumor cells, while the FITC channel is used for the observation of tumor cells.

The assessment of cell growth and viability was conducted through the implementation of a firefly luciferase reporter gene assay. All of the cancer cells that were utilized in the experiment were stably infected, and the firefly luciferase reporter gene (ffLuc2) was constitutively expressed. A volume of 200 μL of stably transduced cells was inoculated into a 96-well plate (5 x 10³ cells per well). Thereafter, PBMC cells (1 x 10⁵) were added, along with the indicated concentrations of bispecific antibodies (0.01-10,000 ng/mL). Following a 72-hour incubation at 37°C, the culture media were aspirated, and 100 µL of phosphate buffered saline (PBS) was added. This was followed by two cycles of freeze-thaw lysis. The release of luciferase activity was measured to represent cell viability. The in vitro assay for measuring cellular toxicity was conducted on three separate occasions, utilizing PBMCs obtained from three distinct donors. Sorafenib (0.0137–10 μM, Solarbio) was tested for synergy using CompuSyn (CI values) (33).

All mice were treated in accordance with the protocol that had been approved by the Animal Care and Use Committee of Huazhong Agricultural University (approval number: HZAUMO-2017-051). The female NOD-SCID mice (8 weeks old) were subcutaneously inoculated with 5 × 10⁶ Huh7 cells. When tumors reached 150 mm³, mice received cyclophosphamide (50 mg/kg, i.p.) to deplete lymphocytes. PBMCs (8 × 10⁶/mouse) and BsAbs (2.5 or 5 mg/kg, i.v.) were administered twice weekly for two weeks. Tumor volume was calculated as (length × width²)/2.

CD16A monoclonal antibodies (mAbs) were generated through murine immunization and phage display. The extracellular domain of CD16A (CD16A(ECD)) fused with a His-tag was expressed in HEK 293F cells. SDS-PAGE confirmed high purity of the purified protein ( Figure 1A ). Following immunization of five BALB/c mice and two New Zealand white rabbits with the recombinant protein, phage-displayed antibody libraries were constructed using splenic RNA. Four rounds of biopanning with CD16A-His ( Figure 1B ) yielded six murine-derived and four rabbit-derived CD16A mAbs (scFv-HisFlag), which were expressed and purified in E. coli. SDS-PAGE confirmed bands corresponding to the expected molecular weights ( Figure 1A ). ELISA revealed specific binding of all scFv-HF antibodies to CD16A(ECD)-His ( Figure 1C ), with EC50 values ranging from 0.08 to 60.73 nM ( Table 1 ). M29, M36, and R19 exhibited the highest affinities.

Cross-reactivity assays demonstrated binding of all CD16A mAbs to human CD16 but not murine mCD16A (OD450 <0.5), confirming species specificity ( Figure 1C ). Western blot further validated that four antibodies (e.g., CD16AM20 and CD16AR1) recognized linear epitopes of denatured CD16A ( Figure 1D ), while flow cytometry indicated strongest binding of CD16AM29 and CD16AR19 to native CD16A conformations ( Figure 1E ), suggesting distinct epitope preferences. Western blot band stacking at ~40 kDa reflects CD16A(ECD)-His glycosylation heterogeneity. The absence of visible bands for conformational epitope-binding mAbs (M5/M7/M29/M36) in non-denaturing conditions aligns with their inability to recognize linearized antigens (validated in Figures 1D-E ). Membrane artifacts (stray dots) resulted from transient air bubble entrapment during transfer and incomplete dissolution of blocking reagents. Flow cytometric validation was performed on CD16-negative A431 cells, CD16-positive primary human NK cells (PBMCs), and an engineered A431-CD16A cell line. As expected, none of the mAbs bound to parental A431 cells. In A431-CD16A cells, most mAbs exhibited weak binding except CD16AM19 and CD16AM20, likely due to challenges in re-screening with A431-CD16A cells ( Figure 1E ). Among human PBMCs (where NK cells account for ~15%), the highest binding activities were observed for CD16AM29, CD16AR19, CD16AM5, CD16AR2, CD16AR60, and CD16AR1. The weak CD16A signal in engineered A431-CD16A cells versus PBMCs arises from two factors: (1) Pre-existing selection markers (LG) in A431-LG cells hindered lentiviral transduction efficiency during CD16A overexpression, yielding lower surface expression than endogenous NK cells; (2) Epitope accessibility differences—the MA4-hFc-CD16A series preferentially binds native conformational epitopes that may be partially masked in the artificial overexpression system, whereas PBMC-derived NK cells present naturally processed CD16A conformers.

A tandem scFv format was employed for bispecific antibody (BsAb) construction, linking GPC3-scFv-hFc(N297A) to CD16A-scFv (N297A mutation in Fc to eliminate glycosylation and ADCC; Figure 2A ). BsAbs were expressed in 293F or CHO-S cells, purified via protein A affinity chromatography, and validated by SDS-PAGE. Under non-reducing conditions, the BsAbs formed ~300 kDa dimers composed of two 78 kDa single chains ( Figure 2A ). ELISA confirmed robust binding of MA4-hFc(N297A)-CD16A BsAbs (including M5, M7, M19, M20, M29, and M36) to both GPC3-His and CD16A-His ( Figures 2B, C ). Calculated EC50 values are summarized in Table 2 .

Flow cytometry assessed BsAb binding to A431 (CD16A-/GPC3-), GPC3+ hepatocellular carcinoma (HCC) lines (HepG2, Huh7, Hep3B), A431-GPC3 (G1), PBMCs, and A431-CD16A cells. Positive controls included hYP7(K)-OKT3(H) (T-cell targeting) and hYP7(K)-LC21(H).

(NK-cell targeting). At 10 μg/mL, all BsAbs bound specifically to GPC3+ HCC lines, G1 cells, and CD16A+ NK cells (comprising ~15% of PBMCs), but not to GPC3-/CD16A- A431 cells ( Figure 2D ). MA4-Fc(N297A)-CD16AM19 showed the strongest binding to Huh7, indicating recognition of native membrane-associated antigens.

Furthermore, fluorescence co-localization studies showed that in all cells examined, CD16AM5 was either free in the cytoplasm or bound to NK cells ( Supplementary Figure S1A ). This observation was also validated in A431 cells, where MA4-hFc-CD16AM5 showed a similar response. In the case of co-culture, the red fluorescence of Cy5 around tumor cells was observed to be in clear agreement with the cellular profile of PBMC (bright-field imaging) in multiple contact areas, especially in G1 and Huh7 cells. This finding demonstrates the efficacy of the BsAb in mediating spatial proximity and interaction between tumor cells and PBMCs.

BsAb activity was evaluated in HCC lines and GPC3-A431 cells (negative control) using a luciferase-based cytotoxicity assay. All target cells were stably transduced with firefly luciferase. In co-culture models (E:T ratio = 20:1), nearly all BsAbs exhibited dose-dependent cytotoxicity against GPC3+ HepG2, Hep3B, Huh7, and G1 cells, with no effect on GPC3- A431 cells ( Figure 3A ). Among six BsAbs, MA4-Fc(N297A)-CD16AM5/19/20 demonstrated potent cytotoxicity against GPC3+ cells (particularly Huh7), with maximal killing observed at ~1,000 ng/mL for Hep3B, HepG2, and Huh7, and ~10,000 ng/mL for G1 cells ( Figure 3A ). IC50 values are detailed in Table 3 . In contrast, MA4-Fc(N297A)-CD16AM7/29/36 showed significantly lower efficacy, particularly MA4-Fc(N297A)-CD16AM36. Subsequent experiments focused on MA4-Fc(N297A)-CD16AM5/19/20.