Engineering Bi-Specific CAR-NK Cells to Restore Antibody-Dependent Cellular Cytotoxicity in Solid Tumors

Highlights

What are the main findings?

• Impairment of NK cells under Tumor microenvironment (TME): Chronic exposure to Transforming growth factor-beta (TGF-β) induces a sustained dysfunctional NK cell phenotype characterized by the downregulation of key activating receptors (CD16 and NKG2D) and a profound inability to infiltrate 3D tumor structures.
• Dual-Targeting Potency: Developed a novel CAR-NK cell architecture incorporating both a tumor-specific scFv (e.g., targeting FRα) and a non-cleavable CD16 domain, which demonstrated ADCC-mediated synergistic cytotoxicity in vitro and tumoroids even when exposed to high concentrations of TGF-β.

What are the implications of the main findings?

• Versatility of CD16 Arming in CAR-NK: The study demonstrates that bi-specific CAR-NK platform can restore the synergy between cellular immunotherapy and monoclonal antibodies, potentially expanding the therapeutic window and preventing antigen escape through ADCC.
• Clinical Translation: These advances provide a compelling proof of concept for next-generation CAR-NK therapies engineered to function effectively within highly suppressive solid tumor environments.

Abstract

Natural Killer (NK) cell-based immunotherapy relies on CD16-mediated Antibody-Dependent Cellular Cytotoxicity (ADCC), yet the ovarian tumor microenvironment (TME) severely compromises this function via Transforming Growth Factor-beta (TGF-β). This study investigated the molecular mechanisms driving this suppression and evaluated a bi-specific Chimeric Antigen Receptor (CAR) strategy to overcome this hurdle. Primary PBNK cells exposed to TGF-β showed sustained canonical SMAD2 phosphorylation, accompanied by a marked reduction in activating receptors such as CD16 and NKG2D and an increase in exhaustion markers such as PD-1. Functionally, these phenotypic alterations led to failed infiltration and cytotoxicity in vitro and within ovarian cancer-derived spheroids. To overcome this limitation, we engineered NK-92 cells with a bi-specific CAR-targeting Folate Receptor Alpha (FRα) and CD16. While TGF-β typically impairs NK cell function, our armed CAR-NK cells successfully infiltrated tumoroids and synergized with Trastuzumab to induce potent ADCC-mediated lysis. Our findings define the TGF-β/SMAD2 axis as a central driver of NK cell dysfunction in ovarian cancer and demonstrate that bi-specific CAR-NK platforms offer a robust therapeutic solution to bypass TME-induced suppression and restore antibody-mediated tumor suppression.

1. Introduction

Natural Killer (NK) cells are critical components of the innate immune system, serving as the first line of defense against virally infected and malignant cells. Unlike T cells, NK cells do not require prior antigen sensitization, allowing for rapid cytotoxicity governed by a delicate balance of activating and inhibitory signals. Among the array of activating surface receptors, CD16 (FcγRIIIa) occupies a unique and pivotal role in cancer immunotherapy [1,2]. As the low-affinity receptor for the Fc portion of IgG antibodies, CD16 enables NK cells to recognize antibody-coated target cells and execute Antibody-Dependent Cellular Cytotoxicity (ADCC). This mechanism is the primary effector function underlying the clinical efficacy of several therapeutic monoclonal antibodies (mAbs), including rituximab, trastuzumab, and cetuximab. Consequently, the surface density and functional integrity of CD16 on Peripheral Blood NK (PBNK) cells are determinant factors in the success of antibody-based cancer therapies [3,4,5,6,7].

Despite this potential, the clinical utility of adoptive NK cell therapy is severely compromised by the immunosuppressive tumor microenvironment (TME). Transforming Growth Factor-beta (TGF-β), frequently overexpressed in malignancies such as ovarian cancer, acts as a potent negative regulator of NK cell biology [8,9,10,11]. Beyond its broad metabolic and anti-proliferative effects [12,13,14], TGF-β specifically targets NK cell effector function by inducing severe metabolic restriction and transcriptional repression mediated by sustained canonical SMAD signaling [15,16,17]. This suppression is compounded by the intrinsic instability of the CD16 receptor, which undergoes rapid downregulation upon activation. This loss of surface expression effectively “disarms” PBNK cells upon recruitment to the tumor site, rendering therapeutic antibodies inert and limiting the durability of conventional NK cell therapies [18,19].

To circumvent these intrinsic limitations, genetic engineering offers a strategy to generate “super-physiological” NK cells resistant to TME suppression. Previous studies have focused on overcoming the natural limitations of CD16 by expressing high-affinity polymorphisms, such as the F158V variant, which displays superior IgG binding affinity and elicits stronger ADCC responses compared to the low-affinity phenotype found in many patients [20,21]. However, while the 158V variant improves initial antibody engagement, it remains susceptible to proteolytic cleavage, leading to progressive receptor loss in the TME [22]. Subsequent strategies have successfully engineered “non-cleavable” CD16 receptors (e.g., containing an S197P mutation) to prevent proteolytic cleavage and sustain high surface density [23]. Yet these approaches remain dependent on the continuous presence of therapeutic antibodies and do not adequately address the challenges associated with antigen escape or low antigen density within the TME.

The development of bi-specific Chimeric Antigen Receptors (CARs) represents a pivotal advancement in adoptive immunotherapy. Here, we engineered a dual-modality receptor by fusing a non-cleavable CD16 with a single-chain variable fragment (scFv) targeting a tumor-specific antigen (Folate Receptor Alpha, FRα, in ovarian cancer). This design serves a synergistic dual purpose: the CD16 component restores ADCC capacity while simultaneously providing direct, antibody-independent tumor recognition. By integrating these two distinct regions into a single construct, this bi-specific targeting ensures NK cell activation and infiltration even in antigen-heterogeneous or antibody-limited conditions.

In this study, we first characterize the mechanisms driving NK cell dysfunction in the ovarian tumor microenvironment, identifying TGF-β/SMAD2-mediated signaling and metabolic restriction as key barriers. We then demonstrate that a bi-specific CD16-CAR NK-92 platform effectively bypasses this dual suppression. We provide evidence that, unlike their primary counterparts, these engineered effectors maintain robust surface expression and cytotoxic function, successfully executing ADCC and infiltrating 3D ovarian tumoroids. Collectively, these findings establish the bi-specific CD16-CAR as a pivotal therapeutic strategy to circumvent TGF-β-mediated immunosuppression and restore NK cell efficacy in solid tumors

2. Materials and Methods

2.1. Culture and Stimulation of Human Peripheral Blood CD56⁺ NK (PBNK) Cells

Cryopreserved human peripheral blood CD56⁺ natural killer (PBNK) cells were obtained from CGT global (Folsom, CA, USA). Cells were thawed according to the manufacturer’s instructions. Briefly, vials were rapidly thawed at 37 °C and washed in pre-warmed medium to remove cryoprotectants. The purity of CD3⁻CD56⁺ PBNK cells was confirmed by flow cytometry. PBNK cells were cultured in AlyS505-NK medium (Cell Science and Technology Institute, Miyagi, Japan) supplemented with 5% Immune Cell Serum Replacement (ICSR) (Thermo Fisher Scientific, Waltham, MA, USA). Cytokine stimulation the medium was supplemented with a cytokine cocktail comprising 600 IU/mL recombinant human IL-2 (Thermo Fisher Scientific), recombinant human IL-15 (10 ng/mL), and recombinant human IL-18 (50 ng/mL). The medium was refreshed every 2–3 days. For TGF-β suppression assays, PBNK cells were treated with recombinant human TGF-β1 (10 ng/mL; R&D Systems, Minneapolis, MN, USA) for defined time points (1 h, 24 h, 72 h, or 144 h) to assess temporal changes in phenotype and function.

2.2. NK-92 and Cancer Cell Line Culture

All cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). NK-92-MI (CRL-2408) were cultured in Alpha-MEM (Thermo Fisher Scientific) supplemented with 12.5% FBS, 12.5% horse serum (Thermo Fisher Scientific), 2 mM L-glutamine (Thermo Fisher Scientific), 0.2 mM inositol (Thermo Fisher Scientific), 0.1 mM 2-mercaptoethanol, 0.02 mM folic acid. SKOV-3 (ATCC, HTB-77) cells were maintained in McCoy’s 5A medium (Gibco, Waltham, MA, USA) supplemented with 10% FBS. OVCAR-3 cells were cultured in RPMI-1640 supplemented with 10% FBS. Lenti-X 293T (Takara Bio, 632180, Kusatsu, Shiga, Japan) cells were cultured in DMEM medium (Gibco) with 10% FBS. All cell lines were cultured under 37 °C and 5% CO₂ conditions. Cell lines were grown in a humidified incubator at 37 °C and 5% CO₂ and maintained with flask confluence not higher than 80%.

2.3. Generation of Bi-Specific CAR Construct

A bi-specific CAR was designed consisting of (i) folate alpha receptor-α scFv derived from the MoV19 antibody, (ii) CD16 extracellular domain, (iii) a CD8α hinge and transmembrane region, and (iv) 4-1BB and CD3ζ intracellular signaling domains. DNA fragments were synthesized (Integrated DNA Technologies, Coralville, IA, USA) and assembled using Gibson Assembly or NEBuilder HiFi DNA Assembly (NEB). To generate the CAR-NK cells, a customized lentiviral transfer vector was constructed using the pCDH-MCS-EF1α-Puro (System Biosciences) backbone as a template. To ensure robust and sustained transgene expression in the lymphoid lineage, the original CMV promoter was replaced with the Spleen Focus-Forming Virus (SFFV) promoter. The cloned CAR transgene encodes a multi-domain Chimeric Antigen Receptor comprising a signal peptide, an anti-FRα scFV for tumor targeting, and the extracellular domain of human CD16a (specifically the high affinity V158 variant to enhance ADCC), which are linked via a CD8α hinge and CD28 transmembrane domain to tandem CD28 and 4-1BB intracellular co-stimulatory domains and a CD3ζ activation domain (Supplementary Figure S2). Plasmids were verified by Sanger sequencing.

2.4. Bi-Specific CAR Lentivirus Production and Quantification

Lentivirus was produced by co-transfecting Lenti-X 293T cells with the transfer vector, along with the packaging plasmids pMDLg/pRRE and pRSV-REV, and the envelope plasmid pCMV-BaEV using Lipofectamine 3000 transfection reagent (Thermo Fisher Scientific). Viral supernatants were harvested at 18, 48 and 72 h post-transfection, filtered through a 0.45 µm filter, and concentrated using Lenti-X Concentrator (Takara Bio) according to the manufacturer’s protocol. Functional titer was determined by flow cytometry. HEK293T cells were transduced with serial dilutions of the concentrated viral stock. Transduction efficiency was analyzed 72 h post-infection by detection of CD16 expression using anti-CD16 antibody (Biolegend, San Diego, CA, USA). The functional titer (transducing units/mL, TU/mL) was calculated based on the percentage of positive cells as:

$$\mathrm{T}\mathrm{U}/\mathrm{m}\mathrm{L}={\displaystyle \frac{\mathrm{C}\mathrm{e}\mathrm{l}\mathrm{l} \mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\times \mathrm{\%} \mathrm{T}\mathrm{r}\mathrm{a}\mathrm{n}\mathrm{s}\mathrm{g}\mathrm{e}\mathrm{n}\mathrm{e} \mathrm{p}\mathrm{o}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e} \mathrm{c}\mathrm{e}\mathrm{l}\mathrm{l}\mathrm{s}}{\mathrm{V}\mathrm{i}\mathrm{r}\mathrm{u}\mathrm{s} \mathrm{v}\mathrm{o}\mathrm{l}\mathrm{u}\mathrm{m}\mathrm{e}}}$$

2.5. Transduction of NK-92 Cells

NK-92 cells were transduced with lentiviral vectors encoding the bi-specific CD16-based CAR construct. NK-92 cells (1 × 10⁵ cells/well) were seeded in 24-well plates in 0.25 mL of complete NK-92 medium as described above. Concentrated lentiviral supernatant was added at a multiplicity of infection (MOI) of 5, together with 8 µg/mL polybrene (Sigma-Aldrich, St. Louis, MO, USA) to enhance viral entry. The cells were mixed gently and subjected to spinoculation at 1800× g for 90 min at 28 °C using a benchtop centrifuge without brake. Following centrifugation, plates were incubated at 37 °C, 5% CO₂ for 24 h. After 24 h, viral supernatant was removed and replaced with fresh complete medium. Cells were expanded for 5–7 days with medium changes every 2–3 days. Cell viability was monitored using trypan blue exclusion with an automated cell counter (Luna counter, Logos Biosystems, Anyang-si, Gyeonggi-do, Republic of Korea).

2.6. Flow Cytometry Analysis of PBNK Cells

PBNK cells were cultured as described above. To examine the effect of TGF-β on PBNK cell phenotype, cells were treated with recombinant human TGF-β1 (10 ng/mL) for 1 h, 24 h, 72 h, or 144 h, while untreated cells cultured in parallel served as controls. At each time point, cells were harvested, washed with FACS buffer (PBS/2% FBS), and stained for surface marker expression. Cells were incubated with fluorochrome-conjugated monoclonal antibodies for 30 min at 4 °C in the dark, washed twice with FACS buffer, and re-suspended in PBS containing live/dead stain (Thermo Fisher Scientific) to exclude non-viable cells. The following panels were used for basic NK cell phenotype: CD3, CD56, CD16. Activating receptors: NKG2D, DNAM-1; and inhibitory receptors: PD-1, TIGIT. After staining, samples were acquired on the Beckman Coulter flow cytometry, and a minimum of 10,000 live lymphocyte events were collected per sample. Expression levels of CD16 and CD56, as well as activating and inhibitory receptors, were compared between control and TGF-β treated PBNK cells at each time point to assess temporal TGF-β-mediated suppression.

2.7. Flow Cytometric Analysis of NK-92 and CAR-NK-92 Cells

NK-92 and CAR-engineered NK-92 cells were harvested 6 days after transduction, washed twice with FACS buffer, and stained with fluorochrome-labeled monoclonal antibodies for 30 min at 4 °C. CAR expression was detected using human FITC-Labeled Human FOLR1 Protein (ACRO Biosystems, Newark, DE, USA) and anti-CD16 antibody. For phenotypic characterization, cells were stained for the following activating receptors: NKp46, NKp30, NKp44, NKG2D, DNAM-1, and 2B4. Inhibitory receptor expression was assessed using antibodies against PD-1, TIGIT, NKG2A, and CD94. A detailed list of antibody clones and fluorophores are provided in Supplementary Table S1.

2.8. Western Blot Analysis

Cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors (Roche, Basel, Switzerland). Protein concentrations were quantified using the BCA assay (Thermo Fisher). Equal protein amounts (20 µg) were resolved by SDS–PAGE and transferred to PVDF membranes (Millipore, Burlington, MA, USA). Membranes were blocked with 5% BSA in TBS-T and incubated overnight at 4 °C with primary antibodies against p-SMAD2, SMAD2, β-actin (Cell Signaling Technology, Danvers, MA, USA), and CD3ζ (Abcam, Cambridge, UK). Following incubation with HRP-conjugated secondary antibodies, protein bands were visualized using ECL substrate (Thermo Fisher Scientific) on a ChemiDoc MP System (Bio-Rad, Hercules, CA, USA).

2.9. Antibody-Dependent Cellular Cytotoxicity (ADCC) Assay

NK cell cytotoxicity was evaluated using a CFSE/7-AAD staining. Target SKOV-3 and OVCAR-3 were labeled with 0.5µM CFSE (Thermo Fisher Scientific) for 20 min at 37 °C followed by quenching with an equal volume of complete medium containing 10% FBS for 5 min at room temperature. Labeled cells were washed twice with complete medium and resuspended at the desired concentration. For ADCC conditions, CFSE-labeled target cells were incubated with the indicated therapeutic monoclonal antibody (trastuzumab, 10 µg/mL) for 15 min at 37 °C. Subsequently, target cells were co-cultured with effector NK cells (NK-92 or CAR-NK-92) at effector-to-target (E:T) ratios (5:1) in a V-bottom 96-well plate at a final volume of 200 µL. Control wells included: (i) target cells alone (spontaneous death), (ii) target cells + antibody only, and (iii) target cells + NK cells without antibody (non-ADCC NK cytotoxicity). Co-cultures were incubated for 4 h at 37 °C, 5% CO₂. At the end of incubation, cells were gently re-suspended and transferred to FACS tubes, washed once with FACS buffer (PBS/2%FBS), and stained with 7-AAD (Thermo Fisher Scientific) for 10 min at room temperature in the dark. Samples were acquired within 1 h on a flow cytometer (Beckman Coulter, Brea, CA, USA). CFSE was detected in the FITC channel, and 7-AAD was detected in a far-red channel according to the manufacturer’s settings. Target cells were identified as CFSE⁺ events, and dead targets were defined as CFSE⁺7-AAD⁺ cells. After exclusion of debris and doublets, the percentage of dead target cells was quantified within the CFSE⁺ gate. The percentage of specific ADCC-mediated lysis was calculated as:

$$\mathrm{\%} \mathrm{l}\mathrm{y}\mathrm{s}\mathrm{i}\mathrm{s}={\displaystyle \frac{(\mathrm{\%}7\mathrm{A}\mathrm{A}\mathrm{D}+\mathrm{i}\mathrm{n} \mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}-\mathrm{\%}7\mathrm{A}\mathrm{A}\mathrm{D} \mathrm{i}\mathrm{n} \mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l})}{(100-\mathrm{\%}7\mathrm{A}\mathrm{A}\mathrm{D}+\mathrm{i}\mathrm{n} \mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l})}}\times 100\mathrm{\%}$$

2.10. CD107a Degranulation Assay

NK cell degranulation was assessed by monitoring CD107a surface expression. Effector (NK-92 or CAR-NK-92) and target cells were co-cultured at a 5:1 ratio for 4 h at 37 °C. Post-incubation, cells were stained with antibodies specific for CD3, CD56, CD16, and CD107a, utilizing a viability dye (Thermo Fisher Scientific) to exclude dead cells. Flow cytometry was performed on a Beckman Coulter system, and degranulation was quantified as the percentage of CD107a⁺ events within the live CD3⁻CD56⁺ gated population.

2.11. TGF-β Challenge and Receptor Stability Assay

To evaluate the resistance of the bi-specific CAR to TME-induced immunosuppression, NK-92 and CAR-NK-92 cells were cultured in complete medium supplemented with recombinant human TGF-β (PeproTech, Cranbury, NJ, USA) at concentrations of 0, 5, or 10 ng/mL for 48 h. Following incubation, cells were harvested and washed. Receptor surface expression was analyzed by flow cytometry using anti-CD16 and anti-FRα antibodies. To assess the functional preservation of the Fc-binding domain, cells were incubated with Human IgG1 for 30 min at 4 °C, washed, and then stained with a PE-conjugated anti-human IgG secondary antibody. Mean Fluorescence Intensity (MFI) was quantified using a FACS and analyzed with FlowJo software (V10.10.0, BD Biosciences, Ashland, OR, USA).

2.12. Multiplex Cytokine Analysis Using LEGENDplex^TM Human 13-Plex Panel

Cytokine secretion by NK cells (PBNK, NK-92, or CAR-NK-92) was quantified using a bead-based multiplex immunoassay (LEGENDplex™ Human Cytokine Panel, 13-plex; BioLegend) according to the manufacturer’s protocol. Briefly, effector (NK-92 or CAR-NK-92) and target cells were co-cultured at a 5:1 ratio, and supernatants were collected at the indicated time points, centrifuged at 300× g for 5 min to remove debris, and stored at −80 °C until analysis. Briefly, 25 µL of each sample or standard was added to wells containing the mixed capture bead solution. Plates were sealed, protected from light, and incubated on a plate shaker at 800 rpm for 2 h at room temperature or overnight at 4 °C. Following incubation, wells were washed twice with wash buffer, and the beads were re-suspended in detection antibody cocktail. Plates were incubated for an additional 1 h at room temperature, followed by the addition of SA-PE reagent and incubation for 30 min in the dark. After final washes, beads were re-suspended in wash buffer and acquired on the Beckman Coulter. A minimum of 4000 bead events per analyte were collected. Cytokine concentrations were calculated from standard curves using LEGENDplex™ Data Analysis Software (BioLegend, San Diego, CA, USA) with 5-parameter logistic regression (5-PL) curve fitting. The 13-plex panel measured multiple NK-associated cytokines and effector molecules, including IFN-γ, Granzyme A, Granzyme B, and Perforin, depending on the selected panel configuration.

2.13. Mitochondria Respiration Analysis (Seahorse Assay)

Mitochondrial respiration was assessed using the Seahorse XF Cell Mito Stress Test Kit (Agilent Technologies, Santa Clara, CA, USA) on a Seahorse XFe96 Analyzer following the manufacturer’s instructions. Briefly, PBNK cells were treated with or without TGF-β (10 ng/mL) for 24 h, harvested, and seeded at a density of 200,000 cells/well onto Poly-L-lysine coated microplates in XF RPMI assay medium (pH 7.4). Oxygen Consumption Rate (OCR) was measured at baseline and following sequential injections of oligomycin (1.5 μM), FCCP (1.0 μM), and rotenone/antimycin A (0.5 μM). Data were analyzed using Wave software version 2.6 (Agilent Technologies).

2.14. Establishment of Orthotopic Xenografts and Tumoroid Generation

Six-to-eight-week-old female NSG mice (JaBio, Suwon, Gyeonggi-do, Republic of Korea) were housed under specific pathogen-free conditions at the Laboratory Animal Research Center of CHA University. All animal experiments were conducted in accordance with the approved IACUC protocol (IACUC 240076) and the relevant guidelines and regulations of the Laboratory Animal Research Center, CHA University. To establish ovarian cancer xenograft model, female NSG mice (8 weeks old) were intraperitoneal (i.p.) injected with 1 × 10⁶ SKOV-3-LUC tumor cells. Tumor growth was monitored via bioluminescence imaging. Upon development, primary tumors were harvested, minced, and enzymatically dissociated to result in single-cell suspension. Matrigel domes were seeded into pre-warmed 24-well plates and allowed to polymerize for 15–20 min at 37 °C. After polymerization, domes were overlaid with 500 µL of ovarian cancer organoid medium, consisting of Advanced DMEM/F12 supplemented with 1 × GlutaMAX, 1 × HEPES, 1 × B27 Supplement, 1 × N2 Supplement, 1 mM N-acetylcysteine, 10 mM nicotinamide, 50 ng/mL EGF, 100 ng/mL Noggin, and 100 ng/mL R-spondin 1, unless otherwise specified. Medium was replaced every 2–3 days. Tumoroids were cultured for 7–10 days until compact spheroid structures formed (diameter: 150–400 µm), confirmed by inverted microscopy.

2.15. Tumoroid Infiltration Assay of PBNK, NK-92, and CAR-NK-92

Tumoroid infiltration by PBNK, NK-92, and CAR-NK-92 cells was assessed using PKH26 fluorescent cell membrane labeling. NK cells were harvested, washed twice with serum-free RPMI medium, and resuspended at 2 × 10⁶ cells/mL. Cell membrane labeling was performed using the PKH26 Red Fluorescent Cell Linker Kit (Sigma-Aldrich) following the manufacturer’s protocol. Briefly, cells were incubated in PKH26 dye solution (2 µM) for 3–5 min at room temperature with gentle mixing. The labeling reaction was quenched with an equal volume of FBS, followed by two washes with complete medium to remove excess dye. Cell viability after labeling was confirmed using trypan blue exclusion. SKOV-3–derived tumoroids were generated as described above (Section 2.12) and used on days 7–10, when compact spheroid structures had formed. Tumoroids were gently harvested from Matrigel domes using dispase (Corning, Corning, NY, USA) and transferred into ultra-low attachment 96-well plates containing 150–200 µL of organoid culture medium. PKH26-labeled NK cells were added to each well at defined effector-to-tumoroid (E:T) ratios (2:1). Co-cultures were incubated for 6 and 24 h at 37 °C, 5% CO₂ to allow NK-cell infiltration. At the end of the incubation period, tumoroids were washed gently with pre-warmed medium to remove non-adherent NK cells. Tumoroid infiltration was assessed by high-content imaging system (ImageXpress Micro Confocal, IXM-C, Molecular Devices, San Jose, CA, USA). PKH26-positive NK cells were visualized using a red fluorescence channel (Ex/Em 551/567 nm). To further analyze the NK cell infiltration, the tumoroids were enzymatically dissociated into single-cell suspensions using Accutase/Trypsin-EDTA for 10 min at 37 °C. The resulting cell suspension was stained with anti-CD3, anti-CD56, and anti-CD16 antibodies. Infiltration was analyzed via flow cytometry. Infiltrating NK cells were identified by gating on the PKH26⁺ populations (pre-labeled effectors) and CD3⁻CD56⁺CD16⁺ population. The Infiltration (%) was calculated as the frequency of these specific NK cells relative to the total viable events acquired from the dissociated tumoroid.

2.16. LDH Assay

ADCC against SKOV-3 tumoroids was quantified using an LDH release assay (Thermo Fisher Scientific). Tumoroids were pre-incubated with Trastuzumab (10 µg/mL) for 20 min, followed by the addition of NK-92 or CAR-NK-92 effectors at a 5:1 E:T ratio. After a 24 h incubation at 37 °C, supernatants were harvested and mixed with LDH reaction buffer for 30 min. Absorbance was measured at 490 nm (reference 680 nm) using an Infinite 2000 promicroplate reader (Tecan, Männedorf, Switzerland)). Percent cytotoxicity was calculated using the standard LDH formula:

$$\mathrm{P}\mathrm{e}\mathrm{r}\mathrm{c}\mathrm{e}\mathrm{t}\mathrm{a}\mathrm{g}\mathrm{e} \mathrm{c}\mathrm{y}\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{o}\mathrm{x}\mathrm{i}\mathrm{c}\mathrm{i}\mathrm{t}\mathrm{y}={\displaystyle \frac{\mathrm{E}\mathrm{x}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{L}\mathrm{D}\mathrm{H}-\mathrm{S}\mathrm{p}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{e}\mathrm{o}\mathrm{u}\mathrm{s} \mathrm{L}\mathrm{D}\mathrm{H}}{\mathrm{M}\mathrm{a}\mathrm{x}\mathrm{i}\mathrm{m}\mathrm{u}\mathrm{n}\mathrm{m} \mathrm{L}\mathrm{D}\mathrm{H}-\mathrm{S}\mathrm{p}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{e}\mathrm{o}\mathrm{u}\mathrm{s} \mathrm{L}\mathrm{D}\mathrm{H}}}\times 100$$

2.17. Data and Statistical Analysis

Flow cytometry data were analyzed using FlowJo v10 software, and image densitometry was performed using ImageJ version 1.54r (National Institute of Health, NIH, Bethesda, MD, USA). All statistical analyses were conducted using GraphPad Prism 10.0 (GraphPad Software version 10.6.1 Boston, MA, USA). Data are presented as mean ± standard error of the mean (SEM) derived from at least three independent experiments or healthy donors. Differences between two groups were evaluated using paired or unpaired Student’s t-tests. Comparisons involving three or more groups were analyzed using One-way or Two-way Analysis of Variance (ANOVA) followed by Tukey’s or Šídák’s post hoc tests for multiple comparisons. A p-value of < 0.05 was considered statistically significant (* p < 0.05, ** p < 0.01, *** p < 0.001).

3. Results

3.1. TGF-β Remodels the PBNK Cell Surface Phenotype by Suppressing Activating Receptors and Promoting Exhaustion Markers

To characterize the impact of the tumor microenvironment cytokine TGF-β on NK cell phenotype, we cultured primary PBNK cells obtained from CGT global in the presence of recombinant human TGF-β. We observed a profound and time-dependent suppression of the key effector receptor CD16 (FcγRIIIa). Flow cytometry analysis revealed that while CD16 expression remained stable in untreated controls, TGF-β treatment induced a significant reduction in the frequency of CD56⁺CD16⁺ cells, particularly evident at 72 h and 144 h (Figure 1A,B). This downregulation was confirmed by quantification of the percentage of positive cells, which showed suppression of the CD16⁺ population compared to controls. To ensure that the observed downregulation of CD16 was not an artifact of selective cell death or donor-specific toxicity, we monitored cell viability and total cell yield throughout the 7-day culture period. As shown in Supplementary Table S2, cell viability remained high (>85%) across all time points for all three donors, with no significant differences observed between the TGF-β treated and untreated groups. This confirms that the reduction in surface CD16 expression reflects specific receptor modulation by the cytokine rather than selective survival of a subpopulation.

To determine whether these immunosuppressive effects were restricted to the CD16 pathway or represented a broader “disarming” of the NK cell, we analyzed the kinetics of other critical activating receptors known to be regulated via the canonical SMAD2 signaling axis. Longitudinal profiling revealed that TGF-β caused a progressive and profound downregulation of key activating receptors. At baseline, all PBNK cells expressed high levels of NKG2D and DNAM-1. While control NK cultures maintained these receptors relatively well over 144 h, TGF-β treated cells exhibited a continuous decline in surface expression, visible as a distinct divergence in the kinetic MFI trajectories (Figure 1J,M). By 72 h, the downregulation was substantial; NKG2D mean fluorescence intensity (MFI) in the TGF-β group was significantly reduced compared to controls, a gap that widened further by 144 h (p < 0.001; Figure 1K). DNAM-1 mirrored this trend. Kinetic analysis revealed a progressive downregulation of surface expression over the time course (Figure 1M). While a statistically significant reduction in DNAM-1 MFI was detected as early as 72 h (p < 0.001; Figure 1N), the profound loss of the DNAM-1 population was most visually apparent in the representative histograms by 144 h (Figure 1L).

Together with the loss of activating signals, TGF-β treatment significantly altered the expression of inhibitory checkpoint receptors, which are also key downstream targets of SMAD2-mediated regulation. Unlike the suppression observed with activating receptors, TGF-β resulted in a significant increase in the expression of the inhibitory receptor PD-1 (Figure 1D,E). In contrast, the expression of TIGIT remained largely unchanged between the treated and untreated groups, showing no significant difference (Figure 1G,H). This phenotype—characterized by CD16^lowNKG2D^lowPD-1^high suggests that TGF-β not only blinds NK cells to antibody-coated targets but also sensitizes them to inhibitory signals present in the tumor microenvironment.

3.2. TGF-β Induces Sustained Canonical SMAD2 Signaling and Metabolic Restriction, Leading to Impaired NK Cell Effector Function and Tumor Infiltration

To elucidate the intracellular signaling pathways driving this suppressive phenotype, we examined the activation of the canonical TGF-β signaling axis. Western blot analysis of PBNK cell lysates revealed a rapid and robust phosphorylation of SMAD2 (pSMAD2) upon exposure to TGF-β. TGF-β treated cells showed a clear pSMAD2 band as early as 1 h, which intensified progressively at 24 h and 72 h, and reached maximal intensity by 144 h of exposure (Figure 2A). Given that sustained TGF-β signaling has been linked to metabolic reprogramming in lymphocytes, we next investigated whether this activation impacted cellular bioenergetics. We assessed mitochondrial functions using a Seahorse XF extracellular flux analyzer to measure the Oxygen Consumption Rate (OCR). As shown in Figure 2C, PBNK cells pre-treated with TGF-β exhibited a significant suppression of mitochondrial respiration, characterized by reduced basal and maximal OCR compared to PBNK control. This indicates that TGF-β induces a state of metabolic restriction, limiting the bioenergetic capacity required for robust effector responses.

We next investigated whether this sustained canonical signaling correlated with functional deficits in NK cell effector molecules (Figure 2B). Consistent with the activation of the immunosuppressive SMAD pathway, TGF-β treatment led to a progressive downregulation of key cytotoxic mediators. In the untreated control group, cells pre-activated with IL-2, IL-15, and IL-18 maintained high constitutive levels of cytokine and granule secretion, which accumulated progressively over the 144 h period. However, while protein levels were comparable between groups at 1 h, a significant divergence emerged by 24 h; TGF-β exposed PBNK cells exhibited markedly reduced levels of IFN-γ compared to untreated controls—a suppression that persisted through 144 h. Similarly, the accumulation of lytic granule components Perforin and Granzyme A/B was significantly blunted in the TGF-β treated condition, indicating that TGF-β signaling effectively dampens the secretory output even of cytokine-primed NK cells. Together, these results demonstrate that TGF-β triggers sustained SMAD2 activation while simultaneously attenuating the production of the essential machinery required for NK cell-mediated lysis.

Finally, to validate the functional impact of this multi-layered suppression in a physiologically relevant model, we assessed the ability of NK cells to infiltrate 3D ovarian cancer tumoroids derived from the SKOV-3 xenograft model (Figure 2D). While PKH26-labeled control PBNK cells effectively infiltrated the tumoroids, accumulating within the core by 24 h, TGF-β pre-treated NK cells were severely impaired, remaining largely restricted to the tumoroid periphery (Figure 2E). Quantification confirmed a significant reduction in the infiltration index (Figure 2F). Collectively, these data demonstrate that TGF-β imposes a comprehensive suppressive program—encompassing canonical SMAD signaling, metabolic restriction, and reduced effector output—that ultimately prevents NK cells from physically penetrating and engaging the tumor mass. Also, this failure to penetrate the tumor mass likely results from the combined loss of activating receptors (CD16/NKG2D) (Figure 1) and adhesion molecules (DNAM-1), rendering NK cells unable to engage with the tumor matrix or target cells, thereby solidifying the immune-evasive state of the ovarian TME.

3.3. Engineering and Characterization of a Bi-Specific CD16-CAR NK-92 Cell Line

To overcome the inherent limitations of PBNK cells specifically their susceptibility to CD16 downregulation in the TME, we developed a genetically engineered NK cell line capable of dual targeting. A bi-specific CAR lentiviral construct (Figure 3A) comprising of (1) an scFv targeting Folate Receptor Alpha (FRα), a tumor antigen highly expressed in ovarian cancer and (2) non-cleavable CD16 was designed to restore ADCC function while simultaneously providing direct tumor recognition independent of CD16 shedding (Supplementary Figure S2 [23,24]). We transduced the CD16-negative NK-92 cell line (CAR-NK-92) with this bi-specific construct and validated expression via flow cytometry. The resulting CAR-NK-92 cells exhibited robust surface expression of the CAR, as evidenced by strong binding to FRα and high detection levels of CD16 (Figure 3B). High transduction efficiency was achieved, with the CAR-NK-92 population showing >96% positivity for CD16 surface expression (Figure 3C). Successful integration was further verified by Western Blot analysis, which revealed a distinct exogenous CD3ζ band at approximately 77 kDa in the CAR-NK-92 lysates, in addition to the endogenous 18 kDa CD3ζ band present in both lines (Figure 3D).

To determine whether CAR transduction affected the global phenotypic profile of the NK-92 cells, we analyzed the expression of key activating and inhibitory receptors. Flow cytometry analysis revealed that the CAR-NK-92 cells maintained a receptor profile comparable to that of the parental NK-92 cells (Figure 3E,F). Quantification of percent positive cells showed no significant alterations in the expression of inhibitory receptors (PD-1, TIGIT, NKG2A, CD94) or major activating receptors (NKp46, NKp30, NKp44, NKG2D, DNAM-1, 2B4) following genetic engineering (Figure 3F). Furthermore, to ensure that this gating strategy did not mask subtle shifts in receptor density, we further quantified the Mean Fluorescence Intensity (MFI) of these markers. As shown in Supplementary Figure S3, no significant differences in surface density were observed between the two groups for any of the receptors tested, confirming that the lentiviral transduction and CAR expression did not alter the intrinsic phenotypic profile of the effector cells. These data suggest that the introduction of the bi-specific CAR confers specific targeting capabilities without compromising the intrinsic cytotoxic potential of the NK-92 cells.

3.4. Bi-Specific CAR-NK-92 Cells Exhibit Synergistic Cytotoxicity Against FRα+HER2+ Ovarian Cancer Cells via ADCC and CAR Signaling

To evaluate the functional efficacy of the engineered bi-specific CAR-NK-92 cells, we performed in vitro cytotoxicity assays against the human ovarian cancer cell lines, SKOV-3 and OVCAR-3, as well as the antigen-negative control A2780. We first verified the surface expression of the target antigens; flow cytometric analysis confirmed that both SKOV-3 and OVCAR-3 cells strongly express FRα (Figure 4A), the target of the CAR scFv, and HER2, the target for Trastuzumab-mediated ADCC (Figure 4B), whereas A2780 cells lack both antigens. The cytotoxic potential of the NK cells was then assessed across diverse effector-to-target (E:T) ratios (5:1, 1:1, 0.5:1) in the presence and absence of Trastuzumab (Herceptin). Parental NK-92 cells, which lack endogenous CD16, exhibited moderate baseline cytotoxicity against both target cell lines, and not significant enhancement upon the addition of Trastuzumab, confirming their inherent inability to mediate ADCC (Figure 4F,J). In contrast, CAR-NK-92 cells demonstrated significantly higher specific lysis even in the absence of antibody (SKOV-3: ~60%, OVCAR-3: ~50%), indicating effective direct recognition of FRα via the CAR domain. Crucially, this cytotoxicity was strictly antigen-specific; CAR-NK-92 cells exhibited negligible killing of the antigen-negative A2780 cells, showing lysis levels comparable to parental controls (Figure 4E). To validate the function of the CD16 domain, we assessed ADCC activity in the presence of Trastuzumab. The addition of Trastuzumab to CAR-NK-92 co-cultures induced a synergistic increase in cytotoxicity against HER2 targets (Figure 4C,D). To confirm that this effect was not due to non-specific Fc interactions, we performed specificity controls using a IgG1 isotype antibody at the optimal 5:1 E:T ratio. While the isotype control failed to induce lysis above baseline (ns), Trastuzumab significantly boosted total lysis to approximately 80% for SKOV-3 and 78% for OVCAR-3 (Figure 4F,J). These results confirmed that the CD16 domain integrated into the CAR construct is fully functional and capable of triggering robust ADCC, effectively complementing the CAR-mediated killing to maximize therapeutic potency against antigen-positive ovarian cancer cells.

This enhanced cytotoxic response was further confirmed by analyzing NK cell degranulation and cytokine secretion. The frequency of CD107a+ degranulating cells was significantly elevated in the CAR-NK-92 + Trastuzumab group compared to CAR-NK-92 alone or parental controls (Figure 4H,L). Furthermore, multiplex cytokine analysis revealed a significant upregulation of pro-inflammatory and cytotoxic mediators. While parental NK-92 cells produced negligible levels of cytokines, CAR-NK-92 cells showed substantial secretion of IFN-γ, Perforin, and Granzyme A upon target engagement. The combination of CAR engagement (FRα) and ADCC (HER2/Trastuzumab) led to the highest secretion levels, with IFN-γ concentrations reaching ~20,000~30,000 pg/mL and perforin levels exceeding 1500~6000 pg/mL (Figure 4I,M). These results demonstrate that the bi-specific CAR-NK-92 cells effectively overcome the limitations of conventional NK therapy by dual-targeting tumor antigens, leading to superior degranulation, cytokine production, and tumor cell lysis.

3.5. Bi-Specific CAR-NK-92 Cells Maintain Functional Stability and Potency Under High-Dose TGF-β Stress

A major limitation of conventional NK cell therapy in ovarian cancer is the downregulation of CD16 receptor expression mediated by TGF-β in the TME. To confirm that our engineering strategy successfully bypasses this suppression mechanism, we challenged CAR-NK-92 cells with increasing concentrations of TGF-β (0, 5, and 10 ng/mL). Flow cytometric analysis revealed that the CAR-NK-92 exhibited remarkable stability. Unlike endogenous CD16, which is prone to shedding, the surface expression of both the FRα-targeting domain (Figure 5A) and the CD16 signaling domain (Figure 5B) remained constant across all TGF-β concentrations tested. Crucially, we assessed the functional integrity of the receptor by measuring its capacity to bind the Fc region of therapeutic antibodies. As shown in Figure 5C, the Human IgG1 binding capacity (MFI) of CAR-NK-92 cells was fully preserved even at the highest TGF-β concentration (10 ng/mL), demonstrating that the proteolytic cleavage site had been effectively eliminated.

To determine if this physical stability translated to functional resilience, we performed ADCC assays under these same stress conditions. While conventional NK cell cytotoxicity was significantly blunted by TGF-β exposure, CAR-NK-92 cells maintained high potency, exhibiting robust lysis of target cells comparable to non-treated controls (Figure 5D). These results definitively demonstrate that the non-cleavable CD16 domain confers resistance to the TGF-β, ensuring sustained therapeutic efficacy in the immunosuppressive TME.

3.6. CAR-NK-92 Cells Exhibit Enhanced Infiltration and ADCC Potency in Tumoroid Model

To further validate the therapeutic potential of the bi-specific CAR-NK-92 cells in a clinically representative setting, we challenged the ovarian cancer tumoroids generated from SKOV-3-LUC cells-which mimic the dense architecture of the TME-with our engineered cells. We utilized a matrigel-embedded culture system to recapitulate the physical barriers and immunosuppressive condition of solid tumors (Figure 6A) [25]. To analyze infiltration capability, we employed an infiltration assay where tumoroids were seeded on top of the matrigel bed. Microscopic analysis revealed that while parental NK-92 cells failed to penetrate or impact the structural integrity of the tumoroids, CAR-NK-92 cells effectively infiltrated the tumoroids (Figure 6B, Supplementary Figure S4). Consistent with the fluorescence imaging, bright-field microscopy revealed that CAR-NK-92 cells formed dense clusters around and penetrated the periphery of the SKOV-3 tumoroids more effectively than parental NK-92 cells. As shown in Supplementary Figure S5, these effector cells established direct physical contact with the tumoroid surface, further confirming the recruitment capability of the bi-specific receptor. Quantitative analysis via specific lysis and LDH release assays corroborated these visual findings. CAR-NK-92 cells exhibited significant higher baseline cytotoxicity against the tumoroids compared to parental controls (~50% vs. ~20%). Most notably, the addition of Trastuzumab resulted in a synergistic enhancement of killing, elevating the cytotoxicity to approximately 75% (p < 0.001; Figure 6C) and significantly increasing LDH cytotoxicity (Figure 6D). This demonstrates that the restored CD16 receptor remains highly functional and resistant to the immunosuppressive physical constraints of the 3D ovarian tumoroid, enabling robust antibody-mediated killing.

This potent antitumor response was accompanied by a robust secretory profile. Analysis of the supernatant from the tumoroid co-cultures revealed that the combination of CAR-NK-92 cells and Trastuzumab elicited the highest levels of cytolytic and pro-inflammatory molecules. We observed statistically significant increases in the concentrations of IFN-γ, Perforin, Granzyme A, Granzyme B compared to the CAR-only and parental groups (Figure 6E). This synergistic activation pattern was further corroborated by Z-score normalization, which illustrated a distinct clustering of the ADCC-enhanced group (Figure 6F). Collectively, these data demonstrate that the bi-specific CAR-NK-92 cells are not only capable of penetrating solid tumor structures but also maintain a hyper-functional state, effectively orchestrating both CAR-directed lysis and potent ADCC within a complex, immunosuppressive TME.

4. Discussion

The clinical success of NK cell-based immunotherapies against solid tumors has been hindered by the immunosuppressive TME, which systematically dismantles NK cell cytotoxic machinery [26,27]. In this study, we elucidate a dual mechanism of NK cell dysfunction driven by the TGF-β/SMAD signaling axis and TGF-β-mediated metabolic restriction and the intrinsic proteolytic vulnerability of CD16. We demonstrate that this suppression leads to the failure of primary PBNK cells to infiltrate 3D tumoroids. To overcome this, we engineered a bi-specific CAR NK-92 cell line targeting FRα and non-cleavable CD16, which successfully bypass these suppressive checkpoints to mediate potent ADCC and tumor infiltration in a physiological 3D ovarian cancer model.

Our investigation into the kinetics of TGF-β signaling reveals that the suppression of NK cell function is not merely a transient response but a sustained “reprogramming” event. We observed that TGF-β induces rapid and prolonged phosphorylation of SMAD2 (pSMAD2), persisting up to 144 h. This sustained canonical signaling directly correlates with the transcriptional downregulation of critical activating receptors [28]. We observed marked downregulation of NKG2D and DNAM-1—receptors crucial for tumor recognition—alongside increased expression of PD-1 on NK cells. This shift toward an “exhausted” profile aligns with known effects of TGF-β in solid tumors: for example, tumor-infiltrating NK cells in TGF-β-rich milieus often acquire an ILC1-like state with high levels of inhibitory receptors (PD-1, TIGIT, TIM-3) and impaired cytotoxic function [29,30]. It also aligns with recent findings by Cordoba et al. [31], who identified SMAD4-dependent pathways as central to TGF-β-mediated “disarming” of NK cells. However, our data extends this understanding by linking sustained SMAD2 activation to severe metabolic restriction. Using metabolic flux analysis (Seahorse assay), we demonstrated that TGF-β significantly suppresses metabolic respiration (OCR) in NK cells. This metabolic paralysis likely restricts the energetic resources required for the synthesis and mobilization of cytotoxic payloads, explaining the marked depletion of intracellular cytotoxic payloads (Perforin/Granzyme B), rendering the cells functionally inert even before they encounter a target.

While TGF-β drives transcriptional suppression, our data highlights that the intrinsic proteolytic instability of CD16 remains a critical vulnerability within the TME. Even without explicit upregulation of metalloproteases, the basal shedding of CD16 acts as a constant drain on NK cell efficacy [31,32]. This proteolytic loss has immediate functional consequences: the NK cell can disengage from the opsonized target cell, terminating the immune synapse. While shedding may allow NK cells to detach and potentially seek new targets, it also limits the duration and efficacy of ADCC, especially if CD16 is lost too early or in a sustained manner. In the TME, chronic TGF-β and hypoxic stress might takeover this pathway to permanently reduce surface CD16, thus blunting NK cell responsiveness to therapeutic antibodies [18,33,34].The consequence of this dual suppression was starkly evident in our infiltration assays: TGF-β-treated PBNK cells, lacking both adhesion molecules (DNAM-1) and activating receptors (NKG2D/CD16), failed to penetrate the tumoroid core. This exclusion from the tumor infiltration represents a primary reason for the clinical failure of adoptive NK therapies in solid tumors.

To circumvent these barriers, we developed a bi-specific CAR strategy. By genetically engineering NK-92 cells which are inherently lacking CD16 and thus resistant to certain TME regulatory loops with a non-cleavable CD16 signaling domain fused to an FRα-targeting scFv, we restored and enhanced ADCC capability. In vitro results demonstrate that these CAR-NK-92 cells exhibit “super-physiological” cytotoxicity. The dual targeting of FRα (via the CAR) and HER2 (via Trastuzumab-mediated ADCC) resulted in synergistic killing of SKOV-3 and OVCAR-3 cells. This is consistent with prior studies demonstrating that introduction of high-affinity CD16 into NK-92 converts it into an “ADCC-competent” cell line. For example, Freitas et al. (2024) created NK-92 clones expressing various FCGR3A alleles and found that all CD16-transfected NK-92 lines exhibited robust ADCC against tumor targets [35]. Importantly, this synergy was accompanied by a massive release of IFN-γ and Perforin, significantly exceeding that of parental lines.

The definitive proof that proteolytic instability is a critical bottleneck in this context is provided by our genetic rescue experiments. By engineering a CAR construct that specifically lacks the cleavage motif (Val196/Ser197), we observed a complete preservation of receptor stability and Fc-binding capacity even in the presence of high-dose TGF-β (10 ng/mL) (Figure 5). Furthermore, this structural stability translated directly to functional rescue: unlike primary NK cells which succumb to TGF-β suppression, our CAR-NK-92 cells retained their full ADCC potency under stress conditions. This ‘genetic bypass’ confirms that preventing proteolytic shedding is sufficient to restore immune surveillance, even in a cytokine-rich environment that suppresses endogenous NK cell metabolism.

Most significantly, the efficacy of this engineering approach was validated in a physiologically relevant 3D tumoroid model. Unlike parental NK-92 cells, the CAR-NK-92 cells demonstrated robust infiltration into the dense tumoroid architecture [36,37] accompanied by tumoroid disintegration and a significant increase in LDH release. This indicates that the engineered high-affinity CD16 receptor remains functionally active in overcoming physical barriers within the 3D architecture. The enhanced retention likely stems from stronger immune synapse formation and antigen-mediated tethering, which allows the engineered cells to navigate and persist within the dense tumoroid matrix more effectively than non-engineered controls. One possible explanation is that engagement of CD16 by antibodies activates inside-out signaling that strengthens integrin-mediated adhesion or triggers chemokine release, aiding NK cell infiltration [38,39]. Another factor could be that effective killing of tumor cells by CD16⁺ NK-92 breaks down physical barriers in the spheroid, allowing deeper penetration. While the precise mechanism needs further investigation, our data clearly indicates that engineered NK cells retain functionality in a hostile 3D tumor environment better than non-engineered parental cells.

While our findings demonstrate robust efficacy in both 2D and complex 3D models, we acknowledge that this study utilized an ex vivo approach to model the tumor microenvironment. To maximize physiological relevance, we employed tumoroids derived from orthotopic xenografts, ensuring that the target cells retained the phenotypic signatures and drug resistance profiles acquired in vivo. This 3D model effectively recapitulated critical barriers to NK cell efficacy, including dense physical architecture and TGF-β-mediated immunosuppression. However, this system does not account for systemic factors such as effector cell biodistribution, long-term persistence in circulation, or potential off-target toxicities in non-tumor tissues. Consequently, future studies will focus on evaluating the therapeutic durability and safety profile of these bi-specific CAR-NK-92 cells in humanized orthotopic mouse models to support clinical translation.

Our results have important translational implications. They demonstrate that the immunosuppressive barriers of the solid tumor microenvironment—specifically the downregulation of intrinsic NK cell machinery—can be effectively circumvented by genetic engineering. By arming NK cells with a non-cleavable CD16, we designed a therapeutic strategy that synergizes with existing tumor-targeting antibodies. In the context of ovarian cancer, we demonstrated that this engineered NK cell platform works synergistically with trastuzumab to kill SKOV-3 and OVCAR-3 cells and effectively infiltrate 3D tumoroids. By uncoupling effector function from the inhibitory signaling axes we identified in primary cells, this strategy offers a robust approach to maintain cytotoxicity in the hostile environment of solid tumors. Notably, such an approach is not limited to HER2; it could be generalized to any tumor antigen for which a monoclonal antibody (mAb) or bi-specific targeting agent is available. For instance, CD20 (rituximab in lymphomas), EGFR (cetuximab in colorectal/head-neck cancer), GD2 (dinutuximab in neuroblastoma), and others are all amenable to NK cell ADCC. An off-the-shelf NK cell line expressing CD16 thus represents a versatile adjunct to antibody therapies, circumventing the variability of patient NK cell Fc receptor genotypes and the often-impaired state of endogenous NK cells in cancer patients.

5. Conclusions

This study defines the TGF-β-mediated metabolic restriction and intrinsic receptor instability as a critical, multi-layered barrier that severely compromises the efficacy of NK cell-based immunotherapies in ovarian cancer. We provide evidence that sustained SMAD2 signaling drives transcription repression and metabolic paralysis, while the proteolytic vulnerability of CD16 further prevents NK cells from mounting an effective ADCC response within the tumor core. To overcome these barriers, we propose the bi-specific CD16-CAR NK-92 platform as a superior therapeutic strategy. By combining direct tumor antigen recognition with high-affinity ADCC capabilities, this “off-the-shelf” product effectively bypasses the intrinsic limitations of primary PBNK cells. These data support the further clinical translation of bi-specific CAR-NK therapies, highlighting their potential to restore the “missing link” between therapeutic antibodies and cellular cytotoxicity in the hostile microenvironment of solid malignancies.