Human primary NK cells were isolated from the whole blood of healthy volunteers with written informed consent as approved by the local ethics committee of Ärztekammer Nordrhein (2020272). NK cell isolation was performed using CD3 MicroBeads followed by CD56 MicroBeads according to the manufacturer’s instructions. Freshly isolated human NK cells were primed at a density of 1 × 10⁶ cells/mL in a 24-well plate using NK MACS Medium supplemented with 5% AB serum (Access Biologicals, Vista, CA, USA), 140 U/mL recombinant human IL-15, 500 U/mL recombinant human IL-2, and 2000 U/mL recombinant human IL-1β and cultured at 37°C in a 5% CO2 incubator. Two days after seeding, BaEV-LV, premixed with Vectofusin-1, were added dropwise to the activated NK cells, and gently mixed by pipetting. Spinoculation was then applied at 32°C and 400 × g for 2 hours (h). Vectofusin-1 was used at a final concentration of 2.5 µg/mL. One day after transduction, NK cells were washed to remove excess BaEV-LV and cultured further in NK MACS Medium supplemented with 5% AB serum, 140 U/mL IL-15, and 500 U/mL IL-2. Regular feeding and medium changes were performed every 2–3 days to maintain the cell density below 2 x 10⁶ cells/mL until harvest on day 14. The effects of IL-1β, serum, spinoculation, and Vectofusin-1 was assessed in this study to establish the optimized transduction protocol outlined here.

For automated generation of genetically modified NK cells, the NK Cell Transduction (NKCT) process was used on the closed and cGMP-compliant CliniMACS Prodigy platform. NK cells were isolated from leukapheresis, acquired from healthy volunteers, with written informed consent, as approved by the local ethics committees of Hannover Medical School (4718NM), University Hospital Ulm (172/99), and University Hospital Cologne (03-055). NK cell separation was performed by automated immunomagnetic selection. First, T cells were depleted with CliniMACS CD3 Reagent, followed by enrichment of NK cells using CliniMACS CD56 Reagent. For cytokine-based activation of NK cells, 1 x 10⁸ NK cells were seeded in 70 mL of NK MACS GMP Medium supplemented with 5% AB Serum, 140 U/mL MACS GMP recombinant human IL-15, 500 U/mL MACS GMP recombinant human IL-2, and additional 5 mL of MACS GMP recombinant human IL-1β to achieve a final working concentration of 2000 U/mL. Cells were cultivated at 37°C and 5% CO₂. The pH value was maintained between 7 and 7.8. The temperature was automatically maintained by Prodigy devices. CO₂ concentration was set by the gassing input controlled by Prodigy devices. Two days after isolation, NK cells were transduced with BaEV-LV encoding BDCA2 CAR (described below). BaEV-LV, premixed with MACS GMP Vectofusin-1, was added to the activated NK cells at a multiplicity of infection (MOI) of 0.4. The final concentration of Vectofusin-1 was 2.5 µg/mL. The NK cells were then subjected to spinoculation at 32°C and 400 x g for 2 h. One day post-transduction, NK cells were washed to remove excess BaEV-LV and subsequently cultured in NK MACS GMP medium supplemented with 5% AB serum, 140 U/mL IL-15 and 500 U/mL IL-2 with automated feeding and medium exchange until harvest on day 14. During the automated cell harvest, NK cells were washed and resuspended in 100 mL of CliniMACS Formulation Solution for subsequent use.

A second-generation BDCA2-specific CAR was generated using a previously described monoclonal antibody against BDCA2 (13, 16). Briefly, the BDCA2-specific scFv was fused in frame with the CD8 hinge and transmembrane domain, the 4-1BB (CD137) costimulatory domain, and the CD3ζ (CD247) activation domain. A leader peptide derived from GM-CSFRα was included to facilitate CAR cell surface expression. Third-generation self-inactivating BaEV-LV encoding the BDCA2 CAR were produced via transient transfection of HEK293T cells, as described previously (10).

Cellular composition and NK cell phenotype were determined by using the following antibodies: anti-human CD45-VioBlue (REA747), anti-human CD14-VioGreen (REA599), anti-human TCRab-FITC (REA652), anti-human CD3-PE (SK7, BioLegend), anti-human CD56-APC (REA196), and anti-human CD19-APC-Vio770 (REA675). BDCA2 CAR expression was analyzed using a biotinylated anti-BDCA2 CAR idiotype antibody, followed by detection with an anti-biotin-Vio Bright B515 antibody (REA746). All staining procedures were performed in the presence of FcR Blocking Reagent. For the exclusion of dead and apoptotic cells, 7-amino-actinomycin D (7-AAD) was used. Data were acquired by using a MACSQuant^® Analyzer 10 and analyzed either manually or automatically with NK Cell Express Mode in the MACSQuantify 2.13 software (Miltenyi Biotec, Bergisch Gladbach, Germany).

After 14 days of culture, genomic DNA was purified from 1 × 10⁶ harvested CAR NK cells using the DNeasy Tissue Kit (Qiagen, Hilden, Germany). LV encoded gag gene copies per genome were determined in relation to the human reference gene PTBP2 by quantitative PCR using the MACS COPYcheck Kit.

For cytotoxicity analysis, RS4;11 cells engineered to express enhanced green fluorescent protein (EGFP) and firefly luciferase (ffluc) (hereafter RS4;11), or RS4;11 cells additionally expressing human BDCA2 (hereafter RS4;11/BDCA2), were seeded at a density of 2 × 10⁴ cells per well in flat-bottom 96-well plates. Subsequently, either untransduced (UTD) or BDCA2 CAR NK cells were added at the designated effector-to-target (E:T) ratios. Samples containing tumor cells without NK cells served as controls. The long-term in vitro antitumor activity of BDCA2 CAR NK cells was assessed using a rechallenge assay. CAR NK cells were co-incubated with 3 x 10⁴ RS4;11/BDCA2 tumor cells at an E:T ratio of 1:2, followed by the addition of 3 x 10⁴ fresh tumor cells every 24 h. Samples were stained with 7-AAD to exclude dead cells after co-incubation and viable GFP-positive RS4;11 cells were quantified using a MACSQuant Analyzer 10. Cytotoxicity of NK cells was determined as (1 - (number of viable GFP-positive cells/number of control GFP-positive cells)) × 100%. To quantify cytokine secretion, tumor cells were co-incubated with either UTD NK or CAR NK cells at an E:T ratio of 1:1 in a flat-bottom 96-well plate for 24 h. Supernatants were then harvested and analyzed using a MACSPlex Cytokine 12 Kit according to the manufacturer’s instructions.

The in vivo study was approved by the state animal research committee (LANUV, North Rhine-Westphalia, Germany) and animals were cared for according to guidelines of the Federation of European Laboratory Animal Science Associations. Eight- to ten-week-old female NOD-SCID IL2R γ^null (NSG) mice (Charles River Laboratories, Sulzfeld, Germany) were intravenously (i.v.) injected with 5 × 10⁵ RS4;11/BDCA2 cells through the tail vein. Seven days later, mice received an i.v. injection of 1 × 10⁷ CliniMACS Prodigy-manufactured BDCA2 CAR NK cells or an equal number of UTD NK cells. Two groups of animals, treated with either UTD NK cells or BDCA2 CAR NK cells, received daily subcutaneous injections of 25,000 IU IL-2 (Proleukin S; Novartis Pharma, Nürnberg, Germany), starting one day before (day -1) and continuing through nine days after NK cell treatment (day 8). A group of mice that did not receive any NK cells following tumor inoculation served as controls. Tumor load was assessed every 2–3 days by bioluminescence imaging in an IVIS Lumina II system (PerkinElmer, Rodgau, Germany). Mice were anesthetized using 1.5-2% isoflurane in oxygen as the carrier gas during imaging. Luminescence was analyzed using the Living Image software (PerkinElmer, Rodgau, Germany). Upon reaching endpoint criteria or the conclusion of the experiment, mice were euthanized using a CO₂ gas flow regulator with a control valve appropriate for the cage size (6 L/min for EU Type II long cages and 13 L/min for EU Type III cages). The persistence of tumor cells, NK cells and CAR NK cells in the bone marrow, blood, and spleen was determined by flow cytometry. The following antibodies were used: anti-human CD45-VioBlue (REA747), anti-human BDCA2-PE (AC144), anti-human CD56-APC (REA196), anti-mouse CD45-APC-Vio770 (REA737), anti-mouse Ter119-PE-Vio 770 (REA847), biotinylated anti-BDCA2 CAR idiotype antibody, and anti-biotin-Vio Bright B515 antibody (REA746). Dead cells were excluded by Propidium Iodide (PI).

All statistical analyses were performed using GraphPad Prism 10 software (GraphPad Software, Boston, MA, USA). Data was analyzed using two-tailed paired or unpaired Student’s t tests, or two-way analysis of variance (ANOVA). A p value < 0.05 was considered statistically significant.

All reagents, tubing set systems, and devices, software used in this study were obtained from Miltenyi Biotec, Bergisch Gladbach, Germany, unless otherwise specified.

To enable automated NK cell transduction on the CliniMACS Prodigy platform, we first optimized the BaEV-LV-based transduction protocol for its compatibility. The impact of IL-1β and human AB serum on NK cell transduction efficiency was evaluated. Freshly isolated NK cells were activated for 2 days with IL-2 and IL-15, with or without IL-1β, in NK MACS Medium supplemented with 5% human AB serum. NK cells were then transduced with BDCA2 BaEV-LV in the presence or absence of human AB serum. Transduction efficiency was assessed 12 days post-transduction. Neither IL-1β nor human AB serum markedly influenced transduction efficiency (Figure 1A, Supplementary Figure 1). However, IL-1β activation enhanced CAR NK cell expansion compared to non-IL-1β activated controls (Figure 1B, Supplementary Figure 1). These results permit the omission of the serum-removal washing step prior to transduction. Subsequently, spinoculation was confirmed as essential to achieve efficient BaEV-LV-mediated NK cell transduction (Figure 1C). Lastly, the concentration of Vectofusin-1 was optimized during transduction. Although Vectofusin-1 substantially increased transduction efficiency, concentrations above 2.5 µg/mL did not yield further improvement (Figure 1D). Higher concentrations did not adversely affect CAR NK cell numbers and viability, indicating no cytotoxicity introduced by Vectofusin-1 (Figures 1E, F).

The automated NKCT process was developed on the CliniMACS Prodigy platform to manufacture clinical-grade CAR NK cells (Figure 2). Before isolation, the concentration, volume, and proportions of CD3+ and CD56+ cells in a donor-derived leukapheresis product are assessed to set parameters for automated separation. NK cells are then selected via immunomagnetic depletion of CD3+ cells, followed by enrichment of CD56+ cells on day 0. After cytokine-based, feeder cell-free activation, NK cells are transduced by BaEV-LV on day 2 and expanded in NK MACS medium supplemented with IL-15 and IL-2 over 14 days. Following separation, NK cells became the predominant population, increasing from an average of 8.12% to 93.65%, with only residual T cells, B cells, monocytes, and other cells including granulocytes remaining (Figure 3A). NK cell purity further increased to 98.72% on day 7 and 99.45% on day 14 during the culture (Figure 3B). NK cell recovery rates after the CD3 depletion and the CD56 enrichment were 83.03% and 36.08%, respectively (Figure 3C). NK cells showed high viability of more than 91% on average throughout the process (Figure 3D). On day 14, an average of 1.04 x 10⁹ total NK cells were harvested at the end of the process (Figure 3E). BDCA2 BaEV-LV produced under cGMP conditions were used to generate CAR NK cells. The average NK cell transduction efficiency and vector copy number (VCN) of the CAR NK cell products were 59.64% and 2.57, respectively (Figures 3F, G).

To target BDCA2-expressing tumor cells, we designed and generated a second-generation BDCA2 CAR composed of a BDCA2-specific scFv, the CD8 hinge and transmembrane domains, and the intracellular domains of 4-1BB and CD3ζ (Figure 4A). The antitumor activity of BDCA2 CAR NK cells generated by the NKCT process on CliniMACS Prodigy was first evaluated in vitro. CAR NK cells were co-cultured with either RS4;11 or RS4;11/BDCA2 target tumor cells for 4 h at different E:T ratios. CAR NK cells displayed significantly higher cytotoxicity against BDCA2-expressing tumor cells, compared to BDCA2-negative tumor cells (Figure 4B). A prolonged coculture period further enhanced tumor cell lysis by CAR NK cells, eliminating 61% of tumor cells after 24 h at the lowest E:T ratio of 0.16:1 (Figure 4C). Furthermore, BDCA2 CAR NK cells secreted significantly higher levels of proinflammatory cytokines, including GM-CSF, IFNγ and TNFα, when cultured with RS4;11/BDCA2 cells compared to RS4;11 cells, demonstrating specific activation of CAR NK cells upon tumor target engagement (Figure 4D). No detectable secretion of IFNα, IL-2, IL-4, IL-5, IL-6, IL-9, IL-10, IL-12A or IL-17A was observed (Figure 4D).

The in vivo antitumor efficacy of CliniMACS Prodigy-manufactured BDCA2 CAR NK cells with the NKCT process was further evaluated in a xenograft mouse model. BDCA2 CAR NK cells and UTD NK cells derived from the same healthy donor were manufactured in parallel on two CliniMACS Prodigy devices using the NKCT process. After 14 days, a total of 4.09 x 10⁹ NK cells with a transduction efficiency of 34.69% were harvested for the CAR NK cell culture, while a total of 6.28 x 10⁹ UTD NK cells were produced (Supplementary Figures 2A, B). To assess the long-term effector potency of BDCA2 CAR NK cells used in the in vivo experiment, UTD NK cells or CAR NK cells were cultured with 3 x 10⁴ RS4;11/BDCA2 tumor cells at an initial E:T ratio of 1:2. NK cells were then rechallenged with 3 x 10⁴ fresh tumor cells every 24 h for 3 days. Over time, BDCA2 CAR NK cells exhibited significantly higher cytotoxicity against RS4;11/BDCA2 tumor cells compared to UTD NK cells (Supplementary Figure 2C).

NSG mice received 5x10⁵ RS4;11/BDCA2 tumor cells via the tail vein. Seven days later, mice were treated with either 1x10⁷ BDCA2 CAR NK cells or an equal number of UTD NK cells. To maintain NK cell persistence, daily subcutaneous injection of IL-2 was given to 2 cohorts of mice treated with UTD NK cells or CAR NK cells (Figure 5A). Prior to NK cell infusion, mice were randomized into five treatment groups, all of which exhibited comparable tumor burden (Supplementary Figure 3). In the untreated and UTD NK cell groups, tumor growth rapidly progressed regardless of IL-2 support. Infusion of CAR NK cells without IL-2 support resulted in only a slight delay in tumor progression (Figures 5B, C). On the contrary, CAR NK cells combined with IL-2 treatment led to significant tumor regression, although all mice eventually relapsed when IL-2 support was withdrawn (Figures 5B, C).

Consistent with these findings, mice treated with CAR NK cells without IL-2 support had high tumor loads in the bone marrow (BM), blood, and spleen at the end of the study, while those receiving CAR NK cells with IL-2 support showed significantly reduced tumor burden (Figure 6A). No pronounced persistence of NK cells or CAR NK cells was observed in the BM, blood or spleen of mice treated with CAR NK cells without IL-2 support. However, a small population of human NK cells was found in the spleens of 4 out of 6 mice treated with CAR NK cells and IL-2 (Figures 6B, C). These persistent NK cells did not exhibit detectable CAR expression on their cell surfaces (Figures 6B, C). No BDCA2 tumor antigen loss were observed in mice treated with CAR NK cells (Supplementary Figure 4). No weight loss or other abnormalities were observed in mice during the experiment, indicating the interventions were well tolerated (Supplementary Figure 5).