Generation of an NFκB-Driven Alpharetroviral “All-in-One” Vector Construct as a Potent Tool for CAR NK Cell Therapy

Immune cell therapeutics are increasingly applied in oncology. Especially chimeric antigen receptor (CAR) T cells are successfully used to treat several B cell malignancies. Efforts to engineer CAR T cells for improved activity against solid tumors include co-delivery of pro-inflammatory cytokines in addition to CARs, via either constitutive cytokine expression or inducible cytokine expression triggered by CAR recognition of its target antigen—so-called “T cells redirected for universal cytokine-mediated killing” (TRUCKs) or fourth-generation CARs. Here, we tested the hypothesis that TRUCK principles could be expanded to improve anticancer functions of NK cells. A comparison of the functionality of inducible promoters responsive to NFAT or NFκB in NK cells showed that, in contrast to T cells, the inclusion of NFκB-responsive elements within the inducible promoter construct was essential for CAR-inducible expression of the transgene. We demonstrated that GD2CAR-specific activation induced a tight NFκB-promoter-driven cytokine release in NK-92 and primary NK cells together with an enhanced cytotoxic capacity against GD2+ target cells, also shown by increased secretion of cytolytic cytokines. The data demonstrate biologically relevant differences between T and NK cells that are important when clinically translating the TRUCK concept to NK cells for the treatment of solid malignancies.


INTRODUCTION
Chimeric antigen receptors (CARs) as synthetic receptors combine antigen recognition and immune cell activation in one molecule and are designed to genetically modify immune cells, like T and NK cells, which find their application in clinical trials for adoptive anticancer cell therapy (1,2). One of the most successful recent approaches in cancer immunotherapy was marked by the first US Food and Drug Administration (FDA) approvals of CAR T cell therapies against CD19 + hematological malignancies, including B cell lymphoma, acute lymphoblastic leukemia, and multiple myeloma (3)(4)(5)(6). However, to date, adoptive cell-based CAR immunotherapies are limited in the treatment of solid tumors due to the immunosuppressive tumor microenvironment (TME), which results in insufficient tissue homing, restricted proliferation, and short-time persistence of modified immune cells within the tumor (7)(8)(9). Moreover, the antigenic heterogeneity of solid tumors limits the number of targetable antigens for CAR cell therapy (10,11). Preclinical and clinical studies using T cells redirected for universal cytokinemediated killing (TRUCKs) demonstrated improved antitumor activity due to additional CAR-mediated cytokine secretion after target antigen recognition in the tumor bed. Specific cytokine deposition promotes formation of an antitumoral and proinflammatory milieu in the TME, supporting the potential of such approaches to advance the treatment of solid tumors (12)(13)(14). As a further extension of the well-established TRUCK principle, we previously generated an "all-in-one" vector construct that combines constitutive GD2-specific CAR expression and inducible nuclear factor of activated T cells (NFAT)-driven cytokine secretion. T cells engineered with the "all-in-one" vector showed improved killing capacity against GD2 + target cells and enhanced migration capacities of macrophages in vitro (15).
However, there are some limitations that negatively impact CAR T cell success in the treatment of solid tumors; some of these limitations may not apply to CAR NK cells due to differences in cell function and signaling (16)(17)(18). Preclinical and clinical trials using CAR NK cells indicate important advantages of CAR NK cells over CAR T cells in adoptive cell therapy, including the option to prepare modified immune cells from healthy donors, potent antitumor effects, and therapeutic safety (19,20). Thus, primary autologous and allogeneic NK cells including available NK cell lines, e.g., NK-92, are sources for CAR NK cell trials and simplify manufacturing of CAR NK cells with reduced treatment cost (21). First CAR NK cell studies showed antitumor effects in both hematologic and solid malignancies as well as low incidences of on-target/off-tumor toxicity and cytokine release syndrome (CRS) (22,23). For instance, an early clinical study of CAR NK cells in hematological malignancies demonstrated their safe toxicity profile and efficiency to persist in peripheral blood of patients for more than 12 months (24). Additionally, Liu et al. reported a complete absence of graft-versus-host disease (GvHD) after treatment with CAR NK cells (24). For solid tumors, several preclinical studies targeting glioblastoma (GBM) (25), breast (26), ovarian (27), and pancreatic cancers (28), and Ewing sarcomas (29) demonstrated the potential of CAR NK cells and led to clinical trials, e.g., NCT03415100, NCT03940820, NCT03941457, and NCT03383978.
These beneficial CAR NK cell properties prompted us to transfer the "all-in-one" vector concept from T cells to NK cells with the aim to improve adoptive CAR cell therapy for solid tumors. As proof of concept, we focused on the target antigen disialoganglioside GD2, which is overexpressed in neuronal malignancies, such as GBM and neuroblastoma (NB) (30,31). GBM and NB are two of the most common malignancies of the nervous system. GBMs account for 16% of all primary brain cancers and are the most frequent and aggressive malignancy of the central nervous system (32). NBs derive from the sympathetic nervous system and are the most frequent extracranial pediatric solid neoplasm accounting for 7% of all cancers in children under 15 years and 15% of all pediatric cancer-related deaths (33), indicating the clinical need for novel treatment strategies. As an additional treatment modality, CAR cell therapies may further increase the probabilities of diseasefree survival.
To transfer the "all-in-one" vector concept to NK cells, we designed alpharetroviral "all-in-one" vector constructs with a GD2CAR and an inducible promoter-driven transgene based on the previously published lentiviral "all-in-one" vector constructs (15). The objective is based on recent studies indicating that alpharetroviral vectors modify NK cells more efficiently than lentiviral vectors (34,35). Although it was previously described that NFAT signaling is relevant for human NK cell activation, other studies also showed that nuclear factor kappa-light-chain-enhancer of activated B cells (NFkB), which plays a key role in cytokine production and granule exocytosis in NK cells, is important for activation of NK cells and, thereby, potentially a superior inducible promoter element (36,37). We generated alpharetroviral "all-in-one" vectors that combine constitutive GD2CAR expression and inducible cytokine expression under the control of an inducible NFkB promoter element. Here, we demonstrate the overall relevance of the NFkB signaling pathway in CARmediated activation in NK cells and show selective induction of the EGFP reporter as well as the anticancer cytokine human interleukin-12 (hIL-12) in modified NK-92 cells and primary NK cells after GD2 recognition. Moreover, we obtained an augmented in vitro cytotoxic capacity of engineered "all-inone" CAR NK cells against GD2 + target cells assessed via lactate dehydrogenase (LDH) assays and against patient-derived primary GD2 + GBM cells shown by an increased release of NK cell-specific cytokines associated with enhanced lytic potential. We also observed that modified "all-in-one" NFkBmIL2.hIL12.GD2CAR NK-92 cells show a trend to recruit monocytic cells in vitro due to the induced IL-12 release after GD2CAR-specific stimulation.
Taken together, our data provide the basis to improve adoptive NK cell therapy approaches using the CAR and TRUCK concepts. Our engineered NFkB-driven "all-in-one" vector construct is a promising advance toward effective NK cell-based clinical strategies to treat solid tumors.

MATERIAL AND METHODS
Cloning of Alpharetroviral "All-in-One" Vectors, Production, and Titration of Viral Supernatants To generate novel alpharetroviral "all-in-one" self-inactivating (SIN) (24) vectors, multiple cloning steps were performed. In brief, previously described well-established codon-optimized lentiviral "all-in-one" SIN vectors (pCCL.PPT.NFATenhsyn. E G F P . P G K . G D 2 C A R . P R E , p C C L . P P T . N F A T m I L 2 . E G F P . P G K . G D 2 C A R . P R E , p C C L . P P T . N F A T e n h s y n .hIL12.PGK.GD2CAR.PRE, pCCL.PPT.NFATmIL2.hIL12. PGK.GD2CAR.PRE) (15), which function well in primary T cells, provided the basis for our work. A second-generation GD2CAR including GD2-scFv (14.G2a), IgG1 hinge, CD28 transmembrane, CD137 (4-1BB) co-stimulatory, and CD3-z signaling domains was used for the "all-in-one" vector construct (38). As an additional CAR, a second-generation CAR targeting the carcinoembryonic antigen (CEA) BW431/ 26-scFv-Fc-CD28/CD3-z (607) was used (39). This CAR harbors an IgG1 (Fc) hinge region fused to a CD28 transmembrane and CD28-CD3-z signaling domain. As inducible genes of interest (iGOIs), EGFP and human IL-12 (single-chain p40-p35) were used. The "all-in-one" gene sequence was removed from the lentiviral backbone of CAR vectors and inserted into an alpharetroviral SIN vector backbone. An additional modification for suitable NK cell application was made by exchanging the inducible 6xNFAT response elements with two generated NFkB response elements (2xNFkB; each 52 bp). The cloned vectors were verified by Sanger sequencing (Microsynth Seqlab, Göttingen, Germany). Cloning details and sequences are available on request.

Primary Human NK Cells and Patient-Derived GBM Samples
Peripheral blood isolated from anonymous healthy donors was kindly provided by Transfusion Medicine (Hannover Medical School, Hannover, Germany). PBMCs were isolated via Leucosep ™ (Greiner Bio-One, Kremsmünster, Austria) in combination with Bicoll Separating Solution (Ficoll Separating Solution, Biochrom) according to the manufacturer's instructions. The separated lymphocytes were used to isolate NK cells or stored at -80°C until separation. Primary CD56 + NK cells were isolated from PBMCs with the human NK Cell Isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's protocol. In brief, a determined PBMC number was labeled firstly with NK Cell Biotin-Antibody Cocktail and secondly, after incubation and washing, with NK Cell MicroBead Cocktail. After a second incubation and washing, labeled cells were separated in equilibrated columns in the magnetic field of a MACS Separator. The collected flow-through consisted of enriched unlabeled CD56 + NK cells. The purity of isolated cells was quantified via flow cytometric staining for CD56 and CD3. Isolated CD56 + CD3 -NK cells were cultivated in NK MACS Medium (MACS Miltenyi Biotec) with 5% human serum (c.c.pro, Oberdorla, Germany), 100 U/ml penicillin (PAN-Biotech), 100 µg/ml streptomycin (PAN-Biotech), 1% NK MACS supplement (MACS Miltenyi Biotec), 500 IU/ml of hIL-2 (Proleukin ® , Novartis), and 70 ng/ml hIL-15 (PeproTech, Hamburg, Germany).
The amount of used viral vector supernatants was calculated regarding the desired multiplicity of infection (MOI) (44). The transduction efficiency was determined via transgene staining and flow cytometric analysis at least 2 days after transduction.

Western Blot
Collected cell pellets were lysed in lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 100 mM NaF, 1% Triton X-100 enriched with 1 mM DTT, 1 mM Na 3 VO 4 , and 1× cOmplete Mini Protease Inhibitor Cocktail (Roche Diagnostics, Mannheim, Germany)), incubated on ice for 20 min, and centrifuged for 15 min at high speed and 4°C. The protein concentration was determined via Bradford assay (Bio-Rad Laboratories, Hercules, CA, USA). Proteins were separated via sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Protein sizes were identified with the help of the loaded Color Prestained Protein Standard ladder (New England Biolabs, Frankfurt am Main, Germany). After blotting separated proteins on a nitrocellulose membrane, membranes were blocked in Trisbuffer containing 5% milk powder and incubated with an anti-CD3-z-HRP-antibody (6B10.2, Santa Cruz Biotechnology, Dallas, TX, USA) to detect the endogenous CD3-z and the CD3-z domain of the CAR. As an additional control, membranes were also incubated with an antibody to detect the common housekeeping protein glyceraldehyde-3-phosphatedehydrogenase (GAPDH) (GeneTex, Eching, Germany). To detect the antibody protein binding, SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific) was used according to the manufacturer's instructions. Protein bands were visualized via a FusionFX instrument (Peqlab GmbH, Erlangen, Germany).

Coculture Assays
Modified and unmodified NK cells were cocultured with GD2 + or GD2adherent or suspension target cells in the desired effector to target cell ratios (E:T) to investigate the specific inducible gene expression after GD2 antigen recognition. NFAT-or NFkB-driven EGFP induction was detected in addition to GD2CAR expression by flow cytometric analysis at defined time points. Inducible cytokine secretion was detected via ELISA, Bio-Plex assay, or LEGENDplex TM .
Cytokine Analysis via ELISA, Bio-Plex Assay, or LEGENDplex ™ Cell culture supernatants of genetically modified and unmodified NK cell monocultures and NK cell cocultures with GD2 + or GD2target cells were collected at desired time points to measure induced cytokine secretion after antigen recognition. Supernatants were analyzed by enzyme-linked immunosorbent assay (ELISA). For NFAT-or NFkB-driven human IL-12 secretion, the Human IL-12 (p70) ELISA MAX ™ Deluxe Sets (BioLegend) was used according to the manufacturer's instructions. The absorbance of the samples was measured with the SpectraMax ® 340PC384 Microplate Reader (Molecular Devices, San Jose, CA, USA). The cytokine amount was calculated regarding measured absorbance and according to the manufacturer's protocol.
In addition, coculture supernatants were analyzed via Bio-Plex Pro Human Cytokine 27-plex Assay (Bio-Rad Laboratories) or LEGENDplex ™ Human CD8/NK Panel (BioLegend) according to the manufacturer's protocols. For Bio-Plex assay, the provided plates were coated with magnetic beads and washed. Standards, blanks, and samples were added and incubated while shaking. After washing, the detection antibody was applied and incubated with shaking to bind desired biomarkers. After repeating washing, the wells were incubated with streptavidin. Then, the wells were washed and incubated with assay buffer while shaking. Afterward, the fluorescence of samples was measured by Bio-Plex 200 System (Bio-Rad Laboratories) and analyzed by Bio-Plex Manager 6.0 (Bio-Rad Laboratories). For LEGENDplex ™ assay, standards, and diluted samples were incubated with premixed beads in a V-bottom plate overnight at 4°C. Afterward, the samples were incubated with detection antibody for 1 h and streptavidin-PE for 30 min at room temperature. Analysis was performed on a CytoFLEX (Beckman Coulter) and analyzed on legendplex.qognit.com (BioLegend).

Cytotoxicity Assay
The cytolytic capacity of modified NK-92 cells against GD2 + / GD2target cell lines (HT1080, HT1080 GD2, and SH-SY5Y) was analyzed by determination of the lactate dehydrogenase (LDH) release using the Pierce ™ LDH Cytotoxicity Assay Kit (Thermo Scientific, Bremen, Germany). Cocultures were performed in 96-well plates with an E:T of 6:1, and 50 µl of cell coculture supernatants were collected after 24 h and used for the assay according to the manufacturer's instructions. Fifty microliters of the provided reaction mix were added, and incubated for 30 min, and the reaction was stopped by adding 50 µl of the provided stop solution. Absorbance was determined at 490 nm and 680 nm (SPECTRAmax 340PC, Molecular Devices, Software SoftMax Pro 4.0) and cytototoxicity was calculated according to the manufacturer's protocol.

Migration Assay
Coculture supernatants were tested for their capacity to induce the migration of human monocytic cells through specific pore polycarbonate membranes using a modified Boyden chamber (Neuro Probe, MBB96). In more detail, a 96-well flat bottom plate was filled with 300 ml/well of 0.6% agarose and was placed inside the modified Boyden chamber after inserting the appropriate spacer. Then, 120 ml of sample was added into each well and a membrane with 8-mm pore size (Neuro Probe) was placed over the plate in direct contact with the cell culture supernatants. The chamber was closed, and 5 × 10 5 MONO-MAC-6 cells/well (in 200 ml RPMI medium) were added to the top wells of the chamber. The entire chamber was incubated at 37°C with 5% CO 2 for 4 h. After the incubation time, the membrane was removed from the chamber and a scraper was used in order to remove the remaining cells from the top of the membrane. Migrated and membrane-attached cells were fixed with methanol for 10 min and stained with Giemsa's solution for 1 h. Non-specific staining was removed via a washing step with water, and the membrane was dried overnight. Representative pictures for each sample were taken using an Olympus iX71 microscope, and all stained cells were counted. Different concentrations of the IL-12 cytokine (0-500 pg/ml) diluted with the same coculture medium were used as positive controls.

Microscopy and Live-Cell Imaging
The Axio Observer fluorescence microscope (ZEISS, Oberkochen, Germany) was used to take bright field cell images and to observe the fluorescence of target and effector cells in coculture experiments. An inducible GD2CARdependent EGFP expression in modified NK cells was detected in 24-h coculture experiments with GD2 + adherent and suspension target cells and patient-derived GBM spheroids. Additionally, the dynamic NFkB-dependent EGFP upregulation over time was visualized by live-cell imaging for modified NK cells cocultured with GD2 + RFP657 + KG-1a cells. RFP657 + KG-1a cells were distinguished from effector cells in cocultures by their red fluorescence. Microscope images were analyzed by AxioVision Software Rel. 4.8.

Statistical Analysis
Presented data were statistically analyzed by GraphPad Prism version 6.0 (GraphPad Software, San Diego, CA, USA). One-way ANOVA and two-way ANOVA tests with Tukey's multiplecomparison tests and paired t-tests were used as indicated. Significance is indicated in all figures and legends (****p ≤ 0.0001; ***p ≤ 0.001; **p ≤ 0.01; *p ≤ 0.05; ns p > 0.05). For technical and biological replicates, mean values ± SD (standard deviation) are shown.

RESULTS
The NFAT-Driven Alpharetroviral "All-in-One" Vector System Was Insufficient in GD2CAR-Triggered Induction of Transgene Expression in NK-92 Cells Previous data of lentiviral "all-in-one" vector-modified T cells demonstrated the potent NFAT-driven upregulation of the inducible gene of interest (iGOI) upon GD2CAR-specific activation (15). The lentiviral "all-in-one" vector constructs for targeted T cell gene therapy were accordingly transferred to an alpharetroviral SIN vector system to create an effective platform for targeted NK cell therapy ( Figure 1A). Analogously to the lentiviral vector architecture, alpharetroviral "all-in-one" SIN vector constructs combine constitutive second-generation CAR expression under the control of the human phosphoglycerate kinase (hPGK) promoter and NFAT-driven inducible expression of an iGOI, e.g., EGFP or hIL-12. Two different NFAT-driven inducible promoters, namely, NFATmIL2 or NFATsyn, both containing six NFAT consensus response elements (6xNFAT) fused to either a minimal IL2 promoter (48) or a synthetic promoter element and a synthetic TATA box (49) were used. Modification of NK-92 cells with these alpharetroviral vectors, which were pseudotyped with the RD114/TR envelope glycoprotein, showed average transduction rates from 50% to 75% ( Figure 1B). The vector copy number (VCN) of modified NK-92 cells was analyzed by quantitative PCR to detect the wPRE sequence of the alpharetroviral vector construct and showed an average of one to two vector integrations per cell ( Figure 1C). Immunoblot analyses confirmed the successful CAR expression of the "all-in-one" vectors in NK-92 cells on protein level as detected by expression of the CD3-z domain of the GD2CAR (about 53 kDa) ( Figure 1D). The expression of endogenous CD3-z in NK-92 cells (about 17 kDa) and the housekeeping protein GAPDH were used as internal controls.
To assess the functionality and feasibility of the NFAT-driven "all-in-one" vector constructs for NK cell therapy, modified NK-92 cells were analyzed for GD2CAR-specific activation in a proof-of-concept cocultivation experiment with GD2 + target cell lines (HT1080 GD2 and SH-SY5Y). HT1080 cells, determined to be GD2 -, served as a control target cell line ( Figure 1E). NK-92 cells were modified with "all-in-one" alpharetroviral SIN vectors harboring an inducible EGFP or IL-12 expression cassette under the control of the different NFAT-driven promoters, NFATmIL2 and NFATsyn, and cocultured with GD2 + or GD2target cells for 24 h. Specific GD2CAR activation upon recognition of the target antigen GD2 resulted in an NFAT-driven transgene upregulation as detected by flow cytometry for modular EGFP expression or by ELISA for IL-12 secretion in coculture supernatants. NK-92 cells transduced with constructs harboring an EGFP expression cassette under the control of the NFATsyn-responsive promoter led to a constant background expression in unstimulated modified NK-92 cells and modified NK-92 cells cocultured with GD2 -HT1080 cells ( Figure 1F). Moderate EGFP expression was observed in NK-92 cells modified with the NFATsyn.EGFP.GD2CAR construct after GD2CAR-specific activation upon coculture with GD2 + HT1080 GD2 or SH-SY5Y cells. The highest EGFP expression was reached after 4 to 8 h of coculture ( Figure 1F). NK-92 cells transduced with constructs harboring an NFATmIL2-driven inducible EGFP expression cassette did not show increased GD2CAR-specific EGFP expression after cocultivation with the GD2 + target cells ( Figure 1F). Comparably, NK-92 cells modified with "all-inone" vector constructs harboring an NFAT-driven IL-12 cytokine cassette did not show an enhancement of GD2CARspecific IL-12 secretion after cocultivation with GD2 + target cells, but did show a high background IL-12 secretion with GD2target cells ( Figure 1G). In addition to the background expression, NK-92 cells modified with NFATsyn-driven "allin-one" constructs showed a slightly better IL-12 secretion after stimulation with GD2 + target cells compared to NK-92 cells expressing the NFATmIL2 promoter, but the level of secreted IL-12 was low, compared to previous results observed in T cells (15) (Figure 1G).
In conclusion, NFAT-driven inducible expression cassettes coupled to constitutive GD2CAR-specific signaling were far less efficient in NK-92 cells compared to T cells, which necessitated further optimization of the vector architecture to provoke "all-inone" vector feasibility in CAR NK cell therapy.
Optimization of Alpharetroviral "All-in-One" SIN Vectors by Incorporation of Inducible NFkB Promoter Elements NFkB is described as a key element in NK signaling pathways (50). Moreover, it was previously shown that NFkB was efficiently used for transgene induction in NK cells (37). Thus, to improve GD2CAR-specific inducible transgene expression in NK-92 cells, the six NFAT response elements within the inducible expression cassette were replaced by two NFkB response elements, as two NFkB consensus elements (2xNFkB) were linked to the minimal IL2 promoter (NFkBmIL2) or the synthetic promoter fused to the TATA box (NFkBsyn) (Figure 2A).
NK-92 cells were modified with the newly designed vector constructs and analyzed for GD2CAR surface expression by staining the scFv region of the GD2CAR with a specific antiidiotype antibody, Ganglidiomab. Average transduction efficiencies achieved with an MOI of 10 ranged between 40% and 70% ( Figure 2B). Similarly to the NFAT-driven "all-in-one" vector constructs, the VCNs within the modified NK-92 cells were approximately 1 for most of the constructs ( Figure 2C). Western blot analysis confirmed GD2CAR expression in NK-92 cells as shown exemplarily for the NFkB-driven "all-in-one" inducible EGFP vectors and visualized by the expression of the CD3-z domain of the CAR ( Figure 2D). Expression of endogenous CD3-z and the housekeeping protein GAPDH served as controls.
The NFkB-driven "all-in-one" constructs pseudotyped with RD114/TR were also tested in primary NK cells that were enriched from human peripheral blood mononuclear cells (PBMCs) by magnetic separation ( Figure S1A). Transduction efficiencies of approximately 30% on average for NFkB-driven constructs harboring the EGFP expression cassette and of <20% on average for NFkB-driven constructs harboring the IL-12 expression cassette were reached ( Figure S1B). GD2CAR integration in primary NK cells was verified by VCN determination with measured VCNs of approximately 1 per cell ( Figure S1C) and the CAR expression by immunoblot detection of the CD3-z domain of the GD2CAR ( Figure S1D).

GD2CAR-Triggered Induction of NFkB-Driven Transgene Expression in Modified NK-92 Cells
To analyze the functionality and feasibility of the alpharetroviral NFkB-driven inducible "all-in-one" vector constructs, coculture experiments of modified NK-92 cells with GD2 -(HT1080, KG-1a) or GD2 + (HT1080 GD2, KG-1a GD2, and SH-SY5Y) target cells as well as with primary patient-derived GD2 + GBM cells were performed. 24 h after cocultivation with GD2 + target cells ( Figure S2A). After a 24-h coculture with HT1080 GD2 target cells, the NFkBsyn promoter led to on average 60% EGFP upregulation in modified NK-92 cells compared with on average 50% EGFP upregulation driven by the NFkBmIL2 promoter. The same trend was observed with KG-1a cells expressing GD2 ( Figure 3A). Coculture experiments with GD2 + SH-SY5Y cells showed 60% EGFP expression in modified NFkBsyn NK-92 cells compared to 40% NFkBmIL2-driven EGFP induction after GD2CAR-specifi c a c tivat ion. The GD2CARspecific stimulation of NFkBsyn.EGFP.GD2CAR-modified NK-92 cells by GD2 + target cells showed a twofold or fivefold EGFP induction compared to non-specific activated modified NK-92 cells (w/o stimulation; HT1080; KG-1a). NFkBmIL2.EGFP.GD2CAR-modified NK-92 cells showed a sevenfold or 10-fold EGFP induction after antigen recognition ( Figure 3A). For all tested constructs, the density of EGFP expression recorded by the mean fluorescence intensity (MFI) was similar ( Figures S2B, C). Regarding off-target leakiness and background expression of the inducible transgene, the NFkBmIL2 promoter element led to less background EGFP expression of unstimulated modified NK-92 cells and modified NK-92 cells cocultured with GD2target cells compared to the NFkBsyn promoter. The NFkBsyn promoter showed an EGFP upregulation of more than 20% in unstimulated NK-92 cells and transduced NK-92 cells cocultured with HT1080 target cells, indicating that this vector configuration was less tightly controlled as compared to that with the NFkBmIL2 promoter element ( Figure 3A). Similar results were observed in 24-h co-cultures of unsorted modified primary NK cells with GD2 + -or GD2 − adherent target cells at an E:T ratio of 10:1 with maximum average EGFP induction of 30% in NFkBsyn-modified cells compared to 25% in NFkBmIL2-engineered cells after GD2CAR-specific activation by HT1080 GD2 cells. The background EGFP expression of around 5% was also lower in the NFkBmIL2 vector configuration compared to the NFkBsyn vector configuration with 10% ( Figure S3A).
The function of NK-92 cells modified with the alpharetroviral NFkBmIL2-or NFkBsyn-driven "all-in-one" EGFP vector constructs was further characterized by co-incubation with primary patient-derived GBM cells that highly express the GD2 antigen ( Figure 3B). Modified NK-92 cells were cocultivated with patient-derived primary GBM cells. Specific GD2CAR-mediated induction of EGFP was assessed via flow cytometric analysis ( Figure 3C) and fluorescence microscopy ( Figure 3D), indicating the capability of modified NK-92 cells to recognize the GD2 antigen on primary GBM cells. In contrast to unmodified NK-92 cells, which did not show any NFkB-driven EGFP upregulation after cocultivation with primary GBM cells, a 13-fold and twofold higher EGFP expression was induced by NFkBmIL2.EGFP.GD2CAR-and NFkBsyn.EGFP.GD2CARmodified NK92 cells after antigen recognition and GD2CAR signal transduction ( Figure 3C). Similar results were observed for engineered primary NK cells in cocultures with patientderived GBM spheroids with CAR-mediated EGFP induction of approximately 40% for NFkBsyn-responsive constructs and 20% for NFkBmIL2-responsive constructs ( Figure S3B). However, NFkBsyn.EGFP.GD2CAR-modified NK-92 cells showed a CAR-independently elevated EGFP expression in up to 40% of modified cells in the absence of GD2 + target cells ( Figure 3C), indicating leakiness of the NFkBsyn promoter element as described for cocultures with other target cell lines ( Figure 3A). The MFI of EGFP confirmed the leakiness of the NFkBsyn promoter vector construct as seen with high fluorescence intensity values already in the non-induced state in comparison to the NFkBmIL2 promoter vector construct in NK-92 cells (Figures S2B-D).
NFkB-driven EGFP induction upon GD2CAR-specific activation was additionally detected by fluorescence microscopy after cocultivation of modified NK-92 cells with primary GBM cells. Activated modified NK-92 cells that induced EGFP expression became visible as green fluorescent cells ( Figure 3D). Moreover, modified NK-92 cells migrated toward the GBM cells and were discernable as single green fluorescent cells on the outer layer and within tumor spheroids ( Figure 3D). A CAR-signaling-independent EGFP expression was again detected in NFkBsyn.EGFP.GD2CAR-modified NK-92 cells in monoculture without GD2 + GBM cells, confirming the overall leakiness of the NFkBsyn promoter vector construct, which was less observed for the NFkBmIL2 construct ( Figure 3D).
In vitro live-cell imaging experiments were performed to record the behavior of the modified NK-92 cells while recognizing the target antigen GD2. KG-1a cells engineered to highly express GD2 ( Figure S3C) in addition to an RFP657 reporter construct were used as target cells, as they were detectable due to their red fluorescence ( Figure 3E). Modified NFkBmIL2.EGFP.GD2CAR-transduced NK-92 cells were cocultivated with the engineered KG-1a cells for the indicated time points. After 20 h of coculture, the NFkB-driven EGFP upregulation was clearly visible due to the green fluorescence signal detected in modified NK-92 cells, indicating engagement of the target antigen GD2 with consequent GD2CAR-mediated EGFP induction. This interaction was also seen by microscopic analysis as migration of the modified NK-92 cells to the GD2 + KG-1a cells and the integration of target cells into the NK cell cluster, which resulted in the additional detection of the merged yellow color ( Figure 3E). Furthermore, comparable in vitro livecell imaging experiments with NFkBmIL2.EGFP.GD2CARmodified primary NK cells and GD2 + KG-1a cells confirmed the previously described observations after 16 h of coculture ( Figure S3D).
The proof-of-concept experiments demonstrated that GD2CAR-mediated activation induced NFkB-driven EGFP transgene expression, confirming the feasibility of the transcription factor NFkB to translate GD2CAR-triggered signaling into transgene expression in the "all-in-one" alpharetroviral SIN vector system in NK cell subsets.
The modular design of the "all-in-one" vector system simplified the introduction of a different CAR construct to allow verification of the CAR-mediated induction of genes of interest (e.g., EGFP). Thus, GD2CAR was exchanged by a second-generation CEACAR. Modified NK-92 cells harboring the newly generated alpharetroviral "all-in-one" CEACAR vector constructs were compared regarding the promoter-dependent EGFP induction (NFkBsyn and NFkBmIL2) after CEA-specific CAR activation. Similar results as seen for the vector constructs with the GD2CAR were observed, indicating the adaptability of these principles to other applications and research questions ( Figure S4).
Together, the presented data indicate the relevance of the choice of transcription factor-responsive elements within the inducible promoter cassette of the "all-in-one" vector construct that is likely cell-type dependent. As the NFkBsyn promoterdriven vector constructs led to high background expression in NK cells, the NFkBmIL2 promoter-driven "all-in-one" constructs were used in further analyses.

Enhanced Cytotoxic Activity and Augmented Cytokine Secretion in Modified NK-92 Cells After Target Engagement
The cytotoxic activity of NK-92 cells modified with the alpharetroviral "all-in-one" vector constructs was assessed via lactate dehydrogenase (LDH) assays ( Figure 4A). NK-92 cells harboring either the NFkBmIL2.EGFP.GD2CAR or the NFkBmIL2.hIL12.GD2CAR construct showed increased cytotoxicity after GD2CAR-specific activation with GD2 + target cells (HT1080 GD2 and SH-SY5Y). While NK-92 cells modified with the NFkBmIL2.EGFP.GD2CAR indicated enhanced killing capacity against GD2 + target cells compared to NK-92 modified with the NFkBmIL2.hIL12.GD2CAR vector construct, they also had higher background activity against GD2target cells.
Due to the modular alpharetroviral "all-in-one" vector architecture, the iGOI EGFP could be exchanged by a cDNA coding for human IL-12 ( Figure 2A). The GD2CAR-mediated activation with GD2 + target cells resulted in NFkB-driven IL-12 secretion after 24 h of coculture ( Figure 4B). A significant increase of IL-12 secretion was detected in modified NK-92 cells after GD2CAR-specific stimulation with HT1080 GD2, SH-SY5Y, or patient-derived GBM cells in comparison to no stimulation or activation by coculture with GD2 -HT1080 cells as controls ( Figure 4B).
To further test the cytolytic potential of the modified NK-92 cells, the secretion of antiviral, antitumoral, and NK cell activation-specific cytokines was determined in cell culture supernatants after cocultivation of unmodified and modified NK-92 cells with indicated target cells ( Figure 4C). Unmodified NK-92 cells and NK-92 cells harboring the NFkBmIL2-driven "all-in-one" EGFP vector construct did not show increased IFN-g expression after GD2CAR-specific activation. In contrast, NK-92 cells transduced with the NFkBmIL2-driven "all-in-one" IL-12 vector construct secreted elevated amounts of IFN-g after cocultivation with GD2 + target cells ( Figure 4C). Similar results were obtained for the secretion of the cytokines IL-2 and IL-10 ( Figure 4C). Despite the absence of an overall leakiness resulting in non-specific IL-12 secretion ( Figure 4B), NK-92 cells harboring the NFkBmIL2.hIL12.GD2CAR "all-in-one" vector already showed a high basal IFN-g, IL-2, and IL-10 secretion even in the control setting-without stimulation or with cocultivation with GD2target cells ( Figure 4C). Furthermore, NFkBmIL2.hIL12.GD2CAR-modified NK-92 cells secreted higher levels of TNF-a, sFasL, and granzyme B after GD2CARmediated activation with GD2 + target cells. Increased secretion of these activation-specific cytokines was also observed after GD2CAR-specific activation of NFkBmIL2.EGFP.GD2CARmodified NK-92 cells ( Figure 4C). As all cocultures were performed under in vitro conditions without any additional immune cells that might be recruited and further help to eliminate the tumor, the experiment was not powered to detect significant differences between the different "all-in-one" vector constructs by a potential added value of IL-12 since the antitumor effect is mediated by the GD2CAR, which is included in all constructs.
In addition, several other cytokines known to indicate NK cell activation showed enhanced secretion levels after GD2CARmediated stimulation of modified NK-92 cells and primary NK cells with patient-derived primary GD2 + GBM cells ( Figure S5). Here, a direct comparison of the cytokine expression profile in unmodified and modified NK-92 cells and primary NK cells stressed the differences between these NK cell subsets ( Figure S5).
We further assessed the cytotoxic potential via Annexin V staining, which is commonly used to identify apoptotic cells. Flow cytometric analyses were performed after coculture of modified NK-92 cells with patient-derived GD2 + GBM cells ( Figure 4E). Compared to cocultures of unmodified NK-92 cells with patient-derived GBM cells, Annexin V was upregulated in GD2 + GBM cells cocultured with modified NK-92 cells harboring either the NFkB-driven EGFP or NFkB-driven IL-12 "all-in-one" vector constructs. Annexin V expression, as an indicator of apoptotic cell death of GD2 + GBM cells, ranged between 30% and 40% ( Figure 4E).
In summary, the increased killing capacity of GD2 + target cells, the enhanced secretion of NK cell-specific cytokines, and the upregulation of CD107a in modified NK-92 cells cocultivated with GD2 + target cells and the upregulation of Annexin V in GBM demonstrated an antigen-specific cytolytic activity of NK cells mediated by GD2CAR.
The functional assays shown above do not allow the investigation of effects of activation-specific cytokine secretion on additional immune cell types, e.g., recruitment of macrophages. To assess the potential of the GD2CARmediated secretion of IL-12, we performed migration assays with coculture supernatants using a modified Boyden chamber  Figure 4. Indicated significance was determined by one-way ANOVA with Tukey's multiple-comparison test; ***p ≤ 0.001; **p ≤ 0.01; *p ≤ 0.05; ns, not significant. Untransduced effector cells were termed MOCK. and the monocytic cell line MONO-MAC-6 (MM6). An IL-12dependent trend to recruit MM6 cells was observed ( Figure S6) in supernatants from cocultures with GD2 + HT1080 cells.

DISCUSSION
The majority of CAR-based therapeutic approaches employ CAR T cells, which are the most commonly applied immune cell type in the treatment of hematological malignancies (3,4,17,(51)(52)(53). A major interest in the field of CAR therapy for cancer treatment is improvements to effectively address solid tumors. One approach is the extension of CAR engineering to alternative immune cell types, such as macrophages, NKT cells, and NK cells. Natural anticancer properties of NK cells might potentiate CAR-mediated cytotoxic activity against heterogeneous tumor populations as CAR NK cells still possess their natural killing capacities, which are not dependent on antigen recognition but rather are regulated via a balanced stimulation of activating and inhibitory receptors (1,18). This may provide more complete elimination of tumors and could be especially advantageous in the case of target antigen loss. Compared to T cells, NK cells have a reduced risk of GvHD in allogeneic settings due to the lack of endogenous antigen-specific receptors, which engage allorestricted peptides presented on the major histocompatibility complex (MHC) (19). Moreover, NK cells showed minimal cytokine release syndrome and neurotoxicity in first clinical studies, attributed to lower association of their cytokine activation profile with these disease patterns (24). Another benefit of CAR NK cells is the feasibility for an "off-the-shelf" product, e.g., by using the NK-92 cell line, for clinical application, thus reducing cost and time limitations (54).
These beneficial CAR NK cell properties, together with the TRUCK strategy to increase local anticancer efficacy in solid tumors via additional cytokine delivery (12,14), have motivated us to the aim of this study to apply the TRUCK-based "all-inone" vector concept to NK cells. Our group recently showed the applicability of an "all-in-one" lentiviral vector system designed to engineer primary human T cells to target GD2 + GBM (15). In line with this, other recent studies took advantage of the strategy to co-deliver an additional cytokine and showed the potential of the armored CARs (14,55). Although the feasibility, efficacy, and selectivity to induce NFAT-driven cytokine secretion in a CARdependent fashion was successfully demonstrated (15), we encountered some challenges and problems when applying the system to NK cells. The switch from lentiviral to alpharetroviral SIN vectors was accomplished to obtain a more efficient transduction of NK cells. This is in line with earlier studies that demonstrated increased transduction rates with RD114/TRpseudotyped alpharetroviral vectors in NK cells (35). Furthermore, alpharetroviral vectors may provide an additional level of safety due to the more neutral genomic integration patterns as compared to gammaretroviral and lentiviral vectors (44,56). The established one-vector system might further decrease adverse therapy effects and reduce the risk of insertional mutagenesis of retroviral gene transfer because only a single transduction is required for transfer of the inducible cassette and constitutive CAR expression cassette (57).
Thus, despite obtaining adequate and stable transduction efficiencies with alpharetroviral "all-in-one" vector gene transfer in NK-92 cells and in primary NK cells, we observed no significant NFATmIL2-driven gene induction and only lowlevel CAR-mediated gene induction with the NFATsyn promoter element accompanied with high background expression levels. These results demonstrate the challenge to directly transfer established CAR T cell vector concepts to alternative types of immune cells and highlight the importance of the interplay between the engineered signaling cascade of the vector system and the overall natural properties of the individual immune cell to be genetically modified for cell therapy.
Due to the phenotypic heterogeneity of NK cells, further investigations regarding NK cell activation mechanisms are needed to fully understand the role of the different transcription factors. While NFAT is a key transcription factor in T and B cell development and activation, NFAT might also influence NK cell reactivity and function (36,58). However, previous studies identified NFAT as a negative regulator of NK cell function and demonstrated its dispensability for NK cell activation (37,50,58). In line with this, Kulemzin et al. recently showed that activation-induced NFAT-driven promoters showed different activation patterns in T cells and NK cells. NFAT-driven promoters, which worked in T cells, were not efficient in NK-92 cells, whereas NFkB promoters were able to induce the activation of inducible gene cassettes in both cell types (37). NFkB is a key regulator in T and B cell differentiation as well as in transcription, immune response regulation, and feedback mechanisms (50,59), but it also plays a role in cytokine production and granule exocytosis in NK cells. Therefore, NFkB is a promising inducible promoter element for the development of NK cell therapeutics. Indeed, after vector modification by the exchange of NFAT to NFkB in the inducible promoter element, comparable transduction efficiencies of the novel NFkB-driven "all-in-one" vectors in NK-92 cells and in primary NK cells were observed and, more importantly, GD2CAR-dependent gene induction was shown in modified NK cell subsets after cocultivation with all tested GD2 + target cell lines and patient-derived GD2 + GBM spheroids. Appropriate GD2CAR-mediated transgene induction was detected for both the EGFP control vector modules and the hIL-12 vector constructs. Here, the NFkBmIL2 promoter combination performed more efficaciously than the NFkBsyn promoter element as evidenced by greater responsiveness to GD2CAR signaling and lower background expression of the inducible transgene. Furthermore, even in coculture conditions with a mix of CAR-modified and unmodified NK cells, used to more closely resemble the clinical application approach, the potency of CAR-mediated gene induction and cell activation was clearly visible.
Kulemzin et al., who tested NFkB-driven lentiviral prostatespecific membrane antigen (PSMA)-CAR vectors for cytokine induction with different NFkB response elements, showed the highest fold activation with promoters containing 10xNFkB and 30xNFkB response elements, but also clear leakiness in nonactivated cells. In contrast, we achieved efficient NFkB-promoter induction with two NFkB response elements, which were in comparison larger in bp size and thus correspond to eight NFkB response elements used by Kulemzin et al. (37,60). Moreover, low background expression of the inducible transgene was obtained by the fusion of the NFkB response elements to the minimal IL2 promoter. This observation might indicate that not only the number of response repeats, but also the choice of the fused promoter elements (mIL2 and syn) is a critical design feature that impacts the successful implementation of the CARcontrolled therapeutic gene delivery.
Although the background expression of the inducible transgene was very low in NK cells modified with the NFkBmIL2-driven "all-in-one" vector construct harboring an IL-12 gene cassette, unexpectedly high IFN-g and CD107a expression levels were detected in non-activated and CARindependently activated modified NK cells. The observation that EGFP and IL-12 levels were specifically increased after GD2CAR-mediated activation with GD2 + target cells argues against the assumption that this is a result of the leakiness of the designed vector construct. This observation might rather be explained by a high auto-activation of the modified NK cells due to the insertion of the construct and a resulting GD2CARmediated positive autocrine feedback loop, which did not occur in the NFkB-driven "all-in-one" construct containing EGFP as an inducible transgene. While high auto-activation may be suboptimal in certain settings and might cause severe side effects as well as non-specific off-target cytotoxicity, higher baseline NK cell activation, further amplified by a specific GD2CAR-mediated stimulation, might be favorable with respect to increasing anticancer potency of applied NK cells. Nevertheless, further investigations are needed to explore possibilities to improve the CAR-construct design with the aim to diminish the non-specific activation profile.
For CAR T cell application, CARs are usually designed with the signaling domains found in their natural activation pathways (2). In the present study, we used a second generation CAR with a 4-1BB costimulatory and CD3-z signaling domain. This combination was shown to work adequately in both T and NK cells (38,41,61,62). Despite this CAR configuration that was designed for T cell studies, we showed the functionality of GD2CAR also in NK cells, which is in accordance with the general importance of CD3-z for NK cell proliferation and activation (63,64). After GD2CAR-specific stimulation, upregulation of the iGOIs (EGFP and IL-12) as well as specific cytokines such as IFN-g, TNF-a, and IL-10 was observed to indicate a clear link from CAR signaling to iGOI and cytokine release. Moreover, expression of the degranulation marker CD107a further supports the functional initiation of cytolytic activity of GD2CAR NK cells, although there was a high basal activation in non-stimulated cells. This high basal activation in non-modified NK cells may be expected, as these cells can recognize tumor cell markers other than the GD2 antigen targeted by the CAR used in this study. This seems to also be reflected in the high levels of Annexin V + GD2 + tumor cells found upon coculture with unmodified or GD2CAR NK cells ( Figure 4E). Although the CD107a expression level as measured by MFI was only slightly increased in GD2CAR NK cells, the percentage of CD107a + GD2CAR NK cells was significantly increased upon coculture with GD2 + target cells ( Figure 4D). The general activation of NK cells upon coculture with tumor cells may also partially explain the minimal, non-significant differences in secretion levels observed for some cytokines (e.g., IL-10, sFas-L, and granzyme B) upon comparison of MOCK and CAR-expressing NK-92 cells. The high effector-to-target ratio (10:1) used for these experiments, which allowed quantification of secretion and expression levels of the iGOI (IL-12 and EGFP), might be inappropriate for the analysis of other cytokines. Since untransduced NK-92 cells secrete cytokines when activated, the high effector-to-target ratio may have obscured potential differences between MOCK and modified NK-92 cells. Thus, lower effector-to-target ratios might increase the possibility to assess differences in cytokine secretion levels between the unmodified and modified NK-92 cells, similar to what was observed in the cytotoxicity analyses. Furthermore, the efficiency of the "all-in-one" vector constructs, including CARmediated transgene induction, was lower in modified primary NK cells than in engineered NK-92 cells. This might be due to donor-specific differences or may indicate inherent differences in components of signal transduction cascades in these NK cell sources. Better delineation of signaling pathways used in NK cells may help direct further strategies to improve CAR-directed cytokine induction in these settings. For example, remodeling the CAR with more NK cell-specific signaling domains, such as DAP-10, DAP-12, and 2B4, might further improve the antitumor cytotoxicity of these cells (65)(66)(67).
Prospectively, the specific "all-in-one" alpharetroviral NFkBmIL2-driven GD2CAR vectors can be further optimized by additional or different cytokine transgenes. Especially, the pro-inflammatory cytokines hIL-15 and hIL-21 are promising candidates for CAR NK cell therapy as previously described (68)(69)(70). These cytokines can create an antitumoral milieu in the tumor bed by immune cell activation and recruitment of other immune cells, thus enhancing CAR-dependent cytotoxicity. Especially IL-15 is known to improve NK cell proliferation and in vivo persistence.
Another possibility to improve CAR NK cell function is to overexpress the chemokine receptor CXCR4 to promote migration to the tumor site. Preclinical studies on CAR NK cells in solid tumors show that CXCR4 overexpression led to complete remissions in mouse models and increased in vivo survival of CAR NK cells (71). The examined cytokine profile of modified NK-92 cells after cocultivation with primary GBM further highlights the importance of chemokine receptors and shows an increased upregulation of the chemokine ligand CXCL10, which might be induced by IL-12 and is associated with an increased migration potential toward tumor cells (72). While in vitro migration assays supported the capacity of "all-inone" CAR NK cells to recruit monocytic cells (i.e., MM6 cells) upon antigen stimulation of CAR NK cells, further in vivo studies would more completely show the potential of secreted IL-12 to remodel the tumor microenvironment. Noteworthily, the additive effect of cytokine (e.g., IL-12, IL-15, IL-18) secretion by CAR T cells was previously shown in in vivo mouse studies and this concept is being evaluated in clinical trials (e.g., NCT03721068) (13,(73)(74)(75).
CAR NK cell therapy might also be expanded to address further solid malignancies or even diseases beyond cancer, for example, autoimmune and inflammatory diseases as well as chronic infections. Such strategies could be developed by exchange of the targeting scFv region of the CAR or even the entire CAR cassette as demonstrated for the CEACAR ( Figure S4) within the described vectors, which can be easily accomplished due to the modular design.
In conclusion, the presented data demonstrate the in vitro proof-of-concept of the transferability of the designed "all-inone" vector constructs from T cells to NK cells via incorporation of the inducible NFkB-driven transgene cassettes. The data especially highlight that biological differences in the respective immune cells substantially impact the outcome of functional reprogramming. Linking CAR signaling to responsive, cellspecific enhancer/promoter elements can be exploited to improve the delivery of gene cargos, such as cytokines. Our developed NFkBmIL2.iGOI.GD2CAR construct demonstrated tightly linked antigen-specific transgene induction and effective cytolytic potential shown by increased NK cell activation-specific cytokine release in NK cell subsets, even after the recognition of primary patient-derived tumor cells. One limitation of our study is the lack of in vivo testing in humanized mouse models that allow tumor formation and investigation of immune responses. Still, the data presented here indicate that engineered alpharetroviral NFkBmIL2-driven vectors represent potentially useful candidates for application in targeted adoptive cell therapy with genetically engineered NK cells and provide a basis for clinical translation of CAR NK cell therapy to treat malignancies.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Nordstadt Cerebral Cancer Study (NoCCA-Study), Register-Nr. 6864. The patients/participants provided their written informed consent to participate in this study. was exemplarily investigated for NFkBsyn.EGFP.GD2CAR and NFkBmIL2.EGFP. GD2CAR constructs on protein level via Western blot analysis. The CD3-z domain of the GD2CAR (expected size approximately 53 kDa) was detected in modified (MOI 10) primary NK cells in addition to the endogenous CD3-z (17 kDa) and the housekeeping gene glyceraldehyde-3-phosphate-dehydrogenase (GAPDH, 37 kDa) as internal controls. Untransduced primary NK cells were termed MOCK.