Peripheral blood was obtained from healthy donors after informed consent. Peripheral blood mononuclear cells (PBMC) were isolated using ficoll separation. CD56⁺CD3^- NK cells were isolated from frozen PBMCs using an untouched NK cell isolation kit (Miltenyi Biotec, Germany). 1 x 10⁶ NK cells were cocultured in a 24-well plate with 0.5 x 10⁶ irradiated K562 cells expressing membrane-bound interleukin-21 and 4-1BB-ligand (K562^F) in culture medium as described previously (5). Culture medium consisted of IMDM (Lonza, Switzerland) supplemented with 5% heat-inactivated FBS (Gibco, ThermoFisher Scientific, USA), 5% human serum (Sanquin, the Netherlands), 1.5% L-Glutamine (Lonza), 1% penicillin/streptomycin (Lonza). When indicated, culture medium was supplemented with a cytokine mix, consisting of 5 ng/mL IL-15 (Miltenyi Biotec, Germany) and 100 IU/mL IL-2 (Novartis, Switzerland). On day 3 post stimulation, NK cells were transduced with retroviral vector encoding for different murinized TCRs (muTCR) combined with CD8αβ. Transduced cells were MACS (Milteny Biotec) enriched for CD8β on day 7, and immediately re-stimulated with irradiated K562^F. After 2 days, enriched muTCR-CD8αβ positive NK cells were transduced with either a retroviral construct encoding for CD3ζγϵδ or CD3ζγϵδ combined with soluble IL-15. Transduced NK cells were enriched for muTCR expression on day 14 by MACS and subsequently re-stimulated. NK cells were expanded for 7 days and frozen down in liquid nitrogen. Functional assays were performed between day 7 and 14 after the last re-stimulation.

BOB1-B7 TCR (TRAV13 and TRBV4) and CMV-A2 TCR (TRAV18 and TRBV13) pLZRS retroviral vectors were constructed by linking the codon-optimized variable α and β chains to cysteine modified murine TCR α and β constant domains (muTCR) as described previously (18). Subsequently, the muTCRs were linked via 2A self-cleaving sites to CD8αβ ( Figure 1A ) (6, 19). Codon optimized CD3 invariant chains and soluble IL-15 were linked via 2A self-cleaving sites in the following order: IL-15, CD3ζ, CD3δ, CD3ϵ, CD3γ into the Moloney murine leukemia virus based pLZRS retroviral vector ( Figure 1A , Supplementary Table 1 ) (20, 21). Subsequently, virus supernatant was made as previously described (5).

24-well flat-bottom non-tissue culture-treated plates (Greiner Bio-One, Austria) were coated with 30 ug/mL Retronectin (Takara, Kusatsu, Japan) and blocked with 2% human serum albumin (Sanquin, Amsterdam, The Netherlands). pLZRS retroviral supernatant was thawed and spun on Retronectin-coated wells at 3000g for 20 minutes at 4°C. Viral supernatant was then removed and 0.5 x 10⁶ NK cells were added to each well in 1 ml of NK-M. After 24 hours, transduced NK cells were transferred to tissue culture-treated culture plates and further expanded in NK-M. This method was previously described by Morton et al. (5, 19).

Cryopreserved NK:TCR and NK:TCR/IL-15 cells were thawed and weekly stimulated with K562^F in culture media with or without a cytokine mix (IL-15 and IL-2). Expansion of the different NK:TCR cells was assessed by live-cell counting using Eosine every 7 days.

To assess cytokine production, 1 x 10⁶/ml NK:TCR and NK:TCR/IL15 were cocultured with or without K562^F cells in culture medium at an 1:1 Effector : Target (E:T) ratio in a 96-well round-bottom culture plate, in triplicate. After 24 hours supernatant was harvested. To assess the IL-15 concentration, the supernatants of the triplicates were pooled and measured with U-PLEX assay for human IL-15 (Meso Scale Diagnositics) following the manufacturer’s instructions and analyzed with corresponding software. In addition, IFNγ production by the different transduced NK cells was measured in the culture supernatant by IFNγ ELISA (Diaclone), according to manufacturer’s instructions.

Transformed EBV-LCLs (JY, HHC, IZA) and tumor cell lines, UM9 (multiple myeloma; MM), U266 (MM) and K562 were cultured in Iscove’s Modified Dulbecco’s Medium (IMDM) (Lonza, Switzerland) supplemented with 10% FBS (Gibco, Thermo Fisher Scientific, USA) 1.5% 200mM L-glutamine (Lonza, Switzerland), 1% 10,000U/ml penicillin/streptomycin (Lonza, Switzerland). The acute B-cell lymphoblastic leukemia (B-ALL) cell line, ALL BV, was obtained from patient material and cultured in IMDM (Lonza, Switzerland) with serum free supplement (22).

Cytotoxicity was assessed by flow cytometry-based cytotoxicity assay. Target cells were transduced with retroviral vector encoding for dTomato-Red, and dTomato-Red positive cells were subsequently sorted by fluorescence-activated cell sorting (FACS). A total of 30.000 dTomato-Red positive target cells were cocultured with 30.000 NK : TCR/IL15 or NK:TCR cells in round-bottom 96-well plates. After overnight coculture, cells were harvested, washed, and resuspended in 80μL FACS buffer containing Sytox blue live-dead marker (ThermoFisher Scientific). After a five minute incubation on ice, 40μL of sample (pre-set fixed volume criteria) was measured on a 3-laser Aurora at a consistent high flow rate of 65 µL/min (Cytek Biosciences). Flow cytometry data was unmixed using Spectroflo (Cytek Biosciences) and analyzed using FlowJo v10.7.1 (BD Biosciences). A live-dead gate was set on living cells and thereafter target cells were gated on dTomato-Red positive (see Supplementary Figure 1 for a gating example). Percentage of cytotoxicity was calculated as the number of living targets cells after co-culture divided by living untreated targets cells, that were not co-incubated, using the following formula: ( 1 − n u m b e r   o f   l i v i n g   t a r g e t   c e l l s   a f t e r   c o − c u l t u r e n u m b e r   o f   l i v i n g   u n t r e a t e d   t a r g e t   c e l l s ) ∗ 100 . Each experiment was performed in duplo.

On day 0, male or female NOD.Cdg-Prkdc(scid)Il2rg(tm1Wjl/Szj (NOD scid gamma, NSG) mice (The Jackson Laboratory) were intravenously (i.v.) injected with 2 x 10⁶ luciferase-positive multiple myeloma U266 cells in PBS supplemented with 2% heat-inactivated FBS (Gibco, ThermoFisher Scientific, USA). After four days, to confirm localized tumor outgrowth in the bone marrow, tumor outgrowth of U266 cells was measured after subcutaneous (s.c.) injection of 150uL 7.5mM D-luciferine (Cayman Chemical) using a CCD camera (IVIS Spectrum, PerkinElmer). Subsequently, mice were treated with 5 x 10⁶ NK:TCR or NK:TCR/IL-15 cells, TCR specificity and number of mice are mentioned in corresponding figures. Tumor outgrowth was measured at regular intervals. Mice were sacrificed at day 40 or when mice reached an average luminescence of 1 x 10⁷ p/s/cm²/sr. Organs were collected and stained with immunohistochemistry when indicated. These studies were approved by the national Ethical Committee for Animal Research (AVD116002017891) and performed in accordance with Dutch laws for animal experiments. The number of mice included was based on a two-tailed, significance level of (α) 0,01, a power (π) of 80% and a standard deviation (σ) of 20%, with an expected effect size of 70%.

Luciferase expression of U266 cells in FFPE tissue slides was analyzed with primary Goat anti-Human antibodies against Luciferase (AB181640, Abcam, Cambridge). A monoclonal Mouse anti-Human CD3 antibody (AB699, DAKO, Netherlands) was used to stain infiltration of NK:TCR/IL-15 cells. Secondary visualization was performed using a 3,3′-Diaminobenzidine(DAB) based standard laboratory procedure.

FACS analysis was performed on either a 3 or 5-laser Aurora (Cytek Biosciences). For each stain, 50.000 NK cells were aliquoted into 96-well V-bottom plates, washed and incubated with 10uL antibody mix for 15 min at RT. An overview of antibodies used can be found in Supplementary Table 2 . Flow cytometry data was unmixed using Spectroflo (Cytek Biosciences) and analysis was done using the OMIQ software from Dotmatics (www.omiq.ai, www.dotmatics.com).

Statistical analysis was performed using GraphPad Prism 7 software. Statistical tests used are mentioned in corresponding figures.

NK cells were isolated from previously frozen PBMC of five different donors. Isolated NK cells were stimulated with irradiated K562^F cells, expressing membrane bound IL-21 and 4-1BB-ligand, and retrovirally transduced with different TCR-CD8 viral vectors, enriched and restimulated, and subsequently transduced with viral vectors encoding the different subunits of the CD3 complex and soluble IL-15 ( Figure 1A ). Following this two-step transduction protocol, as described previously by Morton et al, and depicted in Figure 1B , we genetically engineered highly pure populations of NK cells expressing muTCR, CD8 and CD3, accompanied by insertion of the IL-15 gene (NK:TCR/IL-15) ( Figure 1C ) (5). In these experiments, the constant region of each transgenic TCR was murinized (muTCR) to allow distinction between any contaminating T cells present in the culture, expressing the huTCR, which may influence functional data. This alteration does not affect any function of the TCR (18). Eventually, in the final products, less than 3% huTCR+ T cells were present. Despite the introduction of an additional gene, expression of the TCR/CD3 complex did not differ between NK:TCR and NK:TCR/IL-15 products ( Figure 1D ). Furthermore, no significant differences were observed in phenotype between NK:TCR and NK:TCR/IL-15 cells using a multispectral flow panel, including several activation, inhibition and adhesion markers ( Supplementary Figure 2 ).

To explore the expansion potential of NK:TCR/IL-15 cells, both NK:TCR and NK:TCR/IL15 were re-stimulated weekly with K562^F cells and cultured for at least 14 days in culture medium with or without suppletion of cytokines (IL-15 and IL-2). NK:TCR cells proliferated strongly upon stimulation with K562^F cells when cultured in medium supplemented with cytokines ( Figure 2A ), but were not able to proliferate when cultured in medium only, resulting in no viable NK cells present after 14 days of culture. In contrast, NK:TCR/IL-15 cells were able to proliferate in culture medium without cytokines and showed a similar expansion kinetic compared to NK:TCR cells cultured with cytokines. Additionally, NK:TCR/IL-15 proliferated at a similar rate whether cultured in presence or absence of cytokines in the culture medium ( Figure 2B ). This suggests that NK:TCR/IL-15 cells produce sufficient IL-15 to sustain their proliferation, without the need for suppletion of IL-15 or IL-2.

To determine the IL-15 secretion by NK:TCR/IL-15 cells, 1 x 10⁶/ml engineered NK cells were cultured alone or with irradiated K562^F cells at an E:T ratio of 1:1 in culture medium for 24 hours. As shown in Figure 3A , NK:TCR/IL-15 cells from four different donors demonstrated on average a baseline level of 5pg/mL secreted IL-15, whereas no IL-15 could be detected in the supernatant of NK:TCR cells. Stimulation with K562^F resulted in a doubling of IL-15 production by the NK:TCR/IL-15 cells, although the concentration of IL-15 in the supernatant did not exceed 15 pg/mL ( Figure 3A ). As a control for cytokine production IFNγ secretion was measured. After stimulation with K562^F cells, IFNγ production was strongly increased by both NK:TCR and NK:TCR/IL-15 cell products ( Figure 3B ).

To investigate whether autonomous production of IL-15 would influence the cytotoxic capability of NK:TCR cells, we transduced NK cells with an HLA-B*07:02 restricted TCR targeting a BOB1 derived peptide (herein referred to as BOB1-TCR). BOB1 is a B-cell lineage specific transcription factor uniformly expressed in malignant B cells, including multiple myeloma (MM), and demonstrated to be essential for the survival of MM (6, 23). As previously shown by Morton et al. NK:BOB1-TCR cells successfully target TCR Ag positive (BOB1+, HLA-B7+) tumor cells. To assess the beneficial effect of IL-15 production by engineered NK:TCR cells, NK:BOB1-TCR/IL-15 and NK:BOB1-TCR cell products generated from five different donors were cocultured overnight in culture medium with a panel of various TCR Ag positive (BOB1+, HLA-B7+) target cells: B-cell lines, EBV-JY and EBV-HHC, MM cell line UM9 and B-ALL cell line ALL BV (BOB1 and HLA-B7 expression; Supplementary Figure 3 ).NK:BOB1-TCR/IL-15 cells tested against this panel of TCR Ag positive (BOB1+, HLA-B7+) targets demonstrated to be as effective or even more effective compared to NK:BOB1-TCR cells ( Figure 4A ). As a control, both NK:BOB1-TCR products were cocultured with a TCR Ag negative (BOB1+, HLA-B7-) target, EBV-IZA and the HLA negative K562 cell line, known for its high sensitivity to NK cells. Both products demonstrated cytotoxic activity against the HLA negative K562 target cells ( Figure 4B ), whereas the TCR Ag negative (BOB1+, HLA-B7-) cell line EBV-IZA was not killed by both NK:BOB1-TCR cell products ( Figure 4C ). When HLA-B7 was introduced in EBV-IZA, NK:BOB1-TCR/IL-15 were able to kill this EBV-LCL ( Supplementary Figure 4 ). It is known that supplement of IL-15 to NK cells increases NK functionality, therefore we engineered NK cells with an irrelevant TCR against a CMV-derived peptide presented in HLA-A*02:01 (referred to as CMV-TCR) to assess NK-mediated cytotoxicity (24). The effector functions of NK:CMV-TCR/IL-15 cells were evaluated in comparison to NK:CMV-TCR cells, showing increased NK-mediated cytotoxicity of NK:CMV-TCR/IL-15 against some CMV TCR Ag negative targets ( Figure 4 ). However, the overall trend of enhanced cytotoxicity was more prominent when comparing NK:BOB1-TCR/IL-15 to NK:BOB1-TCR cells. Furthermore, the comparison of NK:BOB1-TCR/IL-15 cells and control NK:CMV-TCR/IL-15 cells clearly confirmed that NK:BOB1-TCR/IL-15 cells mediate TCR-specific cytotoxicity. as previously seen in NK:BOB1-TCR cells when compared to NK:CMV-TCR cells. ( Figure 4A ). These results demonstrate that autonomous production of IL-15 by TCR engineered NK cells mainly improves TCR-mediated cytotoxicity.

To determine whether autonomous secretion of IL-15 enhances TCR-mediated function of NK cells in vivo, we compared the antitumor activity of NK:TCR and NK:TCR/IL-15 cells in an orthotopic MM model, as previously described (25). NSG mice were intravenously injected with the MM cell line U266 positive for the TCR Ag (BOB1+, HLA-B7+) resulting in bone marrow engraftment of the tumor. After four days the mice were treated with NK:BOB1-TCR or NK:BOB1-TCR/IL-15 cells, both NK cell products were generated from donor 2 ( Figure 1C , dotted symbol Figure 3 and Figure 4 ). During the first week after NK:TCR infusion both groups of mice treated with the genetically engineered NK cells (NK:BOB1-TCR and NK:BOB1-TCR/IL-15) demonstrated decrease in tumor burden, however only when mice were treated with NK:BOB1-TCR/IL-15 was the tumor burden significantly decreased in all mice ( Figures 5A, B ). Even, 5 out of 8 mouse remained free of tumor until 30 days post NK infusion, when the experiment was terminated ( Supplementary Figure 3 ). To investigate whether this was due to increased NK- or TCR-mediated function of the engineered NK cells we performed an additional experiment in which the U266 engrafted NSG mice were treated after four days with NK:BOB1-TCR/IL-15 cells or NK:CMV-TCR/IL-15 cells as a negative control, again both NK products originated from donor 2. All mice treated with NK:BOB1-TCR/IL-15 showed a significant decrease in tumor burden directly after NK cell infusion compared with mice treated with NK:CMV-TCR/IL-15 and non-treated mice ( Figures 5C, D ). These results clearly demonstrate that autonomously produced IL-15 enhances the TCR-mediated function of TCR engineered NK cells in vivo.

Despite these promising observations we also encountered an unexpected and unwanted event within one of the in vivo experiments (depicted in Figures 5A, B ). In this experiment, the beneficial antitumor activity of the NK:BOB1-TCR/IL-15 cells was accompanied by death of 3 out of 8 mice approximately 30 days after injection of NK:BOB1-TCR/IL-15 cells. Autopsy of the complete treatment group demonstrated accumulation of large clusters of cells in several organs, mainly lungs, spleen and liver in all eight NSG mice treated with NK:BOB1-TCR/IL-15 cells. Moreover, immunohistochemistry performed on these affected organs demonstrated that the clusters of cells were negative for luciferase, but positive for human CD3, confirming that these cells were NK:BOB1-TCR/IL-15 cells ( Figure 6 ). In contrast, this phenomenon was not observed in mice treated with NK:BOB1-TCR cells that were generated from the same donor but did not secrete IL-15.To specify, these NK:BOB1-TCR/IL-15 and NK:BOB1-TCR cells were derived from the same NK cell isolation and TCR-CD8 transduction and were thereafter split into two groups. Subsequently, these NK cells were retrovirally transduced with either CD3 or CD3-IL15. The in vitro results of these specific donor-derived products are depicted in Figure 1D as donor 2 or depicted as a dotted symbol in all main figures. To investigate whether this phenomenon could reproducibly be observed and to gain more insight into the underlying mechanism, cryopreserved NK:BOB1-TCR/IL-15 cells from this specific experiment were expanded and injected into mice engrafted with U266. While antitumor activity was observed again, dysregulated growth of genetically engineered NK cells was not observed ( Supplementary Figure 5 ). Eventually, within four independent experiments involving NK:BOB1-TCR/IL-15 cells, generated from the same donor ( Figures 6C, D ) as well as a different donor (data not shown), this unexpected and unwanted event was only observed once.