Cancer cell lines were purchased from the American Type Culture Collection (ATCC, USA) and cultured inRPMI 1640 medium (Welgene, South Korea) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Seradigm, USA) and antibiotic–antimycotic (Gibco, Waltham, MA, USA). The human NK cell line NK92 (CRL-2407; ATCC) was maintained in the ATCC-recommended medium. The proliferation rate was assessed using CellTrace™ Violet (ThermoFisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocol. Cell viability was measured by trypan blue staining.

A lentiviral vector(632155; Clontech) encoding α-Cot-Myc-CD8α-CD28TM-DAP10-CD3ζ was constructed (Figure 1A). The Cot-conjugators and scFv against Cot were obtained from a previous study (18).

Flow cytometric analysis was performed using a BD FACSCanto II cytometer(BD Biosciences, Franklin Lakes, NK, USA), and the data were analyzed using FlowJo software(Treestar Inc.Ashland, OR, USA). The cells were washed with cold FACS buffer and incubated with the following fluorochrome-conjugated mouse anti-human antibodies: NKp30 (CD337; 558408), NKp44 (CD336; 558563), NKp46 (CD335; 558051), NKG2D (CD314; 557940), granzyme B (561142), perforin (51-65995X), CD56 (555518); CD107a (555801), EGFR (563577), and ERK1/2 (pT202/pY204; 612592), purchased from BD Biosciences. The antibodies were diluted 1:100 and incubated for 30 min at 4°C in the dark. In addition, mouse IgG2b, κ isotype control (556437, BD Biosciences), Myc-Tag (9B11) (3739S, Cell Signaling Technology, Danvers, MA, USA), and HER-2 (2G11) (BMS120FI, eBioscience, Santa Clara, CA, USA) were used.

NK cell-mediated cytolysis was examined using calcein-AM (C1430, Thermo Fisher Science) as described previously (22) and the xCELLigence real time cell analysis system. Briefly, the target cancer cells were stained with calcein, and the stained cancer cells and NK cells were co-cultured according to the ratio of effector and target cells. The supernatant was taken 4 h later, the amount of calcein present in the supernatant was measured, and the xCELLigence real time cell analysis system was used to measure the extent to which attached target cells were suspended by NK cells by changing electrodes.

NK and cancer cells were mixed at a 1:1 ratio and incubated for 12 h. Cell-free supernatants were assayed for IFN-γ or TNF-α secretion using an ELISA kit (Cat No. 88-7316-88 and 88-7346-88, Invitrogen, Waltham, MA, USA) according to the manufacturer’s protocol.

The NK92 and AU565 cells were incubated with 1 µg/mL LysoSensor Green DND-153 (Invitrogen) for 30 min and 10 µg/mL CellTrace Far Red DDAO-SE (Invitrogen) for 15 min, respectively, at 37°C. The labeled NK92 cells in complete RPMI medium with 10 µg/mL propidium iodide (BD Biosciences) were seeded onto an anti-CD44 coated glass coverslip loaded in a Chamlide magnetic chamber (Live Cell Instrument, South Korea) (22).

All animal experiments were performed in accordance with guidelines of the Institutional Animal Care and Use Committee of the Korean Basic Science Institute (KBSI-AEC 1916). A549-Red-Fluc cells (Perkin-Elmer, Waltham, MA, USA) were intravenously injected into 6-week-old male BALB/c-nu/nu mice. The mice were injected with PBS, 10 mg/kg taxol, and α-Cot-NK92 cells through the tail vein and were subjected to weekly bioluminescence imaging (BLI; IVIS Spectrum, Perkin-Elmer, USA). For ex vivo analysis of BLI, the mice were sacrificed, and peripheral blood and lungs were harvested.

Data are presented as the mean ± SD or SEM of experimental replicates. One-way ANOVA was used to compare three or more independent groups using GraphPad Prism v5.03 software. P<0.05 was considered statistically significant.

To allow the targeting of multiple TAAs through Cot-conjugators, α-Cot-CAR was engineered by fusing the extracellular domain (comprising scFv against Cot, Myc as an identifier of CAR expression, and a CD8-based hinge as a spacer in the formation of an immune synapse), a transmembrane domain originating from CD28, and intracellular signaling domains with cytoplasmic regions of CD28, DAP10, and CD3ζ (Figure 1A).

The intracellular signaling domains were determined by assessing the impact of each domain on CAR expression, cytolytic activity, and apoptotic cell death through the transient expression of CAR vectors comprising TM-10 (with only DAP10), TM-z (with only CD3z), and TM-10z (with both DAP10 and CD3z) together with a ZEGFR affibody instead of scFv (Figure S1A). The ΔEcto-TM-10z (without an affibody) was used as a negative control for the ectodomain. TM-z and TM-10z CARs containing CD3ζ exhibited higher cytolytic activity and CD107a (degranulation marker) expression than the others (Figures S1C–E). TM-z CAR exhibited the highest apoptotic cell death rate with the highest cytolytic activity. However, DAP10-containing TM-10 and TM-10z CARs exhibited reduced apoptotic cell death rates (Figure S1F). Therefore, CD3ζ and DAP10 play essential roles in cytolytic activity and cell survival.

To establish α-Cot-NK92 cells, α-Cot-CAR was transduced into NK92 cells using a lentiviral vector. Transduced NK92 cells were selected using puromycin for 3 days. The expression of α-Cot-CAR was confirmed by Myc in the extracellular domain of CAR using α-Myc-PE(Figure 1B upper) or after incubation of NK92 or transduced NK92 cells with α-HER2-cot, α-HER2 antibody bound to α-Cot-NK92 was identified using a human IgG-PE- conjugated antibody (Figure 1B lower). The α-Cot-NK92 cells expressed 20–25% more CAR in the extracellular region than the parental NK92 cells (Figure 1B). Further, the α-Cot-NK92 cells retained the characteristics of parental NK92 cells including NK-specific activating receptor profiles, such as NKp30, NKp44, NKp46, and NKG2D (Figure S2A); functional markers, such as IFN-γ, granzyme B, and perforin (Figure S2B); proliferation rate (Figure S2C); and viability (Figure S2D).

To assess the antigen-specificity of α-Cot-CAR, α-Cot-NK92 cells were co-cultured in vitro with three tumor cell lines: AU565 (mammary gland adenocarcinoma), SK-OV-3 (ovary adenocarcinoma), and A431 (skin epidermal carcinoma). The surface expression of HER2 and EGFR was evaluated by flow cytometry. The adenocarcinoma cell lines AU565 and SK-OV-3 expressed high levels of HER2 and EGFR, whereas the carcinoma cell line A431 expressed low levels of HER2 and high levels of EGFR (Figure 2A).

Using an affibody(ZEGFR)-based conjugator against EGFR(ZEGFR-Cot) or antibody(trastuzumab)-based conjugator against HER2(α-HER2-Cot), we evaluated the cytolytic activity of α-Cot-NK92 cells. We found that the α-Cot-NK92 cells exhibited cytolytic activity according to the antigen-specificity of the conjugator, regardless of tissue specificity, in various tissue-derived tumor cells. Upon exposure of tumor cells to α-Cot-NK92 cells with α-HER2-Cot, the cytolytic activity of α-Cot-NK92 cells was positively correlated with HER2 expression levels in the tested cell lines. Due to the high levels of EGFR in the tested cell lines, α-Cot-NK92 cells with ZEGFR-Cot exhibited high cytotoxicity (45%–70%; Figure 2B). The α-Cot-NK92 cells, were co-cultured with AU565 or A431 cells with or without a conjugator to evaluate their cytokine secretion and degranulation functions. The secretion of IFN-γ and TNF-α and the expression of the degranulation marker CD107a in α-Cot-NK92 cells increased when α-HER2/ZEGFR-Cot was added in co-culture with AU565 cells expressing both HER2 and EGFR (Figure 2C), and when ZEGFR-Cot was added to co-culture with A431 cells expressing low levels of HER2 and high levels of EGFR (Figure 2D). The increase in activity following the addition of the conjugator was mediated by ERK phosphorylation (Figure S3).

We investigated the cytolytic kinetics of α-Cot-NK92 cells with and without a conjugator using the xCELLigence real-time cell analysis system to measure the impedance of the attached target cell. A dimensionless cell index indicated the number of surviving lung carcinoma A549 cells with a small amount of HER2 and high EGFR expression on the cell surface (Figures 2E, F). A rapid decrease in cell index <4 h after the addition of α-Cot-NK92 cells with ZEGFR-Cot indicated successful A549 cell lysis, whereas no significant cytolytic effect was detected in parental NK92 and α-Cot-NK92 cells in the absence of conjugators (Figure 2F). This suggests that α-Cot-NK92 cells can react with diverse conjugators and exhibit ligand-dependent cytotoxicity, mediated by the conjugator, against tumor cells.

α-Cot-CAR was designed as a tunable receptor with the ability target multiple antigens. Using ZEGFR-Cot and α-HER2-Cot, we evaluated the cytolytic activity of α-Cot-NK92 cells in a dose-dependently manner. The cytolytic activity of α-Cot-NK92 cells increased with the conjugator dose and reached a maximum at approximately 1 μg/mL α-HER2-Cot or 0.05 μg/mL ZEGFR-Cot. (Figure S4). To identify the ability of α-Cot-NK92 cells to target dual antigens in AU565 cells, 50 ng/mL α-HER2-Cot and 5 ng/mL ZEGFR-Cot (which represent 9.5% and 17.6% of maximal cytolytic effects, respectively) were administered individually or in combination. Combination treatment showed an additive effect, with approximately 26.4% of the maximal cytolytic effect in AU565 cells (Figure 3A). The dual-targeting potential of α-Cot-NK92 cells was evaluated in A549-Red-Fluc cells expressing low HER2 and high EGFR levels (Figure S5A). Upon treatment with 5 ng/mL α-HER2-Cot or 2.5 ng/mL ZEGFR-Cot, α-Cot-NK92 cells exhibited 17.7% or 44.8% of the maximal cytolytic effect of each conjugator, respectively. Combination treatment with both conjugators resulted in a cytolytic effect comparable to that of 50 ng/mL ZEGFR-Cot and an enhanced cytolytic effect compared with that of 500 ng/mL α-HER2-Cot alone (Figure 3A). This suggests that α-Cot-NK92 cells can kill tumor cells expressing multiple antigens at different levels through dual or multiple targeting of various antigens using a small amount of conjugators.

Due to the heterogeneity of tumors, recurring tumors exhibit different characteristics compared to the original tumors. For example, exposure of HER2+HER3+ MDA-MB-453 cells to α-Cot-NK92 cells with α-HER2-Cot for 4h, followed by culture for 96h, decreased the HER2+HER3+ population by <10%, and increase the HER2-HER3+population by >90%. (Figures S6, 3B).

Based on these results, a recurrent tumor model system mimicking the heterogeneity of recurrent tumors was constructed by mixing EGFR⁺HER2⁻ MDA-MB-231 and EGFR^-HER2⁺ MDA-MB-453 cells. To evaluate the cytotoxicity of α-Cot-NK92 cells, we determined the percentage of reduced EGFR+ or HER2+ values in the group with or without the conjugator(s) compared to the untreated group (none) (Figure 3B). Following treatment of α-Cot-NK92 cells with ZEGFR-Cot on mixed target cells for 4 h, cytotoxicity was increased compared to the group without the conjugate (Figure 3C), and the population ratio of the remaining EGFR+ target cells was decreased compared to the untreated group After removing the α-Cot-NK92 cells, the remaining target cells were cultured for 2 days, and the α-Cot-NK92 cells were treated with ZEGFR-cot ​​or HER2-cot, respectively, to determine the cell population ratio and cytotoxicity. As a result, compared with the untreated group, the population ratio of HER2+ cells was decreased (Figure 3B) and the cytotoxicity was increased in the α-HER2-Cot treated group, whereas that of the ZEGFR-Cot treated group did not change (Figure 3D). This implies that dual or multiple targetable α-Cot-NK92 cells can be effectively applied to heterogeneous recurrent tumor cells without re-engineering CARs.

The immunological synapse between an NK cell and its target cell is the dynamic interface that allows for the sequential stages, including the initiation, effector, and termination stages of NK cells resulting in the lysis of the target cells. In the initial stage, the NK cells contact the target cells using diffused (Figure 4A, step A) or polarized granules (Figure 4A, step B). In the effector stage, NK cells form a stable cleft wherein cytolytic molecules are secreted, recruit lytic granules to the distal poles (Figure 4A, step C), and transport them into the immunological synapse (Figure 4A, step D). In the termination stage, the NK cells kill target cells with membrane blebs, increase propidium iodide fluorescence, and detach from the target cells (Figure 4A, step E). The percentage of α-Cot-NK92 cells remaining at each step was measured at 2, 4, and 6 h after co-culture with AU565 cells containing α-HER2-Cot or ZEGFR-Cot, with or without a conjugator (Figure 4B). Following the addition of a conjugator. α-Cot-NK92 cells rapidly entered the effector and termination stage 2 h after co-culture with AU565 cells, approximately 80% of α-Cot-NK92 cells were in the termination stage at 4, and 6 h after co-culture, the duration from steps A to E was much shorter (Figure 4C), and the killing ratio of α-Cot-NK92 cells on contacted target cells was approximately 75% (Figure 4D). These results indicate that α-Cot-NK92 cells act in a conjugator-dependent manner on tumor cells and rapidly induce cytolysis (Mov.S1).

We induced the formation of lung metastases, in which tumor cells are trapped within the lung microvasculature, by injecting 2 × 10⁶ A549-Red-Fluc cells into the tail vein of BALB/c nude mice. On days 2, 4, 7, and 9 after cancer cell inoculation, nude mice bearing A549-Red-Luc cells were i.v. injected with 3 × 10⁶ α-Cot-NK92 cells, which were pre-incubated with/without ZEGFR-Cot (0.1 μg/mL) for 1 h, and metastatic lung tumor development was monitored weekly from day 1 after tumor inoculation using in vivo BLI (Figure 5A). PBS and Taxol was administered to mice as negative and positive controls, respectively, to achieve anti-tumor effects. Although α-Cot-NK92 cells without ZEGFR-Cot had little effect on metastatic tumor development compared to the untreated group, the administration of α-Cot-NK92 cells with ZEGFR-Cot or Taxol was effective and significantly reduced the remaining tumor restricted to small areas with low intensity in the lung (Figures 5B, C). For BLI of excised lungs, α-Cot-NK92 cells with ZEGFR-Cot or Taxol had significant regressive effects on the formation and area of metastatic tumor nodules compared to the untreated or α-Cot-NK92 cells without conjugator groups (Figures 5D, E). The results showed that α-Cot-NK92 cells with a conjugator can significantly inhibit lung tumor burden in vivo.

We then investigated whether α-Cot-NK92 cells with a conjugator could migrate toward the tumor site in vivo. A human skin cancer xenograft model was established by subcutaneous (s.c.) inoculation of A431 (skin epidermal carcinoma) cells into the right flanks of nude mice. On day 7 after tumor inoculation, NK92 or α-Cot-NK92 cells with or without ZEGFR-Cot cells were labeled with 1,1’-dioctadecyl-3,3,3’,3’-tetramethylindotricarbocyanine iodide (DiR), and the DiR-labeled NK cells were intravenously (i.v.) injected through the tail vein. Mice were euthanized 72 h after NK cell injection to collect subcutaneous tumors, and DiR fluorescence signal intensity was quantified in the tumors after treatment with α-Cot-NK92 cells with ZEGFR-Cot and α-Cot-NK92 cells injected with ZEGFR-Cot separately was 1.5- and 1.2-fold higher than that with NK92 cells, respectively (Figure S7). This indicates that α-Cot-NK92 cells with ZEGFR-Cot could migrate toward the tumor site and remain there.

The conjugator`s type, dose, and composition are the key factors that control the target reactivity and cytolytic activity of universal α-Cot-NK92 cells. α-Cot-NK92 cells transfected via lentiviral infection are heterogeneous and contain various levels of CAR and surface receptor repertoires. To modulate the killing potential of α-Cot-NK92 cells, monoclonal α-Cot-NK92 cell lines were classified according to CAR expression level through single-cell sorting using flow cytometry. The cytolytic activity of established monoclonal α-Cot-NK92 cell lines with HER2-Cot or ZEGFR-Cot was evaluated and showed a tendency to increase with CAR-Myc expression (black triangles with a line; Figures 6A, B). However, some monoclonal α-Cot-NK92 cell lines with high CAR levels showed lower cytolytic activity than M2 macrophages with moderate CAR levels (Figures 6A, B).

Monoclonal α-Cot-NK92-M2 cells showed higher basal cytolytic activity and potential than monoclonal α-Cot-NK92-L8 cells co-cultured with AU565, A549-Red-Fluc, or MDA-MB-231 cells in a dose-dependent fashion (Figures 6C, D). The MDA-MB-231 cells showed lower HER2 expression than of A549-Red-Fluc cells did (Figure S5). In addition, α-Cot-NK92-M2 cells showed a higher basal cytolytic activity than heterogeneous α-Cot-NK92 cells (Figure 6E). We evaluated the cytolytic activity of the monoclonal cell lines against normal lung (WI38) and breast cancer (AU565) cells. The WI-38 cells expressed 3.08% (Mean Fluorescent Intensity (MFI) 1.82) and 13.1% (MFI 4.79) HER2 and EGFR, respectively, relative to the IgG-stained cells. The AU565 cells expressed 99.5% (MFI 38.99) HER2 and 98.9% (MFI 26.90) EGFR relative to IgG-stained cells (Figure 6F). The α-Cot-NK92-L8 and α-Cot-NK92-M2 cell lines exhibited specific cytotoxicity against AU565 cells, and lower or similar cytolytic activity to parental NK92 cells compared with WI38 cells (Figure 6G), indicating that monoclonal α-Cot-NK92-L8 and α-Cot-NK92-M2 cell lines have tumor-specific killing potential applicable to tumor traits.