Retargeting of NK-92 Cells against High-Risk Rhabdomyosarcomas by Means of an ERBB2 (HER2/Neu)-Specific Chimeric Antigen Receptor

2.1. Cell Lines and Cell Culture

The generation of the continuously expanding effector cell line NK-92/5.28.z by transduction with a lentiviral CAR vector has been described previously [25,41]. These cells stably express a second-generation CAR consisting of an ERBB2-specific scFv antibody fragment derived from the antibody FRP5 [43] with a modified CD8α hinge region and a CD28 transmembrane and intracellular domain as a costimulatory molecule, as well as a CD3ζ intracellular domain [41] (Figure 1).

NK-92/5.28.z cells were cultured in X-Vivo 10 (Lonza, Basel, Switzerland) supplemented with 5% fresh frozen human plasma of blood type AB (DRK-Blutspendedienst, Frankfurt am Main, Germany) and 100 IU IL-2/mL (Proleukin^® S; Novartis, Nurnberg, Germany) [25]. The NK-92 cell line, which is derived from a human NK cell non-Hodgkin’s lymphoma (kindly provided by Prof. Hans G. Klingemann), was cultured under the same conditions as the NK-92/5.28.z cells. The aRMS cell lines RH30 and RH41 were cultured in Roswell Park Memorial Institute medium (Gibco RPMI Medium 1640 (1X) + GlutaMAX™, Thermo Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS, low in endotoxin; Sigma-Aldrich, St. Louis, MO, USA). RMS cell lines were obtained from the DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany). Primary aRMS cells (kindly provided by Sibylle Wehner, University Hospital Frankfurt am Main, Laboratory of Pediatric Hematology, Frankfurt am Main, Germany) were obtained after written and informed consent of the patient and the patient’s parents and cultured according to the RH30 and RH41 cell protocols. The ERBB2-positive and ERBB2-negative human breast carcinoma cell lines MDA-MB-453 and MDA-MB-468 (ATCC; Manassas, VA, USA) were cultured in Dulbecco’s modified Eagle’s medium (DMEM; 1gGluc m.Glutamax/Nabic o.Pyr, Gibco^®; Life Technologies, Darmstadt, Germany) supplemented with 10% FBS [41]. All cell lines were incubated in a humidified atmosphere at 37 °C and 5% CO₂ and kept under antibiotic-free conditions.

2.2. Cell Surface Staining and Flow Cytometric Analysis for the Characterization of Effector and Target Cells

Different cell lines were analyzed for various cell surface markers using a BD FACSCanto II flow cytometer (BD Biosciences, San Jose, CA, USA). Monoclonal antibodies were conjugated herein with phycoerythrin (PE) or allophycocyanin (APC).

aRMS cell lines RH30 and RH41, as well as primary aRMS cells, were tested for ERBB2 expression by staining with an anti-CD340 antibody (anti-CD340, ERBB2/HER-2, PE, BioLegend, London, UK) using the human breast carcinoma cell lines MDA-MB-453 and MDA-MB-468 as positive and negative controls, respectively [41], as well as an isotype control to substantiate the result (IgG1κ Isotype Ctrl FC Antibody, PE, BioLegend, London, UK). In parallel, aRMS cells were assessed for MIC-A/-B (MICA/MICB Antibody, APC, BioLegend, London, UK; isotype control, IgG2aκ Isotype Ctrl FC Antibody, APC, BioLegend, London, UK), as well as for ULBP-1 (ULBP-1 Antibody, APC, Miltenyi Biotec, Bergisch Gladbach, Germany; isotype control, IgG2a Isotype Ctrl FC Antibody, APC, Miltenyi Biotec, Bergisch Gladbach, Germany).

The effector cell line NK-92/5.28.z was analyzed for ERBB2-CAR expression using an ERBB2-IgG-Fc chimera (Sino Biological Inc., Beijing, China) after nonspecific Fc receptor blocking (Human TruStain FcX™, BioLegend, London, UK) and incubation on ice for 20 min. To visualize the chimera, a secondary anti-IgG-Fc monoclonal antibody conjugated with APC (BioLegend, London, UK) was used. Parental NK-92 cells were included as a control.

Cell surface staining was performed according to the manufacturers’ instructions. Aliquots of 1.5 × 10⁶ cells per cell line per FACS^® tube (BD Biosciences, San Jose, CA, USA) were washed twice in phosphate-buffered saline (PBS Dulbecco, Gibco^®, Life Technologies, Darmstadt, Germany) followed by 15 min of incubation with an antibody conjugated to a fluorochrome. The cells were then washed again, and the data for 5–10 × 10⁴ events per tube were acquired by flow cytometry using FACSDiva software (Version 6.1.3, BD Biosciences). The gates were set on viable cells, which were distinguished according to the absence or expression of the marker in question. Kaluza Analysis 3.1 (Beckman Coulter, Krefeld, Germany) was used for data analysis. Single data are given as median fluorescence intensity (MFI) or percentage of gated cells but are given as mean ± SD in case of replicated data.

2.3. Short-Term Cytotoxic Potential of NK-92/5.28.z and Parental NK-92 Cells against aRMS Tumor Cell Suspensions

To analyze the short-term cytotoxicity of cells against aRMS tumor cell suspensions, the europium release assay was used, which is based on bis (acetoxymethyl) 2,2′:6′,2″- terpyridine- 6,6″- dicarboxylate (BATDA) infiltrating the target cell membrane. After the ester bond of the compound is hydrolyzed, hydrophobic 2,2′:6′,2″ -terpyridine-6,6″ -dicarboxylic acid (TDA) is trapped inside the intact cell. If a cell is lysed by an effector cell, TDA will be released, and as it binds to europium, it builds a stable fluorescent chelate complex. The measured fluorescent signal correlates directly with the proportion of lysed target cells (PerkinElmer Inc. Delfia EuTDA Cytotoxicity Reagents).

In brief, target cells were adjusted to 2 × 10⁶/mL, labeled with 5 µL BATDA reagent (PerkinElmer, Waltham, MA, USA), and incubated on a shaker (Shaker SI100; Pharmacia Diagnostics, Uppsala, Sweden) for one hour. To lower the background emission, the cells were then washed four times. In the case of aRMS and human breast carcinoma cells, probenecid (Sigma-Aldrich, St. Louis, MO, USA) was added to prevent BATDA from being transported out of the intact cell by multidrug transporters, thus altering the results. Afterward, the cell densities of both the target and effector cells were adjusted to 5 × 10⁴/mL and 2 × 10⁶/mL, respectively. Cocultures were loaded into a round-bottom plate (96-well plate Nunclon™; Thermo Scientific, Waltham, MA, USA) at effector-to-target (E:T) ratios of 40:1, 20:1, 10:1, and 5:1. Target cells incubated alone were used to determine spontaneous release, and target cells incubated in the presence of 20% Triton™ X-100 solution (Sigma-Aldrich, St. Louis, USA) were used to determine the maximum lysis value. After another incubation period of three hours, the supernatants were transferred into a flat-bottom plate (96-well plate Falcon^®; Corning, New York, NA, USA), and europium solution (Delfia^®, PerkinElmer, Waltham, MA, USA) was subsequently added. The emitted signals were measured using a Fluorometer Victor 3™ 1420 (PerkinElmer, Waltham, MA, USA), and specific lysis of the target cells was calculated as follows: specific lysis (%) = 100 × (experimental release (counts) − spontaneous release (counts))/(maximum release (counts) − spontaneous release (counts)) [44].

2.4. Long-Term Cytotoxicity Assessment of NK-92/5.28.z and Parental NK-92 Cells against aRMS Monolayers

To determine the cytotoxic capacity over a longer time span, a 16-h coculture assay was performed using a Celigo cell cytometer (Nexcelom Bioscience LLC., Lawrence, MA, USA), which enabled assessment of direct cell counts by bright field imaging, as previously described by Merker et al. [15]. In brief, cells at a target density of 2 × 10⁵ per mL per well were plated in a 12-well plate (Falcon^®, flat bottom, Corning, New York, NA, USA). During a six-hour incubation period, the cells were given time to adhere. The supernatants were removed; the cells were washed once in 1 mL PBS, and 0.5 mL culture medium was added. The remaining cells were counted using a Celigo cell cytometer before the addition of effector cells at E:T ratios of 10:1, 5:1, and 1:1, while target cells incubated without effector cells served as controls. Cells were coincubated for 16 h. Thereafter, the supernatants, including the effector cells in suspension, were removed, and the cells were washed twice in PBS. Residual cells were counted again by a Celigo cell cytometer. The specific lysis was calculated for each E:T ratio as follows: specific lysis (%) = (1 − mean of the remaining target cells)/mean of the corresponding negative control [15].

2.5. Analysis of Target Cell Killing of NK-92/5.28.z and Parental NK-92 Cells by Time-Lapse Microscopy

To not only determine the cytotoxic capacity of NK-92/5.28.z cells at a given time point but also follow the kinetics of the killing process, cocultures of target and effector cells were monitored over a period of 18 h. Target cells (1 × 10⁶) were placed on a petri dish (35 × 10 mm, Nunclon™ Delta; Thermo Fisher Scientific, Waltham, MA, USA) and given a six-hour incubation period to adhere. After gently removing the supernatant, effector cells were added at an E:T ratio of less than 1:1, and cocultures were placed under a microscope (Olympus IX71, Research Inverted System Microscope) for video analysis using a 6.3 × magnification. Target and effector cells were discriminated by their different microscopic morphology described in the results (Section 3.1) and discussion section as well as by their different culture behaviors growing as adherent or suspension cells, respectively Figure S1. Viable adherent target cells remained attached to the bottom of the respective culture tissues while effector cells floated and grew suspended in the culture. Killed target cells lost their growth pattern and floated as cell debris in the culture as well. 18 h later, suspension cells were gently removed, and numbers of effector cells were counted using trypan blue staining (Sigma-Aldrich, St. Louis, MO, USA). Then, the remaining tumor cells were chemically removed, and cell numbers were counted.

2.6. Cytotoxic Activity of NK-92/5.28.z and Parental NK-92 Cells against aRMS Tumor Spheroids

Three-dimensional tumor spheroids were generated using an ultralow attachment 96-well round-bottom plate without additional coating (Corning Incorporated, Corning, NU, USA), and 5 × 10³ target cells in 200 µL medium were plated in each well. The plate was then centrifuged for 10 min (1000× g). Half of the culture medium without supplements (100 µL) was replaced every three to four days.

On day four of cell culture, 2 × 10⁵ or 0.5 × 10⁵ effector cells were added to each well. Images were taken every day using a Celigo cell cytometer (with the embryoid body application), starting six hours after the beginning of coculture and ending on day 10. Discrimination of target and effector cells was shown by their different growing behaviors as tumor spheroids and immune effector cell clones/clusters.

2.7. Confirmation of ERBB2 Expression on Tumor Spheroid Cells

The expression of ERBB2 on the tumor cells within the spheroids was investigated by gently resuspending the spheroids, followed by staining with the anti-CD340 antibody and analysis with a flow cytometer using the same protocol as described above (see Section 2.2 Cell Surface Staining and Flow Cytometric Analysis for the Characterization of Effector and Target Cells). Tumor spheroids were harvested on days 1, 3, 5, 7, 10, and 14 of culture for assessment of ERBB2. The results were compared to those of isotype controls and the ERBB2 expression levels of MDA-MB-453 and MDA-MB-468 cells, which were used as positive and negative controls growing in regular culture, respectively.

2.8. Quantitative Analysis of Immunoregulatory Factors Secreted by NK-92/5.28.z and Parental NK-92 Cells in Coculture with aRMS Cells

A bead-based immunoassay using the Multi-Analyte Flow Assay Kit LEGENDplex™ (Human CS8/NK Panel, 13-plex, Cat. No. 740267; BioLegend, San Diego, CA, USA) was performed according to the manufacturer’s instructions to detect the levels of IL-2, IL-4, IL-6, IL-10, IL-17A, interferon (IFN)-γ, tumor necrosis factor (TNF)-α, soluble Fas and Fas ligand (FasL), granzymes A and B, granulysin, and perforin secreted by NK-92/5.28.z and NK-92 cells while they were in contact with tumor cells. The assay uses capture beads that are conjugated to specific antibodies that bind to the target analytes described above. After washing, biotinylated detection antibodies were added, followed by the addition of streptavidin-phycoerythrin (SA-PE). SA-PE binds to the detection antibodies. In these assays, the fluorescent signal intensities measured with a BD FACS Canto™ flow cytometer directly correlate to the quantity of analytes bound. Additionally, standard curves were generated to ascertain the concentrations of the particular analytes.

To assess the responses of NK-92/5.28.z and NK-92 cells while attacking aRMS cells, supernatants of the europium release assay cultures at E:T ratios of 20:1 (absolute cell counts per well, 1 × 10⁵ effector cells and 5 × 10³ target cells) after three hours of coculture were collected and preserved at −80 °C until the immunoassay was performed.

The standard preparation and the assay were both performed according to the information provided by the manufacturer. Evaluation of the immunoassay was performed using LEGENDplex™ Data Analysis Software (V 7.0) (VigeneTech, Carlisle, PA, USA).

2.9. Analysis of Degranulation by CD107a Staining

The CD107a staining was performed as described elsewhere [45]. In short, 1 × 10⁵ RH30 cells were coincubated with the same number of NK-92/5.28.z or NK-92 cells (E:T ratio 1:1) for two hours at 37 °C. For the basal CD107a expression, the effector cells were incubated under the same conditions without RH30 target cells. Thereafter, cell suspensions were washed once with PBS and stained with CD107a-FITC and CD56-PE/Cy7 antibodies (BioLegend, London, UK). Measurements were conducted by BD FACSCanto10c (BD Biosciences, San Jose, USA), and gates were set on CD107a-CD56 double-positive singlets identifying them as degranulating NK-92/5.28.z or NK-92 cells by FlowJo Software 10.3 (BD Bioscience, Ashland, OR, USA).

2.10. Statistical Analysis

For statistical analysis and graphical representation of data, GraphPad Prism 6.0 software (La Jolla, CA, USA) was used. The results are given as the mean ± SD. A two-tailed Student’s t-test was used to evaluate the differences between values. Differences were considered significant for p < 0.05 (*), p < 0.01 (**), p < 0.005 (***), and p < 0.0001 (****). Flow cytometry data are given as MFI or percentage of gated cells but are given as mean ± SD in case of replicated data.

3.1. Alveolar Rhabdomyosarcoma Cell Characterization—ERBB2 Is a Targetable TAA

As a proof of concept, the expression of the growth factor receptor ERBB2 on the aRMS cell lines RH30 and RH41 grown in suspension cultures was verified via flow cytometry using an isotype control as a negative control (Figure 2B,D). To substantiate the results, primary RMS cells from a BM aspirate of, up to that point, an untreated, newly diagnosed adolescent patient with alveolar histopathologic subtype RMS positive for the PAX3-FOXO1 fusion gene were also analyzed for surface expression of ERBB2 (Figure 2A). The heterogeneity in cell size and morphology shown by forward (FSC) versus side scatter (SSC) is a typical feature of RMS. All aRMS tumor cell suspensions displayed low but homogenous ERBB2 expression (RH30, MFI 3.6 ± 0.0, n = 3; RH41, MFI 1.7 ± 0.1, n = 3; primary aRMS cells, MFI 1.9, n = 1) (Figure 2D).

To evaluate whether ERBB2 is downregulated in the process of tumor growth, ERBB2 expression was assessed on RH30 cell suspensions becoming tumor spheroids over a 14-day period (Figure 2C). The ERBB2-negative MDA-MB-468 breast cancer cell line was used as a negative control (MDA-MB-468, MFI 1.2 ± 0.2, n = 3), while the MDA-MB-453 breast cancer cell line served as a positive control (MDA-MB-453, MFI 138.1 ± 26.3, n = 3). Overall, stable expression of ERBB2 was documented after 10 days of culture (Figure 2E).

Considering the potential impact of intrinsic NK-92-mediated tumor cell lysis, the NKG2D ligands MHC class I chain-related proteins A and B (MIC-A/-B) and UL16 binding protein 1 (ULBP-1) were assessed. MIC-A/-B showed no and ULBP-1 very low but consistent expression on RH30 cells (ULBP-1, MFI 0.9 ± 0.0, n = 3).

3.2. Effector Cell Characterization

The expression of the ERBB2-specific second-generation CAR on NK-92/5.28.z cells was assessed via flow cytometry using a recombinant ERBB2-IgG-Fc fusion protein for detection of functional CAR molecules. Stable CAR expression was confirmed via comparisons with unmodified parental NK-92 cells (Figure S2).

3.3. Cytotoxic Potential of NK-92/5.28.z and Parental NK-92 Cells against ARMS Tumor Cell Suspensions

In a 3-h coincubation europium release assay, NK-92/5.28.z cell-mediated lysis of tumor cell suspensions was directly compared to that of unmodified parental NK-92 cells. ERBB2-negative MDA-MB-468 cells were used as a negative control, and ERBB2-overexpressing MDA-MB-453 cells were used as a positive control.

At an effector-to-target (E:T) ratio of 40:1, NK-92/5.28.z cells induced a mean specific lysis of 77.4 ± 10.8% (n = 8) (Figure 3A), 61.4 ± 14.8% (n = 8) (Figure 3B), and 85.0 ± 2.4% (n = 6) (Figure 3C) of RH30, RH41, and MDA-MB-453 cells, respectively.

Altogether, NK-92/5.28.z cells demonstrated significantly increased lysis of ERBB2-positive tumor cells compared to NK-92 cells, even at an E:T ratio of 5:1. Of note, ERBB2-specific NK-92/5.28.z cells achieved killing rates of primary aRMS cells of up to 99% (Figure 3E).

As expected, specific lysis of MDA-MB-468 cells (which was used as a negative control) was minimal in all experiments performed (Figure 3D). In addition, the mean specific lysis of MDA-MB-453, RH30, RH41, and primary RMS tumor cells by parental NK-92 cells remained low but varied depending on the number of NK-92 cells used.

3.4. Cytotoxic Capacity of NK-92/5.28.z and Parental NK-92 Cells against aRMS Cell Monolayers

To evaluate the capacity of NK-92/5.28.z cells to lyse tumor cell monolayers, we performed a 16-h coincubation cytotoxicity assay using Celigo cell cytometry. As described above (see Section 2.4), RH30, RH41, MDA-MB-453, and MDA-MB-468 cells were used as targets, but in these experiments, they were growing as adherent cell monolayers (Figure 3F–I). At an E:T ratio of 1:1, aRMS cell lines were specifically lysed by NK-92/5.28.z cells, with mean lysis rates of 82.4 ± 11.0% and 73.3 ± 14.7% observed for RH30 and RH41 cells, respectively (Figure 3F,G). In contrast, the degree of tumor cell killing by parental NK-92 cells remained low.

3.5. Target Cell Killing by NK-92/5.28.z and Parental NK-92 Cells over 18 Hours

The lysis of tumor cell monolayers was visualized via time-lapse microscopy over an 18-h time period (see the appendix for a movie covering the entire experiment). Therefore, the effects of NK-92/5.28.z and parental NK-92 cells against RH30 tumor cell monolayers were tested in separate experiments, starting with an initial E:T ratio of less than 1:1.

Target and effector cells were discriminated morphologically as well as by their different growing behavior as adherent or suspension cells (see Section 2.5). NK-92/5.28.z cells proliferated in cocultures with target cells (Figure 4A), while NK-92 cells decreased in number (Figure 4B). At the end of the experiment, RH30 cells were almost completely lysed by NK-92/5.28.z cells, while NK-92/5.28.z cells had increased in number (Figure 4A). In contrast, NK-92 cells were not able to lyse aRMS cell monolayers, so RH30 cells overgrew the NK-92 effector cells (Figure 4B).

3.6. Cytotoxic Capacity of NK-92/5.28.z and Parental NK-92 Cells against aRMS Tumor Spheroids

While the 2D experiments mentioned above demonstrated the specific killing capacity of NK-92/5.28.z effector cells, in vivo, these cells would not be acting on single layers of target cells. In 3D tumors, effectors require the ability to interact physically with tumor targets as well as the ability to infiltrate their 3D structures. Moreover, 3D tumorspheres possess potentially inhibitory extracellular matrix and can accumulate anti-effector cell reactivity, and deeper layers of such 3D tumorspheres reproduce the acidity of a typical hypovascular tumor; all of these factors are critical components of an immunosuppressive TME.

To investigate the potential of NK-92/5.28.z cells to lyse aRMS cells growing in 3D culture, we established RH30 and RH41 tumor spheroids and used them to compare the killing capacity of the ERBB2-CAR-NK-92 cells with that of parental NK-92 cells over a time period of 10 days. In these experiments, NK-92/5.28.z cells were able to lyse aRMS tumor spheroids within two days of coincubation (Figure 5A,B). Furthermore, the effector cells proliferated and formed cell clusters, which is in accordance with their natural growth properties (Figure 5D). Parental NK-92 cells displayed similar growth behavior and lysed some of the tumor cells, but the remaining aRMS cells nonetheless proliferated and clustered in steadily growing tumor spheroids.

A similar outcome was observed when cocultures of RH41 cells were initiated with fewer effector cells than shown in Figure 5B; in these experiments, it took five days for NK-92/5.28.z cells to diminish the tumor cell spheroids, while parental NK-92 cells again displayed much weaker activity against the RH41 cells growing in 3D culture (Figure 5C).

3.7. Immunoregulatory Factors Secreted by NK-92/5.28.z and Parental NK-92 Cells in Cocultures with aRMS Cells

Supernatants from 3-h cocultures of NK-92/5.28.z or NK-92 cells with aRMS cells (RH30, Figure 6A, and RH41, Figure 6B) were assessed for the presence of certain immunoregulatory factors using the LEGENDplex™ multianalyte immunoassay.

High levels of different effector molecules were found (Figure 6, orange bars). The mean granulysin concentrations in the NK-92/5.28.z cocultures were approximately 3.4-fold (RH30) to 4.5-fold (RH41) lower than the granzyme A concentrations, showing significant differences between NK-92/5.28.z and NK-92 cells. The mean perforin concentrations were even lower than the granulysin concentrations but remained significantly higher than the concentrations measured in NK-92 cocultures (p < 0.05).

The Fas receptor and its ligand (Figure 6, red bars), which facilitate apoptosis induction upon cell-cell interaction, were also investigated. Significantly more soluble Fas ligand (FasL) was seen after coincubation of aRMS cells with NK-92 cells than after coincubation of aRMS cells with NK-92/5.28.z cells.

The concentrations of tumor necrosis factor (TNF)-α, IL-2, IL-6, and IL-4 were below 10 pg/mL and therefore not quantifiable; consequently, these molecules likely had no impact on effector and target cell interactions, at least in this setting.

Coculture of either NK-92/5.28.z or NK-92 cells with aRMS cells resulted in the secretion of high concentrations of the proinflammatory cytokine IL-17A (Figure 6, blue bars). While the levels of the antitumoral cytokine interferon (IFN)-γ in cocultures with NK-92/5.28.z cells were significantly higher than those in cocultures with NK-92 cells (p < 0.05), the concentrations of the anti-inflammatory cytokine IL-10 (Figure 6, pink bars) in cocultures with NK-92 cells were significantly higher than those in cocultures with NK-92/5.28.z cells (p < 0.05). Nevertheless, IL-10 production was still measurable in NK-92/5.28.z cultures.

Degranulation capacity of immune effector cells without target cell contact remained low (basal CD107a levels for NK-92 (Figure 6C, 1.53 ± 0.33%) and NK-92/5.28.z cells (Figure 6E, 2.02 ± 0.33%) of three representative experiments are shown). After two-hour coincubation with RH30 cells, the parental NK-92 cells showed minimal enhancement of CD107a expression (Figure 6D, 4.34 ± 1.82%), while in NK-92/5.28.z cells, the CD107a expression increased to 35.65 ± 6.63% (Figure 6F).