Chemotherapeutics Used for High-Risk Neuroblastoma Therapy Improve the Efficacy of Anti-GD2 Antibody Dinutuximab Beta in Preclinical Spheroid Models

Simple Summary We investigated the effects of chemotherapeutics used for the frontline treatment of newly diagnosed high-risk neuroblastoma patients in combination with anti-GD2 antibody ch14.18/CHO (dinutuximab beta, DB) in the presence of immune cells in preclinical models of neuroblastoma. The combined treatment showed an up-to-17-fold-stronger and GD2-specific cytotoxic effect compared to the controls treated with chemotherapy alone in the presence or absence of immune cells. These findings further support a clinical evaluation of DB in combination with frontline induction therapy for high-risk neuroblastoma patients. Abstract Anti-disialoganglioside GD2 antibody ch14.18/CHO (dinutuximab beta, DB) improved the outcome of patients with high-risk neuroblastoma (HR-NB) in the maintenance phase. We investigated chemotherapeutic compounds used in newly diagnosed patients in combination with DB. Vincristine, etoposide, carboplatin, cisplatin, and cyclophosphamide, as well as DB, were used at concentrations achieved in pediatric clinical trials. The effects on stress ligand and checkpoint expression by neuroblastoma cells and on activation receptors of NK cells were determined by using flow cytometry. NK-cell activity was measured with a CD107a/IFN-γ assay. Long-term cytotoxicity was analyzed in three spheroid models derived from GD2-positive neuroblastoma cell lines (LAN-1, CHLA 20, and CHLA 136) expressing a fluorescent near-infrared protein. Chemotherapeutics combined with DB in the presence of immune cells improved cytotoxic efficacy up to 17-fold compared to in the controls, and the effect was GD2-specific. The activating stress and inhibitory checkpoint ligands on neuroblastoma cells were upregulated by the chemotherapeutics up to 9- and 5-fold, respectively, and activation receptors on NK cells were not affected. The CD107a/IFN-γ assay revealed no additional activation of NK cells by the chemotherapeutics. The synergistic effect of DB with chemotherapeutics seems primarily attributed to the combined toxicity of antibody-dependent cellular cytotoxicity and chemotherapy, which supports further clinical evaluation in frontline induction therapy.


Introduction
Neuroblastoma is the leading cancer-related cause of death in children [1]. Despite intensive multimodal treatment options, the long-term event-free survival is still only 3-dimensional (3D) spherical aggregation of tumor cells that represent a complex tumor environment and architecture including zones of proliferation at the outside and quiescent cells in the inside [23][24][25]. Therefore, spheroid models provide a more clinically relevant model regarding chemotherapy diffusion, angiogenesis, tumor invasion and chemotherapy resistance compared to 2D models [23,[26][27][28][29][30].
We used the described spheroid model, to test the hypothesis that combining chemotherapeutic agents used in induction therapy regimens for pts with HR-NB with DB can enhance antitumor efficacy and determined their effect on activating and inhibitory receptors and ligands on target and effector cells.

Stable Transduction of Tumor Cells Using Lentiviral Vectors
For recombinant lentivirus production, a second-generation lentiviral vector system was used. The non-confluent Lenti-X™ 293T cells were co-transfected with purified pVSV-G-envelope-expressing plasmid (Addgene, Watertown, MA, USA), psPAX2 (Addgene, Watertown, MA, USA) vector encoding virus polymerase and packaging genes, and lentiviral vector pWPXL (Addgene, Watertown, MA, USA) coding for a near-infrared reporter (NIR) (iRFP680). The transfection of Lenti-X™ 293T cells was conducted by using CalPhos Mammalian Transfection Kit (Takara Bio Europe, Saint-Germain-en-Laye, France) according to the manufacturer's instructions. For transduction, 8 mg/mL Polybrene (Merck KGaA, Darmstadt, Germany) and lentiviral supernatants were added to the target cells. Target cells were cultured under cell culture conditions, and after 72 h, they were tested for the successful transduction of the IncuCyte ® SX5 live-cell analysis system (Sartorius, Göttingen, Germany).

Long-Term Live-Cell Spheroid Viability Assay and Treatment Conditions
To yield three-dimensional (3D) tumor spheroids, we used ultra-low attachment plates (ULA plates, S-BIO PrimeSurface ® , MS-90384UZ) that were precoated with a hydrophilic polymer facilitating spontaneous self-assembly by preventing cellular attachment to the surface. This method was recently shown to be the best approach for studying the efficacy of drugs with respect to spheroid maintenance and reproducibility of results [37]. A total of 3000 iRFP680-positive neuroblastoma cells were seeded into ULA 384-well plates. For CHLA-136, we additionally used 0.5% Matrigel to improve spheroid formation. Cells were centrifuged at 150× g for 10 min and incubated for 72 h at 37 • C and 5% CO 2 . Spheroids were treated with the respective chemotherapeutic compound for 24 h, followed by the addition of 22,500 PBMCs with and without the anti-GD2 antibody DB and incubation for a further 216 h under cell culture conditions. Image acquisition was performed every 8 h for 240 h (10 days), using the IncuCyte ® SX5 live-cell analysis system. Spheroid viability was calculated as the ratio of integrated spheroid fluorescence intensity of every time point to fluorescence at baseline (0 h). Experiments were performed in six replicates, and viability is reported in %±SEM. To validate NIR fluorescence (680 nm) from stable transduced NB tumor cells as the viability marker, flow cytometry analyses were performed by using 40,6-diamino-2phenylindole (DAPI) (Merck KGaA, Darmstadt, Germany). For these analyses, 50% of the cells were lysed (65 • C, 5 min) to obtain samples with live and dead cells in an equal amount. Next, cells were incubated with a 0.1 mg/mL DAPI solution (Sigma-Aldrich, D9542) for 5 min prior to acquisition. For each sample, 20,000 cells were analyzed by using a BD CANTO II cytometer and FACS Diva software (BD Biosciences, San Jose, CA, USA). Data were analyzed with FlowJo V10 software (Ashland, OR, USA) analyzing the frequency of NIR-and DAPI-positive and -negative cells of all single cells.

NK Cell Activation after Chemotherapy Treatment
For the analysis of activating receptor expression by cytotoxic NK cells (CD3-; CD56dim), 5 × 10 6 human PBMCs were treated with the respective chemotherapeutic compound and incubated for 72 h (37 • C/5% CO 2 ). Then cells were harvested and 1 × 10 6 live cells were washed with wash buffer, followed by incubation with 10 µL of Tandem Signal Enhancer (Miltenyi Biotec). Incubation with the following antibodies in a total volume

Statistics
Differences between the groups were assessed by using ANOVA with Dunnett's post hoc test, if the assumption of normality was met (Shapiro-Wilk Test). Due to donordependent variability of PBMCs-dependent antitumor toxicity, the significant difference between the groups was analyzed by using repeated measurement ANOVA for individual data points. Statistical analysis was performed by using GraphPad Prism (version 9.4.1 for Windows, GraphPad Software, San Diego, CA, USA). Viability data are presented as mean ± SEM (standard error of the mean), and flow cytometry data are shown as individual data point with mean and SEM indicated.

Establishment of a Long-Term Real-Time Viability Assay
We developed a long-term viability assay by using live-cell image acquisition and fluorescent tumor cells. First, the neuroblastoma cell lines LAN-1, CHLA-136, and CHLA-20 were transduced by using a lentiviral expression system to yield stable expression of the fluorescent near-infrared protein iRFP680 (NIR) used as viability staining [38]. Stable expression was confirmed by flow cytometry up to one month after transduction. Only cell lines with over 95% NIR + cells were used. The correlation between viability and NIRfluorescence status was confirmed with DAPI staining and analyzed by flow cytometry (Figure 1). NIR-fluorescence was an accurate marker for viability in over 95% of tumor cells analyzed (99.3%, 99.6%, and 97.6% were NIR+ and DAPI-or NIR-and DAPI+, for LAN-1, CHLA-20, and CHLA-136, respectively) ( Figure 1A). Importantly, we found that the effects of chemotherapeutics used in concentrations realistic in the clinical setting were only measurable from day four after the start of the treatment, showing the requirement for long-term assays to assess effects ( Figures 1B and 2A, cisplatin, Video S1-S3 for cisplatin + ADCC and controls in LAN-1).
the fluorescent near-infrared protein iRFP680 (NIR) used as viability staining [38]. Stable expression was confirmed by flow cytometry up to one month after transduction. Only cell lines with over 95% NIR + cells were used. The correlation between viability and NIRfluorescence status was confirmed with DAPI staining and analyzed by flow cytometry (Figure 1). NIR-fluorescence was an accurate marker for viability in over 95% of tumor cells analyzed (99.3%, 99.6%, and 97.6% were NIR+ and DAPI-or NIR-and DAPI+, for LAN-1, CHLA-20, and CHLA-136, respectively) ( Figure 1A). Importantly, we found that the effects of chemotherapeutics used in concentrations realistic in the clinical setting were only measurable from day four after the start of the treatment, showing the requirement for long-term assays to assess effects ( Figures 1B and 2A, cisplatin, Video S1-S3 for cisplatin + ADCC and controls in LAN-1).

Effects of Chemotherapeutics on Antibody-Dependent Cellular Cytotoxicity (ADCC)
We investigated how chemotherapeutics affect the ADCC with DB (10 µg/mL) and effector cells (PBMCs (7.5 × 10 4 cells) against established tumor spheroids generated from the three cell lines, using real-time viability assay over 10 days.
The chemotherapeutics used at clinically relevant concentrations combined with DB and effector cells (ADCC condition) had an up-to-17-fold-higher long-term antitumoral effect in this model (p = 0.0029). ADCC conditions also showed a delayed tumor growth that was stronger compared to the controls of chemotherapeutics combined with effector cells only (without DB; antibody-independent cellular cytotoxicity (AICC); see Figures 2 and 3).
This clearly indicates a DB-dependent and GD2 specific effect. The viability curves of AICC with and without chimeric isotype control (rituximab) were not different.
There was a differential pattern of efficacy depending on the cell line used to establish the spheroids. For instance, the chemoimmunotherapy with platin agents (cisplatin and carboplatin) significantly improved the antitumoral effects compared to chemotherapy alone or compared to chemotherapy with effector cell only (AICC) in LAN-1 and CHLA-20 spheroids, but to a lesser extent in CHLA-136 spheroids (3.6-, 2.8-, and 2.0-fold decrease in viability (at 10 d) versus AICC in combination with cisplatin, respectively; for overview, see Table 1). Cyclophosphamide significantly increased ADCC in LAN-1-and CHLA-20 spheroids but not in CHLA-136 (1.6-, 1.6-, and 0.8-fold decrease in viability compared to AICC + cyclophosphamide, respectively; see Figure 2A, right panel). This might be attributable to higher resistance of CHLA-136 against cyclophosphamide compared to LAN-1 and CHLA-20 ( Figure 2C, right panel). Chemoimmunotherapy with etoposide was highly effective against LAN-1 and CHLA-136 spheroids (6.4-and 2.9-fold decrease in viability compared to AICC in combination with etoposide; see Figure 3A,C), whereas CHLA-20 spheroids were too sensitive to etoposide with PBMCs to allow for a differentiation between chemoimmunotherapy with AICC and ADCC, even at low concentrations of etoposide (0.1 µg/mL, 1.4-fold decrease; see Figure 3B). Chemoimmunotherapy with vincristine was significantly more effective in LAN-1 and CHLA-20 spheroids, but not in CHLA-136-spheroids ( Figure 3A-C) (2.7-, 2.2-, and 1.0-fold decrease in viability compared to AICC + vincristine, respectively; see Figure 3A-C, right panel).
In conclusion, chemotherapeutics used in induction regimens combined with DB and effector cells showed superior effects against neuroblastoma spheroids compared to the respective controls.

Chemotherapy-Induced Stress Ligands on Tumor Cells
To further investigate the reasons for the observed improved antitumor effects of chemoimmunotherapy compared to monotherapy controls, we investigated the induction of stress ligands involved in NK-cell activation (B7-H6, ULBP 1-3 and MICA/B) three days after chemotherapy, using flow cytometry. Chemoimmunotherapy with etoposide was highly effective against LAN-1 and CHLA-136 spheroids (6.4-and 2.9-fold decrease in viability compared to AICC in sensitive to etoposide with PBMCs to allow for a differentiation between chemoim therapy with AICC and ADCC, even at low concentrations of etoposide (0.1 µ g/ fold decrease; see Figure 3B). Chemoimmunotherapy with vincristine was sign more effective in LAN-1 and CHLA-20 spheroids, but not in CHLA-136-spheroids 3A-C) (2.7-, 2.2-, and 1.0-fold decrease in viability compared to AICC + vincristine tively; see Figure 3A-C, right panel).  All cell lines showed a measurable B7-H6 (NKp30 ligand) but low ULBP and MICA/B (NKG2D ligands) baseline cell surface abundance (Figure 4). The pattern of chemotherapydependent induction of stress ligands was cell-line specific. In LAN-1, the NKp30 ligand B7-H6 was significantly increased by cisplatin, etoposide, and cyclophosphamide treatment compared to controls (2.3-, 2.0-, and 1.5-fold; p < 0.0001, <0.0001, and p = 0.0111, respectively), in CHLA-20 by carboplatin, cisplatin, and etoposide (1.2-, 1.2-, and 1.4-fold, p = 0.024, 0.0041, and 0.113, respectively) and in CHLA-136 by cisplatin (1.3-fold, p = 0.0129) ( Figure 4A). This is in line with a higher level of antitumor toxicity by DB, immune cells, and platin compounds compared to the platin compounds and AICC ( Figure 2).   We found a differential induction of the NKG2D ligands ULBP2, ULBP-3, and MICA/B in all three cell lines ( Figure 4B-D). Interestingly, all chemotherapeutics except carboplatin significantly increased ULBP-2 and MICA/B in CHLA-20 and in LAN-1 (up to 4.5-(vincristine) and 3.3-(etoposide) and up to 9-(cisplatin) and 4.5-fold (cyclophosphamide). We mainly observed effects on ULBP3 abundance after cisplatin, etoposide, and vincristine treatment (up to 3.6-fold increase).
Overall, most chemotherapeutics elicited a stress response in LAN-1 and in CHLA-20. However, only cisplatin significantly affected CHLA-136 stress-ligand surface abundance (B7-H6, ULBP-2, and MICA/B, up to 2.3-fold increase). These data indicate that checkpoint ligand expression also correlates with the tumor stress response following chemotherapy.
These data indicate that checkpoint ligand expression also correlates with the tumor stress response following chemotherapy.

Effects of Chemotherapy on Activating NK Cell Receptors
To further evaluate the immunological effects of chemotherapy on NK cells, we determined the percentage of cytotoxic NK cells (CD56 dim ) in lymphocytes and measured activating NK cell receptors (NKp30, NKp44, NKG2D, and CD226) for the reported stress ligands by flow cytometry. Etoposide and cyclophosphamide significantly reduced cytotoxic NK cell abundance compared to the medium control (1.64 ± 0.26%, 0.89 ± 0.17% vs. 8.17 ± 0.73% in live lymphocytes, respectively; see Figure 6A), whereas vincristine treatment significantly increased the NK-cell number (10.04 ± 1.7%, Figure 6A). Most chemotherapeutics did not affect stress-ligand receptors ( Figure 6B,C). However, etoposide and cyclophosphamide significantly increased NKp44 expression (1.66-and 2.2-fold increase; see Figure 6B,D), whereas NKp46 and CD226 were significantly decreased by etoposide and vincristine treatment (1.41-and 1.47-fold decrease, respectively; see Figure 6B,C,E).    In conclusion, etoposide and cyclophosphamide increased the activating receptor NKp44, and vincristine increased the number of cytotoxic NK cells, indicating an immunological impact of etoposide, cyclophosphamide, and vincristine on NK cells.

Role of Stress Ligands in Chemotherapy-Mediated Antitumor Efficacy of the Anti-GD2 Treatment
B7-H6 stress ligand and NKp30 receptor engagement have been shown to play a crucial role in NK-cell activation. Therefore, we deleted the B7-H6 gene in LAN-1 and CHLA-20 cells to investigate the role of B7-H6 interaction in the cytotoxicity of chemoimmunotherapy in our model (Figure 7).
Since chemoimmunotherapy with carboplatin showed a strong effect compared to ADCC controls ( Figure 2B) and carboplatin exclusively increased B7-H6 surface abundance ( Figure 4A, center), we tested the hypothesis that a B7-H6 knockout (KO) in CHLA-20 cells will reverse some of the beneficial effects of the carboplatin-based chemoimmunotherapy. Additionally, we investigated the impact of a B7-H6-KO in LAN-1 cells treated with etoposide-based chemoimmunotherapy, as this was highly effective (Figure 3A   Indeed, we found that the B7-H6-KO of CHLA-20 cells significantly reversed the chemoimmunotherapy effect of cisplatin, carboplatin and vincristine compared to the wildtype control ( Figure 7B, right panel; and Supplementary Figure S1). Since the viability was also improved under ADCC conditions in B7-H6 KO-cells, the effect was mainly attributable to the B7-H6-KO.
In contrast, LAN-1 cells did not show any dependency on B7-H6 ( Figure 7B left panel). In summary, we found a partial B7-H6-dependency in CHLA-20 but not in LAN-1 maybe due to strong checkpoint induction after chemoimmunotherapy (Figure 2A-D, left panel).

Antibody-Mediated NK Cell Activation
To further investigate whether checkpoint-and stress-ligand-induction affect NK cell activation, we measured the activation of NK cells by means of degranulation (CD107a) and IFN-γ production, using flow cytometry, as described in the Materials and Methods section. For that, we cultured the PBMCs of healthy donors for 5 h with DB and LAN-1-and CHLA-20 tumor cells (wild type and B7-H6-KO) pretreated with etoposide and carboplatin, respectively. The ADCC conditions showed a strong NK cell degranulation (CD107a) and activation (IFN-γ) (Figure 7).
The chemotherapy-treated tumor cells did not enhance NK cell activation compared to untreated LAN-1 and CHLA-20 cells ( Figure 7C,D). Indeed, the activation of NK cells against LAN-1 with B7-H6-KO was markedly, but not significantly, decreased, and against CHLA-20, B7-H6-KO was significantly reduced ( Figure 6C, p = 0.0831; and Figure 7D, p = 0.0265). Overall, we could not observe a stronger activation of NK cells by chemotherapytreated compared to untreated tumor cells.

Discussion
We evaluated the effects of chemotherapeutics currently used in the standard induction regimen to treat patients with HR-NB in combination with the anti-GD2 antibody DB against spheroids generated from tumor cells derived from patients with progressive disease. Antitumor efficacy of chemoimmunotherapy was superior compared to the chemotherapy or DB in the presence of immune effector cells (ADCC) alone (up to 17-fold decrease in viability compared to ADCC; see Figures 5 and 6 and Table 1). Our data provide preclinical proof-of-concept for a combined use of chemotherapy with anti-GD2 antibodies against neuroblastoma.
We developed a spheroid viability assay that allowed us to measure the long-term effects of the chemotherapeutic compound at clinically relevant concentrations ( Figure 1). Live-cell microscopy using fluorescent tumor cells provides the advantage of undisturbed long-term viability analysis. Our approach circumvents the common problem of short-term viability assays that lead to EC50 values that are too high to be achieved in patients [22]. Additionally, a spheroid represents a model that is closer to the clinical reality compared to 2D models [26]. The architecture of a spheroid provides a nutrition and oxygen gradient that can result in the development of cancer-stem-like cells that represent a chemotherapy resistant subgroup of high clinical relevance [39]. However, this model can be further improved by incorporating multiple cell types, such as cancer-associated fibroblast and myeloid-derived suppressor cells to mimic an inhibitory tumor microenvironment [30]. The spheroid model used here is limited, as it does not reflect anti-angiogenic effects of chemotherapy and the role of fluidic shear stress in metastasis [28,40,41]. Despite these limitations, we have shown that a long-term spheroid viability assay is an appropriate tool to analyze combined effects of chemotherapy with antibody-dependent NK-cell-mediated tumor-cell lysis (Figures 2 and 3).
NK cells are the main effector cells mediating the effect of DB, and the activation of NK cells depends on an equilibrium of inhibitory receptors [42], such as killer-cell immunoglobulin-like receptors (KIRs) and PD-1, as well as activating receptors, such as NKGD2 and NKp30 [15], binding to the stress ligands ULBPs and MICA/B, as well as B7-H6, respectively.
The induction of stress ligands on neuroblastoma cells (Figure 4) and, to a lesser extent, of activating receptors on NK cells ( Figure 6) might explain the synergistic efficacy of the chemoimmunotherapy (Figures 2 and 3). In line with that, it has been shown that B7-H6 sensitizes HEK293 cells for NK cell-mediated cytotoxicity [16]. Since NKp30 plays a crucial role in NK-cell activation and tumor surveillance, the increased expression of its cognate ligand B7-H6 by chemotherapeutics enhances NK cell-mediated ADCC against tumor cells [43].
However, we also found a chemotherapy-dependent induction of the checkpoint ligand expression on neuroblastoma, namely PD-L1, CD86, CD155, and Gal-9 (up to 5-fold increase vs. control; see Figure 5). Checkpoint ligand expression correlates with poor survival attributed to inhibition of immune surveillance [44]. This observation suggests that we consider checkpoint inhibitors in chemoimmunotherapy concepts. Another limiting aspect for chemoimmunotherapy may be the effects of chemotherapy on NK cell viability that might directly impact the efficacy of an ADCC-based immunotherapy. For instance, in a clinical study in acute lymphoblastic leukemia that also includes vincristine treatment, the total lymphocyte rate was reduced 18 months after maintenance chemotherapy [45]. Here, platin agents did not negatively affect cytotoxic NK cell count and vincristine even increased the NK cell to lymphocyte ratio ( Figure 6A). In contrast, etoposide and cyclophosphamide reduced the number of cytotoxic NK cells ( Figure 6A). However, etoposide and cyclophosphamide significantly increased the activating receptor NKp44 levels on cytotoxic NK-cells ( Figure 6B,D). Importantly, NKp44, but not NKp30 and NKp46, is an activation marker for cytotoxic NK cells ( Figure 6B,D) [46,47].
This underlines the ambiguous effect of chemotherapy with beneficial but also detrimental consequences for immunotherapy, which is dependent on a functional immune effector cells [48]. In light of the encouraging in vitro effects observed here, it remains crucial to evaluate this concept in patients.
In addition to effects of chemotherapy on NK cells and the role of immune checkpoint pathways, the inhibitory tumor microenvironment and inhibitory leukocyte populations have to be considered for a more comprehensive picture of a DB-based chemoimmunotherapy in HR-NB.
Despite a substantial increase of checkpoint ligands during chemotherapy ( Figure 5), we could demonstrate that chemoimmunotherapy with DB improved efficacy in our models (Figures 2 and 3). Regardless of the much weaker stress response in CHLA-136 spheroids compared to LAN-1 and CHLA-20 spheroids, chemoimmunotherapy with cisplatin and etoposide was more effective compared to the monotherapies. CHLA-136 spheroids showed higher resistance toward cyclophosphamide and vincristine, and, consequently, these agents did not further improve the ADCC effect.
Finally, we observed that ADCC was partially B7-H6-stress-ligand dependent ( Figure 7B). However, we could not find evidence to support the hypothesis that chemotherapy-induced stress ligands improved ADCC ( Figure 7C). This might be attributable to the observed strong induction of checkpoint ligand expression found in tumor cells after chemotherapy. Accordingly, we and others showed that PD-L1 was also elevated by ADCC and IFN-γ via JAK/STAT signaling [49]. On top of that, chemotherapy can increase NFκB-signaling, which, in turn, can elevate PD-L1 expression [50,51]. Intriguingly, NFκB is a transcription factor that also positively regulates GD3-synthase and, therefore, GD2-abundance in cancer stem cells [52]. Increased NFκB expression might therefore lead to higher GD2 abundance and increased susceptibility toward anti-GD2 treatment, which is subject to further research.

Conclusions
In conclusion, chemotherapy used in European chemotherapy induction regimens for HR-NB combined with antibody-based immunotherapy can effectively eradicate tumor spheroids derived from relapsed/refractory patients. Our results encourage the implementation of DB in the induction therapy in future clinical trials.