Interleukin-15 Enhances Anti-GD2 Antibody-Mediated Cytotoxicity in an Orthotopic PDX Model of Neuroblastoma

Neuroblastoma is the most common extracranial solid tumor of childhood (1). Most children present with high-risk disease, of which only half are cured with conventional chemotherapy, surgery, and radiotherapy (1, 2). However, the addition of immunotherapy with IL2, GM-CSF, an anti-disialoganglioside (GD2) antibody combined with cis-retinoic acid significantly extends event-free survival in these patients and is now the standard of care (3). Despite this achievement, one-third of children experience treatment failure while receiving immunotherapy or are unable to tolerate IL2-related side effects (2, 4). Improving the current immunotherapeutic regimen is therefore important to maximize survival and decrease therapy-related toxicities.

Anti-GD2 antibodies mediate tumor-directed therapeutic effects through antibody-dependent cell-mediated cytotoxicity (ADCC), for which natural killer (NK) cells are the primary effector cells that bind to the F_c portion of the antibody and eliminate antibody-labeled tumor cells (5). IL2 is a stimulatory cytokine and increases the cytotoxic capacity of NK cells by engaging the IL2 receptor βγ (IL2Rβγ), which is constitutively expressed on the cell surface of NK cells. In addition to IL2, other cytokines such as IL15 can bind to IL2Rβγ and induce overlapping biologic effects in vivo (6).

Preclinical studies established the importance of IL15 on NK-cell maturation and function (7–9). More recently, clinical development of recombinant human IL15 determined tolerability in adults and elucidated the biologic effects of IL15 and NK-cell homeostasis in humans. In patients receiving recombinant human IL15, NK cells hyperproliferate and attain an activated phenotype, leading to NK-cell expansion in vivo and tumor shrinkage in two patients (10). Because NK cells are one of the main effector cells of ADCC (5), we hypothesize that IL15 is equally or potentially more efficient than IL2 in enhancing NK-cell–mediated ADCC against neuroblastoma. Therefore, to compare the immunoadjuvant effects of IL15 versus IL2, we performed ADCC studies in culture and in vivo, contrasting the specific lysis of neuroblastoma patient-derived xenografts (PDX) by IL2- and IL15-activated NK cells. In addition, we examined the cellular properties of IL2- and IL15-activated NK cells and quantified their ability to induce tumor cell death and infiltrate tumorspheres. Altogether, our studies demonstrate that IL15 is more effective than IL2 to enhance ADCC against neuroblastoma and support our current efforts to develop a clinical trial to test the efficacy of IL15 for immunotherapy in pediatric neuroblastoma.

Neuroblastoma cell lines were maintained in RPMI1640 medium (Lonza; CHLA90; ATCC) and DMEM (Lonza; SK-NAS; ATCC) supplemented with 10% heat-inactivated FBS (Biowest), 100 IU/mL of penicillin, 100 μg/mL of streptomycin, and 2 mmol/L of l-glutamine (all Gibco media). Cells tested negative for Mycoplasma by PCR, and short tandem repeat profile confirmed their identity when they were frozen as batches of 20 vials. After thawing, cells were passaged for at least five passages and then used for experiments. New cells were retrieved from the same cryobatch every 2 months after thawing. The PDX lines SJNBL108_X and SJNBL047443_X1 were grown orthotopically in CD1-Foxn1^nu immunodeficient mice (Charles River Laboratories). Karyotype analysis of SJNBL046_X revealed gain on chromosome 17 by fusion to chromosomes 13 and 22, gain on chromosome 1 by fusion to chromosome 7, and gain on chromosome 7 and 15 by fusion to chromosome 1. MYCN amplification was confirmed by FISH (11). All animal studies were approved by the Institutional Animal Care and Use Committee of St. Jude Children's Research Hospital (Memphis, TX). Palpable tumors were harvested and processed into single-cell suspensions for ex vivo testing (5).

CD1-Foxn1^nu mice were used for in vivo immunotherapy testing. We visualized the injection area by using a VEVO 2100 high-frequency ultrasound instrument (Fujifilm Visualsonics) with an MS-700 transducer (50 MHz). Under anesthesia with isoflurane, mice aged 5 to 6 weeks received para-adrenal injections of PDX cells, which were resuspended as a single-cell solution in Matrigel (Corning Inc.), as described previously (11). As described previously, SJNBL046_X tumors grow orthotopically within 4–6 weeks from implantation date (11).

Human NK cells were isolated from residual peripheral blood from heparinized apheresis rings obtained from healthy deidentified donors. Each in vitro experiment was performed with fresh NK cells from a new donor. Peripheral blood mononuclear cells were isolated via density-gradient centrifugation with Ficoll-Paque Plus (GE Healthcare). Red cell lysis was performed with lysis buffer (Qiagen). The RosetteSep Human NK Cell Enrichment Cocktail (StemCell Technologies) and human MACSxpress NK Cell Isolation Kit (Miltenyi Biotec) were used to isolate NK cells with a purity of >95%. RPMI-based media supplemented with 10% heat-inactivated FBS, 100 IU/mL of penicillin, 100 μg/mL of streptomycin, and 2 mmol/L of l-glutamine (all Gibco media) was used to grow NK cells in ex vivo cultures. IL2 (50 IU/mL) and IL15 (10 ng/mL) were provided by the Biological Resource Branch at the National Cancer Institute for preactivation of NK cells in culture.

The anti-GD2 antibody hu14.18K322A (humanized anti-human) was provided to St. Jude Children's Research Hospital (Memphis, TN) and Children's GMP, LLC by Merck Serono and was manufactured by Children's GMP, LLC. Hu14.18K322A was used in all in vitro ADCC experiments because it recognizes human GD2 and contains a human F_c portion that is recognizable by human NK cells. In in vivo experiments, the mAb 14.G2a (mouse anti-human) provided by the Biological Resource Branch at the National Cancer Institute (Bethesda, MD) was used because it recognizes human GD2 but contains a murine F_c portion.

For ADCC assays, PDX were dissociated into a single-cell suspension and grown in culture in 96-well flat-bottom plates (Corning Inc.) at 37°C in 5% CO₂ incubators for 24 hours prior to the experiment. To induce ADCC, hu14.18K322A (10 μg/mL) was added to culture wells 1 hour before coincubating effector cells with tumor cells. The duration of the ADCC assay was 12 hours. The ADCC assays were performed with effector-to-target (E:T) cell ratios ranging from 1:5 to 1:1.25. The CellTiter-Glo luminescent cell viability assay (Promega) was used according to the manufacturer's instructions to quantify specific lysis. A plate reader (SpectraMax M3) measured the luminescence in relative light units (RLU), and specific lysis was calculated with the following formula: |{\rm{specific}}\,{\rm{lysis}}\ ( \% ) = ( {1 - {\frac{{{\rm{RLU}}\,{\rm{treated}} - {\rm{RLU}}\,{\rm{untreated}}\ }}{{{\rm{RLU}}\,{\rm{untreated}}}}} ) \times 100$|⁠.

To test natural NK-cell cytotoxicity (i.e., cytotoxicity in the absence of antibody), the experimental set-up and quantification were identical to the ADCC assays; however, no hu14.18K322A was added.

One day before the assay, NK cells were grown in culture in NK-cell media supplemented with IL2 (50 IU/mL) or IL15 (10 ng/mL) for 12 to 18 hours, and neuroblastoma cells were plated at a density of 50,000 cells per well in 4-well glass bottom chamber slides (Ibidi). On the day of the assay, neuroblastoma cells were labeled with calcein-AM, and NK cells were labeled with CellTracker Orange (Thermo Fisher Scientific), according to the manufacturer's instructions. Hu14.18K322A (10 μg/mL) was added to the wells for 1 hour to allow binding to tumor cells. Dye-labeled NK cells were then added to the wells containing neuroblastoma cells. A C2 Nikon confocal microscope (Nikon) was used to perform live-cell imaging at a magnitude of 10× for 12 hours. Five different areas were imaged per condition. To quantify the imaging data, we extracted the green (8 bit) and red (16 bit) channels and segmented the green and red signal with FIJI software (12) and Weka Trainable Segmentation plug-in (13). We converted the segmented images into a binary mask and applied the 3D ROI manager plug-in to calculate contact area in μm and signal intensity (14).

Yellow fluorescent protein (YFP)-expressing PDX tumors were processed into single-cell suspensions and resuspended in DMEM (Lonza) supplemented with 10% heat-inactivated FBS, 100 IU/mL penicillin, 100 μg/mL streptomycin, and 10 μmol/L rho kinase pathway inhibitor Y-27632 (StemCell Technologies). Tumor cells were plated at a density of 50,000 cells per well into 96-well low-attachment round-bottom plates (Lipidure), centrifuged at 100 × g for 5 minutes, and maintained at 37°C in 5% CO₂ incubators. On day 3 after plating, single tumorspheres were transferred into 24-well low-attachment plates (Thermo Fisher Scientific) and grown for 24 to 48 hours. To determine NK-cell infiltration into the tumorspheres, NK cells were labeled with CellTracker Orange, as per manufacturer instructions. NK cells were coincubated with tumorspheres at an E:T ratio of 5:1 for 12 hours. The tumorspheres were then fixed with 4% paraformaldehyde for 2 hours and placed in 30% sucrose. Cryosections (8-μm thickness) were stained with 4′,6-diamidino-2-phenylindole and imaged with a C2 Nikon confocal microscope. Tumorsphere-infiltrating NK cells were counted for each section, and results were normalized by the sectioned area. To assess cytotoxicity, NK cells isolated by FACS from tumorspheres were tested against K562 for 4 hours. The E:T ratio was 5:1. Residual tumor cells were quantitated by flow cytometry and the results were normalized with particles for absolute cell counting (Spherotech).

Precomplexing of IL15/IL15Rα was performed with recombinant mouse IL15 (PeproTech) and mouse IL15Rα Fc chimera (R&D) by incubating 2.5 μg of IL15 and 12 μg of IL15Rα at room temperature for 30 minutes as described previously (15).

All in vivo testing was performed with tumor-bearing mice with tumor volumes measuring ≥10 mm³ by ultrasound (i.e., an established tumor model). Animals were randomized to four therapy groups: chemotherapy alone (group 1); chemotherapy followed by anti-GD2 antibody and GM-CSF (group 2); chemotherapy followed by anti-GD2 antibody, GM-CSF, and IL2 (group 3); chemotherapy followed by anti-GD2 antibody, GM-CSF, and IL15/IL15Rα complex (group 4). Chemotherapy was given during week 1 of each cycle and consisted of daily intraperitoneal injections of irinotecan (1.6 mg/kg) and oral administration of temozolomide (16.5 mg/kg) for 5 days. All animals in groups 2, 3, and 4 received 100 μg of the mouse anti-human 14.G2a anti-GD2 antibody on days 2 through 5 of week 2. Animals assigned to group 2 received daily 50 μg/kg of intraperitoneal murine GM-CSF for 5 days of week 2. Mice in group 3 were given daily 10,000 IU of recombinant mouse IL2 (PeproTech) intraperitoneally and daily murine GM-CSF for 5 days of week 2. The mouse IL15/IL15Rα complex was administered intraperitoneally to group 4 mice on day 1 of week 2 together with murine GM-CSF. Each therapy cycle lasted for 3 weeks and a total of two cycles were administered to animals. Staff specialized in preclinical imaging were blinded to the therapy groups and conducted weekly acquisition of the index lesions with a VEVO 2100 high-frequency ultrasound instrument (S. Brown) and tumor volume measurements with VEVO LAB software, version 3.1.1 (W.J. Akers).

NK-cell purity after isolation was determined by flow cytometry and defined as CD3⁻ (UCHT1) and CD56⁺ cells (NCAM16.2; BD Biosciences). For phenotypic analysis of murine immune cells, the following antibodies were used: Cd11c (N418), Cd49b (DX5), Ly-6C (1A8), Ly-6G (HK1.4), MHC II (M5/114.15.2; all BioLegend), B220 (RA3-6B2; Tonbo Bioscience), Cd45 (30-F11), Cd11b (M1/70), Cd27 (LG 3A10; all BD Biosciences). For blocking, the purified human FcR-binding inhibitor (Thermo Fisher Scientific) and anti-mouse CD16/CD32 (2.4G2; Tonbo Bioscience) were used. Data were acquired with a BD Biosciences Fortessa instrument with four-laser capacity (17 colors) and analyzed with FCS Express, version 6.

We used D'Agostino–Pearson omnibus and Shapiro–Wilk normality tests to determine normal distributions of variables. Normally distributed groups were compared with two-tailed Student t tests for paired and unpaired data (two groups) and one-way ANOVA and Tukey multiple comparison tests for post hoc analysis of multiple groups. If data were not normally distributed, we log-transformed the data and used parametric tests. Alternatively, we used unpaired Mann–Whitney tests (two groups) and unpaired Kruskal–Wallis tests with Dunn post hoc analysis of multiple comparisons for nontransformed nonnormally distributed data. Parametric Pearson tests were used for correlation analysis. Linear regression analyses were performed for continuous dependent variables. All statistical analyses were conducted with GraphPad Prism 8 or JMP, version 14 (linear regression analysis). P values < 0.05 were considered statistically significant.

To compare the immunoadjuvant effect of IL2 with that of IL15 on NK-cell–mediated lysis of neuroblastoma cells in culture, we preactivated NK cells with decreasing concentrations of IL2 (1–50 IU/mL) and IL15 (1–10 ng/mL) for 24 hours. These concentrations were chosen because they correspond approximately to steady-state levels in humans treated with less than or equivalent to the MTDs of IL2 (4) and IL15 (10). We then added preactivated NK cells to cocultures with the PDX SJNBL046_X at multiple E:T ratios for 12 hours and found that the cytotoxic activity of NK cells was higher after preactivation with IL15 than with IL2 (P < 0.001; Fig. 1A). This effect occurred in a dose-dependent manner (P < 0.001) and was maintained with lower cytokine doses and E:T ratios. When we analyzed the E:T ratios, we observed that higher tumor lysis by IL15-activated NK cells was dependent on cytokine concentration (P < 0.05).

To confirm these findings, we extended our in vitro testing to include four donors and two PDX lines, SJNBL046_X and the ATRX-mutant line SJNBL047443_X1. NK cells were grown in culture in the presence or absence of intermediate doses of IL2 (i.e., 10 IU/mL) or IL15 (i.e., 5 ng/mL), and an E:T ratio of 5:1 was used in the assay. When we performed multiple comparisons of the cytotoxic activity of resting and IL2-and IL15-stimulated NK cells among donors and PDX lines, we found that the most effective tumor cell ablation was achieved by IL15-activated NK cells (Fig. 1B and C). Specific lysis was dependent on the type of cytokine used to induce activation (P < 0.001) and the presence of hu14.18K322A (P < 0.001). Altogether, these results show that the stimulatory effect of IL15 exceeded that of IL2 and led to improved ADCC against neuroblastoma PDX cells in culture.

Because IL15 can alter chemotaxis and adhesion properties of NK cells (16, 17), we compared the cellular interactions of NK cells with tumors after IL2 and IL15 activation in culture. We developed a two-dimensional contact assay and used live-cell imaging to quantitate the contact area between NK cells and tumors (Fig. 2A). For these human experiments, two donors were tested with the neuroblastoma cell lines CHLA-90 and SK-N-AS, which were chosen because they exhibited comparable human leukocyte antigen expression levels (data not shown). The contact between tumor and effector cells, as well as cell death, over time were highly correlated (Spearman r = 0.6; P = 0.005; Fig. 2B and C). Although NK–neuroblastoma cell contact was predicted by time (P < 0.001), donor (P < 0.001), and tumor cell line (P < 0.001), it was independent of the type of cytokine stimulation (P > 0.05). Collectively, these results suggest that the cellular interaction of NK cells with tumors are equivalent after preactivation with IL15 or IL2.

Neuroblastoma is associated with repressed immunity, which is associated with low tumor mutational burden and reflected by a lack of infiltrating T lymphocytes in the tumor microenvironment (18–20). IL15 promotes chemotaxis of NK cells (16) and therefore may improve tumor invasion in neuroblastoma. To study the effect of IL15 on NK cell chemotaxis, we developed a PDX tumorsphere model using the “immune cold” PDX line SJNBL046_X that, despite visible splenic infiltration, lacks the presence of tumor-infiltrating immune cells in a humanized mouse model (Fig. 3A). Initially, we isolated NK cells from the peripheral blood of healthy donors and quantitated the number of IL15–activated NK cells in the tumorspheres. Sphere-infiltrating NK cells (SPHINK) were detectable within 4 hours of incubation (Fig. 3B) and upregulated cell surface expression of NKp44, CD69, CD25, and CD56, as compared with NK cells that were cocultured with but did not infiltrate tumorspheres (Fig. 3C). However, we found that FACS-isolated SPHINKs eliminated significantly less K562 tumor cells in culture than did sphere-associated NK cells (P ≤ 0.001), suggesting an attenuated cytotoxic capacity (Fig. 3D). Determining the expression of exhaustion markers, we noted comparable percentages of PD-1, TIM-3, and LAG-3 expression. However, a higher percentage of SPHINKs expressed TIGIT than did tumorsphere noninfiltrating NK cells (Supplementary Fig. S1). We compared the chemotactic effects of IL15 with those of IL2 and detected higher numbers of SPHINKs after prestimulation with IL15 (P = 0.002; Fig. 3E). Taken together, this PDX-derived spheroid neuroblastoma model recapitulates the immunosuppressive tumor microenvironment and attenuates NK-cell function despite compensatory upregulation of activating NK-cell receptors. We demonstrated that IL15 can partially overcome these barriers by increasing NK-cell invasion into tumors in this model.

Preclinical studies have demonstrated enhanced ADCC in the presence of IL15 (7–9). Therefore, we integrated the IL15/IL15Rα complex into a treatment regimen that is closely aligned with the chemoimmunotherapy regimen given to patients in the clinic (21, 22) and tested whether IL15 can substitute for IL2 in immunotherapy for neuroblastoma. We evaluated the antitumor effects in four experimental groups with PDX-bearing CD1-Foxn1^nu mice (Fig. 4A), which are immunocompromised and have decreased numbers but fully functional endogenous NK cells with xenoreactivity (Supplementary Fig. S2; ref. 23). Before proceeding with the experiment, we also examined whether immune editing occurred when human PDX cells were propagated in nude mice instead of more immunocompromised NSG mice. We injected 1 × 10⁶ SJNBL046_X neuroblastoma xenograft cells orthotopically into nude and NSG mice. These tumors grew within 4–5 weeks and were subsequently processed into a single-cell suspension and analyzed for NK-cell ligands. We then used residual tumor cells and reinjected them into mice, switching the host (i.e., tumor harvested from nude mouse was reinjected into an NSG mouse and vice versa). After about 4–5 weeks, tumor growth was apparent by ultrasound and the analysis for NK-cell ligands was repeated. NK-cell ligands did not change expression when tumors from NSG mice were passaged through nude mice (e.g., ULBP2, DR4, CD112, MICA, and MICB) or vice versa (e.g., DR4, DR5, CD112, ULBP3, MICA, and MICB; Supplementary Fig. S3). There were other ligands that were downregulated. However, downregulation was noted in both subsets of experiments (e.g., NTB-A and CD155), suggesting that this was not due to host-related immune editing.

Mice received chemotherapy alone (group 1); chemotherapy followed by 14.G2a antibody and GM-CSF (group 2); chemotherapy followed by 14.G2a antibody, GM-CSF, and IL2 (group 3); or chemotherapy followed by 14.G2a antibody, GM-CSF, and IL15/IL15Rα complex (group 4; Fig. 4B). Weekly tumor volumes and weights (Supplementary Fig. S4) were obtained.

Animals in group 4 exhibited the most suppressed tumor growth among all four therapy groups (Fig. 4C). In direct comparison, mice in group 4 had smaller tumors than those in group 1 (P = 0.011), group 2 (P = 0.016), and group 3 (P = 0.035). Five of 7 (71%) mice in group 4 achieved partial (>20% volume reduction) or complete tumor regression, whereas tumor regression did not occur in any group 1 mice (0/8; 0%), occurred in 3 in group 2 (3/6; 50%), and occurred in 1 in group 3 (1/7; 14%; Fig. 4D). None of the mice did lose more than 20% of their body weight during therapy, demonstrating tolerability of the chemoimmunotherapy regimens (Supplementary Fig. S4). Altogether, we found that the substitution of IL15/IL15Rα for IL2 leads to significantly improved tumor regression in vivo.

In animals that completed therapy, we harvested the spleens and tumors (if present) to analyze the proportions of tissue-infiltrating monocytes/macrophages and NK cells. By using a gating strategy as described previously (24), we quantitated the percentage of CD11b⁺ Ly6G⁻Ly6C⁺MHC II⁻ (monocytes) and CD11b⁺ Ly6G⁻Ly6C⁻MHC II⁺ tumor-associated macrophages and used the spleen as a reference (Fig. 5A). Generally, higher percentages of CD11b⁺ Ly6G⁻ macrophages were found in tumors than in spleens for all therapy groups, except group 2 (Fig. 5B). These macrophages primarily comprised a CD11b⁺Ly6G⁻Ly6C⁻MHC II⁺ population (Fig. 5C and D). When we quantified the percentage of DX5⁺CD27⁺CD11b⁻ immature NK cells and DX5⁺CD27⁻CD11b⁺ mature NK cells in tumors and spleens (ref. 25; Fig. 5E), we did not find any significant differences between the NK-cell percentages in the two tissues or within tissues among the different therapy groups (Fig. 5F). However, we observed significantly decreased numbers of DX5⁺CD27⁺ immature NK cells in tumors from mice treated with the IL15/IL15Rα complex than in other therapy groups (P = 0.012), whereas no difference was noted in spleens (Fig. 5F and G). This suggests that IL15/IL15Rα promotes NK-cell maturation, thereby decreasing the proportions of immature tumor-infiltrating NK cells.

To investigate transcriptional changes and activation of functional pathways leading to enhanced cytotoxicity of NK cells after IL15 stimulation compared with IL2, we performed differential gene expression of NK cells from CD1-Foxn1^nu mice after stimulation with either cytokine. Analyzing bulk RNA-sequencing data of two experimental replicates for each condition, we noted that IL15 activation led to a 117-fold (or 6.9-log fold) upregulation of Gzmd compared with IL2 (Supplementary Fig. S5). Granzyme D is expressed in murine cytotoxic lymphocytes and functionally homologous to human granzyme B or H (26). Thus, an elevated expression of Gzmd could contribute to the enhanced antitumor activity observed in IL15- compared with IL2-treated mice.

A seminal immunotherapy trial in pediatric solid tumors established the efficacy of an anti-GD2 antibody, IL2, and GM-CSF for neuroblastoma (3). Although this regimen is now the standard of care, treatment failure and IL2-related toxicities remain a clinical challenge. Because of the known biologic effects of IL15 on NK-cell function and development, we investigated the immunoadjuvant properties of IL15 in the context of neuroblastoma immunotherapy. We found that IL15, integrated in the current immunotherapy regimen for neuroblastoma, exhibits notable preclinical antitumor activity in vitro and in vivo that exceeds the effect of IL2 in a murine model with established tumors. The antitumor effect of immunotherapy with IL2 in vivo was comparable with tumor shrinkage induced by chemotherapy with or without antibody and GM-CSF. Although this observation appears to be at odds with the current treatment consensus, it is important to point out that immunotherapy with anti-GD2 antibody, IL2, and GM-CSF has proven to be effective in the state of minimal residual disease (27). There is emerging evidence that the combination of chemo-immunotherapy also induces shrinkage of large tumors in the clinic (2). However, studies of chemo-immunotherapy in an established xenograft model have not been performed prior to our report. On the basis of these findings, we propose that IL15 is a feasible substitute for IL2 and should be included in future clinical studies to test the tolerability of IL15 for immunotherapy in children with relapsed/refractory neuroblastoma, with antitumor effects as a secondary endpoint.

Recombinant IL2 was the first cytokine used in the clinic for cancer therapy (28). However, the therapeutic benefit of IL2 for high-risk neuroblastoma is controversial. In the most recent randomized controlled trial by the European SIOP Neuroblastoma Group, there was no survival benefit at 3 years with the addition of subcutaneous (SC) high-dose IL2 (6  ×  10⁶ IU/m² per day for 10 days) to ch14.18/CHO and GM-CSF. Nevertheless, cautious interpretation of these results is warranted given that one-third of patients abrogated therapy due to IL2-related toxicities, thus confounding the study results (27). Low or ultralow doses of IL2 are better tolerated but can induce regulatory T-cell proliferation, which suppresses NK-cell expansion and function in preclinical studies (29).

IL15 has unique pharmacokinetic, can directly engage IL2Rβγ, and requires IL15Rα to form a complex for trans-presentation on the cell surface of activated monocytes to exert its immunologic properties (30). In the presence of GM-CSF, IL15 is mobilized to the plasma membrane of monocytes (31), which may increase its bioavailability. Patients with neuroblastoma receive GM-CSF because it improves ADCC in preclinical studies (32). Therefore, by replacing IL2 with IL15 in the current therapy regimen, this effect may be further exploited, leading to a reduction in overall cytokine doses required for inducing the desired biologic antitumor effects and improve the tolerability of chemoimmunotherapy.

The IL15 and IL15Rα concentrations used in our in vivo experiments were based on previously published dosing in mice that yielded a biologic effect in NK cells (15). In preclinical studies with ALT-803, an IL15/IL15Rα complex, mice received 0.2 mg/kg of ALT-803 on day 1 and 8 of therapy that corresponds to 0.06 mg/kg of IL15 (33). We administered IL15 at a dose of about 0.08 mg/kg on day 8 of each cycle. Therefore, we believe that the chosen regimen is in the range of previously tested in vivo concentrations that in case of ALT-803 were successfully translated into the clinic.

Recombinant human IL15 has been tested as intravenous bolus infusion (10) and subcutaneous injection (34) in adult patients with cancer. The MTD was found to be significantly lower when given as intravenous bolus infusion (0.3 μg/kg/day) than as subcutaneous injection (2.0 μg/kg/day). Grade 3 or 4 adverse events at these dose levels were reported to be aspartate aminotransferase elevation, leukopenia, lymphopenia, neutropenia, and anemia (intravenous) as well as hypophosphatemia, hypertension, and lymphopenia (subcutaneous), respectively. In patients who received IL15 daily by bolus infusion for 12 consecutive days, correlative analysis of NK-cell kinetics in the peripheral blood revealed decreased cell numbers as early as 20 minutes after start of the infusion, influx, or demarginalization of NK cells over the subsequent 2 days, followed by hyperproliferation until daily infusions were ceased (10). In patient who underwent daily subcutaneous injections of IL15 for 5 days of two consecutive weeks, highest NK-cell numbers were found 3 days after completion of this regimen (day 15). However, cytotoxicity against K562 was higher after the 10th dose of IL15 compared pretreatment and day 15 testing (34). Considering this knowledge regarding NK-cell function, NK-cell expansion, and spectrum of adverse events with IL15 treatment, there are several ways to incorporate either intravenous or subcutaneous IL15 after confirming the pharmacokinetic and dynamic data obtained in adults in a pediatric phase I trial. For example, one could administer daily intravenous IL15 and start anti-GD2 antibody and GM-CSF on day 3–7. Alternatively, adopting the subcutaneous injection regimen, patients may be pretreated with 5 days of subcutaneous IL15 and commence anti-GD2 antibody and GM-CSF therapy during week 2 (day 8–12). Cytokine-producing CD56^bright NK cells are hypothesized to be the precursors of terminally differentiated cytotoxic CD56^dim NK cells (35) and to mature in response to IL15 (36). However, the predominant subtype in tumors comprise CD56^bright-infiltrating NK cells (37). In our study, the number of immature murine CD27⁺CD11b⁻ NK cells decreased more in tumors after treatment with IL15 than with other therapy groups, constituting a potential mechanism for improved ADCC with IL15. This should be confirmed with more detailed phenotypic characterization, such as CD16 expression, in the future.

In this study, we used athymic immunocompromised CD1-Foxn1^nu mice that have low numbers of functional NK cells. We performed phenotypic, cytotoxic, and molecular studies to assess the functionality of NK cells derived from these animals. NK cells produced IFNγ, perforin, and granzyme B, and were able to degranulate upon contact with target cells. In addition, they showed significant upregulation of Gzmd and activation of the granzyme A pathway that corresponded to increased cytotoxicity and cytokine production after activation with IL15 in culture. Altogether these findings suggest that NK cells from CD1-Foxn1^nu mice can execute common NK-cell functions and respond accordingly to stimulation with IL2 and IL15.

Paradoxically, we found increased tumorsphere invasion by IL15-activated NK cells but less cytotoxic efficacy in these cells compared with tumor-noninfiltrating NK cells. We hypothesized that this effect may be induced by the immunosuppressive tumor microenvironment. Compared with resting NK cells, we found that IL2 and IL15 activation can partially preserve the cytotoxic function of tumorsphere-infiltrating NK cells.

Although nude mice have been previously utilized to model ADCC in neuroblastoma and led to the discovery of the monoclonal anti-GD2 antibody for therapeutic purposes (38), there are several limitations to this model system. The passaging of tumor cells in nude mice did not results in immune sculpting of common NK-cell ligands. However, in our ex vivo experiments, we noted significantly less NK-cell cytotoxicity against human than mouse tumor cells. In the presence of anti-GD2 antibody, lysis of human neuroblastoma PDX cells by IL15-activated murine NK cells approximated that yielded by resting murine NK cells against the syngeneic cell line YAC-1. Although NK cells have been regarded as one of the primary effector population in ADCC, it is unclear to what degree murine myeloid cells contributed to tumor shrinkage in our experiments. Nevertheless, the combined effect of myeloid and NK cells led to significant tumor shrinkage in our tumor model when IL2 was replaced by IL15 in the current immunotherapy regimen. Future studies could further clarify the relative contribution of myeloid cells by using clodronate or other agents that temporarily decrease the number of this population (39). Likewise, the use of anti-GM1 antibody that neutralizes murine NK cells could help discern the cytotoxic effect of NK cells (40). Other limitations are the use of only one PDX line in the tumorsphere model.

On the basis of our results, we propose that recombinant human IL15 or the soluble IL15/IL15Rα complex is a feasible IL2 substitute in the current immunotherapy regimen for neuroblastoma. Because IL15 has not been previously tested in pediatric studies, initial efforts to evaluate its tolerability in children should be prioritized but reports of tumor responses as a secondary endpoint should be included, in addition to detailed correlative studies to elucidate the biologic effects of IL15 when given in combination with immunotherapy.

No potential conflicts of interest were disclosed.

Conception and design: R. Nguyen, J.L. Norrie, M.A. Dyer

Development of methodology: R. Nguyen, J.L. Norrie

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Nguyen, A. Moustaki, J.L. Norrie, W.J. Akers, M.A. Dyer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Nguyen, J.L. Norrie, W.J. Akers, A. Shirinifard, M.A. Dyer

Writing, review, and/or revision of the manuscript: R. Nguyen, A. Moustaki, A. Shirinifard, M.A. Dyer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Nguyen, W.J. Akers, M.A. Dyer

Study supervision: M.A. Dyer

Other (assisted with animal handling, imaging, observations, and care for the animals): S. Brown

We thank Asa Karlstrom, PhD, for regulatory support; Nisha Badders, PhD, ELS, for editing the manuscript; Jennifer Peters, PhD, for optimizing the live-cell microscopy protocol; Melissa Johnson, BS, for performing tumor injections; Benjamin Youngblood, PhD, for comments and suggestions; Merck Serono and Children's GMP, LLC for providing hu14.18K322A for our studies; and the NCI Biological Resource Branch for providing IL2 and IL15. This work was supported by the NIH grant P30 CA021765 and ALSAC (all authors); ASCO Conquer Cancer Young Investigator Award (12822; to R. Nguyen); R50CA211481 (to W.J. Akers); and EY014867, EY018599, and CA168875 (to M.A. Dyer). This research was supported by Howard Hughes Medical Institute.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.