Cabozantinib (XL-184; kindly provided by Dana Aftab (Exelixis, Inc., San Francisco, CA, USA) was stored as a solid powder. Axitinib (AG-013736, ref A-1107) was purchased from LC Laboratories (Woburn, MA, USA). For in vitro experiments, cabozantinib and axitinib were dissolved in 100% dimethyl sulfoxide (DMSO) and diluted in complete cell culture. For in vivo experiments, cabozantinib was dissolved daily in water/HCl to a final concentration of 3 and 6 mg/ml; axitinib was suspended in 5% carboxyl methylcellulose to a final concentration of 6 mg/ml.

Luciferase gene transfected IGR-N91 and IMR-32 cells were derived from metastatic neuroblastoma as previously reported (33). IMR-32-Luc and IGR-N91-Luc were maintained in Roswell Park Memorial Institute (RPMI)-1640 medium and Dulbecco's modified Eagle's medium (DMEM), respectively, containing 10% fetal calf serum (FCS; all from Life Technologies) at 37°C and 5% CO₂. Cell lines were regularly tested and found to be free of mycoplasma.

For cell proliferation assay, 15,000 IMR-32-Luc and IGR-N91-Luc cells/well were seeded in 96-well plates and incubated for 72 h with cabozantinib or axitinib at concentrations of 0–10 µmol/l or 0.5% DMSO. Cell confluence was imaged by phase contrast using the IncuCyte HD system (IncuCyte™ live-cell). Frames were captured at 4-h intervals from 2 separate regions/well using a ×10 objective. Cultures were run three times in at least quadruplicates. Proliferation growth curves were constructed by imaging plates using IncuCyte™ Zoom software, where growth curves were built from confluence measurements acquired during round-the-clock kinetic imaging. Cell migration was evaluated by scratch assays. A scratch was made on confluent monolayers using a 96-pin WoundMaker™ and incubated with cabozantinib at 5 µmol/l for 48 h. Wound images were automatically acquired and registered by IncuCyte Zoom software system. Data were processed and analyzed using IncuCyte Zoom 96-Well Cell Invasion Software Application Module (all from Essen BioScience, Inc., Ann Arbor, MI, USA). Data are presented as the wound width closure. The rate of wound closure was compared between treated and non-treated conditions.

Total tumor lysates were separated electrophoretically and proteins were detected using monoclonal mouse antibody anti-human β-actin (S125; diluted 1:1,000; Cell Signaling Technology, Danvers, MA, USA), mouse polyclonal anti-human PARP-1 (Ab-2, 1:600; Calbiochem, San Diego, CA, USA), rabbit polyclonal anti-human p-ERK1/2 (Thr202/Tyr204), ERK1/2, p-AKT (Ser473), AKT, p-VEGFR2 (Tyr1175) (19A10), VEGFR2, p-MET (Tyr1234/1235), p-MET (Tyr1249), MET (1:1,000, all from Cell Signaling Technology) as previously described (34).

Protein lysates of parental IGR-N91-Luc and axitinib-resistant AxiR cells were subjected to the human angiogenesis and phospho-MAPK array kit (R&D Systems, Inc., Minneapolis, MN, USA) according to the manufacturer's recommendations.

Animal experiments were carried out under conditions established by the European Community (Directive 2010/63/UE) and approved by the Comité d'Ethique en Expérimentation Animale n°26 (CEEA26) and the French Ministry (MENESR) (approval number: 00328.01). Antitumor activity was evaluated against orthotopic and systemic neuroblastoma in female Swiss athymic mice, NSG or BalbC RAGγc mice, established by injection of 10⁶ cells into the left adrenal or the tail vein, respectively, under isoflurane anesthesia as previously reported (33,34) and all efforts were made to minimize suffering. Briefly, treatment started when positive bioluminescent signals were detected. Cabozantinib was administered by oral gavage at 30 or 60 mg/kg/day for a minimum of 14 days and axitinib at 30 mg/kg/twice daily (BID) during 29 days; control animals received vehicle. Animals were followed daily for clinical status, weekly for body weight. Antitumor activity was determined using ultrasound and bioluminescence, respectively. Statistical difference was determined using the two-tailed non-parametric Mann-Whitney or Kruskal-Wallis test and Prism^® software version 6.00. For pharmacodynamic analysis, primary tumors were harvested 2 h after the end of treatment at day 29.

Hematoxylin-eosin-safranin staining was performed on paraformaldehyde fixed tumors for morphology; immunohistochemistry for CD34 and caspase-3 expression was determined using rat anti-mouse CD34 antibody (1:20; Hycult Technologies) and rat anti-human cleaved caspase-3 antibody (1:100; Cell Signaling Technology) as previously reported (34).

In order to explore resistance mechanisms to selective VEGFR inhibition we first developed in vitro a resistant cell line, IGR-N91-Luc-AxiR, through chronic exposure of IGR-N91-Luc cells in culture to the VEGFR1-3 tyrosine kinase inhibitor axitinib. Axitinib at the 50% growth inhibiting concentration (IC₅₀ of 0.75 µmol/l) allowed maintaining cell viability as well as selection of resistant clones in the cell line. Constant resistant behavior was achieved after five consecutive passages under continuous axitinib exposure. In regard to cell proliferation inhibition, the IC₅₀ of axitinib in the parental IGR-N91-Luc cells was 0.75 µmol/l when administered once as measured by phase-contrast imaging during 72 h. In contrast, proliferation of the resistant IGR-N91-Luc-AxiR cells was less affected with an IC₅₀ of 3 µM (Fig. 1A). Indeed, growth of IGR-N91-Luc-AxiR cells under axitinib treatment was identical to that of the non-treated parental IGR-N91-Luc cells (data not shown). We further explored the migration capacity of the parental and axitinib-resistant IGR-N91-Luc cells under axitinib treatment. Cells were seeded in monolayers into 96-well plates and a scratch was done when they were at confluence. Axitinib at the IC₅₀ or 0.5% DMSO were added into each well and migration was followed during 48 h when the wound of non-treated parental IGR-N91-Luc control cells was nearly closed. For parental IGR-N91-Luc cells wound repair inhibition of up to 54% was observed under axitinib treatment whereas IGR-N91-Luc-AxiR cells showed no reduction in cell migration under axitinib and had nearly closed the scratch wound similar to non-treated cells (Fig. 1B). Thus, axitinib-resistant IGR-N91 neuroblastoma cells exhibit reduced sensitivity to inhibition of both growth proliferation and cell migration to axitinib. To explore potential escape mechanisms involved in this resistance we subjected the cell lysates of parental IGR-N91-Luc and IGR-N91-Luc-AxiR cells to proteome arrays for angiogenesis protein and MAPK signaling expression which simultaneously detect the relative levels of 55 angiogenesis-related proteins and the relative phosphorylation of 24 kinases captured by 26 different antibodies, respectively, spotted in duplicates on a nitrocellulose membrane (Fig. 1C). IGR-N91-Luc-AxiR cells exhibited overexpression of HGF and phosphorylation of downstream effector ERK1 as compared to the parental cells as well as reduced expression of PEDF, a physiological negative regulator of angiogenesis. In response to axitinib treatment at 0.75 µmol/l, the parental IGR-N91-Luc cells showed inhibition of VEGFR2 phosphorylation up to 120 min with an activation of p-ERK at the latest time-point explored. In contrast, IGR-N91-Luc-AxiR cells exhibited activated/phosphorylated ERK at baseline, consistent with the MAPK array, which diminished at the latest time-point and no reduction of VEGFR2 phosphorylation was observed to axitinib exposure (Fig. 1D). Thus, activation of the HGF/MET and MAPK signaling pathway is involved in resistance to VEGFR inhibition in neuroblastoma.

We next explored the effects of HGF stimulation on the two parental IGR-N91-Luc and IMR-32-Luc neuroblastoma cell lines. Cells cultured under serum-free medium conditions overnight were incubated with 60 ng/ml HGF and harvested at indicated time-points up to 120 min for western blot analyses (Fig. 2A). In the IMR-32-Luc cells HGF exposure resulted in significant activation of its receptor MET with an increase of phosphorylation in sites Tyr1234/1235 and to a lesser extent at Tyr1349 whereas expression levels of the total receptor remained stable. The effects were less prominent in the IGR-N91-Luc cells. Based on the hypothesis raised above that HGF signaling is involved in promoting cell survival under selective VEGFR inhibition, we subjected both parental cell lines to the potent VEGFR2 and MET tyrosine kinase inhibitor cabozantinib to the proliferation assay for 72 h. Cabozantinib inhibited cell growth of IGR-N91-Luc and IMR-32-Luc cells in a concentration-dependent manner, with IC₅₀ doses of 1.4 and 2.8 µM, respectively (Fig. 2B). Despite the similar IC₅₀ of the cells, treatment with cabozantinib at 5 and 10 µM for 2 and 24 h resulted in inhibition of phosphorylation of MET and downstream signaling effectors AKT and MAPK in IMR-32-Luc cells, whereas the inhibition was found only at higher concentrations and at 2 h in the IGR-N91-Luc cells (Fig. 2C). Effects of selective HGF stimulation on cell proliferation capacity were impossible to evaluate due to the lack of growth of these neuroblastoma cells in the absence of serum. When exploring the effects on migration in both cell lines we found no significant increase in cell migration to HGF selective medium as compared to 10% FCS growth conditions although IMR-32-Luc cells seemed stimulated under HGF (ns). Cabozantinib at 5 µM reduced significantly migratory capacity in both cell lines, under both selective HGF and serum conditions, and migration was completely inhibited in IMR-32-Luc (Fig. 2D). Thus, activation of HGF-mediated MET signaling is involved in neuroblastoma cell migration and inhibition of MET results in reduced tumor cell migration capacity (1,8,9).

In vivo we first explored pan-VEGFR inhibition using axitinib against the orthotopic IGR-N91-Luc neuroblastoma xenograft model in BalbC RAGγC mice in order to extend our previous data in the subcutaneous and orthotopic IGR-N91 xenografts that had shown growth inhibition to axitinib (35). Axitinib treatment was initiated 2 days after injection of cells into the left adrenal administered by oral gavage twice daily at 30 mg/kg/dose and was given for a longer time period (i.e. 29 days) than the previous experiments in order to allow potential development of metastases in this model. At day 29, several animals in the axitinib treated group had reduced tumor load although the median tumor volume of primary adrenal tumors was not significantly reduced compared to controls as determined by ultrasound (Fig. 3A). At autopsy at day 29, macroscopically visible metastases were detected in 6 out of 16 animals treated with axitinib, located in the liver, whereas this was the case in 1 out of 8 control animals (Fig. 3B). Bioluminescence imaging ex vivo detected a signal increase in tumor burden in the livers of treated animals as compared to the treated animals (ns; Fig. 3C). Western blot analysis of adrenal tumors of 7 control and 14 treated animals revealed enhanced MET and ERK phosphorylation in some tumors, independent of tumor size and the occurrence of metastasis. AKT phosphorylation was inhibited in smaller samples in comparison to controls (Fig. 3D). In addition 10 out of the 15 analyzed tumors treated showed SRC phosphorylation which was present only in 2 out of 7 controls. Thus, selective VEGFR1-3 inhibition may result in enhanced metastatic spread and SRC activation in neuroblastoma tumors.

Based on our observation in vitro that HGF-mediated MET upregulation could be involved in the invasiveness of neuroblastoma cells, we investigated the dual inhibition of VEGFR and MET against the adrenal IGR-N91-Luc model in NSG mice. Cabozantinib at 30 and 60 mg/kg resulted in a dose-dependent tumor growth inhibition of 60 and 87%, respectively, compared to the control group (P=0.005; Kruskal-Wallis test) with 7 animals in each group (Fig. 4A–C). Protracted administration of cabozantinib at both doses was well tolerated with minor body weight loss in treated animals. To gain insight into the mechanisms by which cabozantinib inhibits tumor growth, we explored anti-angiogenic and pro-apoptotic effects in adrenal tumors using immunohistochemistry and western blot analysis. Cabozantinib treated xenografts showed significantly reduced microvessel density compared to control tumors as determined by CD34 staining (P=0.0116 for 30 mg/kg and P=0.0025 for 60 mg/kg treated tumors; Kruskal-Wallis test; Fig. 4D). A dose-dependent increase in cleaved caspase-3-positive nuclei was noted for cabozantinib at the higher dose as compared to controls (P=0.0091 for 60 mg/kg; Fig. 4E). When investigating the effects on the VEGF and HGF signaling pathways by exploring known key downstream effectors (Fig. 4F), we observed the appearance of ERK1/2 phosphorylation in tumor samples to cabozantinib treatment as compared to controls. No modification was observed in regard to AKT activation levels. In contrast, SRC phosphorylation was found inhibited at both concentrations in comparison to control tumors.

Thus, cabozantinib acts by inhibition of angiogenesis at both dose levels whereas induction of cell death demands higher concentrations; inhibition of SRC appears associated with antitumor activity.

To be able to best explore the inhibition of metastatic spread of neuroblastoma, we used our systemic IMR-32-Luc model in female NSG mice which develop metastases preferably in lungs, liver and bones (33). Cabozantinib orally at 30 and 60 mg/kg/day was initiated at day 10 post tail vein injection, when bioluminescence signals were detectable in all animals and was given for 34 days when the animals were sacrificed. Bioluminescence signals in vivo in control animals were distributed over the liver region, thorax, and bilateral lower extremities consistent with the previous metastatic homing pattern in this model (33). Cabozantinib treated animals in treatment groups exhibited reduced bioluminescence signals (Fig. 5A). Luciferase activity of tumor cells ex vivo was significantly reduced in liver (n=6, P=0.0238; Kruskal-Wallis test) and lung (n=6, P=0.0141) in the 60 mg/kg treated animals as compared with the control group. This was not the case in bones (Fig. 5B). Thus, cabozantinib reduces metastatic spread to visceral organs in IMR-32 neuroblastoma.