Autophagic flux inhibition enhances cytotoxicity of the receptor tyrosine kinase inhibitor ponatinib

Despite reported advances, acquired resistance to tyrosine kinase inhibitors still represents a serious problem in successful cancer treatment. Among this class of drugs, ponatinib (PON) has been shown to have notable long-term efficacy, although its cytotoxicity might be hampered by autophagy. In this study, we examined the likelihood of PON resistance evolution in neuroblastoma and assessed the extent to which autophagy might provide survival advantages to tumor cells. The effects of PON in inducing autophagy were determined both in vitro, using SK-N-BE(2), SH-SY5Y, and IMR-32 human neuroblastoma cell lines, and in vivo, using zebrafish and mouse models. Single and combined treatments with chloroquine (CQ)—a blocking agent of lysosomal metabolism and autophagic flux—and PON were conducted, and the effects on cell viability were determined using metabolic and immunohistochemical assays. The activation of the autophagic flux was analyzed through immunoblot and protein arrays, immunofluorescence, and transmission electron microscopy. Combination therapy with PON and CQ was tested in a clinically relevant neuroblastoma mouse model. Our results confirm that, in neuroblastoma cells and wild-type zebrafish embryos, PON induces the accumulation of autophagy vesicles—a sign of autophagy activation. Inhibition of autophagic flux by CQ restores the cytotoxic potential of PON, thus attributing to autophagy a cytoprotective nature. In mice, the use of CQ as adjuvant therapy significantly improves the anti-tumor effects obtained by PON, leading to ulterior reduction of tumor masses. Together, these findings support the importance of autophagy monitoring in the treatment protocols that foresee PON administration, as this may predict drug resistance acquisition. The findings also establish the potential for combined use of CQ and PON, paving the way for their consideration in upcoming treatment protocols against neuroblastoma.


Background
The third-generation tyrosine kinase inhibitor (TKi) ponatinib (PON) has produced improvements in the treatment of adult patients with chronic myeloid leukemia (CML) [1]. With respect to the first and second generations of BCR-ABL tyrosine kinase protein inhibitors, PON was more successful in eliminating both BCR-ABL wild-type and mutant (BCR-ABL T315I ) CML cells, thus reducing the possible evolution of resistance due to drug exclusion [2]. Since then, PON has been tested in adolescent patients with CML and pediatric patients with leukemia, with encouraging results [3][4][5]. Due to its multiple targets, further evaluations have demonstrated the efficacy of PON in affecting other important tyrosine kinases, including EGFR, FGFR, PDGFR, and VEGFR, which are aberrantly activated in different malignancies [6]. The mechanisms of action of PON include the regulation of several intracellular signaling pathways, such as STAT3, PI3K/AKT, and ERK, which are all involved in supporting tumor cell proliferation and survival [7]. However, some studies have found an increased rate of resistance to PON in either preclinical or clinical settings [8][9][10], and a similar finding may therefore be possible in neuroblastoma, in which the efficacy of PON emerged from both in vitro and in vivo pre-clinical assessments [11,12]. In a previous highthroughput screening (HTS) study, among 349 compounds tested, PON gave the best results in impeding the growth of neuroblastoma cells [13].
Neuroblastoma is the most common extracranial malignancy, which preferentially occurs in pre-school children. It shows a wide-ranging clinical, histological, and biological heterogeneity and manifests as a localized or metastatic disease [14]. Together, these characteristics determine neuroblastoma tumor staging and patient stratification. At diagnosis, patients with neuroblastoma can be classified into very low, low, intermediate, or high risk groups [15]. The therapy regime is determined by this stratification and is particularly aggressive for the high risk patients, who receive multimodal therapy, autologous stem cell transplantation, and immunomodulatory and pro-differentiation therapy [16]. Nevertheless, these patients rarely achieve complete long-term clinical remission and often face disease recurrence due to the acquired resistance to therapy [17].
In recent years, autophagy has emerged as an important cytoprotective mechanism that is triggered in response to many chemotherapeutic agents. Indeed, several studies have shown the relevance of autophagy in allowing the survival of cancer cells upon the use of anti-neoplastic drugs [18]. In particular, different types of tumor cells exposed to TKi engage autophagy in response to chemical insults in order to relieve cellular stress, and the same behavior has been described in neuroblastoma cells [19,20].
Autophagy is an evolutionarily conserved mechanism required for proper cell function [21], although it has been connected with tumorigenesis and drug resistance phenomena in a subset of tumor cells [22]. In malignant tissues, autophagy is often associated with the development of secondary (acquired) drug resistance [23]. Despite these roles, several studies have evidenced that autophagy may also function as a tumor promoter [24]. The molecular background defining the balance between cell survival and cell death is determined by the strict connection between autophagy and apoptosis that properly maintains tissue homeostasis [25]. The link between these two processes is complex, and it is not yet completely understood. One of the cross talks between autophagy and apoptosis is determined by the BCL2/Beclin 1 interaction. Under normal conditions, BCL2 prevents Beclin 1 from triggering autophagy, but, upon stress, Beclin 1 is released, allowing its initiation [26]. Different chemotherapy drugs act as stress stimuli when added to tumor cells; hence, unraveling the extent to which autophagy can pilot cancer cell death upon administration of newly proposed anti-neoplastic therapies may be essential for shaping the development of future treatment protocols.
In this study, we assessed the likelihood of autophagydependent cytoprotection in neuroblastoma cells during treatment with PON, a third-generation TKi. We evaluated, both in vitro and in vivo, the synergy between PON and the lysosomal catabolism inhibitor chloroquine (CQ), noting the remarkable effectiveness of this combination treatment in impeding neuroblastoma cell survival and tumor growth.

Cell lines and reagents
The neuroblastoma cell line SH-SY5Y was purchased from DSMZ (Braunschweig, Germany), while SK-N-BE(2) and IMR-32 cells were obtained from ATCC (Manassas, VA). The cells were maintained in RPMI medium supplemented with 10% fetal bovine serum (FBS, Sigma-Aldrich), 1% antibiotics, and glutamine (Gibco). The cells were then exposed to PON (Sigma-Aldrich), CQ (Sigma-Aldrich), or a combination of PON and CQ (COMBO) for the indicated times and at the specified doses. The genetic background of the cell lines is summarized in Supplementary Table S1. In vivo studies were done with the less toxic analog hydroxychloroquine (HCQ; Sigma-Aldrich) [27]. Torin 1 (Sigma-Aldrich) was dissolved in DMSO before use, and the cell cultures were regularly tested for the presence of mycoplasmas by PCR. Human cell line authentication was done at BMR Genomics S.r.l. (Padova, Italy).

Orthotopic neuroblastoma mouse model
Female athymic Nude-Foxn1 nu mice were purchased from Envigo (Bresso, Italy) and housed under pathogenfree conditions. All the experiments were approved by the ethical committee of the Italian Ministry of Health (n: 661/2016-PR) in compliance with the ARRIVE guidelines (Animal Research: Reporting of In Vivo Experiments). Five-week-old mice were anesthetized with a xylazine-ketamine mix (Xilor 2% plus Imalgene 1000, Merial SpA, Italy), subjected to laparotomy, and inoculated with 1x10 6 IMR-32 cell line into the left adrenal gland capsule, as previously described [28,29].

Metabolic activity assay
To measure the inhibitory concentration of PON that causes 50% cell viability reduction (IC50), 5x10 3 cells were seeded in a 96-well plate 24h prior treatment to ensure exponential growth. The cells were automatically counted with the Trypan blue exclusion assay (Count-ess™ cell counter, Invitrogen). Cell viability was assessed for 24h after drug exposure by means of their metabolic activity using 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT, Sigma-Aldrich) [30]. The results were compared to control samples treated with an equivalent amount of DMSO.

Immunofluorescent antibody staining and autophagy flux detection
Immunofluorescence analyses were used for autophagosome and autolysosome detection upon staining with LC3 primary antibody (Novus Biologicals; 1:100) and LAMP-2 (Flarebio Biotech LLC; 1:200), respectively, overnight at 4°C. For in vitro drug treatment, neuroblastoma cells were exposed to DMSO (control condition), PON (1 μM for SH-SY5Y and IMR-32, 2 μM for SK-N-BE(2)), CQ (25 μM), and COMBO for 24h. The cells were then fixed with 4% paraformaldehyde (PFA, Sigma-Aldrich) for 15 min and permeabilized with 0.25% Triton X-100 in 3% BSA solution for 10 min. Alexa Fluor 488 (Thermo Fisher; 1:1000) was used as a secondary antibody. Autophagy flux was analyzed using a Premo™ Autophagy Tandem Sensor RFP-GFP-LC3B Kit (Thermo Fisher) according to manufacturer recommendations. Upon autophagy induction, the autophagosomes become double positive (both, GFP and RFP resulting in merged, yellow, signal). Once the lysosome has fused, the pH drops and quenches the GFP, making autolysosomes labeled in red (RFP). The nuclei were stained with Hoechst (Thermo Fisher), and images were acquired with a Zeiss LSM 800 confocal microscope and quantified using Fiji software.

In vitro drug combination studies
The cytotoxic activity of PON, alone or in combination with CQ, was compared after 72h of treatment using the MTT colorimetric assay. To that end, 5x10 3 cells were seeded in a 96-well plate the day before the treatment to ensure cell adhesion and growth. For COMBO treatment, the cells were pre-treated for 6h with 25 μM CQ and then treated with the indicated doses of PON or DMSO (control condition). The percentage of cell viability was normalized to the values obtained for the control cells.

Drug toxicity and autophagy activation in zebrafish embryos
Wild-type (AB/TU) zebrafish embryos were staged and maintained as described previously [33] ; their use was approved by the Italian Ethical Committee OPBA (86/ 2016-PR). For the in vivo drug administration, 48h postfertilization (hpf) embryos were treated for 12h at 28.5°C with DMSO (control condition), PON (1 μM), CQ (25 μM), and the autophagy inducer Torin 1 (400 nM) as a positive control treatment [34]. The embryos were then fixed in 4% PFA and stained with LC3 antibody (Novus Biologicals; 1:100) as previously described [35]. DAPI (Sigma-Aldrich, 1:10.000) was used for nuclear staining, and images were acquired with a Carl Zeiss Axio microscope and analyzed using Fiji software.
Efficacy studies and systemic toxicity evaluation in a neuroblastoma mouse model IMR-32-bearing mice were randomized into n=8 per group and evaluated for increased life span and survival. In a second experiment, mice were randomized into n=5 per group for tumor growth inhibition and systemic toxicity evaluations. Treatments started 12 days after tumor cell implantation. HCQ (60 mg/kg [36];) and PON (30 mg/kg [13];)-as single agents or in combination (COMBO)-were administered i.p. and by gavage, respectively, every day for 16 days total. In the combination setting, HCQ was administered 20 min before PON. In each experiment, a group of control mice received vehicle only. All animals received the entire schedule of treatment without any sign of systemic toxicity. They were monitored two to three times weekly and euthanized humanely just before showing signs of illness/suffering, such as paraplegia, dehydration, severe weight loss (>15%), or abdominal dilatation.
In the systemic toxicity experiment, the mice were anesthetized with xylazine 24h after the last day of treatment, and blood was collected through the retro-orbital sinus from each mouse into either anticoagulant-free tubes (samples A, for clinical chemistry hepatic, cardiac, and renal evaluations) or K3EDTA coated tubes (samples B, for hematological evaluations). Samples A were centrifuged at 2500×g for 10 min at 4°C, and the levels of serum albumin (ALB), cholinesterase (CHE), glutamic-pyruvic transaminase (ALT), glutamic oxaloacetic transaminase (AST), creatine phosphokinase (CK), and creatinine (CREA) were quantified. Levels of red blood cells (RBC), hemoglobin (HGB), hematocrit (HCT), mean cell volume (MCV), mean cell hemoglobin (MCH), mean cell hemoglobin concentration (MCHC), platelets (PLT), and white blood cells (WBC) were quantified in samples B. All the reported evaluations were performed at the Mouse Clinic, IRCCS Ospedale San Raffaele (Milan). The mice were finally sacrificed, and the tumors were weighed, recovered, fixed in formalin, and embedded in paraffin for subsequent immunohistochemical analysis.

Histological and immunohistochemical analyses
The paraffin sections were sliced to obtain tumor sections of 5 μm thickness, which were stained with hematoxylin and eosin (H&E) using standard lab protocols. Tumor slices were subjected to immunohistochemical analysis with the following antibodies: LC3 (Novus Biological; 1:200), CD56 (SC Biotechnology; 1:100), active CASPASE 3 (R&D Systems, 1:100), and Ki67 (Dako; 1:200). The sections were incubated with the antibodies overnight at 4°C after blocking the endogenous peroxidase activity with 3% hydrogen peroxide in methanol for 20 min at room temperature and antigen retrieval in citric acid pH 6.0 for 20 min at 90°C. Vectastain ABC horseradish peroxidase anti-rabbit or antimouse detection kit (Vector Laboratories) was applied for 30 min. The sections were incubated with diaminobenzidine substrate (Vector Laboratories) to visualize immunoreactivity.

TUNEL assay
Terminal deoxynucleotidyl transferase dUTP-mediated nick-end labeling (TUNEL) analysis was performed on the mouse tissues using the Click-iT® TUNEL Alexa Fluor® Imaging Assay (Invitrogen), essentially as described by the manufacturer. Briefly, paraffin-embedded tissue sections were de-waxed and incubated with a buffer containing fluorescent nucleotides and the terminal deoxynucleotidyl transferase enzyme for one hour at 37°C. After being washed in PBS, slides were mounted using 80% glycerol. Hoechst (Thermo Fisher) was used to counterstain all the nuclei, and TUNEL-positive signals were determined by counting five randomly selected fields using a Carl Zeiss Axio microscope.

Transmission electron microscopy (TEM)
Samples were fixed with 2.5% glutaraldehyde (Sigma-Aldrich) in 0.1 M sodium cacodylate buffer pH 7.4 overnight at 4°C. The samples were post-fixed with 1% osmium tetroxide in 0.1 M sodium cacodylate buffer for 1h at 4°C. After three washes with water, the samples were dehydrated in graded ethanol series and embedded in an epoxy resin (Sigma-Aldrich). Ultrathin sections (60-70 nm) were obtained with an Ultrotome V (LKB) ultramicrotome, counterstained with uranyl acetate and lead citrate, and viewed with a Tecnai G2 (FEI) transmission electron microscope operating at 100 kV. Images were captured with a Veleta (Olympus Soft Imaging System) digital camera.

Statistics
All in vitro experiments were performed in triplicate, and the data are presented as mean value ± standard error (SEM). Statistical analyses and graphs were performed using GraphPad Prism 8 (GraphPad, La Jolla, CA). Statistical significance was assessed by one-way analysis of variance (ANOVA) followed by a post hoc Dunnett's test and a two-sided Student's t-test.
The in vivo data are expressed as mean ± standard deviation (SD). The analyses were performed with Graph-Pad Prism 5 software-one-way ANOVA with Tukey's multiple comparison test was used to evaluate differences within treatments, survival curves were drawn as Kaplan-Meier cumulative proportion surviving graphs, and corresponding p-values were calculated using the log-rank (Mantel-Cox) test. p<0.05 (95% confidence interval) was considered statistically significant, and significance is indicated as *p<0.05, **p<0.01, or ***p<0.001.

Ponatinib triggers autophagy in human neuroblastoma cells
In a previously established HTS assay among 349 screened small molecule inhibitors, PON was identified as the most promising candidate for abrogating neuroblastoma growth [13]. Here, we assessed a possible functional association between PON and the modulation of autophagy in human neuroblastoma. To this end, three neuroblastoma cell lines-SK-N-BE(2), SH-SY5Y, and IMR-32-were cultured in the presence of increasing concentrations of PON (0.325-10 μM) for 24h to assess the sub-toxic doses (<IC 50 ) of the drug ( Supplementary  Fig. S1A). In all cell lines, PON activity was observed in the μM range, even though the IMR-32 and SH-SY5Y cells were slightly more sensitive to PON than SK-N-BE(2) (Supplementary Fig. S1B). To determine the extent to which sub-toxic doses of PON affect autophagy and apoptosis, we examined the levels of the main regulators of each of these two processes. PON significantly decreased the levels of polyubiquitin-binding cargo protein p62 (SQSTM1) and provoked the accumulation of LC3-II in a dose-dependent manner, implying the activation of the autophagic flux (Fig. 1a). The level of total Beclin 1 protein showed a modest increase (Fig. 1a), and the crosstalk between autophagy and apoptotic cell death [25] was followed by means of BCL2 protein levels, PARP, and CASPASE 3 protein cleavage. These results confirmed a cell type-dependent modulation of BCL2 (anti-apoptotic protein) levels, with the most evident decrease in the SH-SY5Y cells (Fig. 1a). The cleavage of PARP protein and CASPASE 3 were evident for the IC50 doses of PON; more importantly, the activation of autophagy was already observed at the sub-lethal doses of drugs, while the pro-apoptotic signals were visible at doses approaching the IC50 of PON (Fig. 1a). Besides LC3 and p62, other autophagy-related proteins participate in autophagosome formation [37]. As shown in Fig. 1b, treatment with PON significantly modulated the expression of additional autophagy marker proteins, including ATG3, ATG4, and MSK1 (Fig. 1b). These results confirm that sub-toxic doses of PON induce autophagy in neuroblastoma cells and that autophagy anticipates apoptotic cell death triggered by the drug treatment.

Ponatinib primes the formation of autophagic vesicles in human neuroblastoma cells
To further confirm the activation of the autophagic flux in neuroblastoma cells treated with PON, we performed immunofluorescence staining. Using immunocytochemistry, we detected the accumulation of LC3-positive vacuoles in treated cells (Fig. 2a). More precisely, PON markedly increased cytoplasmic LC3 puncta formation in all three neuroblastoma cell lines with respect to the DMSO-treated control cells (Fig. 2a). TEM experiments confirmed the accumulation of autophagic vesicles (Fig.  2b); together, these results corroborate that PON triggers autophagy in neuroblastoma cells.

Autophagy inhibition sensitizes neuroblastoma cells to treatment with ponatinib
To assess whether autophagy triggered by PON could have a cytoprotective effect in neuroblastoma cells, we combined PON with CQ [40]. Increased p62 levels and accumulation of LC3-II in the samples treated with CQ alone suggested a defective autophagic flux (Fig. 3a). Instead, the parallel increase in LC3-II and decrease in p62 levels observed upon treatment with PON alone confirmed that the treatment itself enhanced the autophagic flux (Fig. 3a). Consistently, single treatments with PON or CQ caused a marked accumulation of green signal-a sign of autophagy vesicles accumulation. COMBO-treated cells displayed a reduced amount of LC3-positive green puncta with respect to the single treatments ( Supplementary Fig. S3); more importantly, COMBO treatment potentiated the PON-induced toxicity, as revealed by increased PARP and CASPASE 3 cleavage (Fig. 3a) and the appearance of fragmented nuclei ( Supplementary Fig. S3). These results confirm that the observed autophagy had a cytoprotective role in neuroblastoma cells and imply that its inhibition can be used to increase tumor cells' sensitivity to PON. However, both treatments produced only slight changes in the levels of total BCL2 and BECLIN1 proteins (Fig. 3a), indicating that they cannot be used confidently as markers for immediate autophagy evaluation in neuroblastoma cells when only a single temporal point is available for testing. The metabolic activity of neuroblastoma cells was significantly more attenuated upon COMBO treatment, especially at the lower PON concentrations (Fig. 3b), indicating that the introduction of CQ had a pro-synergistic nature, particularly in the SH-SY5Y and IMR-32 cells. However, the synergistic effect of CQ was less evident in the SK-N-BE(2) cells, which showed an important sensitivity to CQ treatment alone, as confirmed by intense cleavage of not only CASPASE 3, but also CASPASE 8. The latter result implies that death receptors induced apoptosis [41] in the SK-N-BE(2) cells only upon addition of CQ, but not PON. Taken together, these results show that neuroblastoma cells activate autophagy as a pro-survival cue in response to PON exposure and imply that the greatest efficacy in impeding cancer cell survival comes from lowered doses in a combination treatment.

Combination therapy impedes the formation of autolysosomes in human neuroblastoma cells
By performing co-immunostaining for LC3 (green) and LAMP-2 (red) we revealed high amounts of enlarged lysosomes in PON-treated neuroblastoma cells, whereas in DMSO-treated (control condition) cell cultures, lysosomes appeared as smaller punctuate structures (Fig. 4). Enlarged lysosomes were only detectable after PON administration, primarily in IMR-32 cells. Moreover, in the control condition cells, the co-localization of autophagosomes with lysosomes into autolysosomes (fusion events) showed that autophagy is already active at the basal levels in neuroblastoma cells in the absence of a drug, whereas it was significantly enhanced during treatment with PON. Indeed, PON-treated cells showed a significantly enhanced LC3/LAMP-2 co-localization (Fig. 4 ac), indicating a higher rate of autophagosome-lysosome fusion. As expected, in CQ-treated cultures, colocalization of LC3 and LAMP-2 was comparable to (Fig. 4 a, c) or less than (Fig. 4b) the controls, confirming that CQ acts as a late-stage autophagy inhibitor by impeding lysosomes (red puncta) from fusing to autophagosomes (green puncta). The results were further confirmed using a RFP-GFP-LC3B tandem fluorescent construct that allows an enhanced dissection of the maturation of autophagosomes into autolysosomes ( Supplementary Fig. S4A-C). Increase of both, green (GFP) and red (RFP) fluorescence intensity in PON treated cells versus their controls sustained that PON induced autophagy flux. After CQ treatment, the accumulation of yellow (merged GFP-RFP signal) foci was visible confirming the efficacy of CQ treatment in impeding creation of autolysosomes and hence GFPquenching. The COMBO treatment blocked PONdependent autophagy and autophagy flux as sustained by the reduced number of the GFP and RFP dots when compared to PON treated samples (Supplementary Fig.  S4A-C). Taken together, these approaches evidence that COMBO treatment could be sufficient to inhibit the protective role of autophagy while enhancing neuroblastoma cells' vulnerability to PON. Analysis of all neuroblastoma cell lines exposed to PON provided similar results (Fig. 4 a-c and Supplementary Fig. 4A-C), highlighting the autophagy activation as a common cytoprotective mechanism against this drug.

PON treatment activates autophagy in zebrafish embryos
To answer the question of whether PON may induce autophagy in vivo, we first treated wild-type zebrafish embryos with sub-lethal doses. We compared the number of LC3-positive puncta in embryos treated with PON alone or in combination with CQ ( Supplementary Fig.  S5). Torin 1, a small molecule that induces autophagy activation by inhibiting mTOR complex 1 (mTORC1), was used as a positive control. Treatment with PON provoked a significant increase in LC3-positive (green) puncta with respect to DMSO-treated control embryos. This data confirmed the induction of PON-dependent autophagy in vivo. In contrast, treating embryos with CQ alone (25 μM) did not significantly affect the number of LC3-positive events. The COMBO treatment decreased the number of cells that activated autophagy in the presence of PON alone, reaching, in this way, a level of LC3-positive puncta comparable to that found in the DMSO-treated control embryos.  Combination treatment decreases tumor growth and increases survival in mice Next, to investigate the effects of a COMBO treatment on neuroblastoma tumors in vivo, an orthotopic model was established by inoculating IMR-32 cells into the adrenal gland of athymic nude mice (Fig. 5a). In the first set of experiments, a COMBO treatment strategy was evaluated in terms of the percentage of increased life span (ILS%) and survival. A combination of HCQ with PON led to a 22.5% ILS (median survival: control mice, 40 days; COMBO-treated mice, 49 days) and a significant increase in survival versus both the control (p= 0.007) and PON administered alone (p=0.04) (Fig. 5b). Moreover, in IMR-32 xenografts, HCQ alone did not significantly increase the mice's life span compared to the controls, while a significant improvement of the anti-tumor effect was obtained in mice treated with COMBO therapy (Fig. 5b). From a mechanistic point of view, in the second set of experiments, the tumors were analyzed 24h after the last drug administration. As  Fig. 5c, PON treatment significantly decreased tumor weight in comparison with the control animals, while the COMBO treatment further increased the antitumor effects. Notably, treatment with HCQ alone showed a slight, but not significant, increase in tumor weight (Fig. 5c). On the other hand, the reduction in tumor weight in mice treated with PON alone and in the COMBO set was significant in comparison to the control tumor masses (Fig. 5c).
To define the immunohistochemical profile of local tumor cells, we performed histological analyses on paraffin tumor sections. Neuroblastoma tumor masses at the site of inoculation strongly expressed human CD56, confirming their human origin ( Fig. 6a and Supplementary  Fig. S6). This immunoreactivity was significantly decreased in PON-and COMBO-treated mice (Fig. 6a). To test whether the survival observed upon COMBO treatment was a consequence of impaired tumor growth or increased tumor cell death, we analyzed the proliferation and apoptotic rates in the collected tumor masses. Notably, the slight, but not significant, increase in tumor weight found in HCQ-treated mice (Fig. 5c) was sustained by more intense Ki67 immunoreactivity ( Fig. 6b and Supplementary Fig. S6). On the other hand, the reduction in tumor weight in PON-and COMBO-treated mice was a cumulative incidence of the reduction of Ki67-positive (proliferating) cells and the enhancement of active CASPASE 3 and TUNEL-positive (dead) cells (Fig. 6 c-d and Supplementary Fig. S6). To study the role of autophagy in this scenario, the intracellular localization of LC3 was determined by immunohistochemical staining using an anti-LC3 antibody ( Fig. 6e and Supplementary Fig. S6). The control animals exhibited a low level of autophagy activation, whereas the PON-treated mice showed an increase in LC3 positivity, confirming the results obtained in vitro. Concomitantly, HCQ inhibited autophagy by inducing the accumulation of LC3 in treated mice, whereas COMBO treatment caused a more remarkable reduction in LC3-positive signals (Fig. 6e).
Finally, the proposed treatments were evaluated for eventual toxicity induction. Importantly, no weight loss was evidenced in any of the treated groups (data not shown), and no chronic clinical chemistry and hematological toxicities emerged from the quantification of the level of the corresponding markers in plasma and blood, respectively (Supplementary Fig. S7A-B). Collectively, our data suggests that the inhibition of autophagy by HCQ may further enhance the anti-cancer effects of PON in vivo without showing side effects.

Discussion
Several recent studies have proposed PON-a pantyrosine kinase inhibitor-for the treatment of neuroblastoma [11,42]. In different pre-clinical models of this pediatric cancer, PON was more effective in inhibiting tumor cell growth than several other anti-tumor compounds [12,13]. However, PON does not affect neuroblastoma cells through the chimeric protein BCR-ABL, since it is not found in this pediatric malignancy, but rather through other TKs, such as FGFR1 or EGFR [12,38]. The appearance of drug resistance continues to be one of the most critical impediments to cancer therapies that foresee the adoption of TKi.
In the present study, we investigated whether the use of PON, similar to other TKi [9], may correlate with the activation of cytoprotective autophagy, leading eventually toward acquired drug resistance. We therefore assessed, in vitro and in vivo, the levels of autophagy in neuroblastoma cells treated with sub-toxic doses of PON and subsequently examined the effects of combined treatment with PON and the autophagy inhibitor CQ on neuroblastoma cell survival and tumor growth.
Our data revealed PON-dependent activation of autophagy, both in vitro and in vivo, thus highlighting this cell process as an active mechanism that is plausibly involved in the development of drug resistance in neuroblastoma. In the analyzed neuroblastoma cell lines, PON caused a rapid increase in the autophagy levels found in control cells. This scenario may be explained as an attempt by tumor cells to rapidly deal with toxic insults inflicted by PON. To monitor autophagic flux, the levels of LC3-II and p62 proteins were analyzed [43]. In neuroblastoma cells treated with PON, the observed changes in both LC3-II and p62 protein levels were indicative of a dose-dependent activation of autophagy, and PON treatment triggered a formation of LC3/LAMP-2-positive autophagic vesicles in the cell cytosol. Colocalization of LC3/LAMP-2 occurs during the final phases of the autophagy process, during which the selected constituents are captured and degraded by lysosomal enzymes within autolysosomes. Detection of LAMP-2 protein levels and positioning allowed us to monitor the autophagosome-lysosome fusion process in PON-treated neuroblastoma cells. This result, along with the data obtained by electron microscopy, immunoblotting and the pH-dependent LC3 color-code changes, gave a definite confirmation that PON caused autophagic flux activation in neuroblastoma cells.
It is not rare for tumor cells to trigger autophagy in order to assure their survival after chemotherapy and radiotherapy treatments [44]. Autophagy-dependent mechanisms of acquired resistance to TKi that involve AXL signaling have recently been reported [45]; these mechanisms are activated after EGFR inhibition, upon which AXL takes over the transduction of the interrupted signaling. The cytoprotective role of AXL has been recognized in different tumor types [46], including non-MYCN-amplified neuroblastoma [47]. The AXL targeting in neuroblastoma in vitro models appears to be an effective cytotoxic approach in neuroblastoma, without causing significant variations in autophagy levels [30], but further investigations are required to delineate a possible link between changes in the AXL expression and the observed cytoprotective autophagy in PONtreated neuroblastoma cells. Based on current knowledge, pharmacological inhibitors of autophagy could be an effective adjuvant therapy for enhancing the cytotoxic effects of current chemotherapy protocols. Several autophagy inhibitors have so far been proposed [48]; of these, chloroquine (CQ)-and its analog, hydroxychloroquine (HCQ)-inhibit lysosome fusion to the autophagosome and impair further maturation into degradative autolysosomes. Indeed, they are considered to be late-phase autophagy inhibitors [43,49]. CQ and HCQ are also effective anti-malaria drugs, having anti-inflammatory cues as well [50], and several clinical trials have reported the possible effectiveness of CQ and HCQ in cancer-related therapies [51]. Neuroblastoma cells were successfully tested for the combined use of PON and either CQ or HCQ; CQ hindered autophagosome and lysosome fusion in PON-treated neuroblastoma cells, thus prompting apoptotic death induction. Moreover, another important outcome of the adopted combination treatment was that lower concentrations of PON were sufficient to induce apoptosis in neuroblastoma cells.
Autophagy is a biochemical process that remains active at a basal level of physiological conditions [37]. We exploited the advantages of the zebrafish in vivo model to determine whether PON causes changes in the level of autophagic vesicles with respect to control wild-type embryos. The results confirmed the significant upregulation of LC3-positive puncta in PON-treated embryos with respect to the control group, implying that PON triggers pro-autophagic events. In the neuroblastoma orthotopic mouse model, the reduction in tumor size was potentiated in mice treated with a combination (PON and HCQ) with respect to single (PON) therapy. The combination strategy was associated with the increased expression of active CASPASE 3 and TUNELpositive cells with respect to a single treatment, without causing any systemic toxicity. These data further corroborate the results obtained in vitro, in which the synergistic effects of the proposed combination treatment were determined.

Conclusions
In summary, we demonstrated that PON activates cytoprotective autophagy in neuroblastoma cells if used as a single treatment. The inhibition of autophagy by compounds already available, such as CQ and HCQ, could help in lowering the doses of PON required to affect tumor cell growth while improving its anti-neoplastic effects. These findings are of great relevance, since significant side effects have been reported in leukemia patients upon treatment with PON [45], and we may therefore need to decrease the concentration of such TKi in therapeutic protocols adopted for childhood malignancies, while maintaining their efficacy. One option could be an adjuvant therapy with CQ. Collectively, our results highlight the potential of the combined action of PON and CQ, giving a preliminary insight into the efficacy of this treatment strategy as a novel targeted therapy approach in neuroblastoma.

Supplementary information
Supplementary information accompanies this paper at https://doi.org/10. 1186/s13046-020-01692-x.  Fig. 6 Histopathological examination of post-therapy neuroblastoma tumors confirms the efficacy of combination treatment in vivo. a Hematoxylin & Eosin (H&E) of tumor tissues and immunohistochemistry staining of CD56 proteins were carried out. Red arrows indicate positive cellular labeling. Scale bar, 150 μm. b Immunohistochemistry staining of Ki67, c active CASPASE 3, and d TUNEL assay for the detection of proliferating and apoptotic cells, respectively. The percentage (%) of positive events is reported in the histograms on the right. Scale bar, 150 μm. e Immunohistochemistry staining of LC3-positive cells in resected post-therapy tumors. Red arrows indicate positive cellular labeling. Scale bar, 150 μm. The percentage (%) of LC3-positive cells is reported in the histogram on the right. *p < 0.05, **p < 0.01, ***p < 0.001