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

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).

Female athymic Nude-Foxn1^nu mice were purchased from Envigo (Bresso, Italy) and housed under pathogen-free 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 1x10⁶ IMR-32 cell line into the left adrenal gland capsule, as previously described [28, 29].

To measure the inhibitory concentration of PON that causes 50% cell viability reduction (IC50), 5x10³ 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 (Countess™ 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.

Protein levels were analyzed as previously described [31]. Briefly, 20 μg of proteins was loaded for each sample on precast 4–20% gradient SDS-PAGE gels (BioRad), transferred to nitrocellulose membrane, and probed with the following primary antibodies: LC3 (Novus Biologicals; 1:1000), p62 (Cell Signaling; 1:500), BECLIN 1 (Novus Biologicals; 1:500), PARP (Cell Signaling; 1:1000), BCL2 (Cell Signaling; 1:1000), VINCULIN (SC Biotechnology; 1:2000), CASPASE 3 (Cell Signaling; 1:200), CASPASE 8 (Cell Signaling; 1:200), ERK total (Cell Signaling; 1:1000), and phospho-ERK (Cell Signaling; 1:1000). The phospho-kinase array was performed as described previously [30], and a human autophagy array (RayBiotech) was performed using 500 μg of total protein extracts, following the manufacturer’s instructions. Protein quantification and signal detection for each assay were performed as previously described [31]. Arrays were analyzed using the ImageJ protein array plugin [32]. The data is presented as fold change relative to the controls.

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, 5x10³ 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.

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 post-fertilization (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.

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.

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 anti-mouse detection kit (Vector Laboratories) was applied for 30 min. The sections were incubated with diaminobenzidine substrate (Vector Laboratories) to visualize immunoreactivity.

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.

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.

The in vivo data are expressed as mean ± standard deviation (SD). The analyses were performed with GraphPad 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.

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 (<IC₅₀) 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.

Several pathways have been reported as being affected by PON [13, 38]. To assess how the pathway profiling differs in the three neuroblastoma cell lines treated with sub-IC₅₀ PON, we performed a phospho-kinase array (Supplementary Fig. S2A, left panel). We identified 16 out of 39 (41%) kinase phosphorylation sites (downregulated: CREB^S133, EGFR^Y1086, ERK1/2^{T202/Y204, T185/Y187}, HSP27^S78/S82, LCK^Y394, LYN^Y397, MSK1/2^S376/S360, P38A^T180/Y182, SRC^Y419, WNK1^T60, YES^Y426, STAT3^S727, PDGFRβ^Y751, and β-CATENIN total protein; upregulated: CHK-2^T68, P53^S392) with changed levels in at least two different cell lines. The expression in these samples varied more than 25% with respect to the corresponding controls (data not shown). Seven phospho-kinases (44%) were downregulated in all three cell lines and formed a protein–protein interaction network (Supplementary Fig. S2B). Moreover, pathway enrichment analysis (www.​pathwaycommons.​org) determined the involvement of these kinases in the regulation of PI3K/AKT signaling. The remaining nine modifications (56%) had the same phosphorylation changes in two out of three cell lines. In particular, ERK1/2 was found to be downregulated in the SK-N-BE(2) and IMR-32 cells but upregulated in the SH-SY5Y cells (Supplementary Fig. S2A, right panel), implying a possible MYCN-dependent response to PON [39].

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.

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.

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 a-c), indicating a higher rate of autophagosome–lysosome fusion. As expected, in CQ-treated cultures, co-localization 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 GFP-quenching. The COMBO treatment blocked PON-dependent 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.

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.

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 shown in Fig. 5c, PON treatment significantly decreased tumor weight in comparison with the control animals, while the COMBO treatment further increased the anti-tumor 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.