Autophagy Inhibition Improves Sunitinib Efficacy in Pancreatic Neuroendocrine Tumors via a Lysosome-dependent Mechanism Free

Pancreatic neuroendocrine tumors (PanNET) show a neuroendocrine differentiation analogous to the cells of islets of Langerhans and represent 3% of pancreatic tumors. PanNETs have significant malignant potential and surgery is the only curative option. Despite increasing availability of therapy options, stable disease is achieved in only 30% of stage IV PanNET patients and partial remission is observed only in rare cases (1–4). Sunitinib is a multi-tyrosine kinase inhibitor targeting mainly VEGF and platelet-derived growth factor (PDGF) receptors leading to inhibition of angiogenesis and destabilization of existing vasculature (5). Sunitinib has a dual antitumor effect in PanNETs targeting the vasculature as well as targeting tyrosine kinases expressed by tumor cells (6, 7). Sunitinib is approved for treatment of advanced PanNETs and has been shown to prolong progression-free survival (4). However, its efficacy is limited by intrinsic and acquired resistance. In Rip1Tag2 mice, a transgenic mouse model for PanNETs, mechanisms of acquired resistance towards sunitinib have been described (8–10). Sunitinib is a lysosomotropic drug, which freely diffuses the lysosomal membrane. In the lysosome, it is protonated due to the low pH, and thus accumulates due to the inability to diffuse through the membrane (11). Ellegaard and colleagues have shown that sunitinib inhibits acid sphingomyelinase (ASM), which influences lysosome stability (12). Sunitinib further leads to lysosomal leakage and later to mitochondrial membrane permeabilization and ultimately to cell death (12). Interestingly, cancer cell lysosomes are less stable than normal lysosomes due to transformation-associated changes in lysosome-associated membrane protein 1 (LAMP1) and LAMP2 levels as well as in sphingolipid metabolism (13–15). Lysosomotropic drugs were shown to trigger lysosomal cell death even in apoptosis- and multidrug-resistant cancer cells (15). Therefore, targeting lysosomes is considered a promising yet rather unexplored therapeutic strategy. More recently, it has been proposed that leaky lysosomes do not necessarily lead to cell death. Instead, the release of lysosomal enzymes might affect physiological functions, for example influencing invasion by degradation of focal adhesions (16). In addition, lysosomes undergoing lysosomal membrane permeabilization (LMP) induce autophagy and are subsequently cleared by lysophagy (17–19).

Macroautophagy, hereafter named autophagy, is mainly a cell survival mechanism activated upon cellular stress to sequester and degrade long-lived proteins and damaged organelles and to recycle nutrients (20). Autophagy can promote tumorigenesis by increasing cell survival in unfavorable environment such as hypoxia or anticancer treatments (21). Many anticancer treatments have been shown to affect autophagy in vitro and in vivo. Previously, it was found that sunitinib modulates autophagy in other cellular systems in vitro (22–25). However, whether the modulation of autophagy by sunitinib is due to a block or an induction of autophagy is controversial. As autophagy can promote tumorigenesis and resistance toward anticancer treatments, it has been exploited as therapeutic target in vitro, in vivo and in early clinical trials. In clinical trials, the most commonly used treatments to inhibit autophagy are the antimalaria drugs chloroquine (CQ) and hydroxychloroquine (HCQ; ref. 26). Both are lysosomotropic compounds, which inhibit autophagy via blocking lysosomal function. The role of autophagy and response toward its inhibition depends on tumor stage. Indeed autophagy functions are initially often tumor suppressing and at later stages tumor promoting. In addition, the role of autophagy differs between tumor types (21). The role of autophagy in mediating sunitinib response has not been studied in PanNET yet.

In this study, we investigated whether sunitinib upregulates autophagy in PanNETs and whether this is important in mediating therapy response or resistance. We assessed whether combination with the autophagy inhibitor chloroquine enhanced the antitumor effect of sunitinib. Moreover, we questioned how sunitinib accumulation in lysosomes and LMP affect autophagy and treatment response.

Bafilomycin A1 (B-1080) was purchased from LC Laboratories, chloroquine (C6628) from Sigma, and sunitinib (S1042) from Selleckchem.

Antibodies against ATG5 (#2630), cleaved caspase-3 (#9661 and #9664), cleaved PARP (#9541) were purchased from Cell Signaling Technology, ATG7 (ab52472), and CD31 (ab28364) from Abcam, galectin 3 (#556904) from BD Biosciences, GAPDH (#MAB374, clone 6C5) from Millipore, LAMP2 (sc-18822) and SV40 (sc-20800) from Santa Cruz Biotechnology, Ki67 (#RM-9106) from Thermo Fisher Scientific, and LC3B (NB600-1384) from Novus Biologicals.

BON1 cell line was provided by E.J. Speel, Maastricht, Netherlands in 2011. QGP1 cell line was purchased from the Japanese Health Sciences Foundation, Osaka, Japan in 2011. Short tandem repeat (STR) analysis by PCR was performed for both cell lines (QGP1 in 2011 and 2016, BON1 in 2014 and 2016). QGP1 cells were authenticated. A BON1 profile does not exist yet but the profile of these cells did not match any known profile of cancer cell lines thus excluding contamination from other lines. In addition, expression of the specific neuroendocrine markers chromogranin A and synaptophysin were routinely tested by IHC. BON1 cells were cultured in DMEM/F12 medium (Sigma), QGP1 cells in RPMI1640 medium (Sigma). For both cell lines medium was supplemented with 10% FBS, 100 U/mL penicillin and 0.1 mg/mL streptomycin and cells were kept in a humidified incubator at 5% CO₂ and 37°C. After thawing, cells were cultured for approximately two months.

Short hairpin RNA (shRNA) against ATG5 (NM_004849.1-420s1c1, NM_004849.1-915s1c1, NM_004849.1-1170s1c1), ATG7 (NM_006395.1-268s1c1, NM_006395.1-491s1c1, NM_006395.1-2173s1c1), and LAMP2 (NM_002294.1-291 and NM_002294.1-561) as well as a nontargeting shRNA control (SHC002) were delivered with a lentivirus expressing vector pLKO.1 (all from Sigma, MISSION shRNA). Lentivirus production and transduction were performed as described previously (27). Cells were selected with 1.5 μg/mL puromycin for 3–4 days. Knockdown efficiency was validated by immunoblotting of respective proteins.

Fresh human PanNET tissue was obtained from patients who underwent surgery at the Inselspital Bern (Bern, Switzerland) and have signed an institutional informed consent. Ethical approval was given by the cantonal authorities (Kantonale Ethikkommission Bern, Ref.-Nr. KEK-BE 105/2015). Fresh tissue was digested with collagenase IV (Worthington) and trypsin for 1 hour at 37°C followed by red blood lysis with ACK buffer (Gibco). Single-cell suspension was plated in 96-well plates and viability was measured after 48-hour treatment with ATPlite assay kit (Perkin Elmer) according to manufacturer's instructions. Protein extracts were collected from 96-well plates using urea lysis buffer (8 mol/L urea, 0.5% Triton X-100) and immunoblotting was performed as described previously. Primary tumor was from a male stage II patient (age 62, patient I) with a well-differentiated grade G2 tumor showing no hormone secretion. Liver metastasis was from a male stage IV patient (age 65, patient II), with a well-differentiated grade G3 tumor showing no hormone secretion. Despite sunitinib is currently only approved for treatment of G1 and G2 tumors, patient II was included in this study due to the well-differentiated state of the tumor. The coefficient of drug interaction (CDI) was calculated as CDI = AB/(A × B). AB is the relative viability after the combination treatment of drugs A and B, while A and B are the relative viability after single treatment with drug A and B, respectively. CDI < 1 indicates a synergistic effect, CDI = 1 an additive effect and CDI > 1 an antagonistic effect (28).

Cells were cultured and treated in 96-well plates. To perform the MTT assay, cells were incubated with medium plus MTT (0.5 mg/mL) at 37°C for 30–45 minutes. Medium was removed and cells were lysed in DMSO. Sorensen solution (0.1 mol/L glycine, 0.1 mol/L NaCl, pH 10.5 in water) was added and absorbance measured at 570 nm. For the clonogenic assay, cells were trypsinized after 72 hours of treatment and replated at equal numbers in 6-well plates. Colonies were allowed to grow for two weeks in normal growth medium followed by fixation and staining of the colonies with 0.05% crystal violet in 30% ethanol. Colony numbers were measured with ColCount (Oxford Optronix).

Cells were seeded in 24-well plates, radiolabeled with 0.1 μCi ¹⁴C-valine per mL (L-[U-14-C] valine; code CFB.75, Amersham) for 48 hours and treated with sunitinib for 24 hours. Remaining ¹⁴C-valine was removed and cells were incubated for 1 hour in medium supplemented with 10 mmol/L l-valine. During this incubation time, short-lived proteins were degraded. After washing, cells were incubated with medium supplemented with 10 mmol/L l-valine and with or without BafA1 for 5.5 hours, in which long-lived proteins were degraded. Supernatant and trypsinized cells were collected and acid precipitation and centrifugation were performed to separate supernatant and pellet. Fractions were mixed with Microscint40 and radioactivity was measured by liquid scintillation counting. Percentage of proteolysis of long-lived proteins was calculated from supernatant and pellet fractions.

Cells were lysed in urea buffer (8 mol/L urea, 0.5% Triton X-100) supplemented with protease inhibitors (Roche). Forty to 60 micrograms of protein was loaded on a 12% SDS-PAGE gel (Bio-Rad) and then transferred to a PVDF membrane (Bio-Rad). Incubation with primary antibodies was performed overnight at 4°C, followed by washing steps and incubation with secondary antibodies (DyLight 650 conjugate goat anti-rabbit and DyLight 550 conjugate goat anti-mouse, ImmunoReagents and peroxidase-conjugated AffiniPure donkey anti-rabbit and donkey anti-mouse, Jackson ImmunoResearch) for 1 hour at room temperature. Chemiluminescent or fluorescent signal was detected using ChemiDoc MP System (Bio-Rad). Total protein expression for quantification of specific protein expression was measured by use of the stain-free gel technology and imaged with the Chemidoc MP System (29).

For immunofluorescence, cells were plated on poly-l-lysine coated coverslips and treated for 24–48 hours. Fixation and immunofluorescence for galectin 3 was performed as described previously (30). As secondary antibody, goat anti-mouse Alexa Fluor 568 conjugate (Invitrogen) was used. Analysis was performed at a microscope Zeiss Axioplan 2 at ×40 magnification. Samples were blinded and percentage of positive cells was calculated from 500–600 counted cells on randomly chosen fields in the DAPI channel.

Cells were incubated in coated glass-bottom plates and treated with sunitinib for 72 hours. Medium was replaced by medium containing 100 nmol/L Lysotracker Red DND-99 (L7528, Molecular Probes) and cells were incubated for 1 hour at 37°C. Medium was exchanged to normal medium and cells were imaged at a confocal microscope Olympus FluoView-1000 at ×63 magnification.

Rip1Tag2 (C57BL/6) mice were kindly provided by G. Christofori (Basel, Switzerland). All experimental protocols were reviewed and approved by the Cantonal Veterinary Office of Bern (Bern, Switzerland). Mice were fed with food enriched in glucose starting from 10 weeks of age. Rip1Tag2 mice were treated daily from week 10 of age with 40 mg/kg sunitinib orally, 50 mg/kg chloroquine intraperitoneally or the respective vehicle control for three weeks. Animals were monitored daily. At 13 weeks of age, animals were euthanized, tumor number (>1 mm) and size was measured, and tissues were fixed in formalin overnight and embedded in paraffin. Tumor volume was calculated using the formula V = 0.52 × (width²) × length to estimate the volume of a sphere (31). Tumor volume per mouse was calculated as the sum of single tumors.

For IHC and TUNEL (TdT-mediated dUTP nick end labeling) on mouse pancreas, tissues were cut in 4-μm thick serial sections followed by deparaffinization, rehydration, and antigen retrieval using an automated immunostainer (Bond RX, Leica Biosystems). Antigen retrieval was performed for LC3B with Tris for 30 minutes at 95°C, for Ki67, SV40, and CD31 with citrate for 30 minutes at 95°C. Antibodies were diluted as follows: LC3B 1:4000, Ki67 1:400, SV40 1:100, CD31 1:100. For TUNEL, slides were incubated with TdT enzyme (Promega) for 20 minutes. Slides were counterstained with hematoxylin and eosin (H&E). For liver metastasis assessment, four step sections of 75 μm were cut from liver tissue and stained with SV40 as described above. Number of metastasis was counted and summarized from all sections. Microvessel density (MVD) was evaluated by a staff pathologist (A. Blank) based on IHC for CD31 and classified as high or low. Classification of tumors was performed based on H&E staining and according to classification criteria defined by Lopez and Hanahan (32). Necrosis was evaluated by a staff pathologist (A. Blank) based on H&E staining, as percentage of area showing prenecrotic condensed nuclei (regional cell death) and necrosis (geographic necrosis) compared with total tumor volume. Representative pictures of analyses for classification and necrosis on H&E staining are shown in Supplementary Fig. S1A and S1B, respectively. Proliferation was measured as percentage of Ki67-positive cells compared with tumor volume in all tumors >1 mm and average per mouse is shown. Number of TUNEL-positive cells was counted in one high-power field chosen in the area of highest labeling. Autophagy was assessed by IHC for LC3B, classified according to the abundance and number of LC3B punctae, representative of autophagosomes (33). Mice were classified as either LC3B punctae low or LC3B punctae high. IHC for LC3B on 30 matched samples of human primary PanNET and liver metastasis (patient collective II as reported in ref. 34) was performed as described above. Tumors were classified on the basis of the abundance and number of LC3B punctae in absent, low, or high LC3B punctae.

Statistical analyses were performed with GraphPad Software. Unpaired or paired t test was used to compare groups. Contingency tables were analyzed using Fisher exact test. Sample size (n) refers to biological replicates unless otherwise stated. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

We assessed sunitinib effects on autophagy in PanNET cell lines by immunoblotting for microtubule associated protein 1 light chain 3B (MAP1LC3B/LC3B). The cytosolic form LC3B-I is lipidated to LC3B-II, which is incorporated into autophagosomes and represents autophagosome numbers. Sunitinib treatment alone increased LC3B-II levels in both BON1 and QGP1 cells. To measure the autophagic flux we cotreated the cells with the lysosomal autophagy inhibitors chloroquine (CQ) or bafilomycin A1 (BafA1; ref. 35), which block the fusion of autophagosomes with lysosomes via elevation of lysosomal pH. In both cell lines, LC3B-II levels further increased upon combination treatment of suntinib and autophagy inhibitors compared with single treatments alone, indicating increased autophagic flux. The effect was more prominent with CQ (Fig. 1A). Autophagy was also monitored by long-lived protein degradation assay (LLPDA; ref. 20). Sunitinib increased autophagy-specific proteolysis in both cell lines (Fig. 1B; Supplementary Fig. S2A). We assessed basal levels of autophagy in PanNET human tissue by IHC for LC3B. LC3B punctae, representative for autophagosomes, significantly increased in liver metastasis compared with matched primary tumors, potentially indicating high levels of autophagy in stage IV patients (Supplementary Fig. S2B). Importantly, also in primary cells freshly isolated from PanNET patients, sunitinib increased LC3B-II levels, which were further elevated with autophagy inhibition (Fig. 1C). In conclusion, sunitinib upregulates autophagy in PanNET cell lines and primary cells as measured by LC3B-II and proteolysis of long-lived proteins.

We hypothesized that sunitinib-induced autophagy may provide a prosurvival mechanism limiting sunitinib antitumoral effect and that combination with autophagy inhibitors might have a beneficial effect in reducing cell viability. To address this hypothesis, we combined increasing doses of sunitinib with fixed concentrations of CQ and BafA1, the latter two not affecting cell viability (Fig. 1D). Sunitinib decreased BON1 and QGP1 cell viability in a dose-dependent manner. The effect on viability was similar in both cell lines and seemed to be independent of target receptor expression levels (Supplementary Fig. S3A and S3B). CQ further reduced cell viability when combined with 10 μmol/L sunitinib (Fig. 1D; Supplementary Fig. S3A). Further experiments in the cell lines were performed with these concentrations of sunitinib and CQ. In primary cells from two PanNET patients, both sunitinib and CQ decreased viability in a dose-dependent manner. Also, in primary PanNET cells, their combination led to an increased effect (Fig. 1E). Calculation of the coefficient of drug interaction (CDI) as described previously (28) indicated an additive effect of sunitinib and CQ treatment in patient I (CDI = 1.01) and a trend towards a synergistic effect in patient II (CDI = 0.83). Cell-cycle analysis by FACS revealed that sunitinib induced a G₁ cell-cycle arrest in PanNET cell lines, while the cell cycle was not affected by CQ (Supplementary Fig. S3C). To assess the long-term effect and recovery from treatments, a colony formation assay was used. Sunitinib combined with autophagy inhibition led to a significant 50% reduction in colony numbers (Fig. 1F).

We next assessed whether autophagy inhibition increases the antitumoral effect of sunitinib in vivo in Rip1Tag2 mice, a transgenic mouse model of PanNET. We started treatment of mice at 10 weeks of age, when tumors have developed and can be detected macroscopically (Fig. 2A). Mice were treated for three weeks with either vehicle, CQ (50 mg/kg, daily i.p.), sunitinib (40 mg/kg, daily orally), or a combination thereof (Fig. 2A). A dose of 50 mg/kg CQ corresponds to the human equivalent dose of 250 mg, which is the starting dose of phase I trials of CQ and HCQ in cancer (36, 37). Fifty mg/kg CQ led to an increase of LC3 in a melanoma xenograft model and to intratumoral concentrations of CQ, which blocked autophagy efficiently in vitro (38). Tumor number was significantly reduced after combination treatment compared with single treatments (sunitinib P = 0.0045, CQ P = 0.0480). Tumor volume was reduced by 50% after combination treatment compared with single treatments (sunitinib P = 0.0527, CQ P = 0.0050; Fig. 2A). Importantly, tumor number and volume significantly increase during sunitinib treatment compared with the start at 10 weeks of age, but not during the combination treatment (Fig. 2A). Microvessel density was clearly reduced upon sunitinib and combination treatment and unchanged upon CQ treatment (Supplementary Fig. S4A). Proliferation did not change upon CQ treatment but decreased significantly with sunitinib (P = 0.0224) and the combination treatment (P = 0.0017) compared with vehicle treatment (Fig. 2B and C). In addition to reduced proliferation, necrosis was elevated in tumors treated with sunitinib (P = 0.0005) or sunitinib and CQ (P = 0.0003) compared with vehicle-treated tumors (Fig. 2D). The number of TUNEL-positive cells increased significantly upon sunitinib (P = 0.0222) and the combination (P = 0.0163) treatment (Fig. 2E; Supplementary Fig. S4B). Tumor classification revealed that invasiveness decreased upon sunitinib and combination treatment (Supplementary Fig. S4C), while incidence of metastasis and number of metastases per mouse did not change (Supplementary Fig. S4D). Punctate LC3B positivity by IHC was classified into low and high. The distribution of LC3B punctae low versus LC3B punctae high increased toward LC3B punctae high in single and combination treatments compared with vehicle (Fig. 2F), indicating an upregulation of autophagy by sunitinib also in vivo. In summary, combination treatment of suntinib and CQ leads to an increased antitumor effect compared with single treatments in a preclinical setting.

To prove a direct role of autophagy in limiting sunitinib antitumoral effect, we knocked down key autophagy proteins, autophagy-related protein 5 (ATG5), autophagy-related protein 7 (ATG7), and lysosome-associated membrane protein 2 (LAMP2) by lentiviral delivery of shRNA. ATG5 and ATG7 are involved in vesicle formation and lipidation of LC3, while LAMP2 is a glycoprotein localized on the lysosome membrane, which is involved in the fusion between autophagosome and lysosome as well as in chaperone-mediated autophagy. Upon knockdown, protein levels and starvation-induced autophagic flux were both reduced (Supplementary Fig. S5A–S5D). Proliferation and colony formation were slightly reduced in cells with shATG5, shATG7, or shLAMP2 compared with control-transduced cells (Figs. 3B and 4B). Cells with shATG5 and shATG7 had reduced viability upon sunitinib treatment compared with control-transduced cells (Fig. 3A). In addition, recovery of sunitinib-treated shATG5 and shATG7 cells was significantly reduced compared with control-transduced cells to a similar extent as observed with the combination of sunitinib and CQ (Fig. 3B and C). While neither sunitinib nor the autophagy inhibitors CQ and BafA1 alone induced apoptosis in BON1 cells, their combination led to cleavage of caspase 3 and PARP (Fig. 3D, Supplementary Fig. S6A and S6B). Sunitinib treatment did not induce apoptosis in BON1 ATG5 knockdown cells, whereas in QGP1 cells knocking down ATG5 already induced apoptosis, which was further increased by sunitinib treatment. In both cell lines, sunitinib treatment of ATG7 knockdown cells induced apoptosis, but at lower levels compared with pharmacologic inhibition of autophagy (Fig. 3D and E). In summary, knockdown of early autophagy genes ATG5 and ATG7 only partially phenocopied the effect observed with pharmacologic inhibitors.

Interestingly, knockdown of the late autophagy gene LAMP2 better recapitulated the effect of pharmacologic inhibition. Sunitinib treatment reduced viability of shLAMP2 cells compared with control-transduced cells (Fig. 4A) as well as cell recovery similarly to the combination treatment of sunitinib and CQ (Fig. 4B and C). Sunitinib treatment of shLAMP2 BON1 and QGP1 cells induced apoptosis to a level comparable or even stronger than combination with autophagy inhibitors (Fig. 4D and E). The strong effect observed on apoptosis upon sunitinib and LAMP2 knockdown compared with ATG5 and ATG7 knockdown prompted us to investigate closer the involvement of lysosomes in sunitinib-mediated autophagy induction.

LAMP2 has been shown to contribute to lysosomal stability (14). Knockdown of LAMP2 sensitized BON1 and QGP1 cell lines toward CQ (Fig. 5A–C) and the sensitivity toward CQ seemed to be dependent on LAMP2 levels. BON1 cells with lower levels of LAMP2 were more sensitive to CQ than QGP1 (Fig. 5A–C). On the basis of these results, we hypothesized that lysosomal stability is an important factor in the response towards sunitinib and chloroquine treatment. Sunitinib and chloroquine are both lysosomotropic drugs, which due to their structure accumulate in lysosomes (11). We confirmed the lysosomotropic property of sunitinib by demonstrating that it colocalized with Lysotracker Red in BON1 and QGP1 cells (Fig. 5D). Lysosomotropic drugs can induce lysosomal membrane permeabilization (LMP) leading to release of lysosomal content in the cytoplasm, which in turn can activate apoptosis (39–42), and autophagy (17–19). Immunofluorescence for galectin 1 or 3 is considered the gold standard for detecting LMP (19). Upon permeabilization of lysosomes, galectins translocates from the cytoplasm and bind to the glycocalyx in lysosomes, which can be detected by immunofluorescence as a dot-like staining (Fig. 5E). CQ alone did not induce LMP, while sunitinib alone significantly induced LMP after 48-hour treatment, quantified as percentage of galectin 3–positive cells. Combination treatment of BON1 cells with sunitinib and CQ significantly increased LMP after 24 and 48 hours (Fig. 5F). Accordingly, sunitinib-treated shLAMP2 cells showed a significant increase of LMP and to the same degree as observed in control-transduced cells treated with sunitinib and CQ (Fig. 5G). Together, our data indicates that reduction of lysosomal stability, as in shLAMP2 cells or upon combination of the lysosomotropic drugs sunitinib and CQ, promotes LMP-induced damage, which then triggers apoptosis.

Options for anticancer therapy are limited for patients with stage IV PanNETs. Since 2011, sunitinib has been approved for treatment of advanced PanNETs but it has a limited efficacy (4). Here we investigated whether autophagy is involved in resistance to suntinib and therefore could be a potential therapeutic target. We evaluated our hypothesis on the only two PanNET cell lines available, although these cell lines only partially represent PanNET biology (43, 44). However, no better in vitro PanNET model is currently available. Because of this limitation, we studied our hypothesis additionally in primary cells from two PanNET patients and in vivo in a well-studied transgenic mouse model of PanNET. In PanNET cell lines and a transgenic PanNET mouse model, we demonstrated that sunitinib combined with the autophagy inhibitor CQ has a significantly enhanced antitumor effect compared with single treatment alone. In addition, the beneficial effect of the combination treatment was confirmed in human primary cell culture, suggesting that these findings might be relevant for PanNET patients.

First, we showed that sunitinib increases autophagy in vitro in PanNET cell lines, in cells isolated from human PanNET tumors and in vivo in the Rip1Tag2 transgenic PanNET mouse model. Sunitinib alone led to an upregulation of LC3B-II levels corresponding to an increased number of autophagosomes in PanNET cell lines and primary cells. LC3B-II levels further elevated when autophagy was blocked with lysosomal inhibitors indicating increased autophagic flux. In vivo, the number of mice with high LC3B punctate staining, representative for autophagosomes, increased upon sunitinib. Previous studies on other tumor types in vivo and in patient samples also reported modulation of autophagy upon sunitinib treatment (45, 46). In tumors, autophagy might be influenced by additional factors, which may also result from sunitinib's antiangiogenic activity, such as hypoxia (47), and inflammation or acidosis (48, 49). These can be only partially mimicked in vitro. To our knowledge autophagy has not been assessed in human PanNET yet. We showed by immunohistochemistry for LC3B that autophagy is present in PanNETs with increased levels in liver metastasis compared to primary tumors, suggesting autophagy as a potential target in stage IV patients.

We observed reduced cell recovery and induction of apoptosis in cell lines treated with the combination of sunitinib and CQ. Similarly, Eng and colleagues have shown a synergistic effect of sunitinib and CQ in reducing cell proliferation in a lung adenocarcinoma cell line (50). Combination of sunitinib with CQ in the Rip1Tag2 model led to an increased antitumoral effect compared with single treatments alone, as measured by reduced tumor burden, reduced proliferation, and increased necrosis. Importantly, tumor burden increased during sunitinib treatment compared with the start of the treatment, but not during combination treatment, indicating that the combination treatment strongly impairs tumor development. All the sunitinib-treated tumors were less vascularized; however, there was no difference in CD31 staining upon combination with CQ. This suggests that the increased antitumoral effect by the combination therapy is not explained by an effect on angiogenesis, but rather due to a direct effect on the cancer cells. Indeed, sunitinib has been shown to reduce tumor cell proliferation in vitro (23, 24, 50).

Sunitinib treatment reduced histologic signs of tumor invasiveness but the incidence and number of liver metastasis remained unchanged. Reported data on sunitinib-induced invasiveness and metastasis in the Rip1Tag2 model are however inconsistent. Sunitinib has been shown to lead to increased invasiveness and incidence of metastasis (8, 9), while a recent publication reports no change in invasiveness and number of metastasis per mouse upon sunitinib treatment (51). One important difference among these studies including ours is the age of the mice and the duration of the treatment, which might affect invasiveness and metastasis formation. A possible genetic drift of the Rip1Tag2 mice after breeding in different laboratories for a long time might be another reason for the different results (51, 52). In summary, our findings show an antitumor effect of the combination treatment in vivo.

Sunitinib combined with ATG5 or ATG7 knockdown resulted in little or almost no apoptosis activation, while sunitinib combined with lysosome-targeting autophagy inhibitors such as CQ, BafA1, or LAMP2 knockdown, induced apoptosis. Our data suggest that lysosomes are centrally involved in the combination effect of sunitinib and CQ. This is in agreement with other studies showing that sunitinib as well as CQ and LAMP2 knockdown destabilize lysosomes (12, 14, 40). Destabilization of lysosomes can lead to lysosomal membrane permeabilization (LMP), which in turn can trigger cell death or autophagy (17–19). In BON1 cells, sunitinib alone induced LMP, which was further increased by addition of CQ or upon knockdown of LAMP2. In QGP1, only the combination of sunitinib and a high concentration of CQ induced LMP, while single treatments did not (Supplementary Fig. S7A). This different susceptibility toward LMP seems to correlate with autophagic flux. Indeed, when autophagic flux upon sunitinib treatment is assessed by calculating the difference with and without CQ or BafA1, the two cell lines showed different results. In both cell lines, autophagic flux upon sunitinib increased if LC3B-II levels were measured upon CQ treatment, while the flux was reduced when measured upon BafA1 treatment (Supplementary Fig. S7B). Interestingly, in QGP1 cells, a higher concentration of CQ than the one used in BON1 increased flux, while a lower concentration of CQ also reduced the flux. (Supplementary Fig. S7C). These data suggest that BON1 cells are more susceptible than QGP1 toward LMP and show sunitinib-induced autophagy, while in QGP1 cells, sunitinib rather blocked autophagy. Indeed, a recent study showed that sunitinib has a dual, concentration-dependent effect on autophagy in BON1 cells and that the threshold concentration differs between cell lines (53). This difference between BON1 and QGP1 cell lines might be due to the different levels of LAMP2 which influence lysosomal stability and lysosomal membrane permeabilitation (14). Indeed, our data showed that LAMP2 knockdown sensitized both cell lines toward sunitinib or CQ treatment. Similarly, LAMP1 or LAMP2 knockdown sensitized osteosarcoma cells toward siramesine, a lysosomotropic drug (14).

Our in vitro data suggest that sunitinib has a dual effect on autophagy by both blocking and inducing autophagy. We propose that the determining factor for the net effect of inhibition and induction is lysosome stability, with lysosome instability favoring cell death. Of note, in this context CQ and LAMP2 knockdown, which block autophagy, also contribute to autophagy induction via their destabilizing effect on lysosomes. The effect of BafA1 on LMP seems to be controversial in our results as well as in the literature and is probably context-dependent. In contrast to CQ, BafA1 does not accumulate in lysosomes but elevates the pH via inhibition of vacuolar ATPase (20): BafA1 treatment was shown to induce a release of cathepsin D from lysosomes followed by activation of caspase-3 (54). On the other hand, pretreatment with BafA1 can limit LMP induced by another drug (40, 42, 55). Interestingly, BafA1 could revert sunitinib resistance resulting from lysosomal sequestration in renal and colon cancer cell lines (11).

On the basis of our data, we propose that sunitinib accumulates in lysosomes and induces LMP. Autophagy is upregulated for clearance of damaged lysosomes, leading to recovery of the cells, therefore playing a prosurvival role. Combination of sunitinib with CQ or shLAMP2 leads to a higher destabilization of lysosomes and increased LMP, which is associated with apoptosis-dependent cell death (Fig. 6).

Lysosomes are an attractive therapeutic target because cancer cell lysosomes are less stable and more prone to permeabilization (13). Many drugs approved for other diseases target lysosomes. Because of this property, they might be exploited also as anticancer therapeutics (15). CQ, approved for treatment of malaria, is already tested in several clinical trials in combination with various anticancer treatments. In phase I clinical trials, hydroxychloroquine (HCQ) led to a significant antitumor effect in combination with other therapeutics (36, 37). A clinical phase I trial is ongoing for the combination of sunitinib and HCQ in advanced solid tumors that did not respond to chemotherapy (NCT00813423). Our study strongly supports the rationale of this combination for treatment of PanNET patients and their inclusion in such clinical trials.

No potential conflicts of interest were disclosed.

Conception and design: T. Wiedmer, M.P. Tschan, I. Marinoni, A. Perren

Development of methodology: T. Wiedmer, P. Krebs, A. Perren

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Wiedmer, S. Pantasis, A. Perren

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Wiedmer, A. Blank, R. Bill, A. Perren

Writing, review, and/or revision of the manuscript: T. Wiedmer, A. Blank, S. Pantasis, P. Krebs, M.P. Tschan, I. Marinoni, A. Perren

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Pantasis, P. Krebs

Study supervision: M.P. Tschan, I. Marinoni, A. Perren

Other (laboratory technician; assistance of treatment of animals, laboratory work): L. Normand

We would like to thank PD Dr. Deborah Stroka for critical comments on the manuscript, Dr. Mario Noti for help and advice with animal experiments, and the Translational Research Unit for cutting the slides and IHC. Tissues were provided by the Tissue Bank Bern.

This work was supported by the Swiss Cancer League under Grant KFS-3360-02-2014 (to A. Perren, M. Tschan, and I. Marinoni). Desirée and Niels Yde Foundation, grant no. 466-10 to I. Marinoni. I. Marinoni was supported by a Swiss National Science Foundation career grant (SNF-PMPDP3_164484).

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.