The tubulin inhibitor MG-2477 induces autophagy-regulated cell death, ROS accumulation and activation of FOXO3 in neuroblastoma

Neuroblastoma is the most frequent extra-cranial solid tumor in children with still high mortality in stage M. Here we studied the tubulin-inhibitor MG-2477 as a possible therapeutic agent for neuroblastoma therapy and uncovered that MG-2477 induces death in neuroblastoma cells independent of PKB-activation status and stage. MG-2477 triggers within 30 minutes extensive autophagosome-formation that finally leads to cell death associated with mitotic catastrophe. Autophagy is critical for MG-2477-induced death and is regulated by the BH3-only protein PMAIP1/NOXA which sequesters the anti-apoptotic BCL2-protein BCLXL and thereby displaces and activates the autophagy-regulator BECN1/beclin1. Knockdown of NOXA or overexpression of its pro-survival binding partners MCL1 and BCLXL counteracts MG-2477-induced cell death. MG-2477 also rapidly induces the repression of the anti-apoptotic protein Survivin, which promotes autophagy and cell death. We further observed the accumulation of reactive oxygen species (ROS) that triggers autophagy induction suggesting a change of the PI3 kinase-III/BECN1 complex and activates the transcription factor FOXO3, which contributes to final cell death induction. The combined data suggest that MG-2477 induces a sequential process of ROS-accumulation, autophagy and FOXO3-activation that leads to cell death in neuroblastoma cells.


INTRODUCTION
Neuroblastoma is a pediatric malignancy that develops from undifferentiated precursors of the sympathetic nervous system. Depending on stage, patients with neuroblastoma face a poor prognosis and survival rate, especially children older than one year with metastatic stage M (old classification stage IV) tumors [1]. Due to the broad range of genetic alterations, like genomic amplification of NMYC, the gain of chromosome 17q or hyperactivation of pro-survival phosphatidylinositol-3kinase (PI3K)/protein kinase B (PKB) signaling, high stage neuroblastoma tumors poorly respond to radiation therapy and are frequently resistant to chemotherapeutic agents [2][3][4]. Especially the gain of chromosome 17q leading to enhanced BIRC5/Survivin expression [3] contributes to resistance against DNA-damaging agents, since Survivin efficiently prevents reactive oxygen species (ROS) signaling and the further activation of downstream death signaling cascades [5,6]. For many years, anti-mitotic drugs have been used for treating a variety of malignancies. These compounds are frequently of natural origin and possess high antiproliferative effects [7]. In principal, there are two major classes of compounds which interfere with microtubule dynamics: Inhibitors of the tubulin assembly into microtubule structures or the inhibitors of microtubule disassembly. In principal, they can bind to three different sites of tubulin, the paclitaxel site, the vinca-domain or into the colchicine site [8]. Both classes inhibit mitosis and induce therefore either mitotic cell death or cell death in G1-phase after slippage of tetraploid/aneuploid daughter cells into the cell cycle [9]. Since cancer cells are fast dividing cells, improvement of these substances is still of high scientific interest to overcome resistance. Moreover understanding their modes of action might also improve combination therapies and reduce effective doses. Antimitotic drugs possess high cytotoxic potential, since they target a large number of signaling pathways and proteins and many reports suggest the involvement of autophagy related genes as well as members of the BCL2 family [10,11].
Programmed cell death can be divided into four major forms: in type I -apoptosis, which is induced either by extrinsic or intrinsic signaling pathways that converge on caspase activation and DNA fragmentation [12], in type II -autophagic cell death, in mitotic catastrophe and in regulated necrosis [13].
In most cases cell death execution is not the result of single pathway activation, but the consequence of crosstalks between different cell death signaling pathways and pro-survival signals. The interplay of these pathways is complex since they can be activated simultaneously and frequently the involved key players participate in more than one signaling cascade. This especially includes the family of BCL2 proteins, which is critically involved in the regulation of all forms of cell death. The BCL2 protein family is composed of pro-apoptotic BH3only proteins, such as BCL2L11/BIM, NOXA and the autophagy regulator BECN1/beclin1 [14], pro-apoptotic BCL2 proteins of the BAX-BAK family that control outer mitochondrial membrane permeabilization and the pro-survival BCL2 partners BCL2, BCL2L1/BCLXL, BCL2L2/BCLW, MCL1 and BCL2A1. Different models on how these proteins interact with each other and steer death decision have been proposed [15]. Importantly, pro-survival proteins such as BCL2 and BCLXL, but also pro-apoptotic BIM interact with BECN1 and other autophagy regulators and thereby directly modulate autophagy induction [16,17]. Since many stimuli trigger apoptosis and autophagy, autophagy is often observed before apoptosis induction as a rescue mechanism of the cell to adapt to stress [18]. In this case, however, autophagy inhibits apoptosis execution [19]. On the other hand, apoptosis can also inhibit autophagy by cleavage of autophagy-related genes like ATG3, BECN1, or AMBRA1 [14,[20][21][22]. The existence of autophagic cell death, a form of cell death mediated by autophagy in mammals is still under debate, since it does not have a role in mammalian development, but it was shown to contribute to drug-induced cell death execution in some cancer cells, especially in those lacking BAX and BAK1 or CASP3/ Caspase 3 [23,24]. Autophagy may facilitate apoptosis by degradation of inhibitory proteins, like IAPs (Inhibitor of Apoptosis Proteins) [25] or even induce necrosis by degradation of enzymes involved in stress detoxification [26].
The tubulin-inhibitor 3-cyclopropylmethyl-7phenyl-3H-pyrrolo[3,2-f]quinolin-9(6H)-one also named MG-2477 was obtained as described in Gasparotto et. al. [27]. MG-2477, which shares the three-ring system with flavones and colchicines, binds with high affinity to the colchicine site of tubulin and is cytotoxic to different cancer cell lines like non-small cell lung carcinoma, ovarian or breast cancer [11,27]. Although there are some reports which demonstrate that anti-mitotic drugs are ineffective in cancers with increased/activated PKB levels [28][29][30], Viola and colleagues demonstrated that MG-2477 treatment de-phosphorylates and inactivates PKB [11]. Since this pathway is hyperactivated in human neuroblastoma [4], we investigated the potential cytotoxic effect of MG-2477 in different neuroblastoma cell lines [31]. The project was designed to study the initial phase of MG-2477-induced cell death and to investigate whether this drug may be a potential agent for the treatment of neuroblastoma. MG-2477 efficiently induced cell death in neuroblastoma cells independent of the PKB status, but did not affect differentiated neuronal cells. Upon addition of MG-2477 we observed rapid autophagy associated with accumulation of ROS, induction of NOXA and repression of Survivin. This initial phase was followed by activation of the transcription factor FOXO3 and cell death associated with signs of mitotic catastrophe and slippage into apoptosis.

MG-2477 triggers cell death in neuroblastoma cells independent of the PKB status
MG-2477 was originally developed as an antimitotic drug suggesting that it affects mitotic spindle formation in neuroblastoma. As expected from its ability to bind to the colchicine site of tubulin, MG-2477-treated cells failed to organize polymerized microtubule network and further chromosome orientation and segregation (Supplementary Figure 1A). To investigate the effect of MG-2477 on deregulated PKB survival signaling, we chose a set of human neuroblastoma cell lines which vary in the expression and activity of PKB (Supplementary Figure 1B). The cell lines NB1, NB8, NB15 and SH-EP were treated with 20 and 50 nM MG-2477 for 24 and 48 hours. All tested cell lines showed increasing amounts of cells in sub-G1 already after 24 hours and reached www.impactjournals.com/oncotarget a maximum of 30 to 50% death after 48 hours ( Figure  1A). Additionally, MG-2477 treatment caused Annexin V positivity within 18 to 24 hours ( Figure 1B) and was associated with CASP3 and PARP cleavage in all tested cell lines ( Figure 1D) suggesting that a significant part of cells underwent apoptotic cell death. To exclude cytotoxic effects of MG-2477 on differentiated neuronal cells, we differentiated NB8 and NB15 cells with all-trans retinoic acid (RA) and pheochromocytoma PC12 cells with nerve growth factor (NGF) and treated undifferentiated as well as differentiated cells with increasing amounts of MG-2477. PC12 cells are derived from rat adrenal medulla and represent an established model for neuronal differentiation [32,33]. Differentiated NB8, NB15, and PC12 cells remained resistant to MG-2477 up to a concentration of 100 nM, whereas undifferentiated cancer cells underwent significant cell death (Supplementary Figure 2). This suggests that MG-2477 does not impair the viability of differentiated neuronal cells, but specifically targets undifferentiated cancer cells.
In parallel to flow cytometric analyses we also analyzed the viability of MG-2477-treated SH-EP and NB15 cells by an AlamarBlue assay. Viability was significantly more reduced than what we expected from flow cytometric analyses ( Figure 1C). When analyzing the morphology of MG-2477-treated SH-EP cells by live cell fluorescence microscopy, we observed cockling and detachment of cells already after 3 hours, which continued up to 24 hours when hardly any attached, viable cells were left ( Figure 1E). Nuclei fragmentation, visualized by Hoechst33342 staining, was rarely seen, which was consistent with relative low numbers of fragmented nuclei observed by propidium iodide (PI) flow cytometry. From the combined data we conclude that MG-2477 rapidly induces morphological changes and cell detachment in neuroblastoma cells, which eventually leads to cell death.

MG-2477 induces formation of autophagosomes
As anti-mitotic drugs, including MG-2477 [10,11,34] have been reported to cause cell death associated with autophagy we investigated in a next step also the formation of autophagosomes during MG-2477 treatment and infected SH-EP, NB1, NB8, and NB15 cells with a retroviral vector coding for EYFP-LC3. Live cell fluorescence imaging of SH-EP/YFP-LC3 (Figure 2A), NB1/YFP-LC3, NB8/YFP-LC3, and NB15/ YFP-LC3 cells (Supplementary Figure 3) demonstrates a rapid and extensive formation of autophagosomes already 30 minutes after MG-2477 addition in SH-EP cells and after one hour in NB1, NB8, and NB15 cells. Autophagosome formation was accompanied by the conversion of MAP1ALC3/LC3-I to LC3-II as analyzed by immunoblot ( Figure 2B). To assess autophagic flux we overexpressed a DsRed-LC3-GFP fusion protein in SH-EP cells to visualize both, autophagosome formation and further fusion of autophagosomes with lysosomes in the presence or absence of BafilomycinA1 (BafA), an inhibitor of autophagosome-lysosome fusion [35]. As shown in Supplementary Figure 4, MG-2477 promotes autophagolysosome formation, which can be readily blocked by BafA. This implies that the large number of autophagosomes observed after 30 minutes of treatment results from increased de novo autophagosome formation and is not the consequence of autophagosome accumulation due to reduced fusion between autophagosomes and lysosomes. One essential trigger and key player of autophagosome formation is BECN1 which is normally bound to and thereby inactivated by members of the BCL2 protein family and by the inhibitor of apoptosis protein Survivin in healthy cells [16,[36][37][38]. We therefore analyzed the steady state expression of different pro-and anti-apoptotic proteins during MG-2477 treatment. Immunoblot analyses revealed that MG-2477 leads to a rapid decrease of Survivin, starting already after one hour. At the same time the pro-apoptotic BH3-only protein NOXA increases continuously, whereas BIM that sequesters BECN1 at dynein light chains [16] was repressed ( Figure 2C and Supplementary Figure 5). MCL1, BCLXL and BECN1 levels remained largely unaffected during MG-2477 treatment. Interestingly, NOXA was recently described as rate-limiting BH3-only protein in the regulation of mitotic cell death [39] and Survivin was found to be degraded during autophagy in neuroblastoma [38]. Together, these results suggest that MG-2477 induces an immediate early autophagic response associated with increased expression of the BH3-only protein NOXA, repression of BIM and antiapoptotic Survivin.

NOXA displaces BECN1 from BCLXL and contributes to MG-2477-induced cell death
In a next step we determined whether autophagy induction by MG-2477 is critically influenced by NOXA as NOXA may neutralize the autophagy-inhibiting capacity of pro-survival BCL2-proteins. The pro-survival BCL2 proteins BCLXL as well as MCL1 which are both bound by NOXA in neuroblastoma cells [40] inhibit autophagy by sequestration of BECN1 [41]. Therefore we precipitated endogenous BECN1 from MG-2477treated SH-EP cells and analyzed BECN1-associated candidate proteins in neuroblastoma cells. As shown in Figure 3A, in untreated cells BCLXL binds to BECN1 and this interaction is markedly reduced already within 30 minutes in the presence of MG-2477. In contrast, no interaction between BECN1 and MCL1 was detected in SH-EP cells. Vice versa immunoprecipitation of BCLXL confirmed that 30 minutes after MG-2477-addition BECN1 disappears from BCLXL protein complexes, whereas the amount of bound NOXA strongly increases. This supports the hypothesis that early during MG-2477treatment BECN1 is displaced from BCLXL by increased amounts of cellular NOXA, which triggers autophagy initiation in neuroblastoma cells ( Figure 3B). To determine whether this induction of autophagy is necessary for the further cytotoxic effects of MG-2477, we monitored cell morphology/detachment as well as Hoechst33342-stained nuclei by live cell microscopy in the presence or absence of the autophagy inhibitor 3-Methyladenine (3MA) which inhibits class III PI3-kinases and thereby blocks the first steps of the autophagic process. As shown in Figure 3C and Supplementary Figure 6 3MA effectively prevents the formation of autophagosomes and rescues SH-EP cells from detachment/dying after treatment with 50 nM MG-2477. Together, these results suggest the sequestration of anti-apoptotic BCL2 proteins by NOXA and the release of BECN1 as an inducer of autophagy and potential cell death trigger. Therefore, in a next step we knocked down BECN1 in SH-EP and NB15 cells [38] by stable expression of shRNAs directed against BECN1 ( Figure  4 and Supplementary Figure 7). As shown in Figure  4A, the almost complete repression of BECN1 strongly reduced cell death in SH-EP ( Figure 4B) and NB15 cells (Supplementary Figure 7A) as measured by flow cytometry and also reduced CASP3 as well as PARP cleavage up to 24 hours post addition of MG-2477 ( Figure 4C). Also ROS accumulation, as demonstrated in Figure 6A, was completely inhibited in SH-EP/shBECN1-49 cl15 and SH-EP/shBECN1-Pool cl12 cells (Supplementary Figure  7B). This implies that BECN1 is one critical player in the death-inducing signaling cascade triggered by MG-2477. To further analyze the relevance of the BCL2-protein/ BECN1 and Survivin/BECN1 complexes in the initial phase of MG-2477-induced cell death, we overexpressed Survivin (SH-EP/Surv), MCL1L (SH-EP/MCL1L) and BCLXL (SH-EP/BclxL) and analyzed the effect of these proteins on MG-2477-induced cell death. Since both, Survivin and BCLXL, were described as targets for autophagolysosomal degradation [38,42], we additionally pretreated the cells with BafA. As demonstrated in Figure  3D the ectopic expression of all three pro-survival proteins lowered MG-2477-induced apoptosis. Consistent with their degradation by autophagy, co-treatment with BafA further enhanced the inhibitory effects of Survivin and BCLXL, but not of MCL1. To directly assess whether NOXA is the critical trigger, its induction was prevented by shRNA expression: Knockdown of NOXA (Supplementary Figure 8A) markedly reduced MG-2477-induced cell death ( Figure 3E), CASP3 and PARP cleavage (Supplementary Figure 8B), whereas knockdown of BIM [43] had no effect on cell death induction. This implies that although BIM might sequester BECN1 at dynein light chains [16], its repression does not contribute to autophagy and subsequent death induction. Of note, the knockdown of NOXA lowered MG-2477-induced cell death to a similar extent as ectopically expressed MCL1, without being affected by the co-treatment with BafA ( Figure 3E). The combined data suggest that the induction of NOXA by MG-2477 ( Figure 2C) displaces BECN1 from anti-apoptotic BCLXL, which increases the amount of unbound BECN1 ( Figure 3A and 3B) and thereby triggers autophagy. Both, NOXA and BECN1 are essential players in this initial phase and their knockdown strongly reduces the efficacy of MG-2477. Transgenic expression of MCL1, however, indirectly slows down BECN1 activation by sequestering NOXA away from BCLXL, as NOXA binds MCL1 with higher affinity than BCLXL in neuroblastoma cells [44].

MG-2477 inhibits the PI3K-PKB axis and causes FOXO3 activation
Autophagy is regulated via class II and class III PI3-kinases [45]. One critical downstream target of the PI3K is the PKB and this survival pathway is frequently hyperactivated in neuroblastoma [4], which among other targets inactivates the transcription factor FOXO3 [46]. Therefore we analyzed the effect of MG-2477 on the PKB -FOXO3 signaling axis. Within one hour of MG-2477 treatment, the phosphorylation of PKB at serine473 (Ser473) was decreased to 71% and after four hours to 38% in SH-EP cells and almost identical regulations were also observed in NB1, NB8, and NB15 cells upon MG-2477 treatment ( Figure 5A). Simultaneously, also the phosphorylation of FOXO3 was significantly reduced at threonine32 (Thr32) to 49% in SH-EP cells and down to even 10% in NB8 cells, suggesting FOXO3 activation under these conditions. We also analyzed the phosphorylation of c-jun-N-terminal kinase (MAPK8/JNK) which was also shown to participate in autophagy induction after stress signaling [47]. JNK as well as phospho-JNK-threonine183/tyrosine185 (Thr183/ Tyr185) were slightly induced after MG-2477 treatment, but we did not observe a significant change in the phospho-JNK/ JNK ratio (Supplementary Figure 9), therefore we excluded JNK as critical regulator of MG-2477-induced death.
To study, whether reduced PKB activity and reduced FOXO3-phosphorylation during MG-2477 treatment also leads to FOXO3 nuclear accumulation we determined the subcellular localization of an ECFP-tagged FOXO3 allele by live cell fluorescence microscopy. In line with dephosphorylation of FOXO3 at Thr32 after one hour of MG-2477 treatment, we observed nuclear ECFP-FOXO3 accumulation after 60 minutes ( Figure 5B). Since FOXO3 also regulates the transcription of genes which are involved in autophagy-regulation, like LC3 and BNIP3 [48,49] and to clarify the sequence of events which activate FOXO3 after MG-2477 treatment, we analyzed ECFP-FOXO3 translocation also after pre-incubation with either 3MA or BafA. Inhibition of the formation of autophagolysosomes by BafA had only a minor, not significant effect on the nuclear accumulation of FOXO3 ( Figure 5B and Supplementary Figure 10). Upstream inhibition of autophagy at the level of PI3K class III/BECN1 by 3MA, however, significantly reduced the number of cells with nuclear FOXO3 after MG-2477 treatment ( Figure 5B and Supplementary Figure 10). As a control, we next generated YFP-LC3-expressing neuroblastoma cells with FOXO3 being knocked down by shRNA technology (SH-EP/shFOXO3-16-cl13-YFP-LC3 cells) to monitor autophagosome formation. Knockdown of FOXO3 did not affect the formation of autophagosomes during MG-2477 treatment ( Figure 5C), supporting the results from live cell fluorescence imaging that activation of FOXO3 occurred downstream of PI3K class III/BECN1. Together, our results suggest that FOXO3 activation does not trigger immediate early autophagy, but by its later activation may contribute to the perpetuation of autophagy via induction of autophagy key regulators and to cell death.

Early ROS accumulation in MG-2477-treated cells occurs downstream of PI3K class III activation and is independent of FOXO3
FOXO proteins are regulated by increased cellular reactive oxygen species (ROS) during stress signaling [50], but cause also ROS accumulation in neuroblastoma cells [43]. Thus we next assessed whether MG-2477 treatment is associated with increased cellular ROS and investigated how ROS relate to autophagy, FOXO3 activation and subsequent cell death induction. As shown in Figure 6A, MG-2477 treatment caused elevation of cellular ROS levels as early as 15 minutes post addition of the drug. This highly significant accumulation of ROS in the early phase of MG-2477 treatment was also visible in NB1, NB8, and NB15 cells (Supplementary Figure  11). The knockdown of endogenous FOXO3 in SH-EP/ shFOXO3-16-cl13 cells (Supplementary Figure 15A) did not influence ROS production ( Figure 6A), suggesting that hours. Cell lysates were subjected to immunoblot analyses for CASP3 and PARP cleavage. GAPDH served as loading control. Densitometry was done using Labworks software and expressed as cleaved/full length ratio. www.impactjournals.com/oncotarget neither autophagosome formation (as shown in Figure 5C) nor ROS accumulation in this early phase of MG-2477 treatment depend on FOXO3. However, pre-treatment of SH-EP/ECFP-FOXO3 cells with the ROS inhibitor N-acetyl-L-cysteine (NAC) for 30 minutes reduced nuclear accumulation of FOXO3 in response to MG-2477 ( Figure 6B) and NAC treatment also rescued MG-2477treated SH-EP cells from detaching and dying ( Figure   6C), comparable to 3MA ( Figure 3C). To investigate this effect in more detail we studied the impact of NAC on autophagosome formation and on the regulation of NOXA, BIM, Survivin, and BECN1 in MG-2477 treated/untreated cells. Rather surprisingly, NAC efficiently prevented the formation of autophagosomes (Supplementary Figure  12), inhibited CASP3 cleavage (Supplementary Figure  13B) and attenuated, but did not prevent the regulation  Figure 13A). This suggests that ROS are generated in the initial phase of cell death and contribute to the perpetuation of the whole process. As shown in Supplementary Figure 7B, the knockdown of BECN1 completely prevented ROS accumulation. The combined data from these experiments suggest that ROS accumulation is downstream of BECN1 and that FOXO3 is activated downstream of cellular ROS accumulation and does not affect autophagosome formation in this initial phase of MG-2477-induced death.
To further analyze how ROS accumulation is connected to autophagy initiation, the cells were preincubated for 30 minutes with either 3MA or BafA before MG-2477 was added for additional 30 minutes ( Figure   7A). Whereas BafA did not show a statistically significant protective effect, 3MA almost completely prevented ROS accumulation suggesting that ROS accumulation can be prevented by inhibition of the PI3K class III/BECN1 complex. This is in line with the pronounced effect of BECN1 knockdown demonstrated in Supplementary Figure 7B and defines the time point of ROS formation downstream of the PI3K-III and BECN1 activation, but upstream of autophagosome formation. Since BCLXL and MCL1 are both, directly or indirectly, involved in BECN1 inactivation, we also assessed the influence of the ectopic expression of these pro-survival BCL2 proteins on ROS formation. In line with competitive displacement of BECN1 from BCLXL by NOXA, the ectopic expression of BCLXL as well as MCL1 and the knockdown of NOXA expression blocked ROS formation ( Figure 7B and Supplementary Figure 14). Also Survivin overexpression, which reduced MG-2477-induced apoptosis in an autophagy-dependent manner, efficiently reduced ROS formation ( Figure 7B), either by shutting down mitochondrial respiration at complex I or by sequestering BECN1, as we have described before [5,38,51].

FOXO3-activation by MG-2477 enhances cell death signaling
Besides the reassortment of BECN1, our results suggest a critical role for NOXA, BCLXL, and Survivin in MG-2477-induced cell death. All three proteins are direct or indirect targets of the transcription factor FOXO3 in neuroblastoma cells [43,46,52,53] implicating that, if not in the initiation phase, the secondary induction and repression of these key players by FOXO3 might contribute to cell death. Since MG-2477 triggers nuclear accumulation of FOXO3 ( Figure 5B and 6B), we next tested the relevance of FOXO3 for MG-2477induced cell death. SH-EP/shFOXO3-16 cl13 cells were subjected to MG-2477 treatment and analyzed for cell death. Knockdown of FOXO3 in SH-EP cells reduces cell death from 21.4% to 12.2% after 24 hours and from 40.3% to 23.7% after 48 hours, suggesting that FOXO3 significantly participates in cell death induction ( Figure  8A). On the level of NOXA, BIM, and Survivin steady state expression it becomes evident that reduction of FOXO3 significantly attenuates gene regulation of NOXA and Survivin, whereas the downregulation of BIM or steady state expression of BECN1 was not affected by FOXO3 knockdown (Supplementary Figure 15B). The putative role of FOXO3 as an "apoptosis amplifier" during MG-2477 treatment is shown in immunoblots measuring CASP3 and PARP cleavage (Supplementary Figure  15C): at 16 hours post addition of MG-2477 hardly any difference is visible between SH-EP/shCtr and SH-EP/ shFOXO3-16 cl13 cells, whereas at 24 hours the cleavage of CASP3 and of PARP is significantly reduced in the FOXO3 knockdown cells.
In a next step, as vice versa experiment, we also tested whether the activation of ectopic FOXO3 affects MG-2477-induced cell death. For that purpose we used SH-EP/FOXO3 cells that express an ectopic, phosphorylation-independent FOXO3(A3)ERtm fusion protein [46] and treated these cells with increasing amounts of 4-hydroxy-tamoxifen (4OHT). FOXO3activation by 4OHT enhances MG-2477-induced cell death from 21.3% to 30.7-35.0% ( Figure 8B).
We recently identified the FOXO3 target gene Survivin as both, a critical regulator and a target of autophagosomal degradation [38,51]. Survivin was rapidly downregulated during MG-2477 treatment and this repression is attenuated in FOXO3-knockdown cells (Supplementary Figure 15B). However, when overexpressed, Survivin counteracts MG-2477-induced cell death raising the question, whether this protective effect can be overcome by FOXO3-mediated repression of Survivin and if increased cell death after FOXO3 activation depends on Survivin levels. Therefore we pretreated either SH-EP/FOXO3-Ctr or SH-EP/FOXO3-Surv cells for 24 hours with 20 nM 4OHT to activate FOXO3 and then added MG-2477 (35 nM) alone or in combination with BafA. As demonstrated in Figure 8C Survivinexpressing cells are protected from FOXO3-induced cell death as well as show reduced apoptosis after MG-2477 treatment. Combined treatment of MG-2477 and 4OHT, however, induces cell death in SH-EP/FOXO3-Surv cells to the same amount as in SH-EP/FOXO3-Ctr cells (26.4% and 26.8%). Co-treatment with BafA slightly lowers this effect to 20.9%, suggesting that inhibition of autophagymediated Survivin degradation partially counteracts MG-2477-induced cell death. Immunoblot analyses of SH-EP/ FOXO3-Surv cells reveal that the additional activation of the ectopic 4OHT-regulated FOXO3 allele causes a rapid decay of Survivin steady-state levels (Supplementary Figure 16), explaining the increase of cell death. The combined data support the model that ROS-driven FOXO3 activation enhances the cytotoxic effect of MG-2477 by modulating the regulation of NOXA, Survivin and possibly other death regulators, which triggers additional death pathways and eventually drives the tumor cell into cell death.

DISCUSSION
Anti-mitotic drugs have been used for a long time in cancer therapy but, as also true for almost all other chemotherapeutics, their therapeutic efficacy is challenged by cancer cells that develop drug resistance. Therefore much effort is put into the improvement and patient-oriented re-design of such "old" drugs. In the present study we demonstrate that MG-2477, a derivate of phenylpyrroloquinolinone, efficiently kills neuroblastoma cells without affecting differentiated neuronal cells. The cytotoxic effects of MG-2477 have also been demonstrated in liver, ovarian or non-small lung carcinoma cell lines [11,27]. However, anti-mitotic drugs are thought to be inactive in cancer cells with high levels of active PKB [30], a situation present in most malignant neuroblastoma cells [54,55]. Since MG-2477 was shown to reduce PKB phosphorylation [11], we hypothesized that it might be a suitable drug for neuroblastoma therapy. To analyze the effect of MG-2477 on the survival of neuroblastoma cells, we chose a set of cell lines that are derived from different tumor stages [31,40] and vary in PKB expression and PKB activity (phosphoPKB/PKB ratio from 1.2 to 8.6, Supplementary Figure 1B) and MYCN status [56,57]. MG-2477 treatment resulted in similar cell death induction and reduction of viability in all tested neuroblastoma cell lines, independent of the stage derived from, the PKB or the MYCN status (Figure 1 and Supplementary Figure  1B) [31,40,46,58]. As a control for the selectivity on cancer cells, we further tested MG-2477 on all-trans retinoic acid (RA) differentiated NB8 and NB15 cells and on NGF-differentiated PC12 cells. All three cell lines were highly resistant to higher concentrations of the drug than undifferentiated cells (Supplementary Figure  2), suggesting that this compound selectively acts on undifferentiated, malignant neuronal cells and shows low neurotoxicity on differentiated neurons.
The observed reduction of cell viability induced by MG-2477 ( Figure 1C) was accompanied by detachment of neuroblastoma cells after 15 hours without typical DNA-fragmentation ( Figure 1E), although at 24 and 48 hours DNA fragmentation, Annexin V positivity, CASP3 and PARP cleavage were detectable ( Figure 1A, 1B  and 1D). We therefore investigated the initial phase of MG-2477 treatment and uncovered that death induction was preceded by immediate-early autophagy, with autophagosome formation and LC3-conversion being already extensively detectable 15 to 30 minutes after MG-2477 treatment (Figure 2A and 2B, Supplementary Figures  3 and 4). In this immediate early phase we observed the induction of the BH3-only protein NOXA, repression of the BECN1-scavenging BH3-only protein BIM [16,59] and repression of the BECN1-interacting Survivin ( Figure 2C) [38,51]. This differs from results of Viola and colleagues who observed autophagy simultaneously to signs of apoptosis in MG-2477-treated lung carcinoma cells at later time points [11]. Since increasing amounts of NOXA co-immunoprecipitated with BCLXL and BECN1 in parallel disappeared from BCLXL during treatment, these re-assortments suggest that autophagy is induced via increased levels of non-sequestered BECN1 that is displaced from BCLXL by NOXA ( Figure 3A and 3B). Further flow cytometric analyses support this assumption of a critical role of NOXA in autophagy and subsequent cell death induction, since ectopic expression of the NOXA binding partners MCL1 and BCLXL, as well as NOXA and BECN1 knockdown reduced MG-2477-induced death ( Figures 3D and 3E, Figure 4, Supplementary Figures 7A and 8B). In our model NOXA therefore induces autophagy by binding to BCLXL, which displaces BECN1 from BCLXL. Autophagy triggered by this re-assortment further causes degradation of Survivin, a recently identified key-player of autophagy in neuroblastoma [38], which further releases BECN1 from Survivin/BECN1 [38] complexes and leads to a feed-forward enhancement of autophagic signaling. This explains why ectopic expression of BCLXL or Survivin efficiently counteract autophagy induction similar to ectopically expressed MCL1, which sequesters NOXA with higher affinity than BCLXL [44] and also prevented ROS accumulation in response to MG-2477 treatment (Supplementary Figure 14). 3MA prevented autophagy (Supplementary Figure 6) and rescued SH-EP cells from MG-2477-induced death ( Figure 3C), which raises the question how autophagy contributes to cell death induction. A crosstalk between autophagy and cell death is in line with a previous report on another mitotic inhibitor, where parallel induction of autophagy and apoptosis after treatment with paclitaxel was observed [34]. Neither autophagy nor apoptosis inhibitors alone were able to rescue cells from paclitaxel-induced death suggesting that one pathway fills in for the other in case of failure to eliminate cells. In another recent paper on the ALK inhibitor entrectinib, Aveic et al demonstrated that autophagy reduces the efficacy of ALK inhibition and that partial resistance to this ALK inhibitor can be overcome by autophagy-inhibition via chloroquine [60]. However, in neuroblastoma cells, autophagy by MG-2477 is induced rapidly after treatment (within 30 minutes, Figure 2), while cell death occurred delayed between 24 and 48 hours. As BECN1 knockdown and 3MA efficiently reduced cell death, in neuroblastoma cells a sequential process of autophagy followed by death induction can be proposed. This also fits to the observation that FOXO3 activation is downstream of initial autophagy signaling, since 3MA was able to prevent FOXO3 translocation ( Figure 5B) and ROS accumulation after MG-2477 treatment ( Figure 7A), whereas the late autophagy inhibitor BafA failed to do so ( Figure 5B and 7A). Knockdown of endogenous FOXO3 had no effect on autophagosome formation by MG-2477 ( Figure 5C) further supporting this notion, but attenuated NOXA and Survivin regulation (Supplementary Figure  15B) and reduced CASP3 and PARP cleavage at 24 hours post addition of the drug (Supplementary Figure 15C) as well as DNA fragmentation ( Figure 8A). This suggests that FOXO3 significantly contributes to the apoptotic part of MG-2477-induced cell death. One critical trigger for FOXO3 activation besides reduced PKB survival signaling is the accumulation of cellular ROS, which are frequent byproducts of autophagy. Some substances such as compound K, safingol or buralin were described to induce autophagy in a ROS-dependent manner [61][62][63]. Also in the case of MG-2477 the ROS inhibitor NAC reduced autophagosome formation (Supplementary Figure 12). Interestingly, when we analyzed ROS levels in BECN1-knockdown cells (Supplementary Figure 7B) or after pre-incubation with 3MA and BafA, knockdown of BECN1 and 3MA treatment efficiently prevented ROS accumulation ( Figure 7A). This suggests that ROS accumulation is triggered downstream of the PI3K-III/ BECN1 complex and might be further amplified as a consequence of autophagy-dependent degradation of Survivin [38] as well as by FOXO3 activation [43]. Therefore, ROS are rather a byproduct of autophagy initiation and not the initial cause, but likely further perpetuate this initial autophagy, a notion that is further supported by the attenuation of NOXA and Survivin regulation in NAC-treated cells (Supplementary Figure  13). In the case of MG-2477, the re-organization of the PI3K-III/BECN1 complex by induction of NOXA and decay of Survivin seems to be critical for the observed ROS accumulation that further shifts the signaling towards cell death induction, since it contributes to the activation of the transcription factor FOXO3 ( Figure 5A, 5B and Figure 6B), which further enhances cell death, whereas knockdown of FOXO3 reduces death ( Figure  8A, Supplementary Figure 15C). In this intermediate phase ( Figure 8D) especially the FOXO3 target gene Survivin may play a critical role in pushing autophagic signaling toward cell death, as it was rapidly degraded after MG-2477 treatment and, when overexpressed, inhibited MG-2477-induced cell death in an autophagydependent manner ( Figure 2C, Figure 8C, Supplementary  Figure 16). This is consistent with our recent discovery that Survivin is a target for autophagosomal removal in neuroblastoma [38,51] and our results are in line with data on another IAP, BIRC6/BRUCE, which is also degraded via autophagosomal removal [25]. Importantly, FOXO3-mediated repression of Survivin at least partially abrogated the death-protective effect of overexpressed Survivin ( Figure 8C, Supplementary Figure 16). Since Survivin expression is frequently elevated in human highstage neuroblastoma due to a gain of chromosome 17q [3], MG-2477 represents an interesting substance for the treatment of this disease, as it might allow to neutralize death-protection by elevated endogenous Survivin.
The combined data suggest autophagy induction and ROS accumulation as critical early events during MG-2477 treatment, whereas downstream of these initial events FOXO3 is activated, which is critical for cell death induction (summarized in Figure 8D). By uncovering the sequence of regulations which finally lead to the elimination of cancer cells we were able to identify BECN1, NOXA and Survivin as initial key players in MG-2477-induced cell death. Especially the repressive effect of MG-2477 on Survivin will be important in novel treatment strategies to overcome Survivin-mediated therapy-resistance in aggressive neuroblastoma.
Images were collected at an Axiovert200M microscope equipped with a 20x and a 63x objective (Zeiss, Vienna, Austria). Fluorescence intensity was quantified using Axiovision Software (Zeiss, Vienna, Austria) and relative ROS levels were expressed as % of untreated controls.

Statistics
Statistical significance of differences between controls and treated cells were calculated using unpaired t-test. All statistical analyses were performed using Graph Pad Prism 4.0 software.