FOXO3-induced reactive oxygen species are regulated by BCL2L11 (Bim) and SESN3

The FoxO subfamily of forkhead transcription factors consists of FOXO1 (FKHR), FOXO3 (FKHRL1), FOXO4 (AFX), and FOXO6 (Calnan and Brunet, 2008; Ho et al., 2008). The regulation of these transcription factors is under the control of the phosphatidylinositol-3-kinase (PI3K) and protein kinase B (PKB) signaling pathway (Arden and Biggs, 2002) and depends on the availability of growth factors. Phosphorylation of FOXOs by PKB on distinct threonine and serine residues causes their export into the cytoplasm. Growth factor withdrawal shuts down PI3K–PKB signaling and consequently leads to nuclear accumulation and activation of FOXO proteins. In addition to survival signaling, FOXO proteins are also subject to a complex regulation by stress signaling that overrides inactivation through PKB. Jun-N-terminal kinase (JNK) (Essers et al., 2004) and mammalian-Ste20-like kinase-1 (MST1) (Lehtinen et al., 2006) activate FOXO proteins in response to oxidative stress even in the presence of growth factor signaling. FOXO3 targets are involved in the regulation of apoptosis, the cell cycle and in the defense against oxidative stress (Calnan and Brunet, 2008; Ho et al., 2008).

Apoptosis can be induced either by the extrinsic pathway, where death ligands bind on cell surface death receptors, or by the intrinsic pathway, which is controlled by the balance of pro- and anti-apoptotic BCL2 proteins at the mitochondria (Coultas and Strasser, 2003; Danial and Korsmeyer, 2004; Green and Kroemer, 2004). The family of BCL2 proteins is divided into three subgroups depending on their pro- or anti-apoptotic function and are classified by their number of BH domains. The pro-survival BCL2, BCL2L1 (BclxL), MCL1, BCL2L2 (Bcl-w) and A1 share four BH domains, whereas the ‘multi-domain’ proteins BAX and BAK1 (Bak) contain only three BH domains and are pro-apoptotic. The third group are the pro-apoptotic ‘BH3-only’ proteins e.g. PMAIP1 (Noxa), BCL2L11 (Bim), BID and BBC3 (Puma), which contain only one BH domain. The ratio of pro- and anti-apoptotic proteins determines cell fate at the level of mitochondria. Specific ‘BH3-only’ proteins, such as BID or Bim bind all pro-survival BCL2 proteins and thereby displace BAX and Bak from anti-apoptotic BCL2 family members (Danial and Korsmeyer, 2004). This results in activation and oligomerization of BAX and Bak, which further leads to the permeabilization of the outer mitochondrial membrane (MOMP), the release of cytochrome c and subsequent activation of the caspase cascade.

One apoptotic stimulus of the intrinsic death pathway is the accumulation of reactive oxygen species (ROS) at the mitochondria. Hydrogen peroxide (H₂O₂), superoxide (O ^–₂) or hydroxyl radicals (OH) are short-lived molecules (Scherz-Shouval and Elazar, 2007) that are generated as by-products of mitochondrial respiration (Kim et al., 2007). They might function as specific messengers by modifying target molecules or by changing the intracellular redox state (Thannickal and Fanburg, 2000). High levels of ROS can also damage proteins, nucleic acids and intracellular membranes, leading to oxidative stress and impairing cellular functions (Hasegawa et al., 2008; Menon et al., 2007). Under normal conditions, low amounts of ROS levels are eliminated by glutathione metabolism, scavenging enzymes such as SOD2 (MnSOD) and CAT (catalase), peroxisomes or by members of the sestrin family (Kim et al., 2007; Kopnin et al., 2007; Scherz-Shouval and Elazar, 2007). The mitochondrial enzyme MnSOD converts O ^–₂ to H₂O₂, which is further dismutated to H₂O and O₂ by catalase. MnSOD and catalase are transcriptional targets of FOXO3 and NF-κB (Chiribau et al., 2008; Liu et al., 2005; Storz et al., 2005). SESN3 (Sestrin3), a member of the sestrin family, is also regulated by FOXO3 (Nogueira et al., 2008). Insufficient antioxidant defense results in high levels of ROS, which severely damage the cell and cause changes in cellular ATP and Ca²⁺ levels. This eventually leads to the release of cytochrome c and the induction of apoptosis (Satoh et al., 2004; Kuznetsov et al., 2008a).

High stage neuroblastoma tumor cells frequently show increased phosphorylation of PKB on Ser473 and Thr308, which leads to hyperactivation of PKB signaling, inactivation of FOXO3, the most prevalent FOXO transcription factor in neuronal cells, and correlates with a poor prognosis (Opel et al., 2007). We have previously shown that activation of FOXO3 induces apoptosis in neuronal cells by induction of the pro-apoptotic proteins Bim and Noxa and sensitizes to chemotherapy by repression of the anti-apoptotic protein BIRC5 (survivin) (Obexer et al., 2007; Obexer et al., 2009). However, although FOXO3-mediated induction of Bim and Noxa and repression of survivin occur within 8 hours, apoptosis execution is delayed and starts between 24 and 48 hours. Therefore the purpose of the study was to dissect the molecular mechanisms and the precise sequence of events that lead to activation of FOXO3 by chemotherapeutic agents, and to apoptosis by FOXO3 in neuronal tumor cells. In this study, we focused on the early events at the mitochondria after FOXO3 activation and uncovered the downstream targets of FOXO3 that contribute to ROS formation and neuronal cell death.

Etoposide treatment is associated with the production of ROS, the induction of the prosurvival antioxidant enzyme superoxide dismutase and the activation of the proapoptotic BH3-only protein Bim through FOXO3 in breast cancer cells (Liu et al., 2005). To study whether ROS also lead to FOXO3 activation during chemotherapy-induced apoptosis in neuronal cells, the neuronal tumor cell lines SH-EP and STA-NB15 were treated with etoposide and doxorubicin and ROS production was measured by reduced MitoTrackerRed (CM-H₂XROS). Both chemotherapeutic agents caused a transitory increase of ROS at the mitochondria. In both cell lines, ROS reached a maximum after 2 hours of etoposide treatment and declined after 4 hours. Doxorubicin-induced ROS accumulated later and reached a maximum after 6 hours treatment (Fig. 1A,B). We have shown before that low-dose activation of an ectopic FOXO3 sensitizes neuronal cells to death by DNA-damaging agents such as doxorubicin and etoposide (Obexer et al., 2009). To investigate whether DNA-damaging drugs affect steady state expression of FOXO3, we analyzed FOXO3 protein levels after treatment with etoposide and doxorubicin in neuroblastoma cells. We found that, upon treatment, endogenous FOXO3 protein levels increased in SH-EP (Fig. 1C, left panels) and STA-NB15 cells (Fig. 1C, right panels) after 2 and 4 hours. To determine whether the observed accumulation of FOXO3 correlates with its activation, we measured the protein levels of its apoptosis-inducing downstream targets Bim and Noxa. Both pro-apoptotic BH3-only proteins were rapidly induced within 2 to 4 hours of drug treatment in SH-EP and STA-NB15 cells. Interestingly, Bim was phosphorylated at Ser69, a recognition site for JNK, which stabilizes Bim during cellular stress in neuronal cells (Becker et al., 2004). Phosphorylation by JNK is also thought to release Bim from dynein light chain-1 and to contribute to its activation (Lei and Davis, 2003). The stress-induced kinases JNK and MST1 (Essers et al., 2004; Lehtinen et al., 2006) or the interaction with ataxia telangiectasia mutated (ATM) (Tsai et al., 2008) might trigger nuclear accumulation of FOXO3 during the DNA-damage response. To investigate whether FOXO3 also translocates to the nucleus in etoposide- and doxorubicin-treated neuronal cells we used SH-EP cells designed to express an ECFP–FOXO3 fusion protein (Obexer et al., 2009) and analyzed the subcellular localization of the fusion protein by live-cell fluorescence imaging. In untreated cells, the ECFP–FOXO3 fusion protein was mainly localized in the cytoplasm. Consistent with earlier studies in non-neuronal cells (Essers et al., 2004), H₂O₂ caused rapid nuclear accumulation of FOXO3 within 20 to 40 minutes in neuroblastoma cells (supplementary material Fig. S1A). Etoposide induced nuclear localization of FOXO3 in 72% of the cells within 90 minutes and doxorubicin activated FOXO3 in 62% of all cells within 2 hours (Fig. 1D). Although H₂O₂ is an efficient activator of FOXO3, in drug-treated neuroblastoma cells nuclear accumulation of ECFP–FOXO3 preceded ROS accumulation. We therefore addressed whether activation of FOXO3 occurs downstream or upstream of ROS production during chemotherapy and pretreated SH-EP cells stably transfected with ECFP–FOXO3 with the antioxidant N-actetyl-L-cysteine (NAC) for 1 hour before they were incubated with etoposide and doxorubicin. Although NAC treatment diminished ROS levels (Fig. 3C), the inhibition of ROS production did not prevent shuttling of FOXO3 into the nucleus (Fig. 1D), implying that activation of FOXO3 is not a consequence of ROS accumulation. To verify whether the effects of etoposide and doxorubicin on ROS generation and increase in Bim depend on FOXO3 activation, SH-EP and STA-NB15 cells were infected with lentiviruses for FOXO3-specific shRNA expression. After infection, bulk-selected NB15 cells and individual clones from SH-EP cells were subjected to immunoblotting (supplementary material Fig. S1B) and analyzed for ROS production and the expression of the FOXO3 target Bim after etoposide and doxorubicin treatment. Live-cell fluorescence images of etoposide-treated SH-EP/shFOXO3-17-clone7 (Fig. 2A) and NB15/shFOXO3 cells (Fig. 2B) showed no significant ROS accumulation compared with the controls, SH-EP/shCtr and NB15/shCtr, respectively. Next, we determined the Bim expression level in SH-EP/shFOXO3-17-clone7 cells expressing FOXO shRNA, bulk-selected NB15/shFOXO3-17 cells and control cells, respectively. Knockdown of FOXO3 completely prevented induction of Bim by etoposide for up to 16 hours in both neuroblastoma cell lines (Fig. 2C,D). In a parallel experiment, we infected SH-EP and STA-NB15 cells with a retrovirus coding for a dominant-negative FOXO3 mutant (FOXO3-DBD). Ectopic expression of the Myc-tagged FOXO3-DBD construct was verified by immunoblotting (supplementary material Fig. S1B). ROS accumulation during treatment with etoposide and doxorubicin was prevented by FOXO3-DBD in both cell lines (Fig. 2A,B). These data suggest that in contrast to previous findings, the chemotherapeutic agents etoposide and doxorubicin activate FOXO3, increase the expression of FOXO3 target genes and subsequently lead to ROS accumulation in neuronal cells.

FOXO transcription factors are regulators of cell death but are also reported to protect cells against ROS. These contrasting abilities might depend on post-translational modifications and cell type. To assess how FOXO3 contributes to ROS production in neuronal cells, we used SH-EP and NB15 cells that express a conditional, PKB-phosphorylation-independent FOXO3(A3)-ERtm allele, which can be activated by 4OH-tamoxifen (4OHT) (Obexer et al., 2007). SH-EP/FOXO3 cells or mock-infected controls were treated with 100 nM 4OHT for 18 hours or with H₂O₂ as a positive control. Both, H₂O₂ and FOXO3 activation strongly increased ROS staining as measured by reduced MitotrackerRed CM-H₂XROS (supplementary material Fig. S1C). To specify the time points of ROS accumulation, we performed time course analysis of 4OHT-treated SH-EP/FOXO3 and NB15/FOXO3 cells (Fig. 3A,B). FOXO3 activation resulted in a highly significant increase of ROS after 16 to 18 hours (P<0.0001) in SH-EP/FOXO3 cells, which marks the onset of apoptosis in these cells. In NB15/FOXO3 cells, the climax of ROS accumulation was observed 48 hours after FOXO3 activation, again coinciding with the onset of programmed cell death (Obexer et al., 2007). In both cell lines, FOXO3 activation induced a transient highly significant accumulation of ROS, which peaked at 4 hours in SH-EP cells and at 12 hours in NB15 cells, respectively, with significantly decreased ROS levels at the following time points (Fig. 3B). To investigate the relevance of the first ROS burst for cell death induction by FOXO3, SH-EP/FOXO3 cells were treated with 4OHT for 16 hours alone or after pre-incubation with the ROS-inhibitor NAC (1 hour) to scavenge ROS. Nuclei of cells were stained with Hoechst 33342 dye and analyzed by fluorescence microscopy. NAC efficiently prevented the first ROS burst at 4 hours of 4OHT treatment in SH-EP/FOXO3 cells as measured by CM-H₂XROS (Fig. 3C). As shown in Fig. 3D, combined treatment with NAC and 4OHT delayed the onset of apoptosis, because only 9.9% of cells showed fragmented nuclei compared with 37.5% in 4OHT-treated SH-EP/FOXO3 cells. This indicates that cellular ROS levels critically contribute to FOXO3-induced programmed cell death in neuronal cells and implies that FOXO3 triggers a primary ROS increase that induces a secondary, apoptosis-inducing ROS burst.

The enzymes MnSOD and catalase were reported to be FOXO3 target genes that critically participate in ROS detoxification (Chiribau et al., 2008; Kops et al., 2002; Yalcin et al., 2008). MnSOD converts superoxide into hydrogen peroxide, which is further converted by catalase to oxygen and water. To determine the possible involvement of these enzymes in the biphasic ROS accumulation during FOXO3 activation we determined expression of MnSOD and catalase protein by immunoblotting. No significant induction of these enzymes by FOXO3 was observed in SH-EP and STA-NB15 cells (Fig. 4A), indicating that these two enzymes are not responsible for cellular ROS fluctuations. Consistently, ectopic MnSOD expression also did not prevent ROS induction by FOXO3 (supplementary material Fig. S1D).

Because MnSOD and catalase were not critical for biphasic ROS accumulation during FOXO3 activation, we analyzed the regulation of SESN3, a member of the sestrin family, which was recently shown to be regulated by FOXO3 (Nogueira et al., 2008) and which we identified as a FOXO3 target by Affymetrix gene expression profiling in neuroblastoma cells (data not shown). As shown in Fig. 4B, SESN3 mRNA was strongly induced in SH-EP/FOXO3 cells (about 50-fold after 6 hours, P=0.004) and to a lesser but significant extent in NB15/FOXO3 cells. In parallel, SESN3 protein levels also increased continuously after FOXO3 activation in both cell lines (Fig. 4B, right panels). This induction of SESN3 correlates with the first decrease of ROS after 4 hours in SH-EP/FOXO3 and after 12 hours in NB15/FOXO3 cells. To investigate whether SESN3 participates in ROS detoxification, SH-EP/FOXO3 and NB15/FOXO3 cells were infected with lentiviruses coding for different SESN3-specific shRNAs. Knockdown of SESN3 in bulk-selected NB15/FOXO3-shSESN3 cells (supplementary material Fig. S2A) and in single clones of SH-EP/FOXO3-shSESN3 cells was verified by quantitative RT-PCR (Fig. 4C). Knockdown of SESN3 prevented the ROS decline after 6 hours of FOXO3 activation (Fig. 4C) and increased FOXO3-induced apoptosis from 12% in mock-infected control cells to 29% and 24% in FOXO3-shSESN3-28-clone12 and FOXO3-shSESN3-46-clone16 SH-EP cells, respectively, as determined by propidium iodide (PI) FACS analyses (Fig. 4D). In NB15/FOXO3-shSESN3-46 cells, SESN3 knockdown increased apoptosis from 19% to 37% after 48 hours of 4OHT treatment (supplementary material Fig. S2B). This implies that FOXO3-induced upregulation of SESN3 scavenges ROS after a primary ROS accumulation and thereby defines a cell-protective, apoptosis-delaying ROS threshold.

We have shown before that FOXO3 controls the expression of the pro-apoptotic BH3-only proteins Bim and Noxa, which cause loss of mitochondrial membrane potential and eventually lead to cytochrome c release in neuronal cells (Obexer et al., 2007). Because accumulation of ROS can be triggered by opening of the mitochondrial permeability transition pore (MPT) (Zorov et al., 2006), we determined the steady state expression of anti-apoptotic BCL2 proteins after FOXO3 activation. BclxL protein levels decreased within 8 hours in SH-EP/FOXO3 cells and declined in NB15/FOXO3 cells after 24 hours of 4OHT treatment (Fig. 5A), whereas BCL2 was not regulated and MCL1 was slightly induced (supplementary material Fig. S3A). Because ectopic expression of BclxL significantly prevented FOXO3-induced cell death for up to 72 hours (Fig. 5B), we assessed the interaction of BclxL with the pro-apoptotic BCL2 proteins Noxa, Bim, BAX and Bak in mitochondria in NB15/FOXO3 cells before ROS accumulation could be detected (Fig. 5C). The BH3-only proteins Bim and Noxa rapidly increased in the mitochondrial fraction and co-purified with BclxL within 3 to 6 hours of FOXO3 activation. We recently provided evidence that Noxa binds to and neutralizes BclxL during proteasome-inhibitor-induced apoptosis, suggesting that in neuronal cells, BclxL is a preferential pro-survival binding partner of Noxa (Hagenbuchner et al., 2010). Whereas in untreated cells, neither BAX nor Bak associated with BclxL, the activation of FOXO3 induced a transitory interaction of Bak and BAX with BclxL at 3 hours. At 6 hours, however, BclxL was mainly associated with Noxa and Bim and to a lesser extent with BAX. This suggests that, although nuclear fragmentation is first detectable after 24 to 36 hours in STA-NB15 cells (Obexer et al., 2007), neutralization of the pro-survival BclxL occurs within 6 hours of FOXO3 activation and thereby precedes the first accumulation of ROS in STA-NB15 cells.

In line with the hypothesis that ROS are triggered by the disturbance of the BCL2 rheostat, we analyzed whether BclxL, Bim or Noxa are crucially involved in the first ROS induction by FOXO3 in neuronal cells. To address this, we used SH-EP/FOXO3 and NB15/FOXO3 cells that either overexpress BclxL or prevent induction of Bim or Noxa by shRNA-mediated gene knockdown (Obexer et al., 2007). As shown in Fig. 6A, ectopic expression of BclxL or knockdown of Bim completely prevented ROS accumulation after 2 and 4 hours of FOXO3 activation (P<0.0001), whereas knockdown of Noxa had an attenuating effect only at the 4 hour time-point. Similar results were achieved in NB15/FOXO3 cells, i.e. ectopic BclxL or knockdown of Bim markedly reduced ROS accumulation after 12 hours of FOXO3 activation (Fig. 6B). As in SH-EP cells, knockdown of Noxa in NB15 cells reduced the climax of ROS generation at 12 hours, but did not prevent the ROS peak. These results indicate that the initial ROS production is triggered downstream of Bim and BclxL and that changes in the expression of these two BCL2 proteins might impair mitochondrial function. To address whether these events are accompanied by changes in mitochondrial function, we measured endogenous mitochondrial respiration during FOXO3 activation using high-resolution respirometry. Within 4 hours of 4OHT treatment, mitochondrial respiration decreased (Fig. 6C), which correlated with ROS accumulation (Fig. 3A and Fig. 6A) and induction of Bim by FOXO3 (supplementary material Fig. S3B). Ectopic BclxL significantly prevented the decline of mitochondrial respiration (P<0.05), suggesting that the balance between pro- and anti-apoptotic BCL2 proteins directly affects the mitochondrial activity during FOXO3 activation (Fig. 6A,B).

Bim is a key regulator of FOXO3-induced apoptosis of neuronal cells (Obexer et al., 2007) and its knockdown prevents FOXO3-induced ROS accumulation (Fig. 6A,B). To study whether Bim causes cellular ROS on its own, we conditionally expressed Bim in a tetracycline-regulated manner (Fig. 7A) (Ausserlechner et al., 2005; Hagenbuchner et al., 2010). Induction of Bim rapidly induced ROS within 2 hours (Fig. 7B) and lowered mitochondrial respiration from 100% in mock-infected control cells to 67.6±4.6% and 49.8±9.4% after 2 and 4 hours (P<0.01), respectively. From the combined data, we conclude that induction of Bim by FOXO3 triggers FOXO3-induced ROS accumulation during DNA-damaging drug-induced cell death in neuronal cells.

In this study we demonstrate for the first time that the activation of the transcription factor FOXO3 leads to biphasic ROS accumulation and we provide evidence that ROS are critical mediators of FOXO3-induced cell death in neuronal cells. Others have shown that etoposide activates FOXO3 as a consequence of intracellular ROS accumulation in breast cancer cells (Liu et al., 2005). However, in neuronal tumor cells, in primary neurons and in NGF-differentiated PC12 cells (supplementary material Fig. S4), we demonstrate that etoposide- and doxorubicin-induced cell death depends on the activation of FOXO3 and the subsequent induction of cellular ROS. Chemotherapeutic agents cause nuclear accumulation of FOXO3 before and independent of ROS accumulation, as evidenced by live-cell fluorescence imaging (Fig. 1, Fig. 2A,B, supplementary material Fig. S4B,D). Although SH-EP and STA-NB15 cells differ in their p53 status (Hagenbuchner et al., 2010), FOXO3 was activated by etoposide and doxorubicin in both cell lines (Fig. 1C,D), suggesting that induction of FOXO3 does not depend on a functional p53 pathway. Shuttling of FOXO3 might be triggered by JNK or by the downstream kinase MST1, which both induce nuclear accumulation of FOXO transcription factors during cellular stress (Essers et al., 2004; Lehtinen et al., 2006; Bi et al., 2010). The tyrosine kinase c-Abl which is also a transducer of signals for oxidative stress and DNA-damage acts upstream of MST1 and phosphorylates MST1 at Tyr433 causing stabilization, activation and enhanced interaction of MST1 with FOXO3. Similarly to our experiments, the authors provided evidence that activation of FOXO3 by the c-Abl–MST1 pathway promotes apoptosis in mammalian neurons (Xiao et al., 2011). Bim as the critical mediator of FOXO3-induced ROS was not only induced at the protein level, but also showed increased phosphorylation at Ser69 (Fig. 1C). In neurons, Bim Ser69 can be phosphorylated by JNK or by the pro-survival kinases ERK1/2 leading either to Bim activation (Becker et al., 2004) or, in the case of ERK1/2 to proteasomal degradation. Because the activation status of JNK was not altered during drug treatment as measured by phospho-specific antibodies directed against Thr183 and Tyr185 (supplementary material Fig. S4E), and inhibition of JNK did not prevent induction of Bim, it is possible that Ser69 phosphorylation results from counter-regulatory activation of ERK1/2. Tsai and colleagues demonstrated that during DNA damage the ATM kinase physically interacts with FOXO3 and can recruit FOXO3 to the nucleus (Tsai et al., 2008). In neuronal cells, both etoposide and doxorubicin induce auto-phosphorylation of the ATM kinase at Ser1981, which indicates ATM activation. The inhibition of ATM during drug treatment also blocks FOXO3-dependent induction of Bim (our unpublished results). We therefore believe that ATM contributes to FOXO3 activation during drug treatment of neuronal cells.

When investigating the relationship between FOXO3, cellular ROS levels and cell death in neuronal cells we found that FOXO3 on its own increased intracellular ROS and induced two consecutive waves of ROS accumulation (Fig. 3A,B). To further investigate the underlying molecular reasons for this biphasic ROS burst, we evaluated the expression of the ROS-detoxifying enzymes MnSOD and catalase, both of which are reported to be downstream targets of FOXO3 (Liu et al., 2005; Tan et al., 2008). These enzymes were not regulated by FOXO3 in the examined neuronal cells and therefore cannot contribute to the fluctuation of ROS levels during FOXO3 activation (Fig. 4A). However, we identified SESN3 as a strongly induced FOXO3 target that is critical for the decline of ROS levels after the first ROS accumulation. The sestrin family is comprised of three members, SESN1, SESN2 and SESN3, which have a dual biochemical function: as antioxidant proteins they are involved in the regeneration of peroxiredoxins that scavenge ROS (Budanov et al., 2004) and they also act as inhibitors of target of rapamycin complex 1 (TORC1) (Chen et al., 2010; Budanov and Karin, 2008). SESN1 and SESN2 are induced upon DNA damage and thereby diminish the activity of TORC1. SESN3 is induced by FOXO1 and also represses the activity of TORC1 and anabolic metabolism. This results in reduced protein and lipid synthesis and thereby saves energy for DNA repair. Neuronal cells rely on thiol-reducing systems based on thioredoxin and glutathione, which act as reducers of cellular peroxides. This detoxification occurs by the transfer of reducing equivalents from NADPH to peroxides involving thioredoxin, thioredoxin reductase, and peroxiredoxins (Prx). If peroxiredoxins are overoxidized (Prx-SO_2/3H) during oxidative stress, they cannot be reduced and thereby regenerated to the active form by thioredoxin. This conversion of overoxidized peroxiredoxins to active scavengers requires ATP-dependent reductases of the sestrin family (Budanov et al., 2004). As a result of this function, sestrins are crucial for keeping ROS-scavenging peroxiredoxins active.

Because knockdown of SESN3 prevented the transitory decline in ROS during FOXO3 activation and accelerated FOXO3-induced cell death in these neuronal cells (Fig. 4C,D, supplementary material Fig. S2A,B), FOXO3, through SESN3, seems to activate both ROS production and a ROS-detoxifying pathway.

Zorov and colleagues showed that mitochondrial ROS trigger a secondary enhanced ROS production leading to significant mitochondrial and cellular injury (Zorov et al., 2006). Mitochondrial stress or the opening of channels in the outer or inner mitochondrial membrane might result in a collapse of the mitochondrial membrane potential (ΔΨ) and could transiently increase ROS levels by the electron transfer chain. Because reduced respiratory activity (Fig. 6C) and increase in ROS are observed during initial FOXO3 activation, we hypothesized that FOXO3-induced changes in expression and activation of BCL2 proteins cause a transient opening of the MPT pore. FOXO3 induces the pro-apoptotic proteins Noxa and Bim and leads to release of cytochrome c in neuroblastoma cells (Obexer et al., 2007). The induction of Bim and Noxa within the first 6 hours correlates with a transitory association of BAX and Bak with BclxL at the mitochondria (Fig. 5C). This dynamic interaction might be due to Noxa- and Bim-induced redistribution of BAX and Bak between different pro-survival BCL2 proteins (Willis et al., 2005). Between 3 and 6 hours, however, BAX and Bak are partially displaced from BclxL, allowing them to activate MPT formation. Importantly, conditional expression of Bim triggers ROS accumulation (Fig. 7B) and remarkably reduces mitochondrial respiration (Fig. 7C). Moreover, FOXO3-induced ROS production is prevented by knockdown of Bim or ectopic expression of BclxL, and is attenuated by Noxa knockdown (Fig. 6A,B). This suggests that ROS accumulation during FOXO3 activation occurs downstream of Bim and BclxL. The repression of anti-apoptotic BclxL (Fig. 5A) by FOXO3 might further impair the ability of the cell to cope with increasing levels of pro-apoptotic BH3-only proteins and ROS levels (Fig. 5B).

In this study, we demonstrate in neuronal cells that FOXO3 represses mitochondrial respiration and induces cellular ROS in response to chemotherapy. FOXO3 triggers a biphasic ROS burst by perturbation of mitochondrial function through Bim and BclxL, and in parallel, increases ROS-detoxifying capacity through SESN3 (Fig. 8). The first increase in ROS might cause cellular damage that, in combination with prolonged neutralization of pro-survival BCL2 proteins, results in a secondary stronger ROS burst, which leads to apoptosis. Therefore, ROS can be considered as critical mediators of FOXO3-induced apoptosis, where on one hand FOXO3 activates detoxification of ROS, but on the other hand it also triggers ROS production by induction of pro-apoptotic and repression of pro-survival BCL2 proteins. This dual effect of FOXO3 seems critical in the control of stress sensitivity and chemotherapy-induced cell death in neuronal tumor cells. The combined data clearly demonstrate that FOXO3-mediated induction of Bim disrupts mitochondrial respiration, leading to ROS, which are critical downstream mediators of FOXO3-induced cell death in neuronal cells.

The neuroblastoma cell lines SH-EP and STA-NB15 (Narath et al., 2005) as well as Phoenix packaging cells for helper-free production of amphotropic retroviruses (Grignani et al., 1998) were cultured in RPMI1640 (Lonza, Basel, Switzerland) containing 10% fetal calf serum, 100 U/ml penicillin, 100 μg/ml streptomycin and 2 mM L-glutamine (Gibco BRL, Paisley, UK) in 5% CO₂. HEK293T cells for production of lentiviruses were cultured in DMEM (Lonza, Basel, Switzerland). All cultures were routinely tested for mycoplasma contamination using the Venor GeM-mycoplasma detection kit (Minerva Biolabs, Berlin, Germany). All reagents were purchased from Sigma (Vienna, Austria) unless indicated otherwise.

The vectors pBABE-MnSOD-Puro, pLIB-FOXO3(A3)-ER-iresNeo, pLIB-MCS2-iresPuro, pLIB-rtTA-M2-iresTRSID-iresPuro, pQ-tetH1-shBim-SV40Puro, pQ-tetH1-shNoxa-SV40Puro, pQ-tetH1-SV40Puro, pLIB-ECFP-FOXO3wt-iresPuro, pLIB-BclxL-iresPuro, pQ-tetCMV-Bim-SV40Neo and pQ-tetCMV-SV40Neo have been described previously (Kuznetsov et al., 2008a; Ausserlechner et al., 2006; Hagenbuchner et al., 2010; Obexer et al., 2007; Obexer et al., 2009). The lentiviral vectors coding for FOXO3-specific shRNA (pLKO1-shFOXO3-91617), SESN3-specific shRNA (pLKO-shSESN3-141228 and pLKO-shSESN3-143446) and the control vector pLKO.1 were obtained from Thermo Scientific (Huntsville, AL). pLIB-mycTag-FOXO3-DBD-iresPuro was constructed by inserting the FOXO3-DBD fragment from pSG5-MycTag-FOXO3-DBD (Dijkers et al., 2002) into the EcoR1 and Sal1 sites of pLIB-MCS2-iresPuro (Ausserlechner et al., 2006).

6×10⁵ Phoenix cells were transfected with 2 μg of retroviral vectors and 1 μgofa plasmid coding for VSV-G protein using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) (Ausserlechner et al., 2006; Riml et al., 2004). For production of lentiviruses 6.5×10⁵ HEK293T cells were transfected with 1.6 μg pLKO.1 vectors expressing FOXO3-specific or SESN3-specific shRNA (Thermo Scientific) and the packaging plasmid pCMV 8.91 (kindly provided by Didier Trono, EPFL, Lausanne, Switzerland). After 48 hours, the virus-containing supernatants were filtered through 0.22 μm syringe filters (Sartorius, Goettingen, Germany) and incubated with the target cells for at least 6 hours. SH-EP/ECFP-FOXO3, SH-EP/FOXO3-shBim-clone1, SH-EP/FOXO3-shNoxa-clone3, NB15/FOXO3-shBim, NB15/FOXO3-shNoxa, SH-EP/tetCtr and SH-EP/Bim cells have already been described (Hagenbuchner et al., 2010; Obexer et al., 2007; Obexer et al., 2009). pLIB-BclxL-iresPuro supernatants were used to generate SH-EP/FOXO3-BclxL and NB15/FOXO3-BclxL cells. pBABE-MnSOD-Puro (gift from Jakob Troppmair, Medical University Innsbruck, Austria) supernatants were used to infect SH-EP/FOXO3 cells (SH-EP/FOXO3-MnSOD).

pLKO-shFOXO3-191617, pLKO-shSESN3-141228 and pLKO-shSESN3-143446 lentiviral supernatants were used to infect SH-EP and STA-NB15 cells (SH-EP/shFOXO3-17-clone7, SH-EP/shSESN3-28-clone12, SH-EP/shSESN3-46-clone16, NB15/shFOXO3-17 and NB15/shSESN3-46 cells). The empty vector pLKO.1 served as a control (SH-EP/shCtr and NB15/shCtr). pLIB-mycTag-FOXO3-DBD-iresPuro supernatants were used to infect SH-EP and STA-NB15 cells (SH-EP/FOXO3-DBD and NB15/FOXO3-DBD).

For immunoblot analyses total protein was prepared from 5×10⁶ cells resuspended in lysis buffer (50 mM HEPES-NaOH, 1% Triton X-100, 150 mM NaCl, 2 mM EDTA, 10% glycerol) with protease and phosphatase inhibitors. Equal amounts of total protein (50 μg/lane) were separated by SDS-PAGE. For subcellular fractionation, 5×10⁷ cells were resuspended in MSH buffer (210 mM mannitol, 70 mM sucrose, 20 mM HEPES, 1 mM EDTA, pH 7.4) with protease and phosphatase inhibitors and left on ice for 1 hour. Nuclei and non-lysed cells were removed by centrifugation at 500 g for 5 minutes. To obtain the mitochondrial and cytoplasmic fraction, the supernatant was centrifuged at 25,000 g for 30 minutes. The membrane pellet was resuspended in MSH buffer containing protease inhibitor and 1% CHAPS, lysed on ice for 30 minutes and centrifuged at 25,000 g for 30 minutes. Protein concentration was measured using Bradford Reagent (Bio-Rad Laboratories, Munich, Germany). For immunoprecipitation, 1 μg of rabbit anti-BclxL antibody (Cell Signaling Technology, Boston, MA) or rabbit immunoglobulin, as a negative control, were added to 200 μg mitochondrial lysate and incubated on ice for 2 hours. For pull-down of the immunocomplexes Tachisorb Immunoadsorbent (Calbiochem, Nottingham, UK) was added to the mitochondrial lysate overnight. Tachisorb immunocomplexes were washed four times in MSH buffer, resuspended in SDS sample buffer and subjected to SDS-PAGE and blotting. Equal amounts of total protein and cleared supernatants were loaded as controls. After gel separation and blotting, the membranes were blocked, incubated with primary antibodies specific for pBim(Ser69), BclxL, BAX, Bak, CoxIV, JNK, pJNK-T183/T185 (Cell Signaling Technology), Bim, c-Myc (BD-Pharmingen, USA), Catalase (Calbiochem), Noxa (Alexis Biochemicals, San Diego, CA), MnSOD, FOXO3 (Upstate Biotechnology, Lake Placid, NY), GAPDH (Acris Antibodies, Herford, Germany), SESN3 (Abcam, Cambridge, UK) and α-tubulin (Oncogene Research Products, La Jolla, CA), washed and incubated with anti-mouse or anti-rabbit horseradish-peroxidase-conjugated secondary antibodies. The blots were developed by enhanced chemiluminescence (GE Healthcare, Vienna, Austria) according to the manufacturer’s instructions and analyzed with a AutoChemi detection system (UVP, Cambridge, UK). Signal intensities were measured using LabWorks software (UVP).

For live-cell analyses, cells were grown on LabTek Chamber Slides (Nalge Nunc International, Rochester, NY) or on 35 mm optical plastic dishes (Ibidi, Marinsried, Germany), coated with 0.1 mg/ml collagen. ROS measurements were performed by incubating the cells with MitoTracker Red CM-H₂XROS (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions (final concentration 500 nM). Images were acquired with an Axiovert200M microscope (Zeiss, Vienna, Austria). Fluorescence intensity was quantified using Axiovision Software (Zeiss) and relative ROS levels were expressed as a percentage of that in untreated controls.

To quantify sestrin-3 (SESN3) mRNA levels, we designed ‘real-time’ RT-PCR assays, using GAPDH as reference gene. SH-EP/FOXO3 and NB15/FOXO3 cells were cultured in the presence of 100 nM 4OHT for the times indicated. Total RNA was prepared from 5×10⁶ cells using TRIzol Reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. cDNA was synthesized from 1 μg of total RNA using the RevertAid H Minus First Strand cDNA Synthesis Kit (MBI Fermentas, St Leon-Rot, Germany). Real-time RT-PCR was performed on the iCycler instrument (Bio-Rad) using sestrin-3 (forward, 5′-CCGCCAGTAACTATCATACATGCG-3′ and reverse, 5′-GAGGATGTTGACACAACCATGCTG-3′) and GAPDH oligonucleotides (forward, 5′-TGTTCGTCATGGGTGTGAACC-3′ and reverse, 5′-GCAGTGATGGCATGGACTGTG-3′). All reactions were done in triplicate using a thermal profile of an initial 3 minute melting step at 95°C, followed by 40 cycles comprising 95°C for 20 seconds and 55°C for 45 seconds. To verify the presence of only a single amplicon, a melting curve was processed after each run. After normalization to GAPDH expression, regulation was calculated between treated and untreated cells.

Apoptosis was assessed by staining the cells with propidium iodide (PI) and forward and sideward scatter analysis using a CytomicsFC-500 Beckman Coulter. 2×10⁵ cells were harvested and resuspended in hypotonic PI solution for 2–4 hours at 4°C. Stained nuclei in the sub-G1 marker window were considered to represent apoptotic cells (Hagenbuchner et al., 2010; Obexer et al., 2007; Obexer et al., 2009). Statistical analysis was performed using GraphPad Prism 4.0 software.

Oxygen consumption of the cells and mitochondrial function were analyzed by high-resolution respirometry at 30°C (Kuznetsov et al., 2008b), using a two-channel titration-injection respirometer (Oroboros Oxygraph, Innsbruck, Austria) in RPMI1640 growth medium, assuming an O₂ solubility of 10.5 μmol l^–1 kPa^–1. The software DatLab (Oroboros) was used for data acquisition and analysis. Respiration rates were expressed in pmol of O₂ per second, per 10⁶ cells. Mitochondria respiratory chain specific inhibitors for complex I (rotenone, 0.5 μM) and complex III (antimycin A, 5 μM) blocked respiration to 7–8% and 4–6% of control, respectively, confirming that measured oxygen consumption was due to the respiratory chain activity. For comparison, control cells were set as 100%.

The authors thank Paul Coffer, Department of Cell Biology, University of Utrecht, The Netherlands; Jakob Troppmair, Department of Operative Medicine, Medical University Innsbruck, Austria; and Didier Trono, School of Life Sciences, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, for donating plasmids.