Curcumin induces crosstalk between autophagy and apoptosis mediated by calcium release from the endoplasmic reticulum, lysosomal destabilization and mitochondrial events

Curcumin, a major active component of turmeric (Curcuma longa, L.), has anticancer effects. In vitro studies suggest that curcumin inhibits cancer cell growth by activating apoptosis, but the mechanism underlying these effects is still unclear. Here, we investigated the mechanisms leading to apoptosis in curcumin-treated cells. Curcumin induced endoplasmic reticulum stress causing calcium release, with a destabilization of the mitochondrial compartment resulting in apoptosis. These events were also associated with lysosomal membrane permeabilization and of caspase-8 activation, mediated by cathepsins and calpains, leading to Bid cleavage. Truncated tBid disrupts mitochondrial homeostasis and enhance apoptosis. We followed the induction of autophagy, marked by the formation of autophagosomes, by staining with acridine orange in cells exposed curcumin. At this concentration, only the early events of apoptosis (initial mitochondrial destabilization with any other manifestations) were detectable. Western blotting demonstrated the conversion of LC3-I to LC3-II (light chain 3), a marker of active autophagosome formation. We also found that the production of reactive oxygen species and formation of autophagosomes following curcumin treatment was almost completely blocked by N-acetylcystein, the mitochondrial specific antioxidants MitoQ10 and SKQ1, the calcium chelators, EGTA-AM or BAPTA-AM, and the mitochondrial calcium uniporter inhibitor, ruthenium red. Curcumin-induced autophagy failed to rescue all cells and most cells underwent type II cell death following the initial autophagic processes. All together, these data imply a fail-secure mechanism regulated by autophagy in the action of curcumin, suggesting a therapeutic potential for curcumin. Offering a novel and effective strategy for the treatment of malignant cells.


INTRODUCTION
Curcumin, extracted from the plant Curcuma Longa, has long been recognized as a very promising anticancer drug because of its many beneficial properties, including anti-inflammatory, antioxidant and anti-carcinogenic activities. 1,2 Curcumin is a potent bioactive compound that is used to treat cancer, 3 atherosclerosis 4 and neurodegenerative diseases, such as Alzheimer's 4,5 and Parkinson's disease. 6 Curcumin is particularly appealing as a therapeutic agent because of its low toxicity, although its intracellular bioavailability is quite low. 7 Curcumin has antiproliferative and anti-carcinogenic properties and may be useful for cancer prevention. The mechanisms underlying curcumininduced cancer cell death have not been clearly defined, although available evidence suggests that curcumin downregulates NF-κB signaling, which suppresses proliferation and induces apoptosis. 8,9 Although both caspase 8-mediated and/or caspase 9-dependent apoptosis have been reported to occur upon the exposure of cancer cells to curcumin, 10,11 discrepant results exist for caspase-8. 12 Indeed, it is still unclear which caspase functions as the initiator caspase. Curcumin induces mitochondria-mediated apoptosis 13,14 and reactive oxygen species (ROS) have been implicated in this process. 15 Complex interplay between autophagy and apoptosis has come to light. Autophagy is involved in the anti-proliferative and apoptotic activities of curcumin, 16 both in vitro and in vivo. Curcumin inhibit the growth of malignant cells by activating autophagic-cell death via the Akt/mTOR/p70S6K signaling and ERK1/2 pathways. 17 Here, we show that curcumin with its multifaceted action induces different cell death pathways together with autophagy in a dose-dependent relationship. An intricated crosstalk between autophagy and apoptosis is initiated that correspond to the hormesis characteristics of curcumin. We discuss these results in light of the recent use of curcumin as an anticancer therapy. amount of curcumin was found within the cells (at 5 μM; Figure 1a upper left panel). Intracellular curcumin levels were maximal within 10 min and decreased slowly over the following hours ( Figure 2f).
Curcumin induces apoptosis and necrosis and an increase in mitochondrial membrane potential, ROS production and calcium levels After incubation with curcumin, cells were double stained with YO-PRO-1 (Y)/propidium iodide (PI) and analyzed by flow cytometry to discriminate between viable (Y − /PI − ), apoptotic (Y + /PI − ) and necrotic (Y + /PI − + ) cells (Figure 1c). Curcumin concentrations of 20-25 μM induced apoptosis, whereas higher concentrations favored necrosis over apoptosis. And as the curcumin concentration increases, apoptosis is progressively replaced with necrosis, probably because curcumin is cytotoxic at high concentrations ( Z25 μM).
Analysis of the mitochondrial membrane potential (ΔΨ m ) and viability (PI; Figure 2a) revealed that curcumin decreased the ΔΨ m after 24 h of incubation in viable cells. Indeed, at 15 μM of curcumin, 21% of viable PI − cells showed a low ΔΨ m as opposed to only 5% of viable cells with low potential in untreated samples.
This was associated with a small proportion of dead (PI + ) cells, which was 4.5% in untreated conditions and 14% at 15 μM of curcumin. At curcumin concentrations of 25 μM, more than half of cells had a low ΔΨ m (52%). Moreover, cells started to become permeabilized (high intermediate PI − /+ staining and PI + staining) and the proportion of PI +++ cells with a low ΔΨ m was 19%. At 50 μM of curcumin, 51% of PI + cells and 92% of all cells had a low ΔΨ m . Longer incubation time (48 h) resulted in fewer viable cells exhibiting a high ΔΨ m (Figure 2b).
In general, a ΔΨ m drop is associated with ROS production resulting from mitochondrial destabilization. 18 We therefore assayed superoxide anion (O − ) and hydroperoxide (H 2 O 2 ) production in curcumin-treated cells (Figures 2c and d). Consistent with a mitochondrially induced apoptotic pathway, the decline in ΔΨ m was associated with a raise of O − and H 2 O 2 (Figures 2a and b). ROS production increased as a function of incubation time (Figure 2e). Experiments involving diverse anti-oxidants, including N-acetylcysteine (NAC), Trolox, vitamin E and the mitochondrially targeted MitoQ10 or SQK1, indicated that ROS were produced by mitochondria (Supplementary Figure S2). MitoQ10 or SKQ1, which specifically target the mitochondrial matrix, were more effective than NAC and Trolox at limiting ROS levels.
These data and calcium fluorescence measured with Fluo3-AM are plotted as a timecourse for cells treated with 25 μM curcumin for up to 24 h (Figures 2e and f). After curcumin treatment, there was an increase in the calcium concentration followed by a ΔΨ m loss and ROS production ( Figure 2f). Thus, curcumin enters Huh-7 cells rapidly (Figure 2f) where it provokes a calcium pulse and the early apoptotic events, followed by a drop in ΔΨ m and an increase in ROS production ( Figure 2e). Cellular calcium content appeared to be linked to curcumin fluorescence ( Figure 2f). Indeed, both decreased similarly as a function of time after an initial peak in concentration within the first minutes of curcumin addition. Closer inspection of the early time points reveals that curcumin fluorescence increased before calcium fluorescence increased (Figure 2f enlargement). We therefore investigated the mechanisms of the calcium increase that rapidly follows curcumin uptake and precedes the mitochondrial events ( Figure 2e).
Opening of the mitochondrial transition pore in isolated mitochondria As it has been demonstrated that isolated mitochondria where sensitive to curcumin we tried to test this in our conditions. We incubated a mitochondrial suspension with 25 μM of curcumin. This inhibited state 3 respiration (Figure 3a), uncoupled state 4 respiration and led to a drop in ΔΨ m in state 4 ( Figure 3b). We tested permeability transition pore (PTP) opening in presence of 25 μM curcumin (Figure 3c). Calcium induces the opening of the PTP in a similar fashion as ter-butylhydroperoxide and pore opening is inhibited by cyclosporin A (Figures 3c and d). The presence of 25 μM curcumin opened the PTP, but the initial mitochondrial shrinkage associated with calcium uptake was not observed and the curve of PTP opening was identical to that in the presence of 5 μM of ter-butylhydroperoxide. The minimal curcumin concentration that caused the PTP to open in isolated mitochondria was Z 2.5 μM (Figure 3d); higher concentrations promoted PTP opening and slightly modified the kinetics of opening (Figure 3d).
Curcumin treatment leads to mitochondrial swelling, cytochrome c release and the accumulation of autophagic vacuoles in vitro Although isolated mitochondria are sensitive to curcumin, it is unclear whether this is relevant to whole cells, because the amount of curcumin internalized may be insufficient to trigger PTP opening at the cellular level. Even at an external curcumin concentration of 25 μM, the intracytoplasmic concentration was only 1.5 μM and thus below the minimal concentration needed for    mitochondria was 0.8-1 μm, whereas after 24 h of curcumin treatment it was 1.5-2 μm. Treated cell mitochondria had a clear matrix and fewer cristae membranes than those in untreated cells. This was accompanied by a pronounced increase in the size of the endoplasmic reticulum (ER) lumen, which seemed to swell in parallel with mitochondria ( Figure 4a). Mitochondrial swelling was accompanied by the release of cytochrome c into the cytoplasm (inhibited by cyclosporin A (CsA); Figure 4b), EGTA-AM (a slow calcium chelator) and BAPTA-AM (a fast calcium chelator; Figure 4c). This indicates that intracytoplasmic calcium is involved in mitochondrial swelling and that the calcium pool is accessible to chelation (Figures 4b and c). Cells treated with 25 μM curcumin exhibit autophagosomes (AV) containing digested mitochondria and cytosolic fragments ( Figure 4d). These AV could be distinguished from classical vacuoles, which were also present in the cytoplasm but remained translucid.
Intracellular curcumin is associated with the ER and a subpopulation of lysosomes The effects of curcumin, including the drop in ΔΨ m and the production of ROS, do not appear to be consequences of its direct action on mitochondria. Indeed, the very early release of calcium into the cytoplasm following curcumin treatment (Figures 2e and f) prompted us to investigate potential interactions between curcumin and the ER, which contains the main cellular pool of free calcium.
We therefore used confocal microscopy and the mitochondrial dye, deep red, to analyze simultaneously curcumin fluorescence and the mitochondrial compartment ( Figures 5A-D). The mitochondrial network was clearly visible around nuclei and within the cytoplasm ( Figure 5A), and this analysis confirmed the punctated pattern of curcumin fluorescence ( Figure 5B). However, the curcumin signal did not colocalize with the mitochondria: curcumin fluorescent spots were close to, but did not overlap with, mitochondria ( Figure 5C). Transected images confirmed that the distribution of curcumin and mitochondria was different ( Figure 5C).
However, curcumin fluorescence colocalized perfectly with the ER, which was stained with ER red stain (ER-red; Figures

Curcumin induces autophagy followed by apoptosis
We observed that curcumin treatment leads to the formation of autophagic vacuoles containing degraded mitochondria (see above; Figure 4d). Given that dysfunctional mitochondria are usually subject to mitophagy, we examined the formation of autophagolysosomes by fluorescence microscopy following staining with the lysosomotropic agent acridine orange (AO). 19 Control cells showed mostly green fluorescence with very little cytoplasmic red staining corresponding to lysosomes; however, curcumin-treated cells displayed many spots of red fluorescence, indicating the formation and accumulation of acidic vesicles (Figures 6a and b). At 25 μM curcumin, AO red punctate staining was significantly higher than in control conditions ( Figure 6b). Consistent with the findings of the confocal image analysis (Figures 6a and b), flow cytometry analysis showed that autophagosome formation increased with curcumin concentrations until 25 μM (Figure 6d). However, at higher curcumin concentrations, that is Z 50 μM, the numbers of AO-positive cells ( Figure 6d) and the AO spots were lower (data not shown), suggesting that autophagy is inhibited and/or the vesicular system becomes more alkaline at these concentrations.
To examine further whether curcumin induces autophagy, we used western blotting to study the autophagy marker, LC3-II. 20 Curcumin stimulated the time-dependent production of the microtubule-associated protein light chain 3 (LC3-I), and the accumulation of its processed form (LC3-II; Figure 6e). A high LC3-II/LC3-I ratio is a marker of autophagy. Quantification of LC3-II and actin in cells revealed that LC3-II levels were fivefold higher following 24 h of treatment with 25 μM curcumin than in control conditions (Figure 6e). Huh-7 cells treated for 24 h with Rapamycin as a positive control showed high levels of LC3-II and those treated with Bafilomycin A1 as a negative control exhibited low levels of LC3-II (Figure 6e).
These findings indicate that curcumin may induce ER stress by promoting the formation of ubiquitinated misfolded proteins, leading to the activation of both autophagy and apoptosis. The autophagy inhibitor 3-methyladenine (3-MA) promotes the accumulation of ubiquitinated proteins in cells, thus exacerbating ER stress. We therefore used 3-MA to test whether ER stress has a role in curcumin-induced apoptosis in Huh-7 cells and whether the inhibition of autophagy stimulates curcumin-induced apoptosis by inducing ER stress. We treated cells with 3-MA and 25 μM curcumin for 24 h (Figure 6f) and examined apoptosis by annexin V-staining and cell viability by PI staining. The inhibition of autophagy by 3-MA indeed stimulated apoptosis (Figure 6f).

Curcumin-induced caspase activation
We examined the activity of several caspases during curcumininduced cell death (apoptosis and/or necrosis) (Figure 7a). Caspases-3/7, -9, -8 and -12 were all activated to different extents, and caspases-2 and -10 were not activated (Figure 7a). Given that the mitochondrial pathway is activated after destabilization of mitochondrial homeostasis and cytochrome c release (Figures 4b and c) caused by ER stress and high cytoplasmic calcium concentrations, it is not surprising that caspases-9 and -3/7 are activated following exposure to curcumin. Caspases-12 and -8 were activated to a lesser extent. Caspase-12 is activated in stress conditions (ROS production), although its precise function in this pathway is unclear. Caspase-8 activation may be related to lysosomal destabilization resulting from intracytoplasmic curcumin, which activates cathepsins. Indeed, cathespins activate caspase-8 and cleave Bid. Furthermore, Ca 2+ -dependent calpain activation may also stimulate the activity of caspase-8. caspase-9 and -3/7 was associated with high caspase-8 activity (Figure 7b). Thus, inhibition of mitochondrial cell death appears to exacerbate the alternative lysosomal pathway. These findings confirm that calcium is involved in curcumin-induced mitochondrial destabilization and the activation of caspase-9 and -3/7 via cytochrome c release. The cathepsin inhibitor cocktail had small effects on caspase-9 and -3/7 (Figure 7b), but clearly impaired caspase-8 activity, which implicates the cathepsin pathway in caspase-8 activation.
Next, we treated cells with 5 μM ruthenium red for 30 min and then with curcumin ( Figure 7b) and examined caspase-3 cleavage by western blotting. No caspase-3 cleavage was detected below 5 μM curcumin, although 5 μM curcumin was sufficient to induce calcium influx, a ΔΨ m drop and ROS production. This concentration may be enough to initiate early autophagic processes but not apoptosis. However, caspase-3 cleavage was detected at 20 μM curcumin but was stronger at 25 μM. Cleavage was completely inhibited by 5 μM ruthenium red.
We also tested whether proapoptotic signaling was intact by examining PARP cleavage (Figure 7c). PARP was efficiently cleaved at curcumin concentrations up to 25 μM (which induce apoptosis, Figure 2a) whereas at higher concentrations, which induce mainly necrosis, PARP was degraded, as observed during cytotoxic death (Figure 7d). Curcumin-induced apoptosis is also mediated via cathepsin and the calcium-dependent activation of calpain Since curcumin is taken up by a subpopulation of lysosomes (Supplementary Figures S1a and b), we examined cathepsin activity by analyzing the cleavage of Rh110-bis-CBZ-Lphe-L-Argamide using flow cytometry because lysosomal membrane permeabilization (LMP) is usually associated with cathepsin release (Figure 8a). Cathepsin activity was higher in curcumin-treated Huh-7 cells than in control cells and was inhibited by a cathepsin inhibitor cocktail (Figure 8a). Cathepsin activity increased with increasing curcumin concentrations (Figure 8b). We also tested whether cathepsin activity was affected by inhibition of the mitochondrial pathway. In cells treated with 25 μM curcumin, cathepsin activity was significantly higher in than presence of the caspase-9 inhibitor Z-LEHD-fmk than in control conditions (P o 0.05; Figure 8b). We used several cathepsin inhibitors to modulate caspase-8 activity (Figure 8c). Pepstatin A, Z-FF-fmk and the cathepsin D inhibitor efficiently abolished cathepsin activities and therefore inhibited caspase-8 activity whereas the cathepsin G inhibitor did not (Figure 8c). Given that the caspase-9 inhibitor Z-LEHD promoted caspase-8 activity, we tested whether inhibition of PTP with either CsA or bongkrekic acid (BA) had the same effect. Both CSA and BA promoted caspase-8 activity, probably because they block mitochondrial events and cytochrome c release.
Given that curcumin induces mitochondrial swelling through a calcium-dependent mechanism (Figures 2e and f), we investigated whether the calcium-dependent activation of calpain is involved  in this process (Figure 8d). Calpains were activated by curcumin and this effect was inhibited by calpain inhibitors. Next, we studied whether caspase inhibition influenced calpain activation and whether the ROS generated by mitochondria interfered with calpain activity. The caspase-9 inhibitor Z-LEHD-fmk stimulated slightly calpain activity, whereas MitoQ10 or SKQ1, which efficiently inhibit mitochondrial ROS production (Supplementary Figure S2) through mitochondrial dysfunction and PTP opening, did not interfere with calpain activity. By contrast, Ca 2+ chelation by BAPTA-AM inhibited calpain activity (Figure 8d).

Antioxidants and modulation of curcumin-induced cell death
We then investigated the effects of various antioxidants on curcumin-induced events. We used three antioxidants targeted to the cytoplasm, NAC, Vitamin E and Trolox, and two antioxidant targeted to mitochondria (mitoQ10 or SKQ1). Each of the three cytoplasmically targeted antioxidants almost completely abolished curcumin-induced ROS production (Supplementary Figure  S2 and Table 1). However, MitoQ10 and SKQ1, mitochondrially targeted antioxidants, were the most efficient on curcumininduced ROS production.

Uptake and localization of curcumin
Many studies have investigated the use of curcumin for therapeutic purposes, although few have quantified curcumin uptake or analyzed its intracellular localization. Descriptions of the intracellular localization of curcumin are poor 21 and in some cases, erroneous.
Here, found that curcumin is rapidly taken up by cells (Figure 1a, upper panel) in a 1/20 ratio if referred to the external concentration. Curcumin is very lipophillic, so it may accumulate at intracellular membranes, with deleterious effects. 22 Nevertheless, our findings conflict with previous reports showing that curcumin accumulates at the plasma membrane and in the nucleus. 21 We detected curcumin at the ER (Figures 5D and E) and lysosomes (Supplementary Figures S1a-c) level. Costaining experiments confirmed that the primary target of curcumin is not the mitochondrial compartment. A further analysis of curcumininduced apoptosis indicated that mitochondrial destabilization was an early event, which manifests as a ΔΨ m drop with ROS production. The calcium increased is almost in parallel with curcumin accumulation, and preceded ROS production and the ΔΨ m decrease (Figures 2e and f).

PTP opening and the induction of cell death by curcumin
Experiments with isolated purified mitochondria demonstrated that the PTP opens at curcumin concentrations above 2.5 μM (Figure 3d). Morin et al. and Ligeret et al. 23 report PTP opening at high concentrations of curcumin. However, it seems unlikely that this PTP opening is caused by the direct action of curcumin on mitochondria in vivo, except at very high doses (30 μM) at which curcumin is toxic and induces necrotic-like cell death. Confocal microscopy ( Figure 4) of cells stained with curcumin and a mitochondrial marker demonstrated that curcumin does not accumulate in mitochondria but is present nearby (Figures 5A-C). Thus, PTP opening is not the result of a direct interaction between internalized curcumin and mitochondria. EM pictures of Huh-7 cells revealed swollen mitochondria after 1 h of treatment with 25 μM curcumin, and big, swollen, translucent mitochondria with few cristae membranes after 24 h (Figure 4a). These swollen mitochondria were nearby huge empty vacuoles or autophagic vacuoles, and others were inside autophagosomes (mitophagy; Figure 4d).
Curcumin destabilizes the ER and promotes calcium release Confocal microscopy revealed that curcumin accumulates at the ER membrane ( Figures 5D-F). Electron microscopy showed a swollen appearance of the ER that is probably caused by intracellular ion imbalance eventually leading to osmotic lysis. Indeed, curcumin activates the unfolded protein response that leads to the the upregulation of C/EBP homologous protein and GADD34, the principal mediators of proapoptotic unfolded protein response. 24 Curcumin induces the classical ER stress pathway accompanied by increased expression of GP78 and C/EBP homologous protein. 25 Curcumin-mediated ER stress involves the sarcoplasmic/endoplasmic reticulum Ca 2+ -ATPase2. 25 Effectively, curcumin-induced ER stress provoke the release of calcium, which is rapidly taken up by the mitochondrial via the mitochondria calcium uniporter. The cytochrome c was released from swollen mitochondria, but CsA, BA, inhibitors of PTP opening, and the intracellular calcium chelators EGTA-AM and BAPTA-AM, prevented this process (Figures 4b and c). This implicates calcium in curcumin-induced PTP opening that leads to cytochrome c release and subsequent caspase activation (Figure 7b).
Curcumin colocalizes with a subpopulation of lysosomes and induces lysosomal membrane permeabilization Various death stimuli can induce this lysosomal pathway which is characterized by permeabilization and, possibly, partial rupture of the lysosomal membrane with the subsequent release of the intralysosomal content into the cytoplasm, including cathepsins, 26 calpains 27 and chemotrypsin. 28,29 Some internalized curcumin localizes in the lysosomal compartment (Supplementary Figure S1) where it induces a LMP content into the cytoplasm and activation of cathepsin (Figures 8a-c). When released into the cytosol, lysosomal proteases, in particular cathepsins B and L and aspartate cathespsin, can contribute to apoptotic processes either themselves or by acting together with caspases. 26,30 LMP may therefore be an early event of apoptosis. 31,32 We report that curcumin affects the lysosomal compartment. Curcumin treatment promoted the accumulation of AO into the cytosol. 33 Bafilomycin A1, a specific inhibitor of the lysosomal proton pump, completely abolished curcumin-induced AO red staining and thus protected lysosomes from permeabilization (Figure 6e). Abnormally high LMP has been reported in various types of tumor cells exposed to putative anti-cancer chemopreventive agents during apoptosis. 34,35 One report in particular suggests that curcumin induces LMP. 35 Curcumin may simply insert itself into the membrane bilayer, like cholesterol, 36 thus enhancing LMP. This may be associated with the release of calcium from the ER which may act synergistically with ROS: calcium activates calpains, which contribute to the disruption of the lysosomal membrane 37 and the formation of macroautophagosome, 38,39 and ROS destabilizes the lysosomal membrane (Figure 8d). Indeed, curcumin-induced LMP activated cathepsin L and D (Figure 8c). These observations are consistent with recent studies showing that the lysosomal translocation of cathepsin D in human retinoblastoma Y79 cells promotes mitochondrial membrane permeability thus resulting in apoptosis. 40 Furthermore, cathepsin B is abnormally active and results in similar events in pancreatic cancer cells. 41 We show that curcumin significantly stimulates cathepsin activity, which may affect the cleavage of Bid and the subsequent delocalization of Bax to the mitochondrial membrane followed by caspase-3 activation. At curcumin concentrations Z 50 μM, cathepsin activity was high although this was probably not the result of a deregulated signaling pathway but the indication that cathepsins actively participate in the dismantling of the cell during necrosis.
However, in our system, where the ER stress is a nearly event, it is possible that calcium release from the ER increases mitochondrial dysfunction, ROS production and cytochrome c release. ROS originating from destabilization of the electron transport chain help to destabilize the lysosomal compartment via peroxidation of membrane lipids. The NPe6 photosensitizer, when located in lysosomal membranes, 42 generates singlet oxygen upon photoactivation, leading to the fast release of cathepsins into the cytosol. Furthermore, apoptosis is initiated by UVA radiation and by H 2 O 2 , both of which trigger cell death mainly by generating oxidative stress associated with LMP. Other inducers of apoptosis, like vacuolar ATPase inhibitors 43 and N-(4-hydroxyphenyl) retinamide, 44 cause ROS-dependent LMP. One or more of these various mechanisms may be involved in curcumin-induced lysosomal destabilization. Our findings clearly confirm the importance of lysosomes in apoptotic processes: curcumin accumulated rapidly in lysosomes (Supplementary Figure S1) and ROS production was an early event inhibited by ROS scavengers like NAC, MnBTAP and more efficiently by mitochondrially targeted MitoQ10 or SKQ1. ROS inhibition by NAC significantly delayed curcumin-induced apoptosis. 45 Autophagy is an early event during curcumin-induced cell death Autophagy is an evolutionarily conserved, multistep lysosomal degradation process distinct from apoptosis in which a cell destroys long-lived proteins and damaged organelles. Autophagy is important for cell survival, and the inhibition of autophagy promotes cell death. However, autophagy can also accompany cell death. Curcumin treatment promotes autophagy before signs of cell death manifest, as evidenced by images of autophagosome formation, AO staining (Figures 6a-c) and LC3II formation assessed by western blotting (Figure 6e). Our results are consistent with the recent publication showing that autophagy is involved when the curcumin derivative bis-dehydroxycurcumin is used on colon cancer cells. 46 Evidence now exists of complex interplay between autophagy and apoptosis. Notably, many genes, such as p53 and Bcl-2 family members, are shared by these two pathways.
The accumulation of poly-ubiquitinated proteins and the induction of ER stress promote autophagy, 47 which in turn may trigger mitochondrial-dependent apoptosis. Pharmacological and RNAi-mediated inhibition of ER-stress and autophagy has revealed that autophagy potentiates the anti-proliferative effect of curcumin. 25 Our findings demonstrate that curcumin is a proautophagic cytotoxic drug, providing further evidence of its therapeutic potential against developing cancer cells.
In conclusion, we report for the first time that the ER and the lysosomal compartment act in synergy to induce apoptosis and cell death in curcumin-treated Huh-7 cells. Moreover, ER stress is also involved in the establishment of mitochondrially linked autophagic processes that precede apoptosis and necrosis. These early events were almost completely blocked by the calcium chelators EGTA-AM and BAPTA-AM, and inhibitors of the mitochondrially generated ROS species, that is, MitoQ10 and SKQ1 (Supplementary Figure 2 and Table 1). Curcumin destabilizes the ER leading to calcium release into the cytosol, which links primary ER-stress-related events with classical mitochondrially induced apoptosis. We also found that curcumin triggers the release of lysosomal cathepsins into the cytosol. These events are blocked by NAC, which shows that lysosomal destabilization is linked to ROS production, although the accumulation/insertion of curcumin in the lysosomal membrane per se may explain LMP and the activation of cathepsin/chemotrypsin. Apoptosis/necrosis were markedly impaired by pre-treatment of cells with NAC or broad spectrum caspase inhibitors. The increase in MMP was partly prevented by cathepsin inhibitors, suggesting that lysosomal destabilization is upstream from mitochondrial destabilization.
We propose that curcumin mediates cell apoptosis through lysosomal destabilization. Upon entry into the cell, curcumin destabilizes the lysosomal membrane, triggering the release of lysosomal enzymes into the cytosol, which in turn induces MMP via Bid cleavage and Bax delocalization, resulting in mitochondrial swelling, cytochrome c release and ROS production. ROS then further destabilize lysosomal membrane integrity, thus enhancing LMP in curcumin-treated cells.
These events are summarized in Figure 9. Curcumin inserts into the ER and lysosomal membrane. Calcium is rapidly released as a result of ER stress, and strongly promotes mitochondrial destabilization, involving the production of superoxide anions and H 2 O 2 , mitochondrial membrane alterations and bioenergetic changes. These events promote MMP and cytochrome c release followed by the activation of executive caspases. However, at low curcumin concentrations and early during this sequence of events, curcumin induces autophagic 'survival' processes, which attempt to discard nonfunctional mitochondria and save cells from further destruction. However, if the initial crosstalk between apoptosis and autophagy is unsuccessful, a third pathway leads to cell death. In this case, LMP results in the release of cathepsins which activate caspase-8 and -3/9, to guarantee the destruction of cells.
Our findings suggest that an intricate crosstalk between apoptosis and autophagy determines the overall fate of the cell. Indeed, the final outcome of autophagy depends on: (i) the stressinducing stimulus and (ii) the cellular context. Indeed, autophagy can help cells undergoing ER stress to survive, by eliminating unfolded proteins, or it can lead to cell death. 46,48 Taken together these results imply a fail-secure mechanism regulated by autophagy in the action of curcumin, suggesting a therapeutic potential for curcumin. Then offering a novel and effective strategy for the treatment of malignant cells.

Confocal microscopy
Cells were seeded onto glass coverslips at a concentration of~5 × 10 5 cells per well in six-well plates. Cells were treated with 25 μM curcumin for 24 h in Dulbecco's modified Eagle's medium without phenol red and supplemented with 3% charcoal-treated calf serum. For immunofluorescence, all steps were carried out at room temperature. The coverslips were washed twice in 1 × phosphate-buffered saline (PBS). Mounted cells were observed and images were recorded using a Zeiss LSM 510 confocal microscope (Carl Zeiss Meditec France SAS, Le Pecq, France) with a 40X Plan-Neofluar 1.3 numerical aperture oil objective and LSM Image Browser, or on a Nikon Eclipse TE-2000 E microscope (Nikon France S.A, Champigny sur Marne, France). Deconvolution and threedimensional reconstitution were performed with Autoquant imaging Autodeblur version X 1.4.1 (AutoDebur and Autovisualize; Mediacybernetics, Bethesda, MD, USA). ImageJ 1.37v software was used for images analysis (imagej.nih.gov/ij).
Determination of ΔΨ m , ROS and cytosolic Ca 2+ levels Huh-7 cells at a density of 2 × 10 6 in six-well plates were treated with 0-100 μM of curcumin for 0 to 48 h as indicated. After treatment, cells were trypsinized, harvested, washed and resuspended, with their supernatant in calcium-free PBS (pH 7.2). To determine ΔΨ m , tetramethyl rosamine methyl ester or 3,3'-dihexyloxacarbocyanine iodide were added to a final concentration of 40 nM. To determine superoxide anion and H 2 O 2 concentrations, 1 μM hydroethidine and 5 μM 2′,7′-dichlorodihydrofluorescein diacetate were used, respectively. The following antioxidants were used: a mitochondria-targeted antioxidant, Mito-Q 10 (500 nM; the coenzyme-Q analog attached to a triphenylphosphonium cation targets the antioxidant to the mitochondrial matrix), SKQ1 (plastoquinonyl-decyl-triphenylphosphonium), Vitamin E (1 mM) and 6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid (Trolox; 1 mM), an artificial vitamin E analog and NAC (5 mM). We tested the effects of ROS scavengers (NAC, Trolox, MitoQ10 and SkQ1) on Huh-7 cells treated with curcumin. Huh-7 cells were plated at~2 × 10 6 cells/well in six-well plates and were incubated for 3 h with or without antioxidants. Curcumin (25 μM) was then added and cell viability and ΔΨ m were determined. Fluo3-AM (stock solution: 3 mg/ml) was used to determine cytosolic concentrations of calcium. EGTA-AM was added to cell cultures at 25 μM and the samples maintained in the dark for 30 min at 37°C before measurements.
Double staining was used to examine cell viability involving propidium iodide (PI; 1 mg/ml) with 3,3'-dihexyloxacarbocyanine iodide, dichlorofluorescein diacetate and Fluo4-AM, or TO-PRO-3 iodide (1 mg/ml) with hydroethine and tetramethyl rosamine methyl ester. Double staining with YO-PRO-1/PI (Molecular Probes) and Annexin-V/PI staining was also used to distinguish between viable, apoptotic and necrotic cells. Annnexin-V FITC (Immunotech, Beckman-Coulter, Beckman Coulter International S.A., Nyon, Switzerland) was used to detect the aberrant exposure of phosphatidyl serine residues on the outer surface of the plasma membrane. All samples were analyzed by flow cytometry as described previously 49 Determination of NAD(P)H autofluorescence NAD(P)H fluorescence was measured under illumination with a multiline ultraviolet light at 400 mW; the light emitted was detected with a 424 ± 40 nm bandpass to select NAD(P)H fluorescence. 50 Changes in the auto-fluorescence of control and curcumin-treated cells were recorded. Q-VD-OPh Figure 9. Schematic representation of curcumin effects. Curcumin mainly targets the ER and lysosomes but to a different extent. The classic apoptotic pathway is mediated by calcium release from the ER. Uptake of this calcium by mitochondria disrupts mitochondrial homeostasis. Calcium alters mitochondrial electron transport causing substantial ROS production (both superoxide anions and hydrogen peroxide), which leads to the opening of the permeability transition pore in the mitochondrial membrane. Consequently, cytochrome c is released and the caspase-9 and -3/7 pathway is activated leading to cell death (mostly by apoptosis). Furthermore, the ER stress pathway leads to the formation of autophagic vacuoles that attempt to eliminate the dysfunctional mitochondria. The cleavage of Beclin-1 is associated with apoptosis and leads to the accumulation of autophagic vacuoles. Despite the activation of autophagy, cells undergo necrotic cell death following these initial apoptotic events. A lysosomal pathway may also be active. Curcumin destabilizes lysosomal membranes leading to lysosomal membrane permeability and the activation of both cathepsins and chemotrypsins. Activated caspase-8 leads to Beclin-1 cleavage. The increase in cytosolic calcium concentrations activates calpains that contribute to the degradation process and accelerate cell death. The various inhibitors used in this work are indicated with the pathways they affect. Curcumin interacts with the ER and lysosomes.