Main

Autophagy is a lysosomal degradation process characterized by the formation of double-membraned autophagosomes. During autophagy, cytoplasmic material and organelles are sequestered by an isolation membrane derived from the preautophagosomal structure. Fusion of this double-membraned vesicle creates an autophagosome, which ultimately fuses with a lysosome. Within the resulting autophagolysosome, the sequestered cytoplasmic materials are degraded by acidic hydrolases. Autophagic degradation generates amino acids and fatty acids that can be used for protein synthesis or oxidized through the mitochondrial electron transport chain (mETC) to produce ATP for cell survival under starvation conditions.1 However when autophagy is prolonged, proteins and organelles essential for basic homeostasis and cell survival are degraded, which can lead to cell death (autophagic cell death or programmed cell death type II, PCD II).2 Therefore, autophagy is both a cell survival mechanism and a cell death pathway depending on the extent of cellular degradation.

Reactive oxygen species (ROS) are molecules or ions that are formed by the incomplete one-electron reduction of oxygen. The major species of ROS include superoxide (O2•−), hydrogen peroxide (H2O2), and hydroxyl radical (•OH).3 These three major ROS are mainly produced through a chain reaction as follows: O2•−, which is mainly produced from mETC complexes I and III, is converted to H2O2 by the superoxide dismutase (SOD) family of enzymes, which mainly include SOD1 (copper-zinc superoxide dismutase, Cu, Zn-SOD) and SOD2 (manganese superoxide dismutase, Mn-SOD). H2O2 can then be converted to •OH by ferrous iron (Fe2+) or copper ion (Cu+), or it can be catalyzed to H2O by catalase, glutathione peroxidase (GPx) and peroxiredoxin III (PrxIII).3 Thus, the levels of O2•−, H2O2, and •OH are constantly in flux until an equilibrium is established.

It is well known that ROS can regulate apoptosis (PCD type I).4 ROS regulation of autophagy has also been demonstrated in many reports.5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22 Most of these studies have implicated ROS in autophagy induction by using nonspecific ROS scavengers or exogenous H2O2.5, 7, 8, 9, 12, 14, 21 Our previous studies suggest that O2•− specifically mediates autophagy induced by mETC inhibitors (complex I inhibitor rotenone and complex II inhibitor 2-thenoyltrifluoroacetone (TTFA)), the SOD inhibitor 2-methoxyestradiol (2-ME),23 and exogenous H2O2.3, 17, 18 Kim et al.13 have demonstrated that sodium-selenite-induced autophagy is also regulated by O2•−. In contrast, several reports have specifically implicated H2O2 in regulating autophagy.11, 16, 19, 20, 22 Thus, it seems that both O2•− and H2O2 are involved, however none of the previous studies has investigated the relative functions of both species in the regulation of autophagy. Thus, the ROS responsible for regulating autophagy remains unclear.

In this study we have investigated the relative functions of O2•− and H2O2 in autophagy induced by starvation, mETC inhibitors, and exogenous H2O2. We found that the levels of autophagy induced by starvation, mETC inhibition, and exogenous H2O2 correlated with increased O2•− levels and decreased H2O2 levels. Furthermore, starvation-induced cell death was increased by the inhibition of autophagy with autophagy inhibitors and O2•− scavengers but decreased by the enhancement of autophagy by knocking down SOD2 expression. In contrast, cell death induced by mETC inhibitors was elevated by the enhancement of autophagy with SOD2 inhibition. These results give significant insights into the regulation of autophagy.

Results

Starvation induces ROS generation

Several reports have suggested that starvation-induced autophagy is mediated by ROS.5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22 In this study, we investigated the generation of O2•− and H2O2 under two starvation conditions: starvation of glucose, L-glutamine, pyruvate, and serum (GP), and starvation of amino acids and serum (AA). Cellular levels of O2•− and H2O2 were measured by flow cytometry using specific fluorescent dyes (see Materials and Methods). Figure 1 shows that over a 72-h time course in HeLa cells, GP or AA starvation increased O2•− levels to 50% compared with 10% in control conditions (also Supplementary Figure S1). GP starvation did not induce H2O2 generation (Figure 1a) whereas AA starvation induced H2O2 generation up to 80% (Figure 1b; Supplementary Figure S1). Generation of O2•− and H2O2 was not significant before 6 h of starvation. Similar results were also obtained in U87 cells (Supplementary Figure S2) and HEK293 cells (data not shown). Thus, both GP and AA starvation increased O2•− levels but only AA starvation increased H2O2 levels.

Figure 1
figure 1

Starvation induces ROS generation in HeLa cells over a 72-h time course. (a) Starvation of GP (glucose, L-glutamine, pyruvate, and serum). (b) Starvation of AA (amino acids and serum). Intracellular O2•− and H2O2 were measured by flow cytometry after cells were stained with DHE (dihydroethidium) and CM-H2DCFDA (5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate acetyl ester), respectively, as stated in Materials and Methods section. Because DHE is oxidized to ethidium by O2•− and CM-H2DCFDA is oxidized to DCF (dichlorofluorescein) by H2O2, the percentages of ethidium fluorescence and DCF fluorescence represent the levels of intracellular O2•− and H2O2, respectively. Error bars represent standard deviation (S.D.) from three independent duplicate experiments

Full size image

Starvation induces autophagy

It is well known that starvation can induce autophagy.1 In this study, we detected starvation-induced autophagy using three methods as in our previous studies.17, 18, 24 We transiently transfected cells with GFP-LC3, which is recruited to autophagosome membranes during autophagy resulting in a punctate distribution that can be visualized using a fluorescent microscope. We measured the formation of acidic vesicular organelles (AVOs, which include autolysosomes) by flow cytometry using the pH-sensitive fluorescent dye acridine orange (AO). Finally, we used western blotting to detect the conversion of cytosolic LC3-I to lipidated, autophagosome-membrane-bound LC3-II. Because LC3-II is continually degraded within autophagolysosomes during autophagy, it is recommended that cells be treated with a lysosomal inhibitor before western blotting for LC3-II to demonstrate autophagic flux.25, 26 In the presence of the lysosomal inhibitor NH4Cl, GP or AA starvation in HeLa cells induced elevated levels of LC3-II at 24 and 48 h but not at 1 and 6 h (Figure 2a), correlating with ROS generation (Figure 1). Similar phenomena were also observed in U87 cells (Supplementary Figure S3). Starvation of GP or AA also induced the formation of GFP-LC3 puncta (Figure 2b) and AVOs (Figure 2c) in HeLa cells. Elevated amount of starvation-induced LC3-II and microtubule-associated protein 1 light chain 3 (LC3) puncta could be observed as early as at 24 h (Figure 2a and b) whereas elevated amount of starvation-induced AVOs could only be seen after 48 h (Figure 2c). This supports the current hypothesis that the formation of autolysosomes (which belong to AVOs) lags behind the formation of autophagosomes.1 In the presence of 3-methyladenine (3-MA, a chemical inhibitor of autophagy), starvation-induced LC3-II formation was significantly reduced (Figure 2d). 3-MA treatment also decreased starvation-induced formation of GFP-LC3 puncta and AVOs in HeLa cells: the percentage of GP or AA starved cells with GFP-LC3 puncta was reduced from 40 to 14%, whereas the percentage of cells with AVOs decreased from 33 to 2% (GP starvation), and from 18 to 2% (AA starvation) (Supplementary Figure S4a). Wortmannin, another autophagy inhibitor, decreased the percentage of cells with AVOs from 45 to 11% with GP starvation and from 37 to 6% with AA starvation (Supplementary Figure S4bi). Furthermore, when the autophagy genes beclin-1 and atg-7 were knocked down by siRNA in HeLa cells (Figure 2ei), starvation-induced LC3-II formation was reduced (Figure 2eii), and the percentage of cells with AVOs was reduced from 30 to 9% with GP starvation and from 23 to 2% with AA starvation (Supplementary Figure S4c). GFP-LC3 puncta was also reduced from 40 to 15% with GP starvation and from 34 to 13% with AA starvation (Supplementary Figure S4c). Therefore, both GP and AA starvation conditions induce autophagy to similar extents.

Figure 2
figure 2figure 2

Starvation induces autophagy and cell death in HeLa cells. (a) Formation of LC3-II after cells were starved of GP or AA for 1, 6, 24, and 48 h. (b) Representative fluorescent microscopic figures of the formation of GFP-LC3 dots (puncta) after cells were transfected with GFP-LC3 cDNA and starved of GP or AA for 1, 6, 24, and 48 h. GFP-LC3 dots were not observed when cells were transfected with GFP cDNA (data not shown). Scale bars=10 μm. (c) Representative histographs of AVOs formation analyzed by flow cytometry after cells were starved of GP or AA for 1, 6, 24, 48, and 72 h. The gated region (M1) in each histogram represents the percentage of cells with AVOs. (d) Effect of autophagy inhibitor 3-MA (2 mM) on formation of LC3-II after cells were starved of GP or AA for 48 h. (e) Effect of beclin-1 and atg-7 siRNAs on formation of LC3-II after cells were starved of GP or AA for 48 h. (i) Expression of beclin-1 and ATG-7 after cells were transfected with control, beclin-1 and atg-7 siRNAs. (ii) Western blot showing formation of LC3-II with β-actin used as a loading control. Cells were treated with the lysosomal inhibitor NH4Cl at 30 mM for the determination of autophagy flux. (f) Cell death induced by GP or AA starvation. Cell death was analyzed by measuring the plasma membrane permeability by flow cytometry using Trypan blue to stain cells as described in Materials and methods section. The percentage of plasma membrane permeability represents the percentage of cell death. (i) Cell death induced by GP and AA starvation in a 72-h time course. (ii) Effect of 3-MA (2 mM) on GP or AA starvation-induced cell death at 24, 48, and 72 h. (iii) Effect of beclin-1 and atg-7 siRNAs on GP or AA starvation-induced cell death at 72 h. All data are representative of three independent experiments

Full size image

Under starvation conditions, autophagy is normally considered to be a cell survival mechanism.1 To test this hypothesis, we investigated the effect of autophagy inhibition on starvation-induced cell death in HeLa cells. Figure 2fi shows that GP starvation induced cell death up to 25% and AA starvation induced cell death only to 13% compared to 7% of cell death with non-starvation treatment after 72 h. When autophagy inhibitor 3-MA was added, GP starvation-induced cell death was elevated to 40% and AA starvation-induced cell death was elevated to 30% (Figure 2fii). When another autophagy inhibitor wortmannin was added in HeLa cells, GP starvation-induced cell death was increased to 36% and AA starvation-induced cell death was increased to 43% (Supplementary Figure S4bii). GP or AA starvation-induced cell death was also significantly enhanced at 24 and 48 h (Supplementary Figure S4bii). When the autophagy genes beclin-1 and atg-7 were knocked down by siRNA in HeLa cells (Figure 2ei), GP starvation-induced cell death was increased from 29 to 41% and 41%, respectively, and AA starvation-induced cell death was increased from 15 to 30% and 23%, respectively (Figure 2fiii). This suggests that autophagy is a cell survival mechanism under GP and AA starvation conditions similar to other published results.1

Starvation-induced autophagy is mediated by O2•−

Because both GP and AA starvation induced autophagy and O2•− production, and as H2O2 production was absent during GP starvation (Figure 1; Supplementary S1, and S2), we hypothesized that starvation-induced autophagy is specifically mediated by O2•−. The production of O2•− and H2O2 is a chain reaction, where O2•− is converted to H2O2 by SOD, and H2O2 can either be converted to H2O by other antioxidant enzymes (such as catalase, GPx, and PrxIII), or to •OH by Fe2+ or Cu+.3 Therefore, the levels of O2•− and H2O2 can be adjusted by altering the antioxidant enzymes mentioned above. To determine the relative functions of O2•− and H2O2 in the regulation of starvation-induced autophagy, we manipulated the cellular levels of these two ROS species by modifying the level of SOD2 through overexpression and siRNA knockdown or by treating cells with exogenous catalase. In Figure 3, ROS generation and autophagy was determined in wild-type (wt) HeLa cells or HeLa cells with stable overexpression of SOD2 (SOD2) in the mitochondria (Figure 3a; Supplementary Figure S5). Because SOD2 catalyzes the conversion of O2•− to H2O2, we expected SOD2 cells to exhibit decreased O2•− levels and increased H2O2 levels. Compared to levels in the wt cells, both GP starvation-induced autophagy and O2•− levels in SOD2 cells were significantly decreased. In addition, the percentage of cells with AVOs and GFP-LC3 puncta decreased from 22 to 1% and from 32 to 13%, respectively, whereas O2•− generation was reduced from 30 to 8% in SOD2 cells (Figure 3b). As expected, GP starvation-induced H2O2 generation was significantly increased by SOD2 overexpression, from 7 to 20% (Figure 3b). Similarly, AA starvation-induced autophagy and O2•− generation were significantly decreased in SOD2 cells compared with that in wt cells. The percentage of cells with AVOs and GFP-LC3 puncta decreased significantly from 18 to 1% and from 38 to 13%, respectively, whereas O2•− generation was decreased from 33 to 4% in SOD2 cells (Figure 3c). AA starvation-induced H2O2 levels were 32% in both wt and SOD2 cells (Figure 3c). GP and AA starvation-induced LC3-II protein levels were also decreased in SOD2 cells compared to that in wt cells (Figure 3d). Because the inhibition of autophagy by 3-MA, wortmannin, and siRNAs against beclin-1 and atg-7 increased GP and AA starvation-induced cell death (Figures 2f; Supplementary Figure S4bii), it is expected that SOD2 overexpression would also elevate GP and AA starvation-induced cell death. This is demonstrated in Figure 3e. In the wt HeLa cells, starvation of GP or AA did not significantly increase cell death compared to control whereas in SOD2 HeLa cells, GP starvation increased cell death from 13 to 45%. Similarly, AA starvation increased cell death from 11 to 29% compared to that in wt cells at 48 h. After 72 h, SOD2 overexpression in cells increased GP starvation-induced cell death from 25 to 62% and increased AA starvation-induced cell death from 13 to 50% (Figure 3e). When autophagy inhibitors 3-MA (Figure 3f) and wortmannin (Supplementary Figure S4biii) were added, cell death induced by GP or AA starvation in SOD2 HeLa cells was not significantly affected. This indicates that when GP and AA starvation-induced autophagy was reduced to a very low level by O2•− scavenger SOD2 (Figure 3b, c, and d), addition of other autophagy inhibitors (3-MA, wortmannin) would not further increase the levels of cell death induced by GP or AA starvation.

Figure 3
figure 3

Overexpression of SOD2 downregulates autophagy but upregulates cell death induced by GP or AA starvation in HeLa cells. (a) Comparison of SOD2 expression between the wild-type (wt) and SOD2 overexpressing (SOD2) HeLa cells. β-Actin was used as a loading control. (b and c) Effect of SOD2 overexpression on ROS generation, formation of AVOs, and GFP-LC3 dots (puncta) induced by GP and AA starvation, respectively. The percentages of ethidium fluorescence and DCF fluorescence represent the levels of intracellular O2•− and H2O2, respectively, as described in Figure 1. (d) Effect of SOD2 overexpression on formation of LC3-II after cells were starved of GP or AA for 48 h. β-Actin was used as a loading control. Cells were treated with the lysosomal inhibitor NH4Cl at 30 mM for the determination of autophagy flux. (e) Effect of SOD2 overexpression on GP or AA starvation-induced cell death at 24, 48, and 72 h. (f) Effect of 3-MA (2 mM) on GP or AA starvation-induced cell death in SOD2 overexpression cells at 24, 48, and 72 h. The percentage of plasma membrane permeability represents the percentage of cell death as described in Figure 2. Error bars represent standard deviation (S.D.). All data are representative of three independent experiments

Full size image

Next, we used siRNA to knock down the expression of SOD2 in HeLa cells (Figure 4a). SOD2 knockdown increased starvation-induced O2•− levels from 60 to 80% (GP starvation) and from 40 to 74% (AA starvation) compared to cells transfected with control siRNA (Figure 4b). In addition, SOD2 knockdown significantly decreased AA starvation-induced H2O2 levels from 24 to 12% whereas GP starvation-induced H2O2 was not affected (Figure 4b), because no significant induction was observed (Figures 1a and 4b). Along with increased O2•− levels, SOD2 knockdown cells had higher levels of GP and AA starvation-induced autophagy: the percentage of cells with AVOs was increased from 31 to 50% and from 23 to 38%, respectively, whereas the percentage of cells with GFP-LC3 puncta was increased from 40 to 54% and from 34 to 55%, respectively (Figure 4b). LC3-II protein levels following GP and AA starvation were also significantly increased by SOD2 knockdown (Figure 4c). Thus, SOD2 knockdown cells had increased O2•− (with decreased or unchanged H2O2), and increased starvation-induced autophagy. In contrast to SOD2 overexpression (Figure 3e), SOD2 knockdown decreased GP starvation-induced cell death from 29 to 16% and AA starvation-induced cell death from 15 to 7% at 72 h (Figure 4d). This indicates that when starvation-induced autophagy is enhanced by SOD2 knockdown, starvation-induced cell death would be reduced, agreeing with the cell survival mechanism of autophagy under starvation conditions. Taken together, these data show that O2•− levels regulated by SOD2 are strongly correlated with starvation-induced autophagy.

Figure 4
figure 4

siRNA knockdown of sod-2 upregulates autophagy but downregulates cell death induced by GP or AA starvation in HeLa cells. (a) Western blot showing that SOD2 expression was reduced by sod-2 siRNA. β-Actin was used as a loading control. (b) Effect of sod-2 siRNA on ROS generation, formation of AVOs, and GFP-LC3 dots (puncta). The percentages of ethidium fluorescence and DCF fluorescence represent the levels of intracellular O2•− and H2O2, respectively, as described in Figure 1. (c) Effect of sod-2 siRNA on formation of LC3-II. β-Actin was used as a loading control. Cells were treated with the lysosomal inhibitor NH4Cl at 30 mM for the determination of autophagy flux. (d) Effect of sod-2 siRNA on cell death. The percentage of plasma membrane permeability represents the percentage of cell death as described in Figure 2. Error bars represent standard deviation (S.D.). All data are representative of three independent experiments

Full size image

Others have shown that exogenous catalase enzyme can be added to cells extracellularly to scavenge intracellularly produced H2O2.8, 15, 22, 27 Because cellular H2O2 is generated by a chain reaction from O2•−, catalase can also indirectly reduce cellular O2•− levels.3 The three major cellular ROS species, O2•−, H2O2, and •OH, are in flux through a chain reaction.2, 3 O2•− that is mainly produced from mETC complexes I and III will be catalyzed to H2O2 by SOD. Then H2O2 can either be converted to •OH by Fe2+ or Cu+ or be catalyzed to H2O by catalase, GPx, and PrxIII. Therefore, we treated HeLa cells with catalase and compared autophagy, ROS induction, and cell death under control and starvation conditions (Figure 5). We found that starvation-induced O2•− was reduced from 52 to 5% (GP starvation) or from 54 to 10% (AA starvation), whereas H2O2 was reduced from 43 to 9% (AA starvation) after catalase treatment (Figure 5a). Although GP starvation did not significantly induce H2O2 production compared to control, the background level of H2O2 was reduced by catalase (Figure 5a). In addition to decreased ROS levels, catalase-treated cells exhibited decreased starvation-induced autophagy. The percentage of cells with AVOs was decreased from 31 to 1% (GP starvation) or from 18 to 1% (AA starvation), and the percentage of cells with GFP-LC3 puncta was decreased from 40 to 11% (GP starvation) or from 38 to 11% (AA starvation) after catalase treatment (Figure 5a). Starvation-induced LC3-II protein level was also significantly reduced by catalase (Figure 5b). Cell death induced by GP and AA starvation was increased from 25 to 60% and from 13 to 39%, respectively, at 72 h of catalase treatment (Figure 5c). When another ROS scavenger, N-acetyl-cysteine (NAC) was used in HeLa cells, autophagy induced by GP or AA starvation was significantly reduced and cell death was increased (data not shown). Thus, catalase treatments lowered both O2•− and H2O2 levels and reduced starvation-induced autophagy but increased starvation-induced cell death.

Figure 5
figure 5

Catalase downregulates autophagy but upregulates cell death induced by GP or AA starvation in HeLa cells. (a) Effect of catalase (1000 U/ml) on ROS generation, formation of AVOs, and GFP-LC3 dots (puncta). The percentages of ethidium fluorescence and DCF fluorescence represent the levels of intracellular O2•− and H2O2, respectively, as described in Figure 1. (b) Effect of catalase (1000 U/ml) on formation of LC3-II at 48 h. β-Actin was used as a loading control. Cells were treated with the lysosomal inhibitor NH4Cl at 30 mM for the determination of autophagy flux. (c) Effect of catalase (1000 U/ml) on cell death at 24, 48, and 72 h. The percentage of plasma membrane permeability represents the percentage of cell death as described in Figure 2. Error bars represent standard deviation (S.D.). All data are representative of three independent experiments

Full size image

Autophagy induced by mETC inhibitors is mediated by O2•−

Besides starvation, many other stimuli are known to induce autophagy.1, 2 We sought to determine whether O2•− mediation of autophagy is specific to starvation conditions, or if it is regulating autophagy through other stimuli. Our previous studies indicate that mETC inhibitors (including complex I inhibitor rotenone, complex II inhibitor TTFA, and complex I/SOD inhibitor 2-ME23, 28) can induce ROS-mediated autophagy and autophagic cell death in cancer cells.17, 18 In these studies, SOD2 overexpression reduced autophagy, supporting the involvement of O2•− in autophagy induction. However, by measuring ROS only with dihydroethidium (DHE), which is reported to detect O2•− specifically,27 we failed to directly address the function of H2O2. Therefore, in the present study we investigated the relative functions of O2•− and H2O2 on autophagy induced by these stimuli. Treatment with rotenone or TTFA causes increased production of O2•−, which can be converted to H2O2 by SOD.3, 17 Because rotenone and TTFA can induce O2•− production at as early as 0.5 h, formation of GFP-LC3 puncta at 16 h, AVOs formation and cell death at 24 h,17 we measured ROS at 6 and 24 h, formation of GFP-LC3 puncta and AVOs at 24 h, and cell death at 24 h, in the present study. When combined with 2-ME, the levels of O2•− induced by rotenone and TTFA were significantly increased from 46 to 57% and from 37 to 54%, respectively, in HeLa cells at 24 h (Figure 6a). Accordingly, 2-ME reduced the levels of H2O2 production from 59 to 38% following rotenone treatment and from 39 to 25% following TTFA treatment, in HeLa cells at 24 h (Figure 6a). Similar results were also obtained when HeLa cells were treated for 6 h (Supplementary Figure S6). In correlation with the increase in O2•− levels, the percentage of cells with GFP-LC3 puncta induced by rotenone and TTFA was also increased from 20 to 39% and from 22 to 34%, respectively, by the addition of 2-ME in HeLa cells at 24 h (Figure 6b). The protein level of LC3-II induced by rotenone or TTFA was also significantly increased with the addition of 2-ME (Figure 6c). Similar results were also obtained in HEK293 cells (Supplementary Figure S7). In HEK293 cells treated with rotenone or TTFA, the levels of O2•− production and autophagy (formation of green fluorescent protein (GFP) puncta and LC3-II) were both increased by the addition of 2-ME. Notably, H2O2 production was not significantly induced by rotenone, TTFA or 2-ME, alone or in combination (Supplementary Figure S7). Taken together, these data suggest that similar to starvation, autophagy induced by mETC inhibitors is regulated by O2•−. In contrast to that starvation-induced autophagy reducing cell death, our previous study showed that rotenone-, TTFA-, or 2-ME-induced autophagy increases cell death.17, 18 In this study, cell death induced by the combination of rotenone and 2-ME or TTFA and 2-ME was also investigated. Figure 6d shows that cell death induced by rotenone and TTFA was increased from 20 to 30% and from 20 to 40%, respectively, following addition of addition of 2-ME. In contrast, 2-ME alone induced cell death of 13% compared to cell death of 8% in control at 24 h. Thus, different from that of starvation-induced autophagy, the enhancement of rotenone- or TTFA-induced autophagy (by 2-ME) leads to the enhancement of cell death.

Figure 6
figure 6

SOD inhibitor 2-ME upregulates autophagy and cell death induced by mETC inhibitors rotenone and TTFA in HeLa cells. Cells were treated with 0.1 mM 2-ME and/or 50 μM rotenone (R) and/or 0.5 mM TTFA (T) for 24 h. (a) ROS generation. The percentages of ethidium fluorescence and DCF fluorescence represent the levels of intracellular O2•− and H2O2, respectively, as described in Figure 1. (b) Formation of GFP-LC3 dots (puncta). (c) Formation of LC3-II. β-Actin was used as a loading control. Cells were treated with the lysosomal inhibitor NH4Cl at 30 mM for the determination of autophagy flux. Numerical values indicate protein quantification by densitometry using QuantityOne software (Bio-Rad), normalized to β-actin. (d) Cell death. The percentage of plasma membrane permeability represents the percentage of cell death as described in Figure 2. Error bars represent standard deviation (S.D.). All data are representative of three independent experiments

Full size image

Autophagy induced by exogenous H2O2 is mediated by O2•−

Exogenous H2O2 is another stimulus known to induce autophagy.7, 9, 12, 13, 14, 15, 16, 21, 29, 30, 31 In many previous in vitro studies, addition of H2O2 has been used to mimic the effects of intracellularly produced H2O2, without actually measuring intracellular H2O2 because of the assumption that H2O2 can freely diffuse through biological membranes.7, 9, 12, 13, 14, 15, 16, 21, 29, 30, 31 Indeed, our previous study showed that addition of exogenous H2O2 to HEK293, U87, and HeLa cells induced ROS production without directly measuring intracellular levels of H2O2.18 In the current study, we investigated the functions of both intracellular O2•− and H2O2 on autophagy induced by exogenous H2O2. We found that the addition of exogenous H2O2 induced the production of intracellular O2•− but failed to increase intracellular H2O2 in HeLa cells (Figure 7a). Similar results were obtained in HEK293 cells: exogenous H2O2 induced O2•− production (Supplementary Figure S8a), formation of LC3-II (Supplementary Figure S8b) but failed to increase the production of intracellular H2O2 (Supplementary data Figure S8a). When SOD2 was overexpressed in HeLa cells (Figure 3a), O2•− production was greatly reduced from 77 to 10% whereas intracellular H2O2 production was significantly increased from 1 to 17% following treatment with exogenous H2O2 (Figure 7a). Along with decreasing intracellular O2•− levels, SOD2 overexpression significantly decreased the percentage of cells with AVOs from 21 to 4% (Figure 7a) and the LC3-II protein level (Figure 7b) following exogenous H2O2 treatment. These results implicate O2•− in regulating exogenous H2O2-induced autophagy.

Figure 7
figure 7

Overexpression of SOD2 downregulates autophagy induced by exogenous H2O2 in HeLa cells. Cells were treated with 1 mM H2O2 for 24 h. (a) Effect of SOD2 overexpression on ROS generation and AVOs formation. The percentages of ethidium fluorescence and DCF fluorescence represent the levels of intracellular O2•− and H2O2, respectively, as described previously. (b) Effect of SOD2 overexpression on LC3-II formation. β-Actin was used as a loading control. Cells were treated with the lysosomal inhibitor NH4Cl at 30 mM for the determination of autophagy flux. Error bars represent standard deviation (S.D.). All data are representative of three independent experiments

Full size image

Superoxide (O2•−) production occurs upstream of PI3K class III and beclin-1 activation during starvation-induced autophagy

Scherz-Shouval et al.15 showed that partial H2O2 was produced from downstream of PI3K class III-beclin-1 complex activation during AA starvation-induced autophagy.15 In this study, we found that GP or AA starvation-induced H2O2 generation was increased by 3-MA (Figure 8a), whereas starvation-induced autophagy was decreased by 3-MA (Figure 2d; Supplementary Figure S4a), in HeLa cells. 3-MA treatment had no effect on O2•− levels in HeLa cells (Figure 8a). When another autophagy inhibitor wortmannin (Supplementary Figure S4bi) was used in HeLa cells, the levels of both O2•− and H2O2 induced by GP or AA starvation were not significantly affected (Figure 8b). In addition, beclin-1 and atg-7 siRNAs knockdown failed to significantly affect GP or AA starvation-induced O2•− and H2O2 production in HeLa cells (Figure 8c). Thus, our results indicate that increased levels of O2•− following starvation occur upstream of PI3K class III-beclin-1 complex activation.

Figure 8
figure 8

Effects of autophagy inhibitors on ROS generation induced by GP or AA starvation in HeLa cells. (a) Effect of 3-MA (2 mM) at 48 h. (b) Effect of wortmannin (Wort, 0.2 μM) at 48 h. (c) Effect of beclin-1 and atg-7 siRNAs at 48 h. The percentages of ethidium fluorescence and DCF fluorescence represent the levels of intracellular O2•− and H2O2, respectively, as described in Figure 1. Error bars represent standard deviation (S.D.) from three independent experiments

Full size image

Discussion

Autophagy has been intensively studied in recent years. However, the mechanism of autophagy induction remains unclear. The involvement of ROS in autophagy has been demonstrated in many studies,5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22 but the relative functions of O2•− and H2O2 are not well established. In this study, we have compared the relative functions of O2•− and H2O2 in the regulation of autophagy induced by AA and GP starvation conditions, mETC inhibitors, and exogenous H2O2. Our findings indicate for the first time that autophagy is selectively mediated by O2•−, and that exogenous H2O2 is effectively converted to intracellular O2•− leading to induction of autophagy.

Reactive oxygen species are regulated by a series of antioxidant enzymes. Because SOD is the antioxidant enzyme to convert O2•− to H2O2,2, 3 the level of O2•− can be controlled by modulating SOD activity. The findings from this study indicate that upregulation of autophagy can be achieved by inhibiting SOD to elevate O2•− levels, and that activating SOD can downregulate autophagy. Activation of downstream antioxidant enzymes that catalyze H2O2 to H2O such as catalase, GPx, and PrxIII can also reduce autophagy levels.2 However by reducing H2O2 level, the levels of O2•− are also reduced through a chain reaction.2, 3 This illustrates the importance for measuring all forms of ROS.

One interesting phenomenon observed from this study is that two different starvation conditions (GP and AA starvation) induce different spectra of ROS production in a same type of cells. GP starvation induces O2•− production but not H2O2, whereas AA starvation induces both O2•− and H2O2 production. Because both starvation conditions induce autophagy, it should be expected that O2•− is the ROS-mediating autophagy. However, the AA starvation system is widely used for autophagy studies, and both O2•− and H2O2 are simultaneously produced in this type of starvation.15 There are several possible reasons why H2O2 is the more commonly investigated ROS, whereas the function of O2•− has been neglected under AA starvation. First, because O2•− and H2O2 are simultaneously produced under AA starvation, they could have redundant function in autophagy regulation. Second, H2O2 is more stable than O2•− and is found to be a signaling molecule in many biological pathways. Third, H2O2 scavengers such as catalase and NAC are readily available.3 Fourth, many studies use exogenous H2O2 as a control to mimic the effect of intracellularly produced H2O2. We show, however, that it is not possible to specify whether O2•− or H2O2 is the mediator of AA starvation-induced autophagy simply based on the effect of H2O2 scavengers or the addition of exogenous H2O2. H2O2 scavengers such as catalase and NAC also effectively scavenge O2•− because O2•− can be quickly converted to H2O2 by SOD in the cell, and exogenous H2O2 can effect on the cells to produce O2•− possibly by damaging mETC.

One interesting question is the source of ROS induced by starvation. Scherz-Shouval et al.15 demonstrated that significant amount of AA starvation-induced H2O2 is located in mitochondria. Our results showed that overexpression or knockdown (by siRNA) of the mitochondria-localized SOD2 decreased or increased starvation-induced O2•− also indirectly indicate that mitochondria are an important source of starvation-induced ROS. The localization of starvation-induced ROS will be the focus for future studies.

Some discrepancies exist on the time of ROS production and autophagy induced by AA starvation between this study and other studies from the literature. For example, several groups have shown that ROS and autophagy can be detected at earlier times (between 2 and 4 h) whereas this study showed that ROS and autophagy can only be detected after 6 h. This could be due to the variations of cell lines used, the relative health of the cells before starvation, and/or differences in detailed experimental procedures from different laboratories. Because earlier studies did not quantify both ROS species, O2•− and H2O2, simultaneously, but emphasized on H2O2 quantification, it is quite possible that these difference could also be due to the type of ROS detected. Nevertheless, the correlation between ROS production and induction of autophagy remains consistent in these studies.

Starvation-induced autophagy is associated with cell survival whereas autophagy induced by various other stimuli (including dopamine,8 sodium selenite,13 mETC inhibitors rotenone and TFFA,17 SOD2 inhibitor 2-ME,18 and exogenous H2O212, 18) is associated with cell death. We and others have found that these inducers of autophagy require ROS production.8, 13, 12, 18 Indeed, we have previously found that rotenone and TTFA induce O2•− production and that overexpressing SOD2 blocks autophagy.17 This was reversed by knocking down SOD2 by siRNA. The data in the current study showed that combining the SOD2 inhibitor 2-ME with an mETC inhibitor increased O2•− production, autophagy, and cell death. 2-ME is currently in clinical trials for treating cancers and might potentiate cell death through increased O2•− production and induction of autophagy. Alternatively, increasing O2•− production by using sod-2 siRNA could lead to cell survival under starvation conditions. In contrast, decreasing O2•− production by SOD2 overexpression, catalase, or NAC reduced autophagy but increased cell death under starvation conditions. Therefore, autophagy can lead to cell survival and cell death under different stimuli even in a same type of cells. The complete mechanisms regulating autophagy-induced cell survival and cell death remain to be determined.

Exogenous H2O2 is frequently used to mimic the effect of intracellularly produced H2O2 and has been shown to induce autophagy.7, 9, 12, 13, 14, 15, 16, 21 However, in these studies, the intracellular levels of O2•− and H2O2 induced by exogenous H2O2 were not directly measured. We found that exogenous H2O2 increases the generation of intracellular O2•− but failed to change levels of intracellular H2O2 (Figure 7; Supplementary Figure S8). These findings indicate that exogenous H2O2 induces autophagy through the generation of intracellular O2•− (Figure 7; Supplementary Figure S8; Chen et al.18). This emphasizes the importance for determining the relative levels of O2•− and H2O2.

Reactive oxygen species induces multiple downstream signals. We showed that O2•− production occurs upstream of the activation of beclin-1 and PI3K class III and leads to induction of autophagy. In contrast, others have shown that blockage of PI3K class III activation with 3-MA or wortmannin prevents H2O2 production following AA starvation.15 This could be due to cell-type-specific differences or the lack of determination of the relative amounts of O2•− and H2O2 produced by AA starvation. Nevertheless, ROS could act further downstream by oxidizing Atg-4 and/or increasing beclin-1 expression leading to induction of autophagy.15, 18 However, exact downstream signaling events regulated by O2•− leading to autophagy are unknown and will be the focus for future investigations.

Although we cannot exclude the possibility that H2O2 may be involved in autophagy regulation in other types of cells and with other stimuli, we propose a model for autophagy induction mediated by O2•− as shown in Figure 9. In our model, starvation, addition of exogenous H2O2, inhibition of SOD, or blockade of the electron transport chain causes increased O2•− leading to autophagy. Because autophagy is involved in many human pathologies,32 our studies of autophagy regulation could provide insights for the development of novel drugs or approaches for controlling autophagy through O2•− regulation in human diseases.

Figure 9
figure 9

Model of autophagy induction mediated by superoxide (O2•−). Autophagy can be induced by various stimuli such as starvation, mETC inhibitors rotenone and TTFA, the SOD inhibitor 2-ME, and exogenous H2O2. I, II, III, and IV represent mETC complexes I, II, III, and IV, respectively. 2-ME, 2-methoxyestradiol; Cyt C, cytochrome c; Q, ubiquinone–ubiquinol pool; GPx, glutathione peroxidase; PrxIII, peroxiredoxin III; SOD, superoxide dismutase; SOD1, Cu, Zn-SOD; SOD2, Mn-SOD. Arrows represent activation. Bars represent inhibition. Catalase can be found in cytoplasm and mitochondria33

Full size image

Materials and Methods

Reagents

Acridine orange, 3-MA, wortmannin, H2O2, 2-ME, rotenone, TTFA, DAPI (4′,6-diamidino-2-phenylindole), catalase, and medium for GP (glucose, L-glutamine, sodium pyruvate, phenol red, NaHCO3, and serum) starvation were purchased from Sigma-Aldrich (Oakville, Ontario, Canada). Medium for AA (amino acids and serum) starvation (EBSS) was purchased from HyClone Laboratories Inc. (Logan, UT, USA). Dihydroethidium (DHE) and 5-(and-6)-chloromethyl-2′, 7′-dichlorodihydrofluorescein diacetate acetyl ester (CM-H2DCFDA) were from Invitrogen (Burlington, Ontario, Canada). DHE, 2-ME, rotenone, TTFA, and DAPI were dissolved in dimethyl sulfoxide (DMSO). Catalase, 3-MA, and H2O2 were dissolved in double-distilled water. The final concentration of DMSO in media was less than 0.1% and it did not have any effect on the activities tested in this study (data of DMSO effect are not shown in this study). The concentrations of reagents used in this study were: H2O2, 1.0 mM; 2-ME, 0.1 mM; 3-MA, 2.0 mM; wortmannin, 0.2 μM; DHE, 3.2 μM; CM-H2DCFDA, 5 μM; rotenone, 50 μM; TTFA, 0.5 mM; and DAPI, 3.0 μM.

Antibodies and siRNAs

Beclin-1 primary antibody and its secondary antibody donkey anti-goat HRP were purchased from Santa Cruz Biotechnology Inc. (Santa Cruz, CA, USA). ATG-7 antibody was purchased from PromoKine Inc. (Germany). SOD2 antibody was purchased from StressGen Biotechnologies (Victoria, Canada). Rabbit anti-actin antibody was purchased from Sigma, rabbit anti-LC3 antibody from Abgent Inc. and their secondary antibody goat anti-rabbit IgG(H+L) HRP from Bio-Rad Laboratories. anti-LC3 mouse monoclonal antibody was purchased from NanoTools and its goat anti-mouse IgG(H+L) HRP from Bio-Rad Laboratories. The siRNAs specific for human beclin-1, atg-7, and sod-2 are same as in our previous study.18

Cell culture

Human embryonic kidney cell line HEK293, human glioma cancer cell line U87, and human cervical cancer cell line HeLa were maintained in a humidified 5% CO2, 37°C incubator in Dulbecco's modified Eagle's medium supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin (Invitrogen). Media used for HEK293 and HeLa were supplemented with 10% bovine calf serum and 10% fetal bovine serum (FBS) (Invitrogen), respectively. Medium for the stabilized HeLa cells with overexpression of SOD2 was also supplemented with 0.2 mg/ml G418 (Life Technologies Inc.). Medium for U87 was supplemented with 10% FBS, 1 mM sodium pyruvate, and 2 mM glutamine without penicillin and streptomycin.

Silencing beclin-1, atg-7, or sod-2 genes by siRNA

The methods for transfection of siRNA into cells and for the treatment of cells have been described in our previous studies.17, 18

Flow cytometric quantification of AVOs

The method is same as described in our previous studies.17, 18

Staining autophagosomes with GFP-LC3

The method is same as described in our previous studies.17, 18 Cells transfected with GFP alone plasmid did not show punctate green dots (vacuoles) when treated with starvation, mETC inhibitors, 2-ME or H2O2, and therefore the results have not been demonstrated in this study. Only the results of cells transfected with GFP-LC3 plasmid have been demonstrated.

Flow cytometric analysis of ROS

Reactive oxygen species generation was determined by flow cytometry after cells were stained with DHE or CM-H2DCFDA . DHE and CM-H2DCFDA are used to specifically detect the generation of intracellular O2•− and H2O2, respectively.27 DHE is oxidized to red fluorescent ethidium by O2•−17, 18, 27 and CM-H2DCFDA is oxidized to green fluorescent DCF (dichlorofluorescein) by H2O2.27 For staining of cells, cells were centrifuged and the pellet was resuspended in 0.2 ml PBS in an Eppendorf tube. DHE with a final concentration of 3.2 μM or CM-H2DCFDA with a final concentration of 5 μM was added into the cell suspension and was gently mixed. The mixture was incubated in dark in a water bath at 37°C for 15 min. Then, the cell suspension was transferred into a 5 ml FALCON FACS tube and analyzed on a flow cytometer using CellQuest software (Becton Dickinson, San Jose, CA, USA) within 10 min. Alternatively, cells can be stained with DHE or CM-H2DCFDA for 15 min before trypsinization (for attachment cells). Actually, these two cell staining methods gave similar results (Supplementary data Figure S1). In this study, ROS have been measured by staining cells with DHE or CM-H2DCFDA after trypsinizing cells, removing trypsin, and resuspending cells in PBS.

Flow cytometric analysis of cell death

Cell death was detected by measuring the plasma membrane permeability to Trypan blue as described previously.18 Briefly, cell pellet was collected and resuspended in 200 μl PBS in an FACS tube. Then, Trypan blue with a final concentration of 0.04% was added into the cell suspension. Then cells were stained for 5 min at room temperature and analyzed on a flow cytometer using CellQuest software (Becton Dickinson). Histogram data on log scale were collected on the red filter (675 nm, FL3-H). Two peaks in a histograph could be observed. The first peak represents viable cells, which were dimly fluorescent and not permeable to Trypan blue. The second peak represents dead cells, which were brightly fluorescent and permeable to Trypan blue. Results were confirmed by staining cells with AO and ethidium bromide and counting cells under a fluorescent microscope as described previously.17, 18

Western blot analysis

Western blot analysis was performed as stated previously.17, 18 Tris-glycine SDS-PAGE was used, except for the detection of conversion of LC3-I (18 kDa, cytoplasmic form) to CL3-II (16 kDa, preautophagosomal and autophagosomal membrane-bound form), where Tris-Tricine SDS-PAGE was used. To determine autophagy flux, we used the lysosomal inhibitor NH4Cl at 30 mM to treat cells before LC3-II western blot was performed. Densitometry was performed using QuantityOne software (Bio-Rad).

Statistical analysis

All experiments were repeated at least three times and each experiment was carried out at least by duplicates. The data were expressed as means±S.D. (standard deviation) (n⩾3). Statistical analysis was performed by using Student's t-test employing at least three independent data points. The criterion for statistical significance was P<0.05. The software used was Excel or Sigma blot.