Hypoxia causes autophagic stress and derangement of metabolic adaptation in a cell model of amyotrophic lateral sclerosis

Authors

  • Sara Cimini,

    1. Laboratory of Molecular Pathology, Department of Molecular Biochemistry and Pharmacology, IRCCS-Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Milena Rizzardini,

    1. Laboratory of Molecular Pathology, Department of Molecular Biochemistry and Pharmacology, IRCCS-Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Gloria Biella,

    1. Unit of Genetics of Neurodegenerative Disorders, Department of Neuroscience, IRCCS-Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy
    Search for more papers by this author
  • Lavinia Cantoni

    Corresponding author
    1. Laboratory of Molecular Pathology, Department of Molecular Biochemistry and Pharmacology, IRCCS-Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy
    • Address correspondence and reprint requests to Lavinia Cantoni, Laboratory of Molecular Pathology, IRCCS-Istituto di Ricerche Farmacologiche “Mario Negri”, Via G. La Masa 19, 20156 Milan, Italy.

      E-mail: lavinia.cantoni@marionegri.it

    Search for more papers by this author

Abstract

Amyotrophic lateral sclerosis is a fatal neurodegenerative disease that affects motor neurons. The recruitment of autophagy (macroautophagy) and mitochondrial dysfunction are documented in amyotrophic lateral sclerosis patients and experimental models expressing mutant forms of Cu, Zn superoxide dismutase (SOD1) protein, but their impact in the disease remains unclear. Hypoxia is a stress closely related to the disease in patients and mutant SOD1 mice; in individual cells, hypoxia activates autophagy and regulates mitochondrial metabolism as fundamental adaptive mechanisms. Our aim was to examine whether mutant SOD1 changed this response. Hypoxia (1% O2 for 22 h) caused greater loss of viability and more marked activation of caspase 3/7 in the motor neuronal NSC-34 cell line stably transfected with the G93A mutant human SOD1 (G93A-NSC) than in the one with the wild-type SOD1 (WT-NSC) or in untransfected NSC-34. In the G93A-NSC cells, there was a more marked accumulation of the LC3-II autophagy protein, attributable to autophagic stress; 3-methyladenine, which acts on initiation of autophagy, fully rescued G93A-NSC viability and reduced the activation of caspase 3/7 indicating this was a secondary event; the metabolic handling of hypoxia was inappropriate possibly contributing to the autophagic stress. Our findings evidentiate that the G93A mutation of SOD1 profoundly altered the adaptive metabolic response to hypoxia and this could increase the cell susceptibility to this stress.

image

Hypoxia activates autophagy and modifies glycolysis and mitochondrial respiration as fundamental cell adaptive mechanisms. This stress is closely related to amyotrophic lateral sclerosis. The recruitment of autophagy and mitochondrial dysfunction are documented in patients and models expressing mutant Cu, Zn superoxide dismutase (SOD1) protein, but their impact in the disease remains unclear. G93ASOD1 cells were more susceptible to hypoxia than wild-type SOD1 cells and showed autophagic stress and inappropriate handling of energy metabolism. Defective adaptation to hypoxia may contribute to neurodegeneration.

Abbreviations used
3-MA

3-methyladenine

ALS

amyotrophic lateral sclerosis

Baf

bafilomycin A1

BNIP3

bcl-2/adenovirus E1B 19-kDa interacting protein 3

FALS

familial ALS

HIF-1

hypoxia-inducible factor 1

MDH

malate dehydrogenase

mTORC1

mammalian target of rapamycin complex 1

PDK1

pyruvate dehydrogenase kinase l

ROS

reactive oxygen species

SALS

sporadic ALS

SOD1

Cu, Zn superoxide dismutase

Amyotrophic lateral sclerosis (ALS) is the most common adult-onset motor neuron neurodegenerative disease; it is rapidly fatal and has no therapy (Robberecht and Philips 2013). ALS is traditionally classified into familial and sporadic ALS (FALS and SALS), which are clinically very similar. FALS is caused by mutations in a heterogeneous group of genes; approximately 2% of ALS patients have mutations in the Cu, Zn superoxide dismutase (SOD1) gene; more than 150 different mutations, distributed in all the exons coding for the protein, have been reported to be pathogenic; their pathophysiological role is not clear, however, the mechanism(s) of toxicity is independent of the dismutase activity (Duffy et al. 2011; Robberecht and Philips 2013). In vitro and in vivo models over-expressing mutant forms of SOD1 are widely used for deciphering the neurodegenerative mechanisms involved in ALS.

Hypoxia is a risk factor for neurodegenerative diseases including Alzheimer's disease, the main cause of dementia, (Zhang and Le 2010) and it is a stress which is closely related to the clinical course of ALS; defects in hypoxic signaling have been seen in SALS patients and G93ASOD1 mice (Büchner et al. 1999; Moreau et al. 2011; Xu et al. 2011; Zhang et al. 2011b). The respiratory status is one of the prognostic factors for ALS patients and respiratory impairment is an early sign in G93ASOD1 mice (Tankersley et al. 2007; Beghi et al. 2011).

Hypoxia activates hypoxia-inducible factor-1 (HIF-1)-mediated systemic, tissue restricted, and cell autonomous homeostatic responses; in individual cells, hypoxia activates macroautophagy (referred to here as autophagy) and changes carbohydrate and energy metabolism as fundamental protective mechanisms (Taylor 2008; Semenza 2011).

The recruitment of autophagy in ALS has been documented, associated with increases in autophagy proteins such as LC3-II in post-mortem spinal cords of sporadic and familial ALS patients (Hetz et al. 2009). Spinal cord of mutant SOD1 mice also showed an increase in LC3-labeled autophagic vacuoles, but, as in ALS patients, it is not clear whether they were the result of autophagy induction or autophagy flux impairment (Nassif and Hetz 2011; Chen et al. 2012). In neurons, which do not divide after differentiation, autophagy is particularly important to maintain cell homeostasis by eliminating unneeded or damaged cellular constituents (Chu 2008; Lee 2009). In ALS and other neurodegenerative diseases characterized by formation of protein aggregates, failure of autophagy, one of the two major components of the cell's degradation machinery, is expected to induce cellular stress and ultimately cell death (Li et al. 2010; Robberecht and Philips 2013). However, attempts to delay disease progression in ALS patients with drugs inducing autophagy have given contradictory and mainly disappointing results (Fornai et al. 2008; Aggarwal et al. 2010; Miller et al. 2011). In the mutant SOD1 mouse, approaches to induce autophagy were beneficial (Hetz et al. 2009) or accelerated the disease (Zhang et al. 2011a) or the effect was modulated by the disease state (Zhang et al. 2013).

The main target of autophagy and metabolic adaptation to hypoxia are mitochondria; these organelles undergo major changes of structure, function, and dynamics through activation of several genes that reduce mitochondrial respiration and reactive oxygen species (ROS) production; the supply of mitochondrial substrates (acetyl-CoA and O2) is reduced while glucose is consumed faster to produce ATP via anaerobic glycolysis to lactate (Taylor 2008; Zhang et al. 2008; Semenza 2011). The role of mitochondrial adaptation to low O2 in ALS has been poorly investigated, although mitochondria are known to be involved in the disease pathogenesis in patients and mutant SOD1 mice mainly through mitochondrial dysfunction, generation of free radicals, and impairment of calcium handling (Duffy et al. 2011; Richardson et al. 2013; Robberecht and Philips 2013).

In this study, we utilized a well-characterized FALS motor neuronal model over-expressing G93A mutant human SOD1, (Rizzardini et al. 2005), the pathogenic mutant SOD1 more extensively studied. We found that when this mutant protein was expressed, exposure to low O2 caused greater cell loss, disabled autophagy, and altered reprogramming of metabolism suggesting that a defective adaptation to this stress may contribute to the neurodegenerative process.

Materials and methods

Motor neuronal ALS model

The NSC-34 cell line (a kind gift from N. R. Cashman) was used to obtain lines stably expressing human wild-type SOD1 (WT-NSC) or G93ASOD1 (G93A-NSC) as described previously (Rizzardini et al. 2005). The G93A-NSC cell line expresses a low level of G93ASOD1 (lower than murine SOD1) and therefore it appears a good model of motor neurons in the disease in terms of expression level since only one allele is mutant in FALS patients with SOD1 mutations. Furthermore, this line has been shown to reproduce aspects of the oxidative and mitochondrial toxicity of this mutant SOD1 (Raimondi et al. 2006; Rizzardini et al. 2006). The cell lines were grown in high-glucose Dulbecco's modified Eagle's medium supplemented with 5% heat-inactivated fetal bovine serum, 1 mM glutamine (all from Lonza, Verviers, Belgium), 1 mM pyruvate, and antibiotics (100 IU/mL penicillin and 100 μg/mL streptomycin) (all from Invitrogen, Carlsbad, CA, USA). The WT-NSC and G93A-NSC cell lines were kept in selection by adding 0.5 mg/mL geneticin (Invitrogen). The cell lines were cultured simultaneously, subcultured in parallel every 7 days and never maintained beyond passage 20 to limit any potential effect of senescence.

Treatments

For each experiment, the different cell lines were harvested at the same passage number, were seeded in multiwell plates (7630 cells/cm2), and were grown for 96 h. Then the medium was replaced with fresh Dulbecco's modified Eagle's medium without phenol red (Invitrogen) and without fetal bovine serum (time 0). Cells were then incubated in a cell culture incubator in normoxia (21% O2; 5% CO2) (controls) or exposed to hypoxia alone or combined with autophagy or caspase inhibitors: 3-methyladenine (3-MA) (Sigma-Aldrich, St. Louis, MO, USA), bafilomycin A1 (Baf), CA-074-Me or Z-VAD-FMK (all from Enzo Life Sciences Inc., Farmingdale, NY, USA). The standard length of treatment was 22 h. Hypoxia was done using the Invivo2 400 Hypoxia workstation (Ruskinn Technology Ltd., Pencoed, UK) equipped with the gas mixer Q to obtain accurate control and stability of the O2 and CO2 concentrations. The standard hypoxic conditions were 1% O2 and 5% CO2. The medium for hypoxia samples was pre-conditioned to hypoxic conditions for at least 6 h. Medium was changed and samples were collected for analysis directly in the hypoxic chamber.

Cell viability assay

Reduction of 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide (MTT) to formazan crystals by cellular dehydrogenases was used as a measure of the number of living cells, as described previously (Rizzardini et al. 2005).

Lactate

Lactate in the culture medium was determined as previously described with a spectrophotometric method adapted to analysis with a 96-well plate (D'Alessandro et al. 2011).

NAD(P)+

Total cellular NAD(P)+ content was determined according to the extraction procedure and the cycling assay described by Billington et al. (2008). The cell lines were grown and treated in 24-well plates; 100 μL of extracted NAD(P)+ was mixed with 60 μL of cycling mixture and incubated in the dark for 3 h at 37°C. Fluorescence (excitation 544 nm, emission 590 nm) was measured with a plate reader and values were normalized with the protein contents determined with the bicinchoninic acid protein assay kit (Pierce, Rockford, IL, USA).

Malate dehydrogenase activity

Malate dehydrogenase (MDH) activity was determined as previously described (D'Alessandro et al. 2011) on a post-mitochondrial supernatant obtained after sonication and centrifugation at 12 000 g for 30 min.

Caspase activity

Caspase 3/7 activities were determined with a luminescence assay (Caspase-Glo 3/7 Assay, Promega, Madison, WI, USA) according to the manufacturer's protocol.

SDS–PAGE and western blot analysis

Cells from six-well plates were lysed for 10 min on ice with 50 μL of radio-immunoprecipitation assay (RIPA) buffer (Sigma-Aldrich) supplemented with protease and phosphatase inhibitor cocktails, following the manufacturer's protocol. Equal amounts of proteins were separated by 10–15% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and blotted onto nitrocellulose membrane for western blot analysis of bcl-2/adenovirus E1B 19-kDa interacting protein 3 (BNIP3), LC3, Lon, pyruvate dehydrogenase kinase l (PDK1), and Tom20. The following primary antibodies were used: BNIP3 (1 : 1000; rabbit polyclonal, Cell Signaling Technology, Beverly, MA, USA); LC3 (1 : 2000; rabbit polyclonal, Abcam, Cambridge, UK); PDK1 (1 : 1000; rabbit polyclonal, Enzo Life Sciences Inc.); Tom20 (1 : 1000, rabbit polyclonal, Santa Cruz Biotechnology, Santa Cruz, CA, USA); Lon (1 : 1000, rabbit polyclonal). The polyclonal antiserum against Lon was produced by Primm (Milan, Italy). Actin (1 : 4000; mouse monoclonal, Millipore Corporation, Billerica, MA, USA) was used as protein loading control. Protein bands were detected with the enhanced chemiluminescence detection system. Films were scanned and band intensities were obtained with QuantityOne (Bio-Rad Laboratories, Hercules, CA, USA).

Autophagic flux after hypoxia

A validated protocol was used to measure the autophagic flux, as a way to disentangle the different possible interpretations of changes in snapshot levels of LC3-II in a cell line after a stress (Rubinsztein et al. 2009). The level of LC3-II—as a functional reporter of autophagy—was determined with or without stress and with or without Baf, which blocks the steps of autophagy involved in LC3-II degradation by raising lysosomal pH consequently inhibiting lysosomal enzymes, and by interfering with autophagosomal fusion with late endosomes and lysosomes (Yoshimori et al. 1991; Mehrpour et al. 2010). The cell lines were cultured as described in ‘Treatments’, then exposed to 21% O2 or 1% O2 for 4 h, without or with Baf (100 nM), added when changing the medium at the beginning of the 4-h treatment. This short time of treatment is required to avoid non-specific effects because of prolonged inhibition of autophagy by Baf (Rubinsztein et al. 2009). For each cell line, the LC3-II levels in the different conditions were measured by western blot. The effect of hypoxia on the synthesis of autophagosomes can be inferred by comparing the LC3-II levels after Baf alone and after the combined treatment with hypoxia and Baf: higher levels of LC3-II after the combined treatment indicate an increase of this initial part of autophagy. The effect of hypoxia on autophagosome degradation can be inferred by comparing the LC3-II levels after hypoxia alone and after the combined treatment: higher levels of LC3-II after the combined treatment indicate an increase in the degradative part of autophagy.

Fluorescence microscopy of LC3 protein

The cell lines were grown and treated in 24-well plates (each well containing a 13-mm glass coverslip). The cells were fixed for 10 min with 4% paraformaldehyde and permeabilized for 10 min with digitonin (100 μg/mL). The LC3 primary antibody (1 : 100; MBL, Nagoya, Japan) was added and incubated overnight at 4°C. A FITC- conjugated secondary antibody (Alexa 488, 1 : 200; Invitrogen) was added and incubated for 30 min at 22°C. Hoechst solution (1 mg/mL, for 5 min) was used to stain DNA. Coverslips were washed with phosphate-buffered saline 1X then mounted, adding FluorSave reagent, and observed with a fluorescence microscope (Olympus IX71, Olympus Corporation, Shinjuku, Tokyo, Japan).

Statistical analysis

Data are presented as mean with standard error. The statistical significance of the data was evaluated by one-way anova, followed by the Newman–Keuls multiple comparison post-test.

Results

The G93A-NSC cell line is more susceptible to hypoxia than NSC-34 and WT-NSC

Exposure to 1% O2 concentration for 22 h caused significantly greater loss of viability (< 0.001), determined by the MTT assay, in the G93A-NSC cell line than in NSC-34 and WT-NSC (respectively, about 35%, 25%, and 10% decreases in comparison to culture in 21% O2, i.e., normoxia) (Fig. 1). In contrast, the viability of the WT-NSC line was the highest showing that over-expression of wtSOD1 was significantly protective also in comparison to the untransfected NSC-34 (p < 0.001). A differential susceptibility to low O2 concentration of the cell lines was observed also after treatment with 0.1% O2 for 22 h (about 58%, 43%, and 34% decreases in comparison to culture in 21% O2, respectively, in the G93A-NSC, NSC-34, and WT-NSC) (data not shown). In this study, the cell lines were exposed to 1% O2 which is in the range of reduction of ambient O2 concentration defined as hypoxia (Semenza 2011).

Figure 1.

Viability of the NSC-34, WT-NSC, and G93A-NSC cell lines exposed to hypoxia. The viability of the NSC-34, WT-NSC, and G93A-NSC cells was evaluated with the 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide (MTT) assay after 22-h exposure to 1% O2. Values (n = 24) are percentages of the MTT conversion in normoxia (21% O2). ***p < 0.001 versus respective 21% O2; ▲▲▲p < 0.001, ▲▲p < 0.01 versus NSC-34 1% O2; ○○○p < 0.001 versus WT-NSC 1% O2.

Hypoxia induces accumulation of LC3-II protein in the G93A-NSC cell line

We first analyzed autophagy activation by measuring LC3 fluorescence in the WT-NSC and G93A-NSC cell lines cultured in normoxia or hypoxia (1% O2 for 22 h) (Fig. 2a). LC3 is a cytosolic protein proteolytically activated and lipidated during autophagy which translocates to the membranes of the growing autophagic vesicles, remaining associated with sealed autophagosomes and mature autophagosomes/autolysosomes until degradation by lysosomal proteases (Rubinsztein et al. 2009). After treatment with hypoxia, in the G93A-NSC cell line bright puncta in the cytoplasm were observed in the whole cell population, consistent with more marked vacuolar localization of LC3 than in normoxia (Fig. 2a). Western blot analysis of cell lysates showed that after hypoxia the levels of the phospholipid-linked form LC3-II, which correlates with the number of autophagosomes and/or autolysosomes at a snapshot in time, rose more than three-fold in the G93A-NSC cell line in comparison to normoxia (p < 0.01) (Fig. 2b, c). LC3-II in the G93A-NSC was also significantly higher than in the hypoxia-treated NSC-34 and WT-NSC cells. Therefore, the greater loss of viability of the G93A-NSC cell line after 22 h of 1% O2 was associated with a larger amount of LC3-II than the other cell lines.

Figure 2.

LC3 protein expression in normoxia and hypoxia in the NSC-34, WT-NSC, and G93A-NSC cell lines. (a) Fluorescence microscopy (40X) of LC3 in WT-NSC and G93A-NSC cells after 1% O2 for 22 h. LC3 positive puncta are indicated by arrowheads. (b) Expression of LC3-I and LC3-II (representative western blots) in the NSC-34, WT-NSC, and G93A-NSC cells after 21% O2 or 1% O2 for 22 h. (c) LC3-II levels normalized to actin (n = 3). ***p < 0.001 versus respective 21% O2; ▲▲▲p < 0.001 versus NSC-34 1% O2; ○○○p < 0.001 versus WT-NSC 1% O2.

Hypoxia causes less induction of the autophagic flux in the G93A-NSC cell line

Snapshot levels of LC3-II can rise if autophagic substrate degradation is enhanced or delayed. To understand the cause of this accumulation in the G93A-NSC, we investigated the effect of hypoxia in each cell line on the synthetic and degradative parts of autophagy using the LC3-II level as a functional reporter of the autophagic flux as described in ‘Materials and methods’. For this analysis, the time of exposure to 1% O2 was 4 h. This short time of hypoxic treatment did not change the viability of the cell lines (Fig. 5b). Baf 100 nM for 4 h did not change the cell viability either in normoxia or hypoxia (data not shown).

To evaluate the effect of hypoxia on autophagosome synthesis, we compared the LC3-II protein levels after treatment with Baf alone and after hypoxia plus Baf. In the G93A-NSC cell line the levels did not differ (only about 20% increase) (Fig. 3c) suggesting that hypoxia did not raise this part of autophagy. In contrast, in the NSC-34 (Fig. 3a) and WT-NSC (Fig. 3b) there was a significant difference (respectively, 60% and 80% increases), indicating that hypoxia in these cell lines increased the synthesis of autophagosomes.

Figure 3.

Effect of hypoxia on the autophagic flux of the NSC-34, WT-NSC, and G93A-NSC cell lines. The effect of hypoxia on the autophagic flux was evaluated in each cell line using the lysosomal inhibitor Baf (100 nM). (a–c) show the levels of LC3-II (normalized to actin) (n = 3), respectively, in the NSC-34, WT-NSC, and G93A-NSC after 21% O2 (normoxia) or 1% O2 (hypoxia) for 4 h with or without Baf. In each cell line, the effect of hypoxia on autophagosome synthesis was evaluated by comparing the LC3-II levels after Baf alone (21% O2) and after the combined treatment (Baf and 1% O2) (respectively, the second and the fourth bars). The effect on autophagosome degradation was evaluated by comparing the LC3-II levels after 1% O2 and after the combined treatment (Baf and 1% O2) (respectively, the third and the fourth bars). Representative western blots are shown. ***p < 0.001, **p < 0.01, *p < 0.05 versus respective 21% O2; ●●●p < 0.001, ●●p < 0.01 versus respective 1% O2; □□□p < 0.001, □p < 0.05 versus respective 21% O2 + 100 nM Baf.

We next investigated how hypoxia affected the degradative part of autophagy. In each cell line, we compared the LC3-II protein levels after hypoxia and after treatment with hypoxia and Baf (Fig. 3a–c). LC3-II levels were significantly higher after the combined treatment, indicating that hypoxia enhanced this part of autophagy in all the cell lines; however, the increase was less marked in the G93A-NSC cell line (about 160%) (Fig. 3c) than in NSC-34 and WT-NSC (respectively, about 275% and 310%) (Fig. 3a and b). Therefore, the degradative part of autophagy was induced less in the G93A-NSC.

Results of the analysis of autophagic flux after 4 h of 1% O2 treatment (Fig. 3) indicated that hypoxia increased autophagosome synthesis and degradation more markedly in the NSC-34 and WT-NSC than in the G93A-NSC. This may explain why in the NSC-34 and WT-NSC after 22 h of 1% O2 treatment (Fig. 2c) LC3-II did not rise significantly in comparison to normoxia; in contrast, the more marked accumulation of LC3-II in the G93A-NSC (Fig. 2c) was not caused by greater activation of autophagy in this cell line than in the others, but rather to a dysregulation of the process.

Furthermore, after 4 h of hypoxia there was a significant increase in LC3-II (Fig. 3c, 162% increase) only in the G93A-NSC in comparison to normoxia, as also after longer treatment (22 h) (Fig. 2c).

Autophagy inhibition at the initiation and completion stages has a different impact on hypoxia toxicity

To investigate whether the lower autophagic flux in the G93A-NSC cell line was linked to its greater loss of viability after hypoxia, we used the MTT assay to check the effect of combined treatments with pharmacological inhibitors of autophagy (3-MA, Baf and CA-074-Me) and hypoxia.

3-MA reduces the initiation and maturation phases of the autophagic process by inhibiting class III PI3K/hVps34 (Wu et al. 2010). The cell lines were pre-treated for 1 h with the desired concentration of 3-MA before the change of medium. 3-MA was again added to the fresh medium without fetal bovine serum at time 0 and was present during the 22 h of treatment with 1% O2. At 2.5 or 5 mM, 3-MA attenuated the loss of viability because of hypoxia in the NSC-34 cell line, and most of all in the G93A-NSC, which was fully rescued by 5 mM; in contrast, the mild toxicity of 1% O2 in the WT-NSC was only slightly modified (Fig. 4a). At 10 mM, the concentration most commonly used in vitro to inhibit autophagy, 3-MA still protected the G93A-NSC cell line, but there was a marked loss of viability in WT-NSC (about 40% decrease in comparison to hypoxia alone) (Fig. 4a). Therefore, 3-MA modified the responses to hypoxia of each cell line differently, the 3-MA concentration affecting each response. As a consequence, when comparing the viability of the different cell lines treated with hypoxia at each 3-MA concentration (anova followed by Newman–Keuls multiple comparison post-test), we found that the viability of the WT-NSC cell line was significantly lower than that of the G93A-NSC at 5 mM and 10 mM 3-MA (p < 0.001) and of NSC-34 at 10 mM 3-MA (p < 0.01). The protective effect of 3-MA on viability at 5 mM was significantly greater in the G93A-NSC than in the NSC-34 cell line (p < 0.001).

Figure 4.

Autophagy inhibitors affect the hypoxia-induced loss of viability of the NSC-34, WT-NSC, and G93A-NSC cell lines differently. Effects of 3-MA (2.5 mM, 5 mM, and 10 mM) (n = 8) (a), Baf (100 nM) (n = 12) (b) and CA-074-Me (100 μM) (n = 4) (c) on viability of NSC-34, WT-NSC, and G93A-NSC cells after 1% O2 for 22 h, shown as percentages of the 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide (MTT) conversion in normoxia (21% O2) indicated by the horizontal line. Only the significant differences owing to hypoxia and to the combined treatments in each cell line are shown. ***p < 0.001 versus respective 21% O2; ●●●p < 0.001, ●●p < 0.01 versus respective 1% O2; ♦♦♦p < 0.001 versus respective 1% O2 + 2.5 mM 3-MA; ■■■p < 0.001 versus respective 1% O2 + 5 mM 3-MA.

We next blocked the completion step of autophagy with Baf (100 nM, added at the change of medium). Baf significantly increased (20–40%) the toxicity of 1% O2 for 22 h in all the cell lines (Fig. 4b). WT-NSC maintained higher viability (about 50% of normoxia, p < 0.01) than NSC-34 and G93A-NSC, like after hypoxia alone (Fig. 1). In the WT-NSC, inhibition of autophagy with Baf or with 10 mM 3-MA increased the toxicity of hypoxia to much the same extent (about 40%).

CA-074-Me, a cell-permeable inhibitor of lysosomal cathepsins (mainly cathepsin B) (Montaser et al. 2002), (100 μM, added at the change of medium) also increased the toxicity of 1% O2 for 22 h in all the cell lines (Fig. 4c). This combined treatment was the most cytotoxic (more than 40% loss of viability) and the viability was very low in all the cell lines (about 20–25% of that in normoxia). Statistical analysis of the effect of CA-074-Me showed that it was more damaging in the WT-NSC and in fact in this cell line the loss of viability (77%) in comparison to hypoxia alone was greater than in the NSC-34 and G93A-NSC lines (p < 0.001) (Fig. 4c).

In conclusion, all the steps of autophagy appear essential for motor neuronal health in hypoxia, as clearly shown in the WT-NSC; in the G93A-NSC cells, the protective effect of 3-MA was unexpected, but it is in agreement with the suggestion that dysregulated autophagy contributes to the hypoxic death of this cell line (see 'Discussion').

Caspase 3/7 is activated by hypoxia mainly in the G93A-NSC cell line

Caspase-dependent and caspase-independent apoptotic death pathways are activated by hypoxia (Walls et al. 2009) and mutant SOD1s have a caspase-inducing activity (Guégan and Przedborski 2003), therefore we investigated the role of caspase 3/7 activation in the greater loss of viability of the hypoxic G93A-NSC cell line. We first analyzed caspase 3/7 activity and viability after 4 h of 1% O2, the same treatment used to measure the autophagic flux. Caspase 3/7 activity did not significantly change in any cell line and the MTT assay did not show toxicity (Fig. 5a, b) indicating that the increase in LC3-II in G93A-NSC after this treatment (Fig. 3c) was an independent and early modification. We then determined the effects on caspase activity after 1% O2 for 17 h, when the MTT assay showed that only the G93A-NSC had a significant loss of viability compared to normoxia and the viability of this cell line was significantly lower than NSC-34 and WT-NSC (Fig. 5b), as after 1% O2 for 22 h (Fig. 1). Caspase 3/7 was activated in comparison to normoxia and G93A-NSC had the highest activity (p < 0.001) and WT-NSC the lowest, while NSC-34 had an intermediate level (Fig. 5a). These results suggest that in the G93A-NSC this death pathway may contribute more markedly than in the other lines to the loss of viability after 1% O2 (Fig. 1). The difference of caspase activity between WT-NSC and G93A-NSC cell lines was confirmed by western blot analysis of cleaved caspase 3 protein level (Fig. S1). Interestingly, also after 17-h normoxia, caspase 3/7 activity in the G93A-NSC was twice that of the NSC-34 and WT-NSC (Fig. 5a).

Figure 5.

The G93A-NSC line has the greatest loss of viability and the highest caspase 3/7 activity after hypoxia and 3-MA reduces the caspase 3/7 activation. (a) Activation of caspase 3/7 of the NSC-34, WT-NSC, and G93A-NSC cells after 21% O2 or 1% O2 for 4 h (n = 5) and 17 h (n = 11) and (b) viability (expressed as percentages of 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide (MTT) conversion in normoxia) (n = 4 and n = 8, respectively, for the 4 h and 17 h) in the same conditions. (c and e) Effects of Z-VAD-FMK (25 μM) (n = 6) and 3-MA (5 mM and 10 mM) (n = 5 and n = 10) on the caspase 3/7 activation, after 1% O2 for 17 h. (d) Effects of Z-VAD-FMK on the MTT conversion (n = 7) after 1% O2 for 22 h. In (c–e) the values are expressed as percentages of the hypoxia. ***p < 0.001 versus respective 21% O2; □□□p < 0.001 versus NSC-34 21% O2; $$$p < 0.001 versus WT-NSC 21% O2; ▲▲▲p < 0.001, ▲p < 0.05 versus NSC-34 1% O2; ○○○p < 0.001 versus WT-NSC 1% O2; ●●●p < 0.001 versus respective 1% O2; ■■■p < 0.001 versus respective 1% O2 + 5 mM 3-MA.

The protective effect of 3-MA against hypoxic cell death in the G93A-NSC is associated with inhibition of the activation of caspase 3/7

There is a complex crosstalk between caspase-dependent death and autophagy (Kang et al. 2011) and 3-MA and caspase-dependent death (see 'Discussion'). As 3-MA differently affected the viability of the cell lines exposed to hypoxia (Fig. 4a), we further investigated the effect of caspase 3/7 activation. We first studied the effects of Z-VAD-FMK, a long-acting broad-spectrum caspase inhibitor. This compound (25 μM, added at the change of medium) more than 80% inhibited the caspase 3/7 activation because of hypoxia (Fig. 5c) and slightly increased the viability after 1% O2 for 22 h in the NSC-34 and G93A-NSC without reaching statistical significance (Fig. 5d). This result suggests that caspase activation was a secondary event and that blocking caspase activity was not sufficient to give significant protection in hypoxia. This conclusion is in agreement with the effect of the treatment with hypoxia plus 3-MA. 3-MA significantly reduced the activation of caspase 3/7 (in a dose-dependent way) (Fig. 5e), however, since it is known that this compound does not directly inhibit caspase activity (Canu et al. 2005), this result indicates that 3-MA modified the upstream processes involved in caspase activation. This effect combined with the inhibition of any toxicity caused by caspase activation may be at the roots of the protection by 3-MA to G93A-NSC (and to a lesser extent to NSC-34) against hypoxia (Fig. 4a). Results also suggest that these signaling pathways were not playing a major role in the hypoxic WT-NSC since Z-VAD-FMK or 3-MA did not have any protective effect (Fig. 5d and 4a).

Furthermore, 3-MA at the highest concentration (10 mM) increased cell death in comparison to hypoxia alone in the WT-NSC and was less protective in the other two cell lines (Fig. 4a), possibly because this 3-MA concentration is routinely used to inhibit autophagy, fundamental for successful adaptation to hypoxia; this toxicity was through caspase-independent pathways evidentiating that other death pathways were activated (Fig. 5e).

Over-expression of SOD1 influences hypoxia-induced adaptation of mitochondria and reprogramming of metabolism is different in WT-NSC and G93A-NSC cells

Inhibition of autophagy can trigger mitochondrial membrane permeabilization and apoptosis (Boya et al. 2005); furthermore, hypoxia induces profound metabolic and redox changes which influence cell survival and require a series of HIF-1-dependent modifications of mitochondrial function (Semenza 2011). These aspects were therefore investigated in our model.

Induction of the gene BNIP3 triggers selective autophagy of mitochondria. After treatment with 1% O2 for 22 h, the expression of BNIP3 protein (hardly detectable in normoxia, data not shown) was significantly higher in the NSC-34 cell line than in WT-NSC and G93A-NSC (respectively, 45% and 36% of NSC-34); the BNIP3 level in the G93A-NSC was also lower than in the WT-NSC, although this difference did not reach statistical significance (Fig. 6a). However, the untransfected and the transfected cell lines did not differ in their mitochondrial mass, at least judging from the expression of the constitutive mitochondrial protein Tom20 (Fig. 6b); this suggests that over-expression of SOD1 might also influence other mechanisms controlling the size of the mitochondrial pool, like regulation of morphology or biogenesis of mitochondria (Zhang et al. 2007; Kim et al. 2011).

Figure 6.

Effect of hypoxia on the expression levels of bcl-2/adenovirus E1B 19-kDa interacting protein 3 (BNIP3), Tom20, and Lon proteins of the NSC-34, WT-NSC, and G93A-NSC cell lines. (a–c) Levels of BNIP3, Tom20, and Lon (normalized to actin) (n = 3) in NSC-34, WT-NSC, and G93A-NSC cells after 1% O2 for 22 h. A representative western blot is shown for each protein. In (a) the arrow indicates the BNIP3 monomer (28 kDa MW) used for densitometric analysis. ▲▲p < 0.01 versus NSC-34 1% O2.

Induction of the protease Lon regulates the switch of the isoform composition of subunit IV of cytochrome c oxidase to adapt respiration to the low O2, limiting the increase in the production of ROS. After 22 h of 1% O2, the expression of Lon protein in both the WT-NSC and G93A-NSC was significantly higher than in the untransfected NSC-34 cell line (about 80% more; Fig. 6c), suggesting that the adaptive switch of the subunit IV was not primarily involved in the higher toxicity of hypoxia in G93A-NSC.

The level of lactate in the medium, which estimates the induction of glycolysis, was significantly higher in G93A-NSC cells after 22 h of 1% O2 (Fig. 7a) and this was accompanied by a higher level of expression of PDK1 protein, which inhibits the activity of pyruvate dehydrogenase (Fig. 7b). Both results suggested that G93A-NSC had a greater requirement to convert pyruvate to lactate, which in hypoxia is the main way to convert NADH back to NAD+. Therefore, we measured NAD(P)+ levels and the activity of cytosolic MDH. Hypoxia reduces NAD+ while the level of NADH rises (Lim et al. 2010); the MDH enzyme is part of the malate/aspartate shuttle for transferring reducing equivalents from the cytosol to the mitochondria (McKenna et al. 2006). NAD(P)+ was significantly lower after culture in 1% O2 for 22 h than in normoxia (Fig. 7c) (33%, 22%, and 43% decreases in NSC-34, WT-NSC, and G93A-NSC, respectively); however, it was significantly lower in the hypoxic G93A-NSC and higher in WT-NSC than in the others (Fig. 7c). There was also a significant increase (20%) in MDH activity in the G93A-NSC cell line compared to the NSC-34 and WT-NSC cell lines, whereas in normoxia the cell lines did not differ (Fig. 7d).

Figure 7.

Effect of hypoxia on metabolic parameters of the NSC-34, WT-NSC, and G93A-NSC cell lines. Metabolic parameters regulated by hypoxia were measured in the NSC-34, WT-NSC, and G93A-NSC cells. (a) Levels of lactate in the medium after 1% O2 for 4 h (n = 4) and 22 h (n = 12). (b) Level of pyruvate dehydrogenase kinase 1 (PDK1) (normalized to actin) (n = 9) after 1% O2 for 22 h. A representative western blot is shown. (c) NAD(P)+ (n = 8) and (d) cytosolic malate dehydrogenase (MDH) activity (n = 4) after 21% O2 or 1% O2 for 22 h. (c and d), percentages of the NSC-34 value in normoxia. ***p < 0.001 versus respective 21% O2 (22 h); ●●●p < 0.001 versus respective 1% O2 (4 h); ▲▲p < 0.01, ▲p < 0.05 versus NSC-34 1% O2 (22 h); ○○○p < 0.001, ○○p < 0.01, ○p < 0.05 versus WT-NSC 1% O2 (22 h).

These findings indicated that the higher caspase 3/7 activation and the autophagy dysregulation of the hypoxic G93A-NSC cells were associated with altered metabolic parameters.

Discussion

Hypoxia is a stress closely related to the clinical course of ALS (Moreau et al. 2011; Zhang et al. 2011b). This study found that the cell autonomous metabolic adaptive program promoting cell survival in hypoxic conditions was inappropriate in a motor neuronal ALS model expressing the G93A mutant SOD1.

Our findings indicated that since very early after exposure to low O2 the cells expressing the mutant SOD1 had a higher level of LC3-II protein than the untransfected and those expressing wtSOD1; however, this was not because of a higher rate of HIF-1-induced autophagy and in fact there was evidence of a less active process.

In neurons, impairment of the autophagic flux can create a condition of autophagic stress associated with death pathways (Chu 2008). In our ALS model, the sequence of toxic changes (autophagic stress, metabolic alterations, caspase 3/7 activation, and cell death) suggests that induction of autophagy by hypoxia might eventually operate as a pathological mechanism.

We therefore further studied the role of autophagy in our model and investigated whether pharmacological inhibition of the different steps of the autophagy machinery modified hypoxia toxicity. Inhibition of the completion step with either Baf or CA-074-Me increased the loss of viability induced by low O2 in all the cell lines, in agreement with the vital role of the lysosomal degradative pathways (Pivtoraiko et al. 2009; Frezza et al. 2011). However, the worsening effect of either Baf or CA-074-Me was greater in WT-NSC than G93A-NSC, suggesting that hypoxia already induced lysosomal dysfunction in cells expressing the mutant SOD1. This also agrees with the tendency to less induction of the degradative part of the autophagic flux in the hypoxic G93A-NSC cell line than in the others. A block of LC3-II degradation could result from a decrease in lysosomal proteolytic activity but also from altered maturation of autophagosomes, their trafficking and fusion with endosomes and lysosomes; interestingly, some of the genetic mutations in ALS patients were reported to disturb these pathways (Chen et al. 2012).

Oxidative stress and free radical damage, which are substantially implicated in motor neuron degeneration in models expressing mutant SOD1s, including ours, and in ALS patients (Rizzardini et al. 2005; Barber and Shaw 2010), can cause lysosomal damage, thus also impairing degradative pathways involving lysosomes other than autophagy (Pivtoraiko et al. 2009; Castino et al. 2010). Lysosomes are also involved in mutant SOD1 degradation and inhibition of their activity might favor accumulation of G93ASOD1 aggregates which may be toxic (Kabuta et al. 2006; Hetz et al. 2009). However, it is hard to assess the role of such aggregates in cell lines stably expressing a low level of G93ASOD1, as in our model.

Lysosomal dysfunction would be consistent with the greater caspase 3/7 activation in the hypoxic G93A-NSC line, according to a sequence of events described in other neuronal models (Castino et al. 2010; Walls et al. 2010). However, triggering apoptosis and caspase activation is not always or exclusively associated with lysosomal damage as a mechanism causing inhibition of autophagy (Boya et al. 2005; Pivtoraiko et al. 2009) and we found that the mutant SOD1 also dysregulated the initiating step of autophagy as hypoxia did not boost the synthesis of autophagosomes in the G93A-NSC as much as in the other lines. The relevance of this effect to hypoxia toxicity was indicated by the protective action in this cell line of 3-MA, an inhibitor of both class I PI3K and class III PI3K/hVps34.

Class I PI3K suppresses autophagy by activating its master regulator, the mammalian target of rapamycin complex 1 (mTORC1) and also activates the anti-apoptotic Akt pathway; class III PI3K/hVps34 is essential for initiation of autophagy as it promotes the nucleation and assembly of the initial phagophore membrane and is involved in the maturation of the autophagosome. Class III PtdInsK/hVps34 is also needed in the amino acid-responsive activation of mTORC1 (Jung et al. 2010; Mehrpour et al. 2010). As 3-MA can either inhibit or promote autophagy (Wu et al. 2010), this makes it difficult to interpret the mechanism(s) of action of 3-MA, although the more evident protection at a low concentration suggests it was multifactorial, affecting not only autophagy but also pathways playing a role in cell fate decisions.

3-MA is protective in models implying autophagic stress and its interplay with dysfunctional endosomal sorting complex required for transport III (ESCRT-III) (Lee and Gao 2009) or necrosis (Higgins et al. 2011) or apoptosis as causative factors of neural death (Xue et al. 1999; Canu et al. 2005; Kunchithapautham and Rohrer 2007; Zhang et al. 2009; Castino et al. 2010) and in these latter models 3-MA affected the status of mitochondria. This compound might be protective in the hypoxic G93A-NSC line up to the concentration maintaining a pool of intact mitochondria by preventing bioenergetic failure, keeping the balance between pro- and anti-apoptotic mediators, or favoring the removal of mitochondria with altered membrane permeability, all mechanisms involved in apoptosis initiation when autophagy is impaired (Boya et al. 2005) and upstream of caspase 3/7 activation. 3-MA is also an inducer of the multifunctional protein p62, (Wu et al. 2010), which contributes to the recognition of dysfunctional mitochondria by the autophagy machinery (Geisler et al. 2010).

The fine tuning of the mitochondrial pool, which requires control of its quality, size, and metabolism (Semenza 2011), is peculiar to successful HIF-1-mediated adaptation to low O2 and is directed at preserving the redox homeostasis and reducing the formation of ROS. In the hypoxic G93A-NSC, there was multiple evidence of dysregulation of metabolic parameters, suggesting there was a pool of poorly adapted mitochondria. Interestingly, treatment of G93ASOD1 mice with the autophagy inducer rapamycin caused more severe mitochondrial impairment and greater caspase activation, in addition to increasing LC3-II accumulation and motor neuron degeneration (Zhang et al. 2011a). Alterations in the metabolic adaptation to hypoxia in the G93A-NSC line might promote cell death, but also contribute to the autophagic stress affecting the activation of the autophagy machinery and/or the clearance of dysfunctional mitochondria. In fact, these alterations might modify the signaling cascade controlling the activity of mTORC1 for autophagy activation by hypoxia which is partly shared with other stimuli such as growth factors, glycolytic flux, cellular redox, and ROS (Jung et al. 2010; Yoshida et al. 2011).

An evidence of these effects might be the very low BNIP3 level in hypoxic cells G93A-NSC. BNIP3 induction by hypoxia is involved in autophagy of mitochondria and in the activation of caspase-independent apoptotic death pathways (Walls et al. 2010; Semenza 2011). BNIP3 was also lower in WT-NSC than in NSC-34, suggesting the possibility that over-expression of SOD1 decreased BNIP3 induction by changing quantitatively/qualitatively the pool of cellular ROS which may serve as a stimulus for autophagy (Wen et al. 2013).

The hypoxic G93A-NSC line had two particularly interesting metabolic alterations: the low level of NAD(P)+, a regulator of the activity of sirtuin 1, which controls autophagy and the activity of HIF-1 (Lee et al. 2008; Lim et al. 2010) and, indirectly, also the activity of Akt, the master of a powerful anti-autophagic pathway (Hui et al. 2008; Jung et al. 2010), and the high expression level of PDK1. The induction of PDK1 shunts pyruvate away from the mitochondria and contributes to the use of reductive carboxylation of glutamine for fatty-acid synthesis (Metallo et al. 2011; Semenza 2011), governing metabolite fluxes from glycolysis and glutaminolysis in the tricarboxylic acid cycle and synthesis of lipids (Metallo et al. 2011; Wise et al. 2011). These pathways control mTORC1 activity and autophagy (Durán et al. 2012) and the activity of prolyl hydroxylases involved in the regulation of stability of the α subunit of HIF-1 (Pan et al. 2007).

Earlier studies in normoxic conditions with NSC-34 cell lines stably expressing high/low levels of human mutant SOD1 transgenes and cultured with/without the addition of oxidants highlighted alterations to metabolic markers, including some of those examined in this study (Mali and Zisapels 2008; Mali and Zisapel 2010; D'Alessandro et al. 2011; Richardson et al. 2013). Reduced O2 concentration is also a stimulus causing enhanced ROS formation; these results as a whole suggest that oxidative stress causes significant metabolic derangements and this occurs when cells express either a high level of mutant SOD1 or a low level but are exposed to oxidants.

Mitochondrial dysfunction is present in models expressing a low amount of G93ASOD1 including the one used in this study (Menzies et al. 2002; Rizzardini et al. 2005; Raimondi et al. 2006). Hypoxia might impinge on antecedent mitochondrial dysfunction; toxic interactions of G93ASOD1 with functional molecules linked to apoptosis and control of mitochondrial membrane potential such as Bcl-2 and voltage-dependent anion channel have been described (Tan et al. 2013). In turn, as discussed above, in the G93A-NSC line, the hypoxia-induced dysregulation of autophagy may decrease the clearance of mutant SOD1 favoring the accumulation of the mutant protein in the mitochondria and this is detrimental to these organelles (Barber and Shaw 2010).

Mutant forms of SOD1 might affect glucose/glutamine metabolism, changing the cell's metabolic adaptation to hypoxia. Interestingly, in yeast SOD1 has a role in casein kinases γ-mediated regulation of fermentative metabolism, which in mammals is through the Wnt signaling pathway (Reddi and Culotta 2013), which involves endosome and lysosome function (Dobrowolski and De Robertis 2011).

In conclusion, our findings highlight within a cellular model of ALS an inappropriate adaptive metabolic response to low availability of O2, which could increase the cell susceptibility to this stress. It is tempting to suggest that in these conditions the G93A mutation of SOD1 might convert autophagy up-regulation from a neuroprotective to a toxic player illustrating the potential difficulties of designing a treatment including enhancement of this pathway. Hypoxia toxicity to G93A-NSC cells was rescued by treatment with 3-MA, so the study also provides a pointer to potential protective pathways.

Acknowledgments and conflict of interest disclosure

We are grateful to Dr N.R. Cashman for providing the original NSC-34 cell line. We thank Dr M. Salmona for the synthesis of the immunizing peptide to obtain the polyclonal antiserum against Lon protease and Dr G. D'Alessandro for testing the activity of this antiserum; Prof J. Orly for helpful discussion; Dr D. Albani for critically reading the manuscript. The research leading to these results has received funding from the European Community's Health Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 259867 (to LC) and from Ministero dell'Istruzione, Università e Ricerca (MIUR), Fondo Investimenti Ricerca di Base (FIRB), Protocol RBIN04J58W (to LC).

The authors have no conflict of interest to declare.

Ancillary