Proteasomal inhibition by misfolded mutant superoxide dismutase 1 induces selective motor neuron death in familial amyotrophic lateral sclerosis


Address correspondence and reprint requests to Ryosuke Takahashi, Laboratory for Motor System Neurodegeneration, Brain Science Institute, RIKEN, 2–1 Hirosawa, Wako, Saitama, 351–0198, Japan. E-mail:


Accumulating evidence indicates that abnormal conformation of mutant superoxide dismutase 1 (SOD1) is an essential feature underlying the pathogenesis of mutant SOD1-linked familial amyotrophic lateral sclerosis (ALS). Here we investigated the role of ubiquitin-proteasome pathway in the mutant SOD1-related cell death and the effect of oxidative stress on the misfolding of mutant SOD1. Transient overexpression of ubiquitin with human SOD1 (wild-type, ala4val, gly85arg, gly93ala) in Neuro2A cells decreased the amount of mutant SOD1, but not of wild-type, while only mutants were co-immunoprecipitated with poly-ubiquitin. Proteasome inhibition by lactacystin augmented accumulation of mutant SOD1 in the non-ionic detergent-insoluble fraction. The spinal cord lysates from mutant SOD1 transgenic mice showed multiple carbonylated proteins, including mutant SOD1 with SDS-resistant dimer formation. Furthermore, the treatment of hSOD1-expressing cells with hydrogen peroxide promoted the oligomerization, and detergent-insolubility of mutant SOD1 alone, and the oxidized mutant SOD1 proteins were more heavily poly-ubiquitinated. In Neuro2A cells stably expressing human SOD1 protein, the proteasome function measured by chymotrypsin-like activity, was decreased over time without a quantitative alteration of the 20S proteasomal component. Finally, primary motor neurons from the mouse embryonic spinal cord were more vulnerable to lactacystin than non-motor neurons. These results indicate that the sustained expression of mutant SOD1 leads to proteasomal inhibition and motor neuronal death, which in part explains the pathogenesis of mutant SOD1-linked ALS.

Abbreviations used

amyotrophic lateral sclerosis


copper chaperone for SOD1




superoxide dismutase 1

Amyotrophic lateral sclerosis (ALS) is a fatal paralytic neurodegenerative disease of unknown etiology. Since the discovery of a genetic defect in superoxide dismutase in a subpopulation of ALS patients (Rosen et al. 1993), the causative mechanism of motor neuron death has been intensively investigated. Protein chemistry, cell biology, and transgenic mice studies have provided a substantial basis for several hypotheses that explain the pathology and pathogenesis of ALS. The major theories include reactive oxygen species (Wiedau-Pazos et al. 1996; Bogdanov et al. 1998), excitotoxic amino acids (Bruijn et al. 1997; Trotti et al. 1999), axonal flow damage (Zhang et al. 1997; Williamson and Cleveland 1999; Nguyen et al. 2001), abnormal copper chemistry (Gabbianelli et al. 1999), and post-translational modifications such as glycation (Kato et al. 1999). However, how these mechanisms are involved in mutant SOD1-linked ALS pathology remains unclear.

Ubiquitinated and SOD1-immunopositive inclusions are the neuropathological features of SOD1-related familial ALS and its model mice (Tu et al. 1996; Kato et al. 2000), indicating that degradation impairment of mutant SOD1 through the ubiquitin-proteasome system are closely related to the pathogenesis of FALS. Recent progress of knowledge about ubiquitin-proteasome system (UPS) provides many insights into the pathogenesis of neurodegenerative diseases. Proteins that do not fold correctly, or that are abnormally modified after translation, are degraded in the ubiquitin-proteasomal pathway (Adamo et al. 1999; Ellgaard et al. 1999; Mori 2000). However, if abnormal proteins overcome the degradation capacity of the proteasomes, they may aggregate or form ubiquitinated cellular inclusions, a process implicated in the pathogenesis of several neurodegenerative diseases such as Alzheimer's disease, polyglutamine disease, Parkinson's disease and ALS (Julien 2001; Sherman and Goldberg 2001). Such abnormally folded proteins are assumed to be cytotoxic (Orr and Zoghbi 2000). However, the exact role of ubiquitinated inclusion or aggregation remains controversial, especially whether they are toxic or protective (Sherman and Goldberg 2001). Several lines of evidence indicate that FALS-related mutant SOD1 proteins are misfolded and unstable. Recombinant mutant SOD1 proteins sediment more easily than wild-type (Okado-Matsumoto et al. 2000), and the solubility of mutant SOD1 proteins (G41S and G93A) from stable cell lines or from the spinal cord tissues of transgenic mice, is decreased in non-ionic detergent (Shinder et al. 2001). Various mutant SOD1 recombinant proteins showed decreased thermal stability and metallation (Hayward et al. 2002; Rodriguez et al. 2002). In contrast, mutant SOD1 proteins are rapidly degraded when transiently expressed in vivo and proteasomal inhibition prolongs their half-lives (Hoffman et al. 1996; Nakano et al. 1996; Johnston et al. 2000). Johnston et al. (2000) showed that mutant SOD1 proteins forms non-native oligomers in the cell-line or spinal cord lysates form the transgenic mice, and that proteasomal inhibition induces aggresome formation. However, the role of the misfolded nature of mutant SOD1 to its toxicity, especially, the temporal relationship between protein misfolding and pathological aggregation, remains elusive.

Oxidative stress has been implicated in the pathogenesis of mutant SOD1-related FALS. Several markers for oxidative damage are elevated in the CNS tissues from both sporadic and familial ALS patients (Ferrante et al. 1997). Mutant SOD1 is reported to generate hydroxyl radicals (Bogdanov et al. 1998) and inhibit GLT-1, a subtype of the glutamate transporter (Trotti et al. 1999). This adverse effect of mutant SOD1 to generate reactive oxygen species is reportedly based on abnormal copper chemistry, and copper chaperone for SOD1 (CCS) plays an exclusive and crucial role for the incorporation of copper into SOD1 (Casareno et al. 1998; Corson et al. 1998). However, CCS knockout in the mutant SOD1-transgenic mice failed to improve the clinical course of mutant SOD1-transgenic mice, which may argue abnormal copper chemistry hypothesis (Subramaniam et al. 2002). Another important finding is the oxidation of mutant SOD1 itself in the spinal cord lysates from mutant SOD1-transgenic mice (Andrus et al. 1998). However, the pathological role of oxidation of mutant SOD1, especially in the context of ubiquitin-proteasome pathway, is unknown.

Here, we show that mutant SOD1 proteins, unlike the wild-type, are specifically degraded through the ubiquitin-proteasome system and oxidative stress to mutant SOD1 exacerbates its misfolding and promotes its ubiquitination. We also provide evidence that proteasomal inhibition caused by mutant SOD1 leads to selective motor neuron death in familial ALS.

Materials and methods


Mouse monoclonal anti-FLAG antibody (M2) and rat monoclonal anti hemagglutinin (HA) antibodies were purchased from Sigma (St Louis, MO, USA) and Roche (Basel, Switzerland), respectively. Rabbit polyclonal antibody against human SOD1, which also reacts with mouse SOD1, was purchased from StressGen (Victoria, BC, Canada). An antibody against 20S proteasome a subunit type I and peripherin were purchased from Calbiochem (San Diego, CA, USA) and Chemicon (Temecula, CA, USA), respectively. Anti-choline acetyl transferase antibody was a gift from Dr Hidemi Misawa (Tokyo Metropolitan Institute for Neuroscience). Lactacystin was purchased from Kyowa Medex (Tokyo, Japan). BDNF, CNTF, and NT3 were from Peprotech (Rocky Hill, NJ, USA) and insulin, sodium selenite, corticosterone, and DAB were from Nacalai Tesque (Kyoto, Japan). Other chemicals were purchased from Sigma. Neuro2A cells were a gift from Guan-Hui Wang (CAG Laboratory, RIKEN, BSI, Japan). Human pCGN-ubiquitin tagged with HA was a gift from Shigetsugu Hatakeyama (Kyushu University, Fukuoka, Japan).

Culture, construct, and transfection of cell-lines

Murine neuroblastoma Neuro2A and human embryonic kidney (HEK) 293T cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS). Transfection was performed using Lipofectamin PLUS (Invitrogen, Carlsbad, CA, USA), according to the manufacturer's protocols. The medium for Neuro2A cells was replaced with nutrient medium containing 5 mm dibutylyl cyclic-AMP 3 h after transfection to differentiate the cells. Wild-type human SOD1 DNA was cloned by RT-PCR from poly A-RNA extracted from the peripheral blood of a normal volunteer, and subcloned into pcDNA3 between the EcoRV/XhoI sites (pcDNA3-hSOD1). The mutant hSOD1 gene was generated by Quick Change site-directed mutagenesis (Stratagene, La Jolla, CA, USA) according to the manufacturer's protocol. FLAG-tagged SOD1 at the carboxyl (C)-terminus (pcDNA3-hSOD1-FLAG) was prepared using conventional PCR and blunt-ligated into pcDNA3 at the EcoRV site, using pcDNA3-hSOD1 as a template. The primer pairs were as follows: 5′-GCCGATATCATGGCGACGAAGGCCGTG-3′ and 5′-CTCTCGAGTTATTGGGCGATCCCAATTACACCA-3′ for pcDNA3-SOD1, 5′-GCCGATATCATGGCGACGAAGGCCGTG-3′ and 5′-GGGGATATCTCACTTGTCGTCATCGTCTTTGTAGTCTTGGGCGATCCCAATTAC-3′ for pcDNA3-hSOD1-FLAG. The underlined cytosine of the forward primer was converted to thymine to construct the A4V mutant. To generate Neuro2A cells stably expressing wild-type or mutant human SOD1, pcDNA3-SOD1-FLAG was digested at BamHI/XhoI sites and subcloned between BglII/SalI sites of the pIRES2-EGFP plasmid in which the insert was followed by IRES-EGFP cDNA (Clontech, Palo Alto, CA, USA). Plasmids were transfected into Neuro2A cells, then fluorescent and G418-resistant clones were selected and cloned (Neuro2A-SG).

Transgenic mice

Transgenic mice harbouring mutant (B6SJL-TgN[SOD1-G93A]1Gur, hSOD1G93A)) and wild-type (C57BL/6-TgN[SOD1]3Cje, hSOD1WT) human SOD1 were purchased from Jackson laboratory (Bar Harbor, ME, USA). Transgenic mice were generated and screened according to conventional methods (Gurney et al. 1996). Both hSOD1G93A and hSOD1WT were bred in the C57BL/6 background. Animals were treated in compliance with the RIKEN Guide for Animal Care for Research Use.


Cells were harvested and lysed in TN buffer consisting of 50 mm Tris-HCl, 150 mm NaCl, and a protease inhibitor cocktail (Roche) containing 1% Triton X-100 (TN-T) for 2 h at 4°C. After centrifugation at 15 000 r.p.m. (20 400 × g) for 30 min, protein in the supernatant was quantified using the Coomassie protein assay reagent or BCA protein assay kit (Pierce, Rockford, IL, USA), and equal concentrations of lysates were mixed with the same volume of 4% SDS sample buffer. The pellet was washed once with TN-T buffer, solubilized with an equivalent volume of 4% SDS sample buffer according to the protein concentration in the soluble fraction, sonicated and boiled for 5 min. The spinal cords from the transgenic or non-transgenic mice were homogenized using a Tephrone homogenizer in TN-T buffer, and were centrifugated at 15 000 r.p.m. (20 400 × g) for 30 min. Lysates were resolved by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to PVDF membranes (Millipore Corporation, Bedford, MA, USA). Immunoreactivity was detected using an enhanced chemiluminescence detection kit (ECL; Amersham Pharmacia Biotech, Piscataway, NJ, USA).

In vivo ubiquitination assay

Neuro2A cells plated on 6-well culture dishes were transiently co-transfected with pcDNA3-SOD1-FLAG and pCGN-HA-ubiquitin. After solubilization with RIPA buffer (20 mm Hepes pH 7.4, 150 mm NaCl, 2 mm EDTA, 1% Nonidet-P40, 1% Sodium deoxycholate, 0.1% SDS, and protease inhibitor cocktail), the cell lysates were immunoprecipitated using monoclonal anti-FLAG M2 agarose affinity gel (Sigma) for 2 h at 4°C, washed five times with RIPA buffer and finally eluted by boiling in 4% SDS sample buffer for 5 min. Cells were exposed to hydrogen peroxide at the indicated concentrations for 60 min before harvest in TNG-T buffer (50 mm Tris-HCl, pH 7.5, 150 mm NaCl, 10% Glycerol, 1% Triton X-100 and protease inhibitor cocktail). Subsequent manipulations proceeded as described above, except for using TNG-T buffer. After protein transfer, membranes were immunoblotted with anti-HA antibody to evaluate poly-ubiquitination.

Expression and purification of recombinant human SOD1 protein

The pcDNA3-SOD1 (wild-type, gly85arg; G85R and gly93ala; G93A mutants) was digested and subcloned into EcoRI/XhoI sites positioned after the GST sequence of the E. coli expression vector, pGEX6p-1 (Clontech). After induction by 1 mm isopropyl-1-thio-s-d-galactopyranoside (IPTG) for 1 h at 30°C in E. coli (BL21 strain), the cells were pelleted through one cycle of freeze-thaw in PBS containing 1% Triton X-100, followed by centrifugation at 10 000 g for 10 min. The supernatant was passed through a glutathione Sepharose 4B column (Amersham), and SOD1 was released from the GST by digestion using Precision (Amersham) in cleavage buffer consisting of 50 mm Tris-HCl (pH 7.5), 150 mm NaCl, 1 mm EDTA, and 1 mm DTT. The cleavage buffer was replaced with PBS by dialysis. This apoenzyme was refolded by adding CuSO4 (200 µm) and ZnCl2 (200 µm) for 1 h at 4°C in the same buffer as described (Volkel et al. 2001). The identity and function of the recombinant protein were verified by SDS-PAGE, followed by immunoblotting with anti-hSOD1 antibody and by the nitro-blue tetrazolium gel activity assay (Ratovitski et al. 1999). SOD activity was obtained at least 30 min after refolding (highest from 24 to 48 h after refolding). The enzymatic activity of wild-type and G93A mutant SOD1 was almost the same, whereas that of G85R mutant was decreased to about 5–10% of them (data not shown).

Protein oxidation assay

Protein oxidation was evaluated by analyzing carbonyl modification using a kit (Oxyblot STM; Intergen, Milford, MA, USA) according to the manufacturer's protocol. In brief, resected spinal cord from hSOD1wild-type, hSOD1G93A, or non-transgenic littermates of hSOD1G93A mice, was solubilized with the TN-T buffer containing 1% 2-mercaptoethanol. Alternatively, 3 µg of bacterially purified recombinant SOD1 was reacted with H2O2 at the indicated concentrations for 60 min at 22°C. The reaction was terminated by adding catalase and 2-mercaptoetanol (Sampson and Beckman 2001). Tissue lysates (20 µg protein) or recombinant SOD1 (2 µg) was denatured in 6% SDS and reacted with 2,4-dinitrophenylhydrazine (DNP) to derivatize carbonyl groups for 15 min at room temperature. After neutralizing the reaction mixture, the products were resolved by SDS-PAGE and transferred to PVDF membranes where carbonylated protein was detected using a rabbit polyclonal anti-DNP antibody.

Measurement of proteasome activity

The 20S proteasomal activity in Neuro2A cells stably transfected with human SOD1 was estimated as a chymotrypsin-like activity as described with minor modifications (Jana et al. 2001). Frozen stock cells were thawed and seeded in 6-cm culture dishes at a density of 1 × 106 cells/dish in nutrient medium containing 0.7 mg/mL G418 (Nacalai Tesque, Kyoto, Japan). The cells were harvested in cold PBS 4 and 7 days after seeding, and pellets were stored at − 80°C. Cultures were homogenized in proteolysis buffer (10 mm Tris-HCl (pH 7.4), 1 mm EDTA, 5 mm dithiothreitol, 5 mm ATP and 20% Glycerol), and separated by centrifugation at 15 000 r.p.m. (20 400 × g) for 15 min. The protein concentration of the supernatant was determined using a Coomassie reagent kit (Pierce). Aliquots of 25 µg/100 µL of supernatant were incubated with 100 µL of proteasome assay buffer (50 mm Tris-HCl, 0.5 mm EDTA) containing 50 µm succinyl-Leu-Leu-Val-Tyr-4-methylcoumaryl-7-amide (suc-LLVY-MCA; Sigma), a fluorogenic substrate of chymotrypsin, for 2 h at 37°C. Proteasomal activity was determined as a linear increase in chymotrypsin activity over 60 min, monitored every 10 min at 355 nm excitation and 460 nm emission using a fluorometer (Fluoroskan Ascent FL; Labsystem, Chicago, IL, USA).

Primary motor neuron culture

Primary cultures of embryonic murine spinal cord were prepared with reference to published methods with major modifications for motor neuron enrichment (Durham et al. 1997; Urushitani et al. 1998). The complete spinal cord from the E12 embryo of C57/B6 mice was resected and incubated in 0.05% trypsin for 20 min at 37°C. Cells were dissociated by gentle pipetting, and plated in culture dishes coated with polyethyleneimine at a density of 4 × 105 cells/cm2 in DMEM/F12 Ham's medium (DF; Sigma) containing 2 ng/mL basic FGF (Toyobo, Osaka, Japan) and N2 supplement (Invitrogen). One hour later, ten-fold volume of neurogrowth medium consisting of 2% horse serum, 10 µg/mL bovine serum albumin, 10 µg/mL insulin, 26 ng/mL sodium selenite, 100 µg/mL conalbumin, 13 ng/mL progesterone, 32 µg/mL putrescine, 20 ng/mL corticosterone, 20 ng/mL triiodotyronine, 0.1 ng/mL brain-derived neurotrophic neurotrophic factor (BDNF), 10 ng/mL ciliary neurotrophic factor (CNTF), and 0.1 ng/mL neurotrophin-3 (NT3) in DF medium was added. Cultured cells were assayed 7–8 days after seeding.

Immunocytochemistry and toxicity assay

Primary cultures were fixed with 4% paraformaldehyde (PFA) for 20 min at room temperature. After incubation in blocking serum containing 0.2% Triton X-100 for 20 min at room temperature, cultures were reacted with primary antibody in 3% blocking serum in Tris-buffered Saline (TBS). After incubation with the biotinylated secondary antibody (1 : 200; Vector Laboratories Burlingame, CA, USA) for 1 h at room temperature, cultures were reacted with avidin-biotinylated horseradish peroxidase complex (ABC Kit Elite, Vector Laboratories), followed by diaminobenzidine (DAB). Several experiments also included the fluorescent secondary antibodies Alexa 488 and 546 (Molecular Probes. Eugene, OR, USA). To stain nuclei, 300 nm 4′,6-diamino-2-phenylindole (DAPI) was added to cultures for 10 min in the dark after the reaction with the secondary antibody. The number of surviving neurons stained with anti-microtubule-associated protein 2 (MAP2) antibody (1 : 400. Chemicon) was taken as the number of the total neurons, whereas the number of large cells stained with SMI32, an anti-nonphospholyrated neurofilament-H (1 : 4000. Steinberger Monoclonals. Baltimore, MD, USA) antibody, was taken as the total number of motor neurons (Carriedo et al. 1996; Urushitani et al. 1998). Large SMI32-positive neurons were also stained with antibodies against peripherin (1: 200) and ChAT (1 : 1000) antibodies. To count the surviving neurons, we observed four randomly selected regions containing three continuous fields using a phase-contrast microscope. Each region contained at least 300 and 100 anti-MAP2-positive and SMI32-positive neurons, respectively, under sham-operated conditions. Thereafter, we counted the number of neurons, corrected each value to the number of cells/cm2, and obtained the number of non-motor neurons by subtracting the number of SMI32-positive neurons from that of anti-MAP2-positive neurons.

Statistical analysis

Statistical significance was evaluated by single factor anova (analysis of variance), or two-way anova, to evaluate the selectivity of toxicity in mutant-type or motor neurons, followed by Scheffe's method.


Involvement of ubiquitin-proteasomal pathway in the metabolism of mutant SOD1

The neuropathological features of mutant SOD1-related ALS and mutant SOD1 transgenic mouse are hyaline inclusions immunoreactive to both SOD1 protein and ubiquitin (Bruijn et al. 1998; Kato et al. 1999). Moreover, it has been reported that mutant SOD1 proteins are unstable and stabilized by proteasome inhibitors, indicating that SOD1 is degraded through ubiquitin-proteasome system (Hoffman et al. 1996; Johnston et al. 2000). Therefore, we first conducted an in vivo ubiquitination study using murine neuroblastoma cell-line, Neuro2A cells, that were transiently co-transfected with plasmids expressing human SOD1-FLAG and haemagglutinin (HA)-ubiquitin. Figure 1a shows that mutant SOD1 proteins (ala4val-A4V, gly85arg-G85R, and gly93ala-G93A), but not wild type, were immunoprecipitated with a high molecular ubiquitin chain in the RIPA buffer, indicating that the mutant SOD1 protein is poly-ubiquitinated. Wild-type SOD1 protein was mono-, but not poly-ubiquitinated (Fig. 1a, right upper). Moreover, mutant SOD1 proteins seemed degraded in the presence of HA-ubiquitin (Fig. 1a, left lower). These results confirmed that mutant SOD1 proteins are degraded through the ubiquitin-proteasome pathway, whereas wild-type proteins are not. We next studied the effect of proteasome inhibition on the expression level of transiently expressed human SOD1 protein in the neuroblastoma cell-line, Neuro2A. The expression of two mutant SOD1 proteins (G85R, G93A) was lower in both the detergent-soluble and -insoluble fractions without lactacystin, indicating their state of degradation. Proteasome inhibition by lactacystin rather decreased the levels of these mutants in the detergent-soluble fraction (Fig. 1b, upper panel). Conversely, lactacystin led to the accumulation of mutants, but not of the wild-type, in the detergent-insoluble fraction in a dose-dependent manner (Fig. 1b, middle panel). These results indicate that mutant SOD1 is contitutively degraded through ubiquitin-proteasome pathway because of its misfolded nature.

Figure 1.

Involvement of ubiquitin-proteasomal pathway in the metabolism of mutant SOD1. (a) Murine neuroblastoma cell-line, Neuro2A cells at 80% confluence in 6-well culture dishes were transiently transfected with pcDNA3-hSOD1-FLAG (wild-type, ala4val; A4V, gly85arg; G85R, gly93ala; G93A) or pcDNA3 as a control, with or without pCGN-HA-ubiquitin. Cells were harvested and lysed in RIPA buffer 24 h after transfection, and lysates were immunoprecipitated on anti-FLAG agarose gels for 2 h at 4°C. Five per cent of input and immunoprecipitated fractions were resolved on gradient SDS-polyacrylamide gels (20–4%), then immunoblotted against anti-HA (left upper and right upper panels) and anti-hSOD1 antibodies (left lower and right lower panels) Arrowhead indicates hSOD1 and asterisk shows endogenous mouse SOD1. (b) Mutant SOD1 decreased detergent-solubility that was enhanced by proteasome inhibitor lactacystin. Neuro2A cells at 70% confluence in 12-well culture dishes, were transiently transfected with pcDNA3-hSOD1 (wild type, G85R, or G93A mutant) or pcDNA3 as a control (1.5 µg total DNA), then incubated with lactacystin (Lacta) 24 h later for an additional 24 h with the indicated concentrations. Cells were lysed in buffer containing 1% Triton X-100, and pellets were solubilized in equivalent volumes of sample buffer containing 4% SDS as described in Materials and methods. Identical volumes of detergent-soluble (sol) and -insoluble (insol) fractions were applied to the 16% SDS-polyacrylamide gels (25 µg protein per lane, soluble fraction), and each fraction was immunoblotted against anti-human SOD1 that also reacts with murine SOD1 (mSOD1). The lowest panel is Coomassie staining (CBB) to verify that equal amounts of proteins were included in each lane of detergent-insoluble fractions.

The self-oxidation of mutant SOD1 proteins formed non-native oligomer, exacerbated detergent-insolubility, and promoted their poly-ubiquitination

Because only mutants were obviously ubiquitinated and we detected poly-ubiquitinated SOD1 in the detergent-soluble fractions, properties other than detergent insolubility may be recognized by the ubiquitin-proteasome pathway. It is reported that mutant SOD1 forms soluble non-native oligomer in the spinal cord lysates from mutant SOD1-trangenic mice, or in the cell lysates from HEK cells expressing mutant SOD1 when they are treated with proteasome inhibitors (Johnston et al. 2000). We focused on the oxidative stress to mutant SOD1 itself, because mutant SOD1 in the spinal cord lysates from the transgenic mice, is reportedly modified by carbonyl moieties, indicating its oxidation (Andrus et al. 1998), and because oxidative stress is one of the major factors to denature and misfold the various proteins (Adamo et al. 1999). Immunoblotting of the spinal cord lysates from mutant (G93A) and wild-type hSOD1 transgenic mice, revealed that mutant SOD1, but not wild-type SOD1 formed an SDS-resistant dimer, which was evident in the older mice (Fig. 2a). Moreover, the lysates of hSOD1G93A mice showed enhanced multiple carbonylated molecules compared with non-transgenic littermates or hSOD1wild–type (Fig. 2b left panel). Furthermore, a 34-kDa band, corresponding to dimeric hSOD1 was also significantly carbonylated in the sample from hSODG93A mice, indicating that mutant SOD1 is oxidized (Fig. 2b right panel). Based on these results, we hypothesized that oxidative stress enhanced misfolding of mutant SOD1 proteins. When human embryonic kidney (HEK) 293T cells or Neuro2A cells, transiently expressing hSOD1 were exposed to hydrogen peroxide (H2O2), shifted bands with a high molecular weight corresponding to the dimer, trimer of hSOD1 proteins appeared (Figs 3a, parts 1 and 2, upper panels). In HEK293T cells A4V mutants similarly but faintly formed oligomers, presumably due to its poor dimer-forming activity (Deng et al. 1993). Moreover, H2O2 augmented the accumulation of mutant SOD1 proteins in detergent-insoluble fractions of Neuro2A cells. These results indicate that oxidative stress promotes the oligomerization and detergent-insolubility of mutant, but not wild-type SOD1 in vivo.

Figure 2.

Non-native dimer formation and carbonylation of mutant SOD1 in the spinal cord lysates from hSOD1G93A transgenic mice. The hSOD1G93A mice of 1, 3 (pre-clinical), and 8 (post-clinical) months of age, corresponding non-transgenic littermates, and 8 months old hSOD1WT mice were sacrificed. (a) Mutant SOD1 protein forms non-native dimer in the spinal cord lysates from the transgenic mice. The lysates (30 µg protein) from the 1, 3 months (hSOD1G93A and non-transgenic littermates), and 8 months (hSOD1WT, hSOD1G93A, and non-transgenic littermates) of age were immunoblotted using anti-hSOD1 and -actin antibodies. (b) Carbonylation of mutant SOD1 in the spinal cord lysates of G93A transgenic mice. After terminating the DNPH-derivative reaction with neutralization buffer, 10 µg protein from the same lysates as those in (a) were electrophoresed on a 4–20% gradient SDS gel in duplicates in the same gel, and followed by immunoblotting by anti-DNP and anti-SOD1 antibodies. *Monomeric hSOD1 proteins; **dimeric mutant SOD1. Each data is a representative data from three independent experiments.

Figure 3.

Mutant SOD1 proteins that were oxidized and oligomerized became more detergent-insoluble upon oxidative stress. (a) Oxidative stress induced by hydrogen peroxide promoted the oligomerization and detergent-insolubility of mutant proteins in transiently transfected Neuro2A (1) and HEK293T cells (2). Cells at 80% confluence in 6-well culture dishes were transiently transfected with pcDNA3-hSOD1-FLAG or pcDNA3 as a control for 24 h. For 60 min before harvest, cells were incubated with hydrogen peroxide (H2O2) at various concentrations. Supernatant of HEK293T as well as Neuro2A cell supernatant and pellet were separated by SDS-PAGE and immunoblotted against anti-FLAG (1) and anti-hSOD1 antibodies (1 and 2). Blots shown here were over-exposed to enhance high molecular weight bands. (1) Arrowheads correspond to dimer of G85R or G93A mutant (sol; upper panel). In detergent-insoluble fractions (insol), only the mutant SOD1 signal increased in the presence of H2O2 (middle panel). The lowest panel in insoluble fraction is Coomassie staining (CBB) to verify that equal amount of proteins were loaded on electrophoretic gels. mSOD (endo) indicates endogenous mice SOD1 protein. (2) Single and double arrowheads in soluble fraction are compatible with trimer and dimer molecular weights, respectively, and triple arrowheads indicate G85R mutant dimer. hSOD (endo) indicates endogenous human SOD1 protein. (b and c) Bacterially purified SOD1 was oxidized by hydrogen peroxide (H2O2). Recombinant SOD1 protein was reacted with H2O2 for 60 min at room temperature. Reactions were terminated by adding catalase (b and c) and 2-mercaptoetanol (b). (b) hSOD1 (2 µg) was used in protein carbonyl assays with anti2,4-dinitriphenylhydrazine (DNP) antibody. (c) After reaction with H2O2 (0, 10, 100, 1000 µm). (c) hSOD1 proteins with (lanes 1–3, 5–7, and 9–11) or without (lanes 4, 8, and 12) metallation with Cu2+ and Zn2+ were separated by SDS-PAGE and immunoblotted against antihSOD1 antibody (upper and lower panels are long and short exposure, respectively) after reaction with H2O2 (0 and 100 µm for lanes 1–2, 5–6, and 9–10, respectively. 1000 µm for lanes 3, 4, 7, 8, 11, and 12). Asterisk shows catalase. Arrowheads indicate dimeric mutant SOD1 (single and double arrowheads indicate trimers of G93A and dimers of wild-type, G85R, and G93A types, respectively).

To investigate the direct effect of reactive oxygen species on the SOD1 protein, we purified recombinant hSOD1 protein from E. coli as described (Volkel et al. 2001), and examined the oxidation of SOD1 protein, because carbonyl modification occurs not only in oxidation but also in glycation (Stefek et al. 1999), which is frequently undergone by both wild and mutant-type SOD1 proteins (Mrabet et al. 1992). However, we directly showed that SOD1 itself was oxidized by H2O2 without the effect of glycation using bacterially purified hSOD1 (Fig. 3b). The G85R mutant was already oxidized in the absence of H2O2, indicating that this mutant is very susceptible to oxidation. Moreover, mutant SOD1, especially G93A, formed oligomer to multimer with a decrease in monomeric form (Fig. 3c, lanes 6, 7, 9–11). This oligomerization crucially depends on the metal because H2O2 gave no effect on mutant apoenzymes (Fig. 3c lanes 7, 8, 11, and 12). We further investigated the effect of mutant SOD1 protein auto-oxidation on poly-ubiquitination. Figure 4a shows that H2O2 remarkably enhanced the poly-ubiquitination of mutant, but not wild-type, SOD1 from Neuro2A cells.

Figure 4.

Poly-ubiquitination of mutant SOD1 was promoted by oxidative stress. Neuro2A cells at 80% confluence in 6-well culture plates were transiently transfected with pcDNA3-hSOD1-FLAG or pcDNA3 as a control, with/without pCGN-HA-ubiquitin (HA-Ub). For 60 min before harvest in 80 buffer, several wells were incubated with 1.5 mm H2O2. Lysates were incubated with anti-FLAG affinity gel for 1 h at 4°C, immunoprecipitates were separated by gradient SDS-PAGE (20–4%), and immunoblotted against anti-HA (a) or anti-hSOD1 (b) antibody. (c) Five per cent input of total lysates, showing the same amount of poly-HA-ubiquitin (shown by Ubn). *Mono-HA-ubiquitin.

Time-dependent decrease in proteasome activity in Neuro2A cells stably expressing mutant SOD1 proteins

We showed that transiently expressed mutant SOD1 proteins are poly-ubiquitinated then rapidly degraded, findings with which detergent-insolubility and oligomerization correlate well. Based on these observations, we surmised that the continuous and preferential degradation of mutant SOD1 proteins would overload proteasomes, resulting in impaired function or ‘choking’ of proteasomes (Cleveland and Rothstein 2001). To test this notion, we generated a Neuro2A cell-line that stably expressed human SOD1 (Neuro2A-SG). The amount of proteasome catalytic activity measured as chymotrypsin-like activity, was significantly decreased in Neuro2A-SG cells carrying mutant SOD1 7 days after seeding, compared with vector control or wild-type on the same day, and with mutants on day 4 (Fig. 5a). We then examined the amount of proteasome components in the cytosol, because proteasome components are recruited with polyglutamine aggregates, which is likely to decrease cytosolic catalytic activity (Jana et al. 2001). Immunoblots using the same lysates used in the proteasome activity measurement revealed that 20Sα proteasome was equally expressed in the cytosol of Neuro2A-SG carrying vector control, wild-type and mutants, on days 4 and 7 (Fig. 5b left panels), whilst mutant SOD1 accumulated in the detergent-insoluble fraction of the lysates on day 7 (Fig. 5b right panels). These results suggest that the continuous degradation of mutant SOD1 protein leads to the proteasomal dysfunction through a mechanism distinct from the sequestration of proteasome into the detergent-insoluble fraction.

Figure 5.

Time-dependent decrease in proteasomal activity of Neuro2A-SG stably expressing human SOD1, without altered proteasome content. Frozen stocks of Neuro2A-SG cells were thawed and seeded onto 6 cm culture dishes at a density of 1.0 × 106 cells per plate. Cells were washed and harvested with PBS 4 and 7 days after seeding and stored at −80°C before the following assays. (a) Proteasomal activity in stable transfectants measured as chymotrypsin-like activity. Cells were thawed and solubilized in proteasome lysis buffer by sonication. After centrifugation, lysates (25 µg protein/100 µL proteasome lysis buffer) were reacted with 100 µL of proteasome assay buffer containing fluorogenic substrate, Suc-LLVY-MCA (50 µm), and monitored at 355 nm excitation and 460 nm emission every 10 min at 37°C to determine linearity of the enzymatic reaction. Proteasomal activity is expressed as altered fluorescence intensity per 60 min in the linear range. Values are means of triplicates ± SD and are representative of three different experiments. *p < 0.01 versus proteasomal activity on day 4. #p < 0.01 versus proteasomal activity of wild-type. (b) The proteasomal component was not quantitatively altered despite decreased activity. Immunoblotting using anti-proteasome 20Sα type-I (20Sα, or anti-hSOD1 antibody. Pellets were incubated with TN-T buffer containing 1% Triton X-100 for 2 h at 4°C and centrifuged. The pellet was solubilized using 4% SDS sample buffer. The same lysates as used in (a) (supernatant; 25 µg protein) and detergent-insoluble fractions (insoluble) were separated by SDS-PAGE (16%) and immunoblotted. Arrowheads indicate hSOD1; *murine endogenous SOD1.

Motor neurons are more vulnerable to proteasomal inhibition than non-motor neurons

As we have found that mutant SOD1 proteins induce proteasome inhibition, we hypothesized that motor neurons are vulnerable to this condition. To test this hypothesis, we established a motor neuron-enriched primary culture system, by using basic FGF in the initial culture, followed by CNTF, BDNF, and NT3. The number of total neurons stained with anti-MAP2 antibody was 163 947 ± 1891 cells/cm2, whereas that of SMI32-positive neurons was 36 896 ± 2137 cells/cm2 (four sister cultures). Large SMI32-positive neurons were exclusively motor neurons in the spinal cord (Carriedo et al. 1996; Urushitani et al. 1998), and they were also immunoreactive to anti-ChAT and peripherin antibodies (Fig. 6a).

Figure 6.

Cultured motor neurons from embryonic murine spinal cords are vulnerable to proteasomal inhibition caused by lactacystin. (a) Immunocytochemistry of cultured motor neuron from embryonic mice embryos at 8 days in vitro. SMI32-positive large neurons (1 and 4) were also immunoreactive with anti-ChAT (αChAT; 2) and anti-peripherin (αperipherin; 5) antibodies. (3) and (6) Merged images from i-ii and iv-v, respectively. Scale bar = 20 µm. (b) Lactacystin was applied at 7 days in vitro for 24 h before fixation in 4% paraformaldehyde. (1) Number of motor neurons was taken as that of SMI32-positive neurons. (2) Number of non-motor neurons was obtained by subtracting the number of SMI32-positive neurons from that of αMAP2-positive cells. *p < 0.05 versus sham-treated neurons. Data represent the mean values ± SEM (n = 4 sister cultures). (3) Photographic representation of lactacystin (0.5 µm)-induced toxicity in motor neurons. (I), (ii) and (iii), Sham operation; (ii), (v) and (vi), 0.5 µm lactacystin for 24 h. (i) and (iv), Anti-MAP2 antibody-positive neurons. (ii) and (v), SMI32-positive motor neurons; (i), (ii), (iv) and (v), Immunoreactive cells detected using DAB. (iii) and (vi), cultures stained with SMI32 followed by 300 nm DAPI. Scale bar = 50 mm.

We used this culture system to study the effect of proteasomal inhibition on motor neuron survival. The viability of cultured motor neurons was decreased by lactacystin at concentrations as low as 0.05 µm, whereas non-motor neurons were not affected (Fig. 6b). Nuclear staining with DAPI revealed that motor neuron death is partially apoptotic, with several shrunken motor neurons having an intact nuclear morphology [Fig. 6b, part 3, panels (iii) and (vi)], indicating that the death of these cells induced by proteasomal inhibition is a result of both apoptotic and non-apoptotic mechanisms (Sperandio et al. 2000).


In the present study, we have provided direct evidence showing that mutant SOD1, but not wild-type is poly-ubiquitinated and degraded by proteasome, and that oxidative stress exacerbates the detergent-insolubility and non-native oligomer formation, resulting in enhancement of ubiquitination of mutant SOD1. Moreover, the perpetual expression of mutant SOD1 in neuroblastoma cell-lines leads to proteasomal dysfunction. Furthermore, motor neurons from embryonic spinal cord were vulnerable to the proteasomal inhibition. These findings strongly support the presumed idea that misfolding of mutant SOD1 is one of the causal features for FALS pathology (Johnston et al. 2000; Kopito and Ron 2000).

The mechanism of ubiquitination of mutant SOD1 proteins remains unclear. Our results indicate that non-native oligomer formation of mutant SOD1 is one of key features for ubiquitination. The non-native oligomer formation and the altered detergent-solubility have been previously pointed out by different groups as two structural abnormalities of mutant SOD1 proteins (Johnston et al. 2000; Shinder et al. 2001). Our results extend their findings and implicate that these features are recognized by ubiquitination machinery of mutant SOD1. As there are more than 90 mutants of SOD1 associated with FALS, the existence of common machinery targeting misfolded SOD1 for ubiquitination is plausible, rather than mutant type-specific recognition system.

The role of oxidative stress in the pathogenesis of mutant SOD1-linked ALS is unclear. Mutant SOD1 itself generates reactive oxygen species (ROS), such as hydroxyl radicals via reaction with copper located in the active sites of SOD1 (Bogdanov et al. 1998; Eum and Kang 1999). However, a recent report showing that CSS-knockout mice expressing mutant SOD1 suffer from same clinical symptoms as mutant SOD1-transgenic mice expressing CCS (Subramaniam et al. 2002). However, regardless of its enzymatic activity, SOD1 is always exposed to substrate superoxide anions by electrostatic guidance formed by amino acid cite chains (Getzoff et al. 1992). We showed that the spinal cord of mutant SOD1 mice is exposed to oxidative stress, where mutant hSOD1 itself is more oxidized than wild-type. Moreover, we have directly shown that excessive oxidation of mutant SOD1 promotes oligomer and multimer formation depending on metals. The predisposition of oligomerization might be associated with the severity of mutant SOD1-related ALS, as G93A mutant is more prone to be oligomerized than G85R mutant, both in vivo and in vitro. Recombinant wild-type hSOD1 indeed formed dimer, but lack of multimer formation is a clear difference from mutant SOD1. These results are consistent with the previous report (Andrus et al. 1998). Although it is unclear whether mutant SOD1, in the absence of CSS, take the misfolded state by oxidative stress in vivo, factors other than CSS might play roles in linking oxidative stress to mutant SOD1. The reason why the level of oxidation in wild-type hSOD1 transgenic mouse is lower than in non-transgenic littermate mice is possibly because wild-type SOD1 function without adverse effect of radical generation. Furthermore, it is reported that proteasomal activity in rat spinal cords decreases in an age-dependent manner, and that this decrease correlates with the presence of ROS products (Keller et al. 2000). Our data, showing that self-oxidation of mutant SOD1 promotes formation of non-native oligomers and decreases detergent-solubility, have provided another possibility about the role of oxidative stress in the pathogenesis of mutant-SOD1-linked ALS.

Mutant SOD1 proteins were rapidly degraded when host cells were highly capable of proteasomal proteolysis. Our data demonstrated that transiently expressed mutant SOD1 proteins were ubiquitinated and degraded, without an obvious loss of cell viability or biochemical abnormality. However, in Neuro2A cells stably expressing mutant SOD1, proteasomal function was depressed in a time-dependent manner (Fig. 5a). Proteasomal impairment is implicated in the pathogenesis of other neurodegenerative diseases such as Parkinson's disease and polyglutamine disease (Bence et al. 2001; Jana et al. 2001; Tanaka et al. 2001). Proteasomes may be recruited into aggregation, resulting in a proteasomal decrease in both amount and function in cell model of polyglutamine disease (Jana et al. 2001). However, immunoblots of stable Neuro2A cell (Neuro2A-SG) lysates did not reveal any significant difference in the expression level of the cytosolic proteasomal component between 4 and 7 days after seeding. Moreover, we did not detect obvious aggregation by immunocytochemistry even when SOD1 was overexpressed using the cytomegalo-virus promoter. These results are consistent with the idea that the perpetual degradation of mutant SOD1 via ubiquitin-proteasome pathway, leads to proteasomal inhibition, as expressed as ‘proteasome choking’ (Cleveland and Rothstein 2001).

The proteasomal inhibition may result in accumulation of mutant SOD1 that exerts toxic functions such as generation of reactive oxygen species by mutant SOD1 (Liu et al. 1999), elevation of intracellular calcium concentration, and relevant mitochondrial dysfunction (Carri et al. 1997; Kruman and Mattson 1999). Moreover, proteasomal inhibition may allow pro-apoptotic molecules to escape proteolysis in the proteasome, resulting in activation of the caspase-dependent cascade, including the mitochondrial release of cytochrome c and mitochondrial membrane depolarization (Qiu et al. 2000; Wagenknecht et al. 2000; Jana et al. 2001).

We showed that cultured motor neurons from the embryonic mouse spinal cord are vulnerable to proteasomal inhibition. Proteasomal impairment is implicated in other neurodegenerative conditions such as Parkinson's, Alzheimer's and polyglutamine disease, in whose pathogenesis misfolded proteins are implicated. Based on these recent findings and our data, we speculate that SOD1-linked familial ALS is also caused by proteasomal dysfunction and that selective vulnerability of motor neurons to proteasomal inhibition provides the molecular basis for selective motor neuronal death. The reason why motor neurons are especially susceptible to proteasomal inhibition, however, remains unknown. We may theorize that unknown proteins that escape from proteasomal degradation may be especially toxic to motor neurons. Alternatively, the effect of proteasomal inhibition on the glial cells is also to be considered. Although we observed no overt toxicity in astrocytes by lactacystin treatment in our culture, some functional disturbance of glial cells such as glutamate transporter impairment, may be involved in motor neuron death.

Taken together, we propose the following working hypothesis. Constant proteolysis of mutant SOD1 proteins via the ubiquitin-proteasome cascade overloads proteasomes and inhibits their function. Oxidative stress augments conformational changes of mutant proteins, which promotes their ubiquitination. Impaired proteasomal function potentiates the accumulation of mutant SOD1, which exacerbates oxidative stress. Proteasomal impairment accelerated by this cycle predominantly injures motor neurons, and might explain the progressive pathology of ALS (Fig. 7).

Figure 7.

Schematic representation of our hypothetical mechanism of motor neuron death associated with mutant SOD1. Mutant SOD1 proteins tend to assume abnormal conformations including altered solubility and oligomerization (a). These features might be preferably recognized by ubiquitin-proteasome degradation pathway (b). Alternatively, auto-oxidation of mutant SOD1 by its substrate superoxide anion (O2) or its product hydrogen peroxide (H2O2) augments its conformational change (c). However, continuity of this process might overload or ‘choke’ proteasomes, which decrease proteolysis under physiological conditions (d). Moreover, motor neurons are vulnerable to proteasomal suppression (e). Furthermore, proteasomal inhibition helps abnormally conformed mSOD1 to accumulate, which might enhance this pathological cascade (f).


We thank Dr Masaharu Ogawa (Brain Science Institute, RIKEN), for critical advice regarding the primary culture of motor neurons from mouse embryos. We also thank Dr Hidemi Misawa (Tokyo Metropolitan Institute for Neuroscience), for helpful discussions. We thank Dr Tomoko Oeda (Kyoto Graduate School of Medicine), Dr Nihar Jana (CAG Repeat Laboratory, RIKEN BSI) for their experimental suggestions. This study was supported by research grants from RIKEN BSI, a Grant-in-Aid from the Japan Society for the Promotion of Science (JSPS), a Grant-in-Aid from the Japan Foundation for Neuroscience and Mental Health, and by grants from the ministry of Health and Welfare, Japan.