Superoxide dismutase-1 and other proteins in inclusions from transgenic amyotrophic lateral sclerosis model mice

Authors


Address correspondence and reprint requests to Stefan L. Marklund, Department of Medical Biosciences, Clinical Chemistry, Umeå University, SE-901 85 Umeå, Sweden.
E-mail: stefan.marklund@medbio.umu.se

Abstract

J. Neurochem. (2010) 114, 408–418.

Abstract

Mutant superoxide dismutase-1 (SOD1) causes amyotrophic lateral sclerosis (ALS) through a cytotoxic mechanism of unknown nature. A hallmark in ALS patients and transgenic mouse models carrying human SOD1 (hSOD1) mutations are hSOD1-immunoreactive inclusions in spinal cord ventral horns. The hSOD1 inclusions may block essential cellular functions or cause toxicity through sequestering of other proteins. Inclusions from four different transgenic mouse models were examined after density gradient ultracentrifugation. The inclusions are complex structures with heterogeneous densities and are disrupted by detergents. The aggregated hSOD1 was mainly composed of subunits that lacked the native stabilizing intra-subunit disulfide bond. A proportion of subunits formed hSOD1 oligomers or was bound to other proteins through disulfide bonds. Dense inclusions could be isolated and the protein composition was analyzed using proteomic techniques. Mutant hSOD1 accounted for half of the protein. Ten other proteins were identified. Two were cytoplasmic chaperones, four were cytoskeletal proteins, and 4 were proteins that normally reside in the endoplasmic reticulum (ER). The presence of ER proteins in inclusions containing the primarily cytosolic hSOD1 further supports the notion that ER stress is involved in ALS.

Abbreviations used:
ALS

amyotrophic lateral sclerosis

ER

endoplasmic reticulum

hSOD1

human SOD1

LC-ESI

Liquid chromatography-electrospray ionization

MALDI-TOF

Matrix-assisted laser desorption ionization time of flight

MS

Mass spectrometry

NP-40

Nonidet P-40

SOD1

superoxide dismutase-1

Amyotrophic lateral sclerosis (ALS) is a devastating neurodegenerative disease characterized by loss of motor neurons in the motor cortex, brainstem, and spinal cord. This results in progressive muscular atrophy, and the patients usually succumb to respiratory failure within a few years. About 10% of ALS cases appear in families (Haverkamp et al. 1995), and in some of these the disease is linked to mutations in the gene of the ubiquitously expressed antioxidant enzyme CuZn-superoxide dismutase (SOD1) (Rosen et al. 1993).

Overall, about 6% of all cases with ALS show SOD1 mutations, and more than 140 such mutations have been identified [(Andersen et al. 2003); http://alsod.iop.kcl.ac.uk/Als/Summary/summary.aspx]. The mutations confer a cytotoxic gain of function of unknown character to the enzyme (Gurney et al. 1994; Andersen et al. 1995). There is now considerable evidence to suggest that the demise of motor neurons is non-cell autonomous and dependent on noxious effects of mutant SOD1s in several cell types in the motor areas (Boillee et al. 2006; Nagai et al. 2007; Henkel et al. 2009).

A hallmark of ALS caused by mutant SOD1s, both in patients and transgenic models, is inclusions that are immunoreactive for SOD1 (Bruijn et al. 1997; Johnston et al. 2000; Jonsson et al. 2004, 2006b). In the animal models, inclusions appear concomitantly with the onset of symptoms and become markedly more abundant in the terminal phase of the disease (Johnston et al. 2000; Wang et al. 2002b; Jonsson et al. 2004, 2006a). They are present in motor neurons, astrocytes, and the neuropil in affected areas of the CNS (Bruijn et al. 1997; Watanabe et al. 2001; Jonsson et al. 2004), but they have also been observed in some peripheral tissues (Jonsson et al. 2008). The SOD1 inclusions might conceivably cause cytotoxicity by blocking essential cellular functions and/or by entrapment of other proteins which might become inactivated, depleted or erroneously activated. Alternatively they are merely markers of failing protein quality control and clearance in terminally injured cells. Studies of these cellular entities have so far been limited to histochemical techniques. To shed further light on these issues, we report the isolation and characterization of inclusions from four transgenic mouse models. The basic separation was by density gradient ultracentrifugation using a protocol developed for separation of organelles (Bergemalm et al. 2006).

Material and methods

Transgenic mice

The G127insTGGG (G127X) human SOD1 (hSOD1) and D90A mouse models were developed in our laboratory and backcrossed with C57/BL6BomTac mice for more than 25 generations (Jonsson et al. 2004, 2006b). The strain expressing G85R mutant hSOD1 was obtained from Dr D. W. Cleveland and was similarly backcrossed for > 15 generations (Bruijn et al. 1997). The strain expressing the G93A mutant (Gurney et al. 1994) was obtained from the Jackson Laboratory (Bar Harbor, ME, USA) and was backcrossed with C57/BL6BomTac for > 20 generations. Mice were regarded as terminal when they were no longer able to reach for food. This happened at 216, 447, 370, and 135 days of age for G127X, D90A, G85R and G93A mice respectively. As controls, 200 days old C57/BL6BomTac (Taconic Europe, Bomholt, Denmark) or non-transgenic littermates from the breeding program were used.

Homogenization of tissue and separation of inclusions

Mice were killed by intraperitoneal injection of pentobarbital. The thorax was cut open and the animal perfused through the left ventricle with 20 mL 0.15 M NaCl. The spinal cord was dissected by cutting the vertebral column at the base of the scull and just above the hip bone. A syringe was inserted at the lower opening and the spinal cord was flushed out with 0.15 M NaCl. The spinal cord was then gently homogenized in a Dounce glass-pestle homogenizer with 12 + 2 strokes using loose-fit and tight-fit pestles respectively, at 4°C in 25 volumes of mitochondrial buffer [0.21 M mannitol, 0.11 M sucrose, 10 mM K HEPES (pH 7.2), 1 mM EGTA, and the antiproteolytic cocktail Complete® without EDTA (Roche Diagnostics, Basel, Switzerland)]. The homogenate was passed through a 5-μm nylon mesh by centrifugation at 20 g at 4°C for 10 min and was then separated by density-gradient ultracentrifugation as described previously (Bergemalm et al. 2006). Briefly, the filtrates were supplied with iodixanol (OptiPrep; Axis-Shield PoC AS, Oslo, Norway) to a final concentration of 8%. The filtrates (1–1.5 mL) were loaded on top of continuous 9-mL 12–38% iodixanol gradients in mitochondrial buffer and centrifuged at 100 000 g for 90 min, at 4°C in a swinging-bucket rotor (45 Ti; Beckman-Coulter, Fullerton, CA, USA). The tubes were punctured with a cannula and tapped from the bottom in 500-μL portions and, if not analyzed immediately, were stored at −80°C. The localization of organelles was determined by marker protein assays as previously described (Bergemalm et al. 2006). For further information, see Appendix S1.

Filter trapping of hSOD1 aggregates and inclusions

For trapping of hSOD1 aggregates, a dot blot apparatus with 0.15-μm cellulose acetate filters (Schleicher & Schuell, Dassel, Germany) was used, essentially as described by Wang et al. (Wang et al. 2002a). In the gradients from the G93A and D90A models, substantial amounts of hSOD1 are present inside mitochondria (Bergemalm et al. 2006). To prevent trapping of such hSOD1, organelles were disrupted by addition of 0.5% Nonidet P-40 (NP-40) to the fractions. Following washing with the mitochondrial buffer, the dots were punched out and extracted in sodium dodecyl sulfate–polyacrylamide gel electrophoresis sample buffer.

For capture of intact inclusions present in pooled dense fractions (fractions 1–4) from terminal G85R and G127X mice, the procedure was carried out without addition of detergent. The punched-out dots were extracted in DeStreak 2-D rehydration buffer (GE Healthcare, Uppsala, Sweden).

Immunocapture and protease susceptibility of inclusions

A dense fraction (fraction 5) from G85R and G127X mice was incubated for 1 h at 22°C with three different anti-hSOD1 antibodies (to residues 24–39, 100–115, and 131–153) immobilized on CNBr-Sepharose gel beads. The beads were washed thrice and the final pellet was incubated in sample buffer for immunoblotting.

For analysis of proteinase susceptibility, fraction 5 was incubated with proteinase K (50 or 10 μg/mL) at 23°C for 30 min with or without the addition of 0.5% NP-40. The reaction was stopped with 40 mM phenylmethylsulphonylflouride.

Reduced and non-reduced immunoblots, in-gel reduction, and quantification

The immunoblots were generally carried out using a human-specific antibody raised against a peptide corresponding to amino acids 24–39 of the human SOD1 sequence (Jonsson et al. 2004). The sodium dodecyl sulfate–polyacrylamide gel electrophoresis step was run both with and without (non-reduced) the reductant mercaptoethanol in the sample buffer. In the cases non-reduced immunoblots were run, 40 mM iodoacetamide was added to the ‘mitochondrial buffer’ used for homogenization to alkylate cystein thiols. Human SOD1 subunits lacking the intrasubunit C57–C146 disulfide bond show much higher immunoreactivity in the blots than subunits with the bond intact (Zetterstrom et al. 2007). To achieve more equal reactivity for all hSOD1 species (subunits with and without the C57–C146 bond, or disulfide-linked to form hSOD1 polymers or complexes with other proteins), after electrophoresis non-reducing gels were immersed for 10 min at 23°C in transfer buffer containing 2% mercaptoethanol prior to electroblotting onto polyvinylidene difluoride membranes (in-gel reduction). The chemiluminescence of the blots was recorded in a ChemiDoc apparatus and analyzed with Quantity One software (Bio-Rad Laboratories, Hercules, CA, USA).

Two-dimensional electrophoresis and protein identification

Filtered-trapped inclusions (fraction 1–4 pools) from pools of 2–3 mice were run on 11-cm 3-11NL IPG strips (GE Healthcare). Analytic 2-D gels were stained with silver stain and preparative with Coomassie Brilliant Blue G-250. Transgene-specific proteins were identified manually on the 2-D gels and the proteins identified by Matrix-assisted laser desorption ionization time of flight (MALDI-TOF) Mass spectrometry (MS) as described previously (Bergemalm et al. 2009). For further details, see Appendix S1.

Liquid chromatography-electrospray ionization (LC-ESI) MS/MS and data analysis

Filter-trapped inclusions (a pool of fractions 1–4) from control (C57/BL6BomTac), G85R and G127X mice were incubated at 55°C for 60 min in 0.1 M Tris, pH. 9.0, containing 10 mM dithiothreitol and 6 M guanidine hydrochloride. Iodoacetamide was then added to a final concentration of 55 mM, followed by incubation at 37°C for 30 min in the dark. The reaction mixture (without cellulose acetate membranes) was filtered through Microcon YM-10 filters (Millipore, Billerica, MA, USA) and retained material was washed twice with 0.2 M NH4HCO3. Trypsin was added at an enzyme-to-substrate ratio of 1 : 50 and digestion was carried out overnight at 37°C on the top of the Microcon YM-10 filter in the presence of 0.05 M NH4HCO3. [Correction added after online publication: 26/05/10: 7°C changed to 37°C]. The resulting peptides were collected by centrifugation, lyophilized, and resuspended in 0.1% formic acid. Peptide identification was performed as previously described (Srivastava et al. 2009). For further details, see Appendix S1.

Results

Characterization of the hSOD1-containing inclusions

Inclusions immunoreactive for hSOD1 are found in motor neurons, glial cells, and the neuropil of both ALS patients and transgenic model mice that carry mutant hSOD1s (Figure S1). To preserve the integrity of the inclusions, we used a protocol developed for organelle separation by density gradient ultracentrifugation (Bergemalm et al. 2006). The homogenization is gentle, no sonication is used, and there is no pelleting by centrifugation. Detergents were avoided, as they disrupt the inclusions (see below) and might also release proteins bound to hSOD1 through hydrophobic interactions. The typical density gradient and the localization of markers for organelles is shown in Fig. 1a. Both pre-symptomatic and terminal G93A, D90A, G85R, and G127X transgenic mice were examined. The G127X mice have very low levels of the mutant hSOD1, and virtually no mutant protein entered the gradient when pre-symptomatic mice were examined (Fig. 1b), as shown previously (Bergemalm et al. 2006). In contrast, a significant proportion of the hSOD1 in spinal cord homogenates from pre-symptomatic G93A and D90A mice entered the gradient (Fig. 1c, Figure S2). These mice have very high levels of mutant hSOD1 (Jonsson et al. 2006b).

Figure 1.

 Density-gradient separations. Homogenates of spinal cords from different mutant human superoxide dismutase-1 (hSOD1) transgenic mouse models were subjected to density-gradient separations. Mutant hSOD1 was detected by immunoblotting (IB). For detection of mutant hSOD1 aggregates, fractions were filter-trapped in a dot-blot apparatus. (a) Distribution of nuclei, cytosol, and mitochondria in the density gradient located by marker proteins (hnRNP, LD, SDH, and respectively). (b) Distribution of hSOD1 in a pre-symptomatic (100 days) low hSOD1-level model (G127X). (c) Distribution of C57-C146 disulfide-reduced and oxidized mutant hSOD1 in a high hSOD1-level model (G93A) at a pre-symptomatic stage (70 days) (upper panel). The middle panel shows the non-reduced western blot. Arrow marks oxidized and arrowhead marks C57–C146 disulfide-reduced species. SE = short exposure of the blot. Note that the weak band at approximately 30 kDa in fractions 17–18 was a cross-reacting protein, only visible when reducing agent was omitted and never with other anti-hSOD1 antibodies. Distribution of aggregated hSOD1 as revealed by filter-trapping (lower panel). (d) Distribution of hSOD1 in terminal low hSOD1-level models (G85R, G127X).

The two identical subunits in hSOD1 contain four cysteins (C6, C57, C111 and C146), two of which (C57 and C146) form a stabilizing intrasubunit disulfide bond. With G93A mice, virtually all of the hSOD1 that entered the gradient carried the C57–C146 disulfide bond, as determined by non-reducing immunoblots (Fig. 1c, upper and middle panels). The distribution of this disulfide-oxidized hSOD1 closely followed the distribution of the mitochondrial marker, and it probably represents the hSOD1 previously shown to artificially overload mitochondria in these models (Fig. 1c) (Bergemalm et al. 2006). There was no evidence for the presence of significant amounts of hSOD1 aggregates in pre-symptomatic G93A mice, as examined by filter-trap analysis in the presence of NP-40 (Fig. 1c, lower panel).

In terminally ill mice from all the transgenic models, much more mutant hSOD1 entered the gradient (Figs 1d, 2 and 3a–f) and hSOD1 was now captured by filter-trapping, showing that the protein was aggregated.

Figure 2.

 Characterization of human superoxide dismutase-1 (hSOD1) inclusions. (a) Homogenate of a spinal cord from a terminal G127X mutant hSOD1 transgenic mouse was subjected to density-gradient ultracentrifugation, for 90 min or 360 min. All fractions were subjected to immunoblotting for hSOD1. To better visualize the hSOD1 pattern in the gradient, the concentrations were multiplied by 5 (5×). (b) To study the effect of detergent on hSOD1 inclusions, a G127X spinal cord homogenate was split in two and 0.5% Nonidet P-40 (NP-40) was added to one of the samples prior to density-gradient separation. After density-gradient separation as before, all fractions were subjected to immunoblotting for hSOD1. To better visualize the hSOD1 pattern in the gradient, the concentrations were multiplied by 5 (5×). (c) Dense inclusions (fraction 5) from terminal G127X and G85R mice were incubated with anti-hSOD1 antibodies immobilized on Sepharose beads with or without the addition of the non-ionic detergent NP-40. The beads were captured by filtration on a 5-μm nylon filter and then incubated in sample buffer for immunoblotting. (d) Dense inclusions (fraction 5) were subjected to treatment with two different concentrations of proteinase K (50 or 10 μg/mL), with or without the addition of non-ionic detergent (0.5% NP-40). Remaining hSOD1 was analyzed by immunoblotting.

Figure 3.

 Mutant human superoxide dismutase-1 (hSOD1) in inclusions lacks the C57–C146 disulfide bond. To allow discrimination between C57-C146 oxidized and reduced subunits, the immunoblots were non-reduced followed by in-gel reduction. To visualize the hSOD1 pattern in fractions 16–18, short exposures (SE) of these wells are shown in addition to longer exposures (a, c, e). The same fractions were also subjected to filter-trapping for isolation of aggregated hSOD1 (b, d, f). The arrow marks C57-C146 oxidized subunits and the arrowhead marks disulfide-reduced species. Note that the hSOD1 band at approximately 30 kDa in fractions 16–18 corresponded to a cross-reacting protein that was only visible when reducing agent was omitted and never with other hSOD1 antibodies. In G127X, one of the bands at approximately 32 kDa (arrow) represents the previously characterized covalently coupled hSOD1 dimer also seen in dense fractions from other models.

The density of proteins is around 1.37 (Schachman 1959). Simple protein aggregates should thus accumulate at the bottom of the ultracentrifugation tube (c.f. gradient in Fig. 1a). In the present transgenic models, the inclusions were distributed all over the gradient and there was almost no hSOD1 in fraction 1 (i.e. at the bottom) (Figs 1d, 2a and 3a–f). One possible explanation could be that the inclusions/aggregates are small and therefore have not yet reached their equilibrium positions in the density gradient after 90 min at 100 000 g. Prolongation of the centrifugation to 360 min, however, did not significantly alter the position of the inclusions (Fig. 2a). In contrast, addition of the detergent NP-40 to the spinal cord homogenate prior to the centrifugation caused a major shift of hSOD1 towards the bottom of the tube as expected for protein aggregates (Fig. 2b). We conclude from this that the inclusions are not simple protein aggregates but more complex structures.

Next, attempts were made to capture hSOD1 from dense inclusions derived from G127X and G85R mice (fraction 5, Fig. 1d) using anti-hSOD1 peptide antibodies immobilized on Sepharose. We used three different antibodies to increase the likelihood of capturing corresponding reactive epitopes exposed in the aggregates. In the absence of detergent, virtually no hSOD1 was captured (Fig. 2c). With the addition of 1% NP-40, the capture of hSOD1 increased markedly. This indicates that the aggregates of hSOD1 in inclusions may be covered by a coating of other proteins or other cellular components. Exposure to proteinase K, however, resulted in near complete degradation of the hSOD1 both in the presence and absence of NP-40 (Fig. 2d).

The nature of hSOD1 in inclusions

Structural disulfide bonds are rare in proteins in the strongly reducing cytosol. The C57–C146 intrasubunit disulfide bond is a potential weak spot in hSOD1, and significant fractions lacking the linkage have been found previously in transgenic models (Jonsson et al. 2006a; Zetterstrom et al. 2007). The resulting loss of stability leads to increased sampling of disordered conformations and such hSOD1s should have an increased propensity to aggregate (Zetterstrom et al. 2007; Chattopadhyay et al. 2008; Furukawa et al. 2008; Karch et al. 2009). We therefore studied the status of the disulfide bond in hSOD1 from our separations. Fig. 3a shows a non-reduced, in-gel reduced immunoblot of ultracentrifugation fractions from a terminal G93A mouse. Most of the hSOD1 remained in the top fractions before the density gradient. The largest proportion of hSOD1 in these fractions, as well as in the minor portion entering the gradient, carries the native C57–C146 disulfide bond. Human SOD1 lacking the bond is much less abundant, and has previously been estimated to account for 8% of the total in this model (Jonsson et al. 2006a). Even in terminally ill G93A mice, aggregated hSOD1 represents only a small subfraction (< 10%) of the total hSOD1 in spinal cords (Jonsson et al. 2006a; Zetterstrom et al. 2007). To specifically study aggregated hSOD1, the fractions were filter-trapped in a dot-blot apparatus. The detergent NP-40 was added to the fractions to solubilize organelles and prevent capture of any soluble hSOD1 in them. From the terminal G93A mice, all the aggregated hSOD1 that had entered the gradient was found to lack the C57–C146 disulfide bond (Fig. 3b). Some hSOD1 with the bond oxidized was found in fractions that remained on top of the gradient (fractions 17–18). These fractions contained the abundant sticky lipid materials present in CNS homogenates that were not completely solubilized by the weak NP-40 detergent. A band with mobility intermediate between that of reduced hSOD1 and C57–C146 disulfide-oxidized hSOD1 also appeared. This hSOD1 apparently carries a non-native disulfide bond and has previously been observed in soluble form in the G93A model (Zetterstrom et al. 2007).

The G85R and G127X mice contain ≥ 20-fold less hSOD1 in the spinal cord than G93A mice (Jonsson et al. 2006a). The aggregates in terminally ill mice of these models accounted for about 50% of the total hSOD1, and the patterns of total hSOD1 and filter-trapped hSOD1 were much more similar (Fig. 3c and d). In G127X hSOD1, there is no native disulfide bond owing to the C-terminal truncation, and only traces of hSOD1 carrying the C57–C146 bond were seen in the G85R model (Fig. 3c). Accordingly, the filter-trapped hSOD1 is disulfide-reduced in these mice (Fig. 3d and f). In the G85R aggregates, we did not detect any band corresponding to the putative non-natively disulfide-coupled hSOD1 subunits of the G93A model (Fig. 3d). A band with increased mobility, which disappeared in common reduced immunoblots was, however, seen in non-reduced in-gel reduced fractions from G127X mice (Fig. 3f). This species with higher mobility is probably a subunit with a non-native disulfide bond.

From terminal mice belonging to all models, there were also hSOD1 bands of lower mobility than SOD1 subunits (Fig. 4). In reduced immunoblots of hSOD1 aggregates, most of the protein appeared at the position of disulfide-reduced subunits (Fig. 4, right panels). The low-mobility bands thus mainly correspond to hSOD1 that is disulfide-linked to form hSOD1 oligomers and complexes with other proteins. We estimate that such disulfide-coupled hSOD1 represents ≤ 30% of the aggregated hSOD1.

Figure 4.

 Disulfide-coupled high molecular-weight human superoxide dismutase-1 species. Fractions from separation of a homogenate from a terminal G85R mouse were subjected to non-reduced and reduced immunoblotting. The arrow indicates C57-C146 oxidized subunits and the arrowhead marks disulfide-reduced species. (a) Non-reduced immunoblots. (b) The same fractions electrophoresed with reducing agent in the sample buffer followed by in-gel reduction.

In immunoblots for detection of G127X hSOD1, a high molecular weight band is present at roughly the position one would estimate for a dimer (33 kDa) (Fig. 5a, arrow). This is a common theme among mutant hSOD1 variants (Jonsson et al. 2006a) and has recently been reported to represent a covalently coupled hSOD1 dimer in G127X mice (Bergemalm et al. 2009). On 2-D immunoblots (which are reducing in the first dimension) of spinal cords from terminally ill G127X mice, in addition to monomers and dimers, spots with higher molecular weights were present (Fig. 5b, arrowheads). These represent ubiquinated hSOD1 (Jonsson et al. 2004). We extended this finding by running 2-D western blots of dense inclusions (fractions 1–4) from G85R mice; we also examined fraction 18 in the top of the gradient for detection of hSOD1 and ubiquitin (Fig. 5c). In the inclusions, trains of faint spots coinciding with hSOD1 and ubiquitin were seen, in addition to hSOD1 dimers. No other ubiquitinated proteins were detected. In fraction 18, no ubiquinated hSOD1 was detected but there was weak immunostaining of some other ubiquitinated proteins.

Figure 5.

 Ubiquitination of human superoxide dismutase-1 (hSOD1) in inclusions. Reducing 2-D immunoblots of hSOD1 revealed a series of spots of high molecular weight. To explore the nature of these entities, dense and light fractions (fractions 1–4 pooled and fraction 18) were analyzed further. (a) The G127X SOD-1 covalent dimer (arrow) previously characterized. (b) From 2-dimensional immunoblotting with anti-hSOD1 on pooled fractions 1–4 of terminal homogenates, a ladder of spots with decreasing mobility was seen (arrowheads). Arrow indicates the covalent G127X dimer. (c) To determine whether these represented ubiquitinated hSOD1, pooled fractions 1–4 and a fraction containing mostly soluble protein (fraction 18) from a terminal G85R mouse, were subjected to 2-D immunoblotting with anti-hSOD1 and anti-ubiquitin antibodies (arrow marks suspected dimer).

Other proteins in inclusions

For this study, detergents were avoided as they disrupt the inclusions (Fig. 2b and c). Furthermore, the association between hSOD1 aggregates and other proteins might partially be of hydrophobic character and thus potentially susceptible to dissolution by detergents. Instead, we took advantage of the fact that some of the inclusions (fractions 1–4) from G85R and G127X mice are denser than those from the other transgenic models. This material was estimated to represent about 10% of the total amount of aggregated protein in the gradient. By filter-trapping pools of fractions 1–4, most of the organelles were avoided except a minor proportion of the nuclei (Fig. 1a). Most nuclei in the homogenates are captured in the nylon-filter step prior to the ultracentrifugation (Bergemalm et al. 2006). In fractions 5 and upwards there was too much contamination of mitochondria to allow delineation of proteins present in the SOD1-containing inclusions.

The captured material was then solubilized and subjected to 2-D electrophoresis, and representative resulting spot patterns are shown in Fig. 6a–c. Four replicate preparations of 2-3 spinal cords from terminally ill G85R and G127X mice and C57BL/6JBomTac controls were analyzed. Some 20 spots, which were only seen in both transgenic strains and never in the controls, were picked for identification by MALDI-TOF analysis. The major train of spots, which represented about 50% of all protein, was found to be hSOD1 subunits. Ten other proteins were identified (Table 1, Fig. 6d). Some prominent spots that were present in material from both controls and transgenic mice were also analyzed and found to be cytoskeletal and nuclear proteins (not shown). Similar preparations from three transgenic and two control mice were also subjected to LC-MS/MS analysis. We found one additional transgene-specific protein (tubulin-beta) and most MALDI-TOF identifications were replicated (Table 1).

Figure 6.

 2-D analysis of inclusions and validation of proteins. Filter-trapped pooled fractions 1–4 of terminal G85R, G127X, and control mice were subjected to 2-dimensional electrophoresis. (a–c) Silver-stained gels for analysis. (d) For identification of proteins, spots from Coomassie-stained gels were subjected to MALDI-TOF MS. Identifiers are in Swissprot formatting as in Table 1. (e) Pools of fractions 1–4 from G85R, G127X, and C57/BL6BomTac control mice were subjected to immunoblotting with anti-vimentin and anti-HSC70 antibodies.

Table 1.   Transgene-specific proteins identified. Fractions 1–4 from terminal G127X, G85R and C57/BL6BomTac control mice were pooled and inclusions filter-trapped. After solubilisation in rehydration buffer the inclusions were separated on 2-D gels and transgene-specific spots picked and identified using Matrix-assisted laser desorption ionization time of flight (MALDI-TOF). Similar preparations (filter-trapped inclusions) from both control and transgenic mice were also subjected to Liquid chromatography-electrospray ionization mass spectrometry (LC-ESI MS)
Protein nameAccession no. (Swissprot)Compartment2-D gelsLC-ESI MS/MS
  1. ER, endoplasmic reticulum; LC-ESI, Liquid chromatography-electrospray ionization; MS, Mass spectrometry.

Alpha-crystallin B chainP23927CytoplasmXX
HSC70P63017CytoplasmXX
VimentinP20152CytoplasmXX
Tubulin beta-5 chainP99024CytoplasmX
Neurofilament light polypeptideP08551CytoplasmXX
Alpha-internexinP46660CytoplasmXX
GRP78P20029ERX
Endoplasmin (HSP90B1/GRP94)P08113ERX
Calreticulin (ERP60)P14211ERXX
Protein-disulfide isomeraseP09103ERX

To validate our findings with a different method, we run 1-dimensional immunoblots for HSC70 and vimentin. The staining was only seen in extracts of pooled dense fractions (fractions 1–4) from transgenic mice (Fig. 6e). The intermediate filament protein vimentin is primarily expressed in glial and mesenchymal cells, but is also expressed in adult motor neurons and found in inclusions from symptomatic G93A mice (Perrin et al. 2005).

Discussion

Here we report the first isolation and partial characterization of hSOD1-containing inclusions from spinal cords of ALS model mice. The inclusions appear to be relatively complex detergent-susceptible structures that equilibrate in the density gradient at lower densities than simple protein aggregates. The cause of the reduced density is not quite clear; the protease-susceptibility suggest that this is not explained by encapsulation by a continuous membrane. The presence of several proteins from the endoplasmic reticulum (ER) suggests that ER membranes may possibly be constituents of the inclusions (see further below).

Similar separations were recently carried out on homogenates of spinal ventral horns from an ALS patient carrying the G127X SOD1 mutation (Jonsson et al. 2008). It is noteworthy that the density patterns of the inclusions were very similar to those found here for spinal cords from terminal G127X mice. This suggests that the inclusions in the transgenic mouse models are of the same nature as those in ALS patients.

The densities of the inclusions were very heterogeneous, covering those of all organelles. This should make it virtually impossible to separate organelles in terminal spinal cords from the inclusions using differential centrifugation. This difficulty is the basis of the (in our view) misconception that the hSOD1 aggregates are mainly associated with mitochondria (Liu et al. 2004; Vijayvergiya et al. 2005; Vande Velde et al. 2008). It may also lead to erroneous conclusions regarding the properties of hSOD1 in mitochondria. The present results do not formally exclude the possibility that some aggregated hSOD1 actually is associated with this organelle, as a proportion of the inclusions cofractionate with the mitochondrial marker. There was, however, no evidence in any of the cases for peaks in aggregated hSOD1 at the location of the mitochondrial marker peak (fractions 9–10); troughs or comparatively low levels were found instead. In D90A and G93A mice, the hSOD1 in the cytosol on the top of the gradient is partially disulfide-reduced (Fig. 1, middle and lower panels, Fig. 3a, Figure S2), and in these models 8–12% of the total hSOD1 has been estimated to lack the native disulfide bond (Jonsson et al. 2006a). In contrast, the hSOD1 that co-fractionates with the mitochondrial marker appears to be completely C57–C146 disulfide-oxidized (Fig. 1c, upper and middle panels, Fig. 3a, Figure S2). Reduced subunits entering the intermembrane space are apparently efficiently oxidized by copper-chaperone for SOD1 (CCS) and the disulfide-oxidizing machinery present in this compartment (Reddehase et al. 2009). As oxidized subunits cannot leave the intermembrane space, this is a likely explanation for the artificial hSOD1 overloading of mitochondria in the D90A and G93A models (Bergemalm et al. 2006). The mitochondrial hSOD1 should have a low propensity to aggregate, as hSOD1 aggregates appear to be derived mainly from reduced subunits (see discussion below).

Human SOD1 in inclusions

We have previously shown that in the present transgenic murine ALS models, spinal cords are enriched with soluble hydrophobic disulfide-reduced hSOD1 throughout life (Jonsson et al. 2006a; Zetterstrom et al. 2007). We found here that the inclusions contained such hSOD1 in aggregated form. There was also some disulfide-coupled hSOD1 of higher molecular weight and minute amounts of ubiquitinated subunits. Human SOD1 has already been shown to be ubiquitinated in transgenic models (Jonsson et al. 2004; Basso et al. 2006), and here we found that all such species are trapped in inclusions and that this modification is restricted to hSOD1. Similar to our findings, Karch et al. recently reported that the hSOD1 in detergent-resistant aggregates from transgenic mice is mainly composed of subunits lacking the native C57–C146 disulfide bond (Karch et al. 2009). Thus, absence of the C57–C146 disulfide bond appears to be the fundamental prerequisite for aggregation of hSOD1 in vivo. This conclusion is supported by studies on hSOD1 aggregation in vitro (Chattopadhyay et al. 2008; Furukawa et al. 2008) and on various cysteine mutants of hSOD1 expressed in cultured cells (Karch et al. 2009). Several studies both in vitro (Banci et al. 2007; Oztug Durer et al. 2009) and in cultured cells (Niwa et al. 2007; Karch and Borchelt 2008) have reported that C6 and C111 can engage in formation of multimers and aggregates. In the present study and in the previous study on detergent-resistant aggregates in mice (Karch et al. 2009), the amounts of disulfide-coupled species were substantial but they were probably formed secondarily to aggregation of large amounts of hSOD1 with reduced cysteines.

Other proteins in inclusions

Human SOD1 was found to account for about 50% of the protein. Around 20 other protein spots were found, which were not detected in fractions from control mice. Out of ten that could be identified, two were cytoplasmic chaperones, four were cytoskeletal proteins, and four were proteins that normally reside in the ER.

The second most abundant protein in inclusions was the small heat-shock protein αβ-crystallin. It has previously been found in detergent-insoluble fractions containing mutant hSOD1 (Wang et al. 2003) and can prevent aggregation of mutant hSOD1 in vitro (Wang et al. 2005). HSC70 has been found to co-immunoprecipitate with mutant hSOD1 (Shinder et al. 2001; Choi et al. 2004) and therapeutic intervention aimed at increasing the amount of heat-shock proteins has been shown to increase the lifespan of an ALS transgenic mouse model (Kieran et al. 2004).

Out of the four proteins detected that normally reside in the ER, one – calreticulin – has not been implicated in ALS before. Protein-disulfide isomerase and GRP78 have been found by immunohistochemistry in hSOD1-positive inclusions in transgenic mice (Tobisawa et al. 2003; Wate et al. 2005; Atkin et al. 2006) and also in inclusions from sporadic ALS patients (Atkin et al. 2008). Endoplasmin has been identified in detergent-insoluble fractions of G93A mice (Basso et al. 2009). Each of these four proteins is involved in recognition of misfolded proteins and targeting as part of a process called ER-associated degradation. It is notable that along with the cytoplasmic chaperones HSC70 and αβ-crystallin, in total six key players in protein quality control can be found in the terminal inclusions.

In a previous study, aggregates in spinal cords from transgenic mice were isolated by two different approaches: as pellets collected by centrifugation following a harsh detergent extraction procedure and by antibody capture (Shaw et al. 2008). The major aggregated protein was found to be mutant hSOD1. The only other protein found with reasonable consistency was vimentin. In another study, symptomatic mice expressing YFP-G85R hSOD1 were compared with mice expressing YFP-wild type hSOD1 (Wang et al. 2009). Homogenates of spinal cords were washed several times with detergent, and final pellets after centrifugation were analyzed using a modified LC-MS protocol. Vimentin was again the only reasonably abundant mutant-enriched protein, which coincided with the present findings. Glial fibrillary acidic protein (GFAP), neurofilament-M and -H, and peripherin were also found at higher levels in YFP-G85R hSOD1 mice. In a recent similar study, detergent-resistant pellets from spinal cords from symptomatic G93A mice and wt-hSOD1 mice were compared. 42 proteins were more abundant in the G93A mutant mice, and among them, similar to our findings, vimentin, αβ-crystallin, protein-disulfide isomerase and endoplasmin were found (Basso et al. 2009). Whereas mutant SOD1 was the major component in our present study, such material was a minor component in both studies on detergent-resistant pellets (Basso et al. 2009; Wang et al. 2009). This suggests that the pellets contain aggregated material from compromised spinal cords that is not present in SOD1 inclusions. Detergent-induced disruption of inclusions might also have caused loss of some proteins coaggregated with hSOD1. Analysis of inclusions and detergent-resistant pellets thus yields different and complementary information on protein perturbations in spinal cords of ALS model mice.

Conclusions

The inclusions analyzed in these experiments have only been found in terminal transgenic mice. The obvious question is whether this late entity is of any major importance for the pathogenic process of ALS. The protein content of the inclusions probably reflects interactions that are established between individual soluble proteins, and these have been shown to be highly regulated (Rajan et al. 2001). As the amounts of proteins apart from hSOD1 are minute, it appears unlikely that a major depletion of the proteins results from entrapment in the inclusions. This study was, however, made on the tissue level and the picture might be different in individual cells. Given that SOD1 is primarily a cytosolic protein, it is remarkable that 4 of the proteins identified normally reside in the ER. SOD1 has recently been shown to be present in the ER and to be secreted by neural cells through an ER-dependent mechanism (Turner et al. 2005; Kikuchi et al. 2006; Urushitani et al. 2006). Furthermore, the unfolded protein response, which follows severe ER-stress, has been shown to be activated in both transgenic mice and humans with ALS (Atkin et al. 2006, 2008; Kikuchi et al. 2006), and interference with ER stress has been found to prolong the lifespan of transgenic mice (Saxena et al. 2009). The present findings lend further support to the idea that the ER, and ER stress and unfolded protein response, are involved in ALS.

Acknowledgements

This work was supported by the Swedish Science Council, the Swedish Brain Fund/Hållsten Fund, the Swedish Medical Society including the Björklund Fund for ALS Research, the Swedish Association of Persons with Neurological Disabilities, Västerbotten County Council, The Kempe Foundations, and Konung Gustaf V’s and Drottning Victorias Foundation. We thank Eva Bern, Karin Hjertkvist, Ann-Charlott Nilsson, Ulla-Stina Spetz, and Agneta Öberg for technical assistance and Dr D.W. Cleveland for supplying the G85R SOD1 transgenic mouse strain.

Ancillary