OUP user menu

Minute quantities of misfolded mutant superoxide dismutase‐1 cause amyotrophic lateral sclerosis

P. Andreas Jonsson, Karin Ernhill, Peter M. Andersen, Daniel Bergemalm, Thomas Brännström, Ole Gredal, Peter Nilsson, Stefan L. Marklund
DOI: http://dx.doi.org/10.1093/brain/awh005 73-88 First published online: 8 October 2003


Mutant forms of superoxide dismutase‐1 (SOD1) cause amyotrophic lateral sclerosis (ALS) by an unknown noxious mechanism. Using an antibody against a novel epitope in the G127insTGGG mutation, mutant SOD1 was studied for the first time in spinal cord and brain of an ALS patient. The level was below 0.5% of the SOD1 level in controls. In corresponding transgenic mice the content of mutant SOD1 was also low, although it was enriched in spinal cord and brain compared with other tissues. In the mice the misfolded mutant SOD1 aggregated rapidly and 20% occurred in steady state as detergent‐soluble protoaggregates. The misfolded SOD1 and the protoaggregates form, from birth until death, a potentially noxious burden that may induce the motor neuron injury. Detergent‐resistant aggregates, as well as inclusions of mutant SOD1 in motor neurons and astrocytes, accumulated in spinal cord ventral horns of the patient and mice with terminal disease. The inclusions and aggregates may serve as terminal markers of long‐term assault by misfolded SOD1 and protoaggregates.

  • aggregation; transgenic mice; SOD; motor neuron disease
  • ALS = amyotrophic lateral sclerosis; DTPA = diethylenetriaminepentaacetic acid; NP40 = Nonidet P‐40; GFAP = glial fibrillary acidic protein; SOD = superoxide dismutase; UCH‐L1 = ubiquitin C‐terminal hydrolase‐L1


Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative syndrome characterized by adult‐onset progressive loss of motor neurons in the motor cortex, brainstem and spinal cord. This loss results in muscular wasting, generalized paralysis and inevitable death. About 10% of ALS cases are familial (Haverkamp et al., 1995) and in ∼20% of affected families the disease has been linked to mutations in the Cu,Zn‐superoxide dismutase gene (SOD1). More than 100 different mutations have been identified in ALS, all but one showing a dominant pattern of inheritance (Andersen et al., 2003). Most mutations are associated with reductions in SOD1 activity in erythrocytes (Robberecht et al., 1994; Andersen et al., 1997), other cell types (Tsuda et al., 1994) and CNS tissue (Bowling et al., 1993). The collective evidence suggests a cytotoxic gain of function conferred by the mutations (Gurney et al., 1994; Andersen et al., 1995; Reaume et al., 1996). SOD1 is ubiquitously expressed, and in several tissues at higher levels than in brain and spinal cord (Marklund, 1984). However, both the nature of the toxic property of the mutant SOD1s and the reason for the particular susceptibility of some parts of the CNS remain to be explained.

We have previously reported a Danish ALS pedigree in which the disease is associated with a mutation in the SOD1 gene, Gly127insTGGG (G127X), which leads to the synthesis of a truncated enzyme subunit with five novel amino acids following Gly127 (Andersen et al., 1997). Using mutant‐specific antibodies directed against the novel epitope in the carboxy‐terminal end, here we were able for the first time to examine mutant SOD1 in the CNS of an ALS patient, and we found very low levels. To study the pathogenesis of the disease, we generated a series of mouse lines transgenically overexpressing G127X mutant human SOD1. These mice also contain very low levels of the mutant enzyme, but still develop motor neuron disease. The scarcity of the G127X mutant SOD1 should facilitate detection of any critically altered toxic forms of SOD1. Here we report biochemical and histopathological studies on an ALS patient carrying G127X mutant SOD1, on corresponding transgenic mice and, for comparison, the widely examined G93A transgenic mouse model (Gurney et al., 1994). We focused on altered forms of the mutant enzyme and also searched for similarities between the human and murine material.

Material and methods

Patient and controls

The ALS patient carrying the G127X mutation has been described previously (Andersen et al., 1997). At the age of 62 years, 32 months after left arm onset, she died with tetraparalysis with both upper and lower motor neuron signs. With informed consent following a 20 h time to autopsy, the whole brain and spinal cord were removed and frozen at –80°C. For analysis, pieces (10–50 mg wet weight) of different areas were dissected out under a microscope on an aluminium plate at –20°C. Pieces for histological studies were immersion‐fixed in 4% paraformaldehyde in 0.1 M sodium phosphate, pH 7.4. Corresponding tissue pieces from five aged‐matched patients who had died from non‐neurological diseases were used as control material for the morphological and biochemical analyses. Skin fibroblast lines were initiated from skin punch biopsies from the patient. They were cultured in Ham’s F10 containing 10% fetal calf serum. The study adhered to the tenets of the Declaration of Helsinki, and was approved by the ethics committees of the Universities of Copenhagen, Denmark and Umeå, Sweden.


A 0.6 kb fragment containing NcoI‐BclI exon 5, amplified from genomic DNA of a G127X ALS patient (primers 5′‐AAA GTA AGA GTG ACT GCG GAA CTA‐3′ and 5′‐CTG GCA AAA TAC AGG TCA TTG A‐3′), was ligated into an NcoI‐BclI‐cleaved PstI‐BamHI subclone of the human SOD1 gene. To restore a complete 11.6 kb EcoRI‐BamHI SOD1 genomic fragment (Levanon et al., 1985) the PstI‐BamHI fragment with mutant exon 5 was isolated and ligated to a PvuII‐PstI subclone of the SOD1 gene. In a final step, the PvuII‐BamHI fragment containing exons 2–5 was isolated and combined with an EcoRI‐PvuII fragment adding exon 1 and upstream sequences to the construct. This 11.6 kb SOD1 genomic fragment was then excised with EcoRI and BamHI, electro‐eluted and used for microinjections into ova from C57BL6/CBA mice. Nineteen lines of transgenic mice carrying the human mutant G127X SOD1 gene were generated. The 716 line was the principal focus of the present study and was backcrossed with C57BL/6JBom mice for about five generations. Transgenic mice were identified by Southern blots.

For purposes of comparison, G93AGur mice (Gurney et al., 1994) and a line with loss of copies of the transgene G93AGurdl were obtained from The Jackson Laboratory (Bar Harbor, ME, USA). They are both heterozygous for the insertion and had been backcrossed with C57BL/6JBom mice for about five generations at the time of this study. SOD1 knockout mice were obtained from A. Reaume (Reaume et al., 1996).

The mice were euthanized at preselected intervals or when so terminally ill that they could no longer feed. Brain, spinal cord and peripheral organs were rapidly dissected out and rapidly frozen in liquid nitrogen. Other mice were perfusion‐fixed with 4% paraformaldehyde in 0.1 M sodium phosphate buffer, pH 7.4. The animal care and experiments were carried out in accordance with the European Communities Council Directive (86/609/EEC) and the experimental protocol was approved by the Ethics Committee for Animal Research at Umeå University.

Southern blotting

DNA was prepared from mouse tails using standard methods. DNA was separated on 0.7% agarose gels and blotted onto Hybond XL nylon membranes (Amersham Biosciences, Uppsala, Sweden). The EcoRI‐BamHI fragment of the SOD1 gene was used as a probe. Labelling and visualization of bands were as described for northern blots. Copy number was determined by quantification against known amounts of the EcoRI‐BamHI fragment.

Northern blotting

Total RNA was prepared from mouse brain using the Trizol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s description. Electrophoresis was performed using the NorthernMax‐Gly kit (Ambion, Austin, TX, USA) and RNA was blotted onto Hybond XL nylon membranes (Amersham Biosciences). Blots were probed using ULTRAhyb hybridization buffer (Ambion) and the MegaPrime random labelling kit (Amersham Biosciences). Bands were visualized with a Bio‐Rad GS‐525 Molecular Imager and quantified using Molecular Analyst software (Bio‐Rad, Hercules, CA, USA). Samples were normalized against β‐actin (Clontech, Palo Alto, CA, USA). SOD1 mRNA was detected using a purified PCR fragment as a probe. Primers were 5′‐GCGTGGCCT AGCGAGTTATG‐3′ and 5′‐ATCCTTTGGCCCACCGTGTTT TCTG‐3′ yielding a 248‐bp amplicon of exons 1–3. The PCR amplicon was purified using High Pure PCR Product Purification kit (Roche Diagnostics, Mannheim, Germany)


The primary antibodies used were polyclonal rabbit antibodies raised against keyhole limpet haemocyanin‐coupled peptides corresponding to amino acids 3–20, 24–39, 43–57, 58–72, 80–96, 100–115 and 131–153 in the human SOD1 sequence. The 24–39 antibody is human SOD1‐specific. The 131–153 sequence is the same in human and murine SOD1 but absent in G127X mutant protein. The others show various degrees of cross‐reaction with murine SOD1. Antibodies were also raised against peptides corresponding to the carboxy‐terminal end of G127X mutant SOD1: CIIGRTLVVHEKADDLGGQRWK (the Agaz antibody) and ADDLGGQRWK (the Inger antibody). The novel epitope is shown in bold. Since the aminoterminal parts of the peptides are also present in the wild‐type enzyme, the antibodies were always adsorbed with human wild‐type SOD1 that had been denatured by exposure to 6 M guanidinium chloride and 3 mM of the chelator diethylenetriaminepentaacetic acid (DTPA) followed by dialysis. Following blocking, no reactivity versus the wild‐type SOD1 subunit was found with the Inger antibody (Fig. 1B), whereas minor reactivity was found with the Agaz antibody (Fig. 2A). The rabbit antisera were purified on protein A–Sepharose (Amersham Biosciences) followed by purification on a Sulfolink coupling gel (Pierce, Rockford, IL, USA) with the corresponding peptides coupled.


Fig. 1 Immunoblots of brain and spinal cord homogenates from the ALS patient, carrying G127X SOD1. (A) Immunoblot of lumbar ventral horn homogenate from the G127X patient and controls using the 58–72 anti‐SOD1 antibody. Note that after a short exposure time only the wild‐type SOD1 subunit was visible, but that after long exposure the 36‐kDa mutant band (arrowhead) was also detected. (B) Immunoblots of homogenates from temporal lobe (TL), precentral gyrus (PG), cervical ventral horn (CVH) and lumbar ventral horn (LVH) from the G127X patient and from a control individual, using the mutant‐specific Inger antibody. The arrow shows the G127X‐mutant SOD1 subunit and the arrowhead an additional 36‐kDa immunoreactive band existing primarily in the ventral horns. Note the heavy smearing in the ventral horn samples. No reactivity versus the wild‐type SOD1 is seen even after very long exposure. A few unspecific bands are seen in the temporal lobe and precentral gyrus samples. The amounts of G127X‐mutant SOD1 were calculated from the sum of the 17‐ and 36‐kDa bands and are expressed as percentages of the content of SOD1 in corresponding areas in controls (Table 1).

Fig. 2 Immunoblots of tissue and cell homogenates from the ALS patient and transgenic mice carrying G127X SOD1. (A) A piece of cervical ventral horn from the ALS patient was homogenized in pH 7.0 PBS containing 0.1% of the detergent NP40. The homogenate was centrifuged at 20 000 g for 30 min and the first supernatant collected. The pellet was washed another four times in the detergent‐containing buffer. The original homogenate (H), the first supernatant (S) and the final pellet (P) were immunoblotted using the Agaz antibody. The G127X mutant SOD1 was not found in the supernatant, but was found in the pellet (arrowhead). The major part of the wild‐type protein (arrow) was found in the supernatant, but some was also present in the washed pellet. (B) Immunoblot of skin fibroblast homogenates from the G127X patient cultured for 2 days in the absence (control) or presence of 20 µM of the proteolysis inhibitor NLVS (3‐nitro‐4‐hydroxy‐5‐iodophenylacetyl‐LLL‐vinylsulfone). For detection, the mutant‐specific Agaz antibody was used. (C) Homogenate of fibroblasts from the G127X ALS patient cultured in the presence the proteolysis inhibitor NLVS (3‐nitro‐4‐hydroxy‐5‐iodophenylacetyl‐LLL‐vinylsulfone). The homogenate was centrifuged at 20 000 g for 30 min and the whole homogenate (H) and pellet (P) were analysed by immunoblotting. The mutant‐specific Agaz antibody indicated that ∼60% of the G127X mutant SOD1 was pelleted. Use of an antibody (131–153) directed against part of the protein absent in G127X SOD1 showed that there was no sedimentation of the wild‐type SOD1 in the homogenate. (D) Spinal cord and brain from G127X transgenic mice of different ages from 2 days until the terminal disease stage were homogenized in the pH 7.0 PBS containing 0.1% of the detergent NP40. The homogenate was centrifuged at 20 000 g for 30 min and the pellet then washed another four times with the NP40 buffer. The final pellet was analysed by immunoblotting using the 58–72 anti‐SOD1 antibody. (E) Spinal cords from G127X mice of different ages were extracted in pH 7.0 PBS without detergent. The extracts were centrifuged at 20 000 g for 30 min and the resulting pellet was analysed by immunoblotting using the 58–72 anti‐SOD1 antibody. (F) Post‐mortem stability. Pieces of brain from a 100‐day‐old G127X mouse were kept warm and at intervals snap‐frozen as described in Material and methods. The pieces were then homogenized, centrifuged and immunoblotted as described under E above. H, S and P indicate whole homogenate, 20 000 g supernatant and pellet respectively.

Homogenization of tissues

Following dissection at near 0°C, the tissue pieces were kept at –80°C. For centrifugation experiments, weighed pieces of tissue were homogenized in 25 volumes of phosphate‐buffered saline [PBS; 10 mM K phosphate, pH 7.0, in 0.15 M NaCl with Complete without EDTA (Roche Diagnostics) antiproteolytic cocktail added], using an Ultraturrax followed by sonication. For immunoblotting and SOD activity analysis, 50 mM K phosphate, pH 7.4, 3 mM DTPA, 0.3 M KBr and Complete with EDTA was used instead.

Immunoblotting and quantification of SOD1

Samples were solubilized 1 : 1 in 2 × SDS–PAGE (sodium dodecyl sulphate–polyacrylamide gel electrophoresis) sample buffer (100 mM Tris–HCl, pH 6.8, 10% β‐mercaptoethanol, 20% glycerol, 4% SDS and bromophenol blue) and heated for 5 min at 95°C, and then separated on 12% SDS–PAGE. The gels were then electroblotted onto polyvinylidine difluoride membranes (Amersham Biosciences). Blots were probed with antibodies and chemiluminescence was generated using Supersignal West Dura substrate (Pierce). Bands were visualized on film or by using a Fluor‐S Multiimager and Quantity One software (Bio‐Rad). Secondary antibodies were horseradish peroxidase‐labelled anti‐rabbit IgG antibody or biotinylated anti‐rabbit IgG antibody and horseradish peroxidase‐labelled streptavidin (Amersham Biosciences). For quantification of SOD1 by immunoblotting, wild‐type human SOD1 with the concentration determined by quantitative amino acid analysis (Marklund et al., 1997) was used as original standard. For quantification of mouse SOD1 the 131–153 antibody was used, and for G127X and G93A mutant SOD1 in murine tissues and human wild‐type SOD1 the 24–39 antibody was used. Finally, in CNS from the human G127X ALS patient the mutant‐specific Inger antibody was used for determination of the mutant enzyme. In this case, G127X mutant SOD1 in a transgenic mouse brain homogenate previously quantified against the original standard was used as standard. Generally, the quantifications by immunoblots were run in triplicate.

2‐D gel electrophoresis

Spinal cords from 100‐day‐old and terminal G127X mice were homogenized in 8 M urea, 4% CHAPS and 2% Bio‐Lyte 3–10, swollen into 7‐cm pH 3–10 immobilized pH‐gradient strips (Bio‐Rad) and electrofocused according to the manufacturer’s instructions. The strips were then top‐loaded onto 12.5% SDS–PAGE gels, electrophoresed, and finally electroblotted onto polyvinylidene difluoride membranes and immunoblotted as described above with the 24–39 anti‐SOD1 antibody and an anti‐ubiquitin antibody (Sigma, St Louis, MO, USA).

Analysis of NP40 insoluble aggregates

Brain and spinal cord samples from mice and the G127X patient were homogenized in the pH 7.0 PBS with 0.1% of the detergent Nonidet P40 (NP40) (Roche Diagnostics) added. Homogenized samples were then centrifuged at 20 000 g for 30 min at 4°C. Supernatants were carefully removed and pellets were resuspended and sonicated in double the initial amounts of homogenizing solution. The resuspended pellets were centrifuged at 20 000 g for 30 min at 4°C followed by removal of the resulting supernatant. This washing step was repeated five times. Following the last wash, the pellets were resuspended and sonicated in 1 × SDS–PAGE sample buffer. Samples were then analysed by immunoblotting, using the 24–39 anti‐SOD1 antibody. The relative amounts (percentages) of pelleted SOD1 were quantified against dilutions of the original homogenate in immunoblots.

Analysis of detergent‐soluble aggregates

Mouse tissues homogenized in the pH 7.0 PBS were centrifuged at 20 000 g for 30 min at 4°C. Supernatants were carefully removed and the resulting pellets resuspended and sonicated in the pH 7.0 PBS with 0.1% NP40. The resuspended pellets were then analysed by immunoblotting, using the 24–39 human‐specific anti‐SOD1 antibody. The relative amounts of G127X mutant SOD1 in the pellets (duplicate analysis) were quantified against dilutions of the original homogenate in immunoblots. The amounts of material apparently trapped in ‘processomes’ in the pellets were similarly estimated by quantification of the mouse SOD1, by probing the blots with the 131–153 antibody. The proportion (percentage) of mouse SOD1 pelleted in processomes was then subtracted from the proportion (percentage) of G127X mutant protein pelleted by the centrifugation, to yield what we have called ‘detergent‐soluble (proto)aggregates’ (Fig. 4C).

Effect of post‐mortem time

To mimic the effect of storage/post‐mortem time of a deceased human body on the various assays carried out on the patient, pieces of brain from G127X transgenic mice were put aseptically into small tubes and kept at 37°C for 6 h followed by room temperature for 6 h, and then at 4°C. Immediately thereafter and after different times, tubes were snap‐frozen to –80°C for subsequent thawing and analysis.

SOD activity analysis

The SOD activity was determined with the direct spectrophotometric method using KO2 (Marklund, 1976). One unit is defined as the SOD activity that brings about a decay of O2. at the rate of 0.1 s–1 in 3 ml buffer. One unit in the assay corresponds to 4.2 ng human SOD1.

Effects of buffer composition, pH and chelators on SOD activity

In the laboratory, tissues are routinely extracted in a buffer containing 3 mM DTPA, which reduces the activity of some proteases and binds contaminating transition metal ions, and 0.3 M of the chaotropic salt KBr to improve extraction of proteins. The SOD activity is analysed with a direct assay at pH 9.5, in medium with a chelator. The truncated G127X mutant protein lacks several parts of the molecule important for conformation and the subunit interaction. A putative SOD activity of the protein might therefore be lost because of metal ion extraction by the chelators and denaturation by the chaotropic salt and high pH. We therefore made primary brain homogenates from a G127X mouse and a wild‐type mouse in 50 mM sodium phosphate, pH 7.4; these were then split into aliquots to which were added nothing, 0.3 M KBr, 1 mM DTPA, or both. Following further homogenization, the extracts were then assayed with the direct KO2 assay at pH 9.5 and the pyrogallol autoxidation assay at pH 7.8 (Marklund and Marklund, 1974) in the presence or absence of chelators in the assay medium.

Isoelectric focusing and SOD staining

Brains from 100‐day‐old mice were homogenized in 10 volumes of pH 7.0 PBS and centrifuged at 20 000 g for 30 min. Twenty, 40 and 60 µl of the supernatant were separated by isoelectric focusing on Immobiline pH 3.5–9 gels according to the instructions of the manufacturer (Amersham Biosciences). The gel was stained for SOD activity by immersion in 0.25 mM nitroblue tetrazolium, 30 µM riboflavin in 200 mM Tris cacodylate, pH 7.8, for 30 min followed by irradiation, as adapted from Beauchamp and Fridovich (1971).


Fixed tissue pieces from the patient and the mice were embedded in paraffin, sectioned and stained with haematoxylin/eosin, Klüver–Barrera, modified Bielschowsky and periodic acid–Schiff. Single‐ and double‐labelling immunohistochemistry using the Ventana immunohistochemistry system was performed using the antibodies mentioned above, as well as the following antibodies: anti‐glial fibrillary acidic protein (GFAP) (Dako, Glostrup, Denmark), anti‐ubiquitin (Dako), ubiquitin C‐terminal hydrolase‐L1 (UCH‐L1) (Ultraclone, UK) and anti‐αB‐crystallin (Novocastra, UK).


Minute quantities of G127X mutant SOD1 in the CNS of the patient and aggregates in ventral horns

Analysis of CNS extracts by immunoblotting using an antibody raised against the middle part of the SOD1 molecule showed a lower content of the wild‐type band but, despite overexposure, only a faint band at ∼36 kDa that might represent mutant enzyme (Fig. 1A). The mutation‐specific Inger antibody, on the other hand, clearly revealed the mutant enzyme (Fig. 1B). The levels were highest in the ventral horn of the cervical and lumbar spinal cord (reaching 2–4% of control brain SOD1 content), followed distantly by the precentral gyrus and then other parts of the CNS. In the ventral horn two major bands were seen, one of apparently native size at 17 kDa and one at ∼36 kDa. The 36‐kDa band appeared almost specific for the areas afflicted by disease. Long exposure revealed additional bands with high molecular weight and extensive high‐molecular weight smearing in the ventral horn samples (Fig. 1B). No SOD1‐immunoreactive material was, however, detected at the position of the stacking gels.

All G127X mutant enzyme extracted with buffer without detergent was sedimented by centrifugation at 20 000 g for 30 min, suggesting that all mutant enzyme was present in aggregates. The detergent‐resistance of the aggregated enzyme was examined and it was found that sonication at room temperature in 1% solutions of NP40, Tergitol NP40 and CHAPS in 10 mM K phosphate pH 7.4, 0.15 M NaCl resulted in <10% solubilization. All G127X mutant SOD1 was, however, solubilized by 1% SDS. All these treatments, including SDS, were relatively gentle and did not inactivate the SOD1 and SOD2 activities of the precentral gyrus homogenate from the G127X patient (data not shown). The composition of the aggregated material was further analysed by five‐fold washing of material pelleted at 20 000 g in pH 7.0 buffer containing 0.1% NP40 (Fig. 2A). No mutant enzyme was solubilized in the first extraction step, but there was some wild‐type SOD1 together with the mutant enzyme in the final washed pellet.

As found for several other SOD1 mutations (Bowling et al., 1993), the SOD1 activity in the CNS of the G127X patient was broadly 50% of the control levels (Table 1). No differences compared with controls were found in the Mn‐SOD (SOD2) and extracellular SOD (SOD3) activities. The SOD1 activity of erythrocytes from G127X carriers was previously found to be ∼40% of the control value, suggesting a dominant negative effect of the mutant protein (Andersen et al., 1997). The present results do not support such a mechanism in the CNS.

View this table:
Table 1

SOD isoenzyme activities (U/g wet weight) in various CNS areas from the G127X patient and controls

G127X patientControls (n = 5)
Frontal lobe67007908916 000 ± 24001060 ± 30090 ± 41
Temporal lobe71005807215 100 ± 2900880 ± 12089 ± 27
Precentral gyrus740071010214 600 ± 1300930 ± 380101 ± 33
Cervical ventral horn73008407112 200 ± 3000470 ± 120154 ± 98
Cervical dorsal horn64004305113 900 ± 3900400 ± 11070 ± 26
Cervical dorsal column46002005310 800 ± 1500250 ± 7076 ± 33
Cervical corticospinal tract5400560164 9200 ± 900280 ± 100115 ± 133
Lumbar ventral horn9000118010414 300 ± 1000830 ± 150106 ± 56
Lumbar dorsal horn84007306513 600 ± 1700540 ± 6098 ± 58
Lumbar dorsal column670026010611 700 ± 1100300 ± 110109 ± 54
Lumbar corticospinal tract68008408112 900 ± 1100480 ± 90153 ± 65

Values are mean ± SD.

No G127X mutant SOD could be detected in extracts of cultured skin fibroblasts from the ALS patient by immunoblotting using the mutant‐specific Agaz antibody. Culture of the fibroblasts in the presence of 20 µM of the proteasomal inhibitor lactacystin for 2 days resulted in a barely visible band (not shown), whereas with 3‐nitro‐4‐hydroxy‐5‐iodophenylacetyl‐LLL‐vinylsulfone (NLVS) the mutant SOD1 was easily detectable (Fig. 2B). This finding is similar to that of previous studies on mutant SOD1s using proteasome inhibitors (Hoffman et al., 1996). The 36‐ or 33‐kDa bands found in the brain and spinal cord of the patient and mice of all ages (below) were not seen, and there was only a single band at 17 kDa, probably representing the full‐length mutant subunit. About 60% of the G127X mutant protein in the fibroblast homogenate was pelleted by centrifugation at 20 000 g. There was no sedimentation of the wild‐type human SOD1 (Fig. 2C)

Inclusions of mutant SOD1 abundant in ventral horns from the patient

Sections from cervical, thoracic and lumbar spinal cord, as well as from frontal, temporal and precentral cortical regions were examined. In all areas there was astrogliosis, especially intense in the corticospinal tract. In the cortical regions there were small inclusions in both astrocytes and neurons that reacted with the mutant‐specific Inger antibody (not shown). No apparent major cell death or spongiosis was seen in the investigated cortical areas. There was a severe loss of motor neurons at all levels of the spinal cord. Bunina bodies were seen in many of the remaining motor neurons (Fig. 3A). Ubiquitin‐immunopositive inclusions were seen in motor neurons (Fig. 3J, K) some of which also stained with the Inger antibody (Fig. 3K). Lewy body‐like hyaline inclusions were seen in both motor neurons (Fig. 3B) and astrocytes in the spinal cord (Fig. 3C). These inclusions often reacted with the Inger antibody (Fig. 3G), and some also showed weak staining with the 131–153 antibody (not shown). This suggests the presence of some wild‐type SOD1, in accordance with the findings with detergent‐washed aggregates (Fig. 2A). Small, round inclusions reacting with the Inger antibody were seen in motor neuron somata and in the neuropil (Fig. 3E–G), as were skeins (Fig. 3F) and Lewy body‐like hyaline inclusions (Fig. 3G). No aggregates of SOD1 were seen in motor axons. The findings show similarities with some previous reports in patients (Shibata et al., 1996; Kato et al., 1997, 2000; Bruijn et al., 1998; Kokubo et al., 1999). In these studies, the anti‐SOD1 antibodies that were used could not distinguish between mutant and wild‐type enzyme. Skein‐like ubiquitin‐immunopositive inclusions were seen in some motor neurons (Fig. 3J). Astrocytes were frequently immunoreactive for antibodies directed against αB‐crystallin (Fig. 3L). Astrocytes and motor neurons were often immunopositive for antibodies against UCH‐L1 (Fig. 3I).

Fig. 3 Histopathology of spinal cord from the ALS patient. (A) Haematoxylin/eosin (HTX) staining showing bunina bodies (arrows) in a motor neuron. (B and C) HTX staining showing Lewy body‐like hyaline inclusions (LBHI) in motor neurons and astrocytes respectively (arrows). (D) GFAP‐stained astrocyte (brown) with an inclusion labelled by the mutant‐specific Inger antibody (red, arrow). (EG) Staining with the mutant‐specific Inger antibody, showing small inclusions, a skein and an LBHI in motor neurons respectively. (H) Motor neuron with small granular inclusions below the nucleus staining with anti‐SOD1 antibodies (not shown) from a homozygous D90A ALS patient. Note the absence of staining with the Inger antibody. (I) UCH‐L1 staining in motor neurons. (J) Motor neuron with skein‐like inclusions labelled by an antibody directed against ubiquitin. (K) Motor neuron with small round inclusions labelled by antibody directed against ubiquitin (brown) and some also labelled by the mutant‐specific antibody Inger (red, arrows). (L) αB‐crystallin in astrocytes. Scale bars = 10 µm.

Survival and disease phenotype of G127X transgenic mice

Nineteen lines of G127X transgenic mice were generated and the two with the highest gene copy numbers, line 832 with 28 copies and line 716 with 19 copies, developed signs of motor neuron degeneration. Mice homo‐ and heterozygous for these two insertions were studied, and mean survivals for the 832 mice were 126 and 213 days respectively, and for the 716 mice 250 and 477 days respectively (Fig. 4A). Homozygous 716 mice were the principal focus of the present study, and are called G127X mice here. Expression of the SOD1 transgene in the brain of G127X mice, as determined by northern analysis, was ∼60% of that in the commonly studied G93AGur mice (Gurney et al., 1994), which survived 124 days in our laboratory. The mRNA content was 50% of that in G93AGur mice in a subline with reduced copy number, G93AGurdl mice. The survival time of the G93AGurdl mice, 253 days, was similar to that of the G127X mice.

Fig. 4 Survival and mutant protein in G127X transgenic mice. (A) Survival of line 716 G127X transgenic mice homozygous and heterozygous for the insertion. The mice were killed when they could no longer feed. (B) Amount of SOD1 protein in mouse tissue homogenates estimated from immunoblots. Mutant SOD1 was determined in line 716 G127X transgenic mice using the24–39 anti‐SOD1 antibody. Wild‐type mouse SOD1 was determined in 100‐day‐old control mice using the 131–153 anti‐SOD1 antibody. G93A mutant protein was determined in pooled extracts from three 100‐day‐old G93AGurdl mice using the 24–39 anti‐SOD1 antibody. (C) Amounts of G127X mutant SOD1 pelleted by centrifugation at 20 000 g expressed as percentages of total amount of mutant enzyme in the mouse spinal cord homogenates. Solid triangles show the aggregated enzyme remaining after five washes in pH 7.0 PBS containing 0.1% NP40 (cf. Fig. 2D and see Material and methods). Open circles show pelleted mutant SOD1 following one 20 000 g centrifugation of spinal cord homogenates in pH 7.0 PBS without detergent (cf. Fig. 2E). To compensate for cytosol apparently trapped in processomes, the amount of wild‐type mouse SOD1 was also analysed, and that percentage was subtracted from the data for the mutant enzyme (see Material and methods).

The disease phenotype differed somewhat from the G93AGur mice. Whereas onset with hindleg paralysis was the rule in the G93AGur mice, about one‐third of the G127X mice showed foreleg onset. There was also a more rapid disease course in the G127X mice: 7–10 days from the first signs versus about 3 weeks in the G93AGur mice.

Low level of inactive mutant SOD1 in the transgenic mice

Immunoblots of whole homogenates of brain and spinal cord of G127X mice showed two heterogeneous major bands, one around 17 kDa and the other at 33 kDa, accounting on average for 60 and 40% of the mutant enzyme respectively (Fig. 5A). This pattern was already seen at 2 days of age and persisted thereafter. The major 17‐kDa band was found in the same position as that in fibroblasts from the patient treated with an inhibitor of the proteasome and probably represents a full‐length G127X subunit (Fig. 2B).

Fig. 5 Analysis of brain and spinal cord from G127X mice by immunoblotting, staining for SOD activity, and 2‐D gel electrophoresis. (A) Immunoblots of brain and spinal cord homogenates from 100‐day‐old G127X transgenic mice (Tg) and wild‐type control mice (C) using the 24–39 anti‐SOD1 antibody. (B) Isoelectric focusing and staining for SOD activity. Twenty microlitres of 20 000 g supernatant of brain homogenates from a SOD1 knockout mouse, a heterozygous G127X mouse on the SOD1 knockout background, a control mouse and a G127X mouse was applied to the gel and analysed. Asterisks mark cyanide‐resistant SOD2 bands and arrowheads cyanide‐inhibited SOD1 bands. (C) 2‐D gel electrophoresis of spinal cord homogenates from a 100‐day‐old and a terminal mouse blotted and stained for SOD1 using the 24–39 anti‐SOD1 antibody and for ubiquitin using an anti‐ubiquitin antibody. Note the lack of coincidence of staining at 100 days and the apparently coinciding diagonal train in the terminal mouse.

The G127X SOD1 protein patterns in spinal cord from 100‐day‐old and terminal mice were further analysed by 2‐D gel electrophoresis and immunoblotting (Fig. 5C). Both the 17‐kDa and in particular 33‐kDa regions were found to be composed of multiple species. To probe for complexing with ubiquitin, replicate gels were blotted with anti‐ubiquitin antibodies. A low molecular mass spot, apparently representing monomers, was found, as well as multiple high molecular weight spots. At 100 days there was no coincidence of staining between SOD1 and ubiquitin. In the terminal mouse, on the other hand, an apparently coinciding train of diagonal spots suggested the presence of some ubiquitinated SOD1.

The total level of G127X mutant SOD1 was estimated from the sum of the two major bands in immunoblots. In 100‐day‐old G127X mice, the levels in brain and spinal cord were low, and corresponded to about 97 and 45% respectively of the levels of SOD1 in wild‐type mice (Fig. 4B). In brain there was a 30% rise between day 2 and day 50 and thereafter a plateau was reached. In spinal cord there was a rise in G127X mutant SOD1 only in the last phase of the disease. However, the relative amounts of mutant protein in the CNS were much higher than those seen in liver and kidney, which corresponded to 6 and 8% respectively of the levels of background mouse SOD1 of the organs (Fig. 4B). In erythrocytes, which contain about as much SOD1 as brain and spinal cord, the G127X level was below our immunoblot detection limit, i.e <0.2% of the wild‐type mouse SOD1 level.

There was no significant difference in SOD1 activity in the CNS between the G127X mice and wild‐type mice of the same age (Table 2). The extraction medium normally used in our laboratory contains a high concentration of KBr and a chelator. This might inactivate the labile G217X SOD1. We therefore also extracted the tissues in medium of physiological ionic strength and pH without chelators. All the different extracts were analysed with the two different KO2 and pyrogallol SOD assays with and without chelators in the assay medium (see Material and methods). Omission of the chaotropic salt and the chelators did not, however, result in increased SOD1 activity of the tissue extracts from the G127X mice (data not shown). The SOD activity was also determined in isoelectric focusing gels stained for SOD activity. Extracts and staining solutions were prepared in the absence of chelators. To increase the possibility of detecting a minor activity band, line 716 G127X transgenic mice heterozygous for the insertion in SOD1 knockout background were also analysed. No activity band that could be attributed to G127X mutant SOD1 was detected (Fig. 5B). The assay has high sensitivity, and if the soluble G127X mutant enzyme had had a specific activity ≥0.5% of that of wild‐type SOD1 it should have been detected.

View this table:
Table 2

Total SOD1 activity (U/g wet weight) in mutant SOD1 transgenic strains and wild‐type controls

Brain14 10013 300105 000
Spinal cord10 30010 90087 600
Liver90 80087 000383 000
Kidney37 90040 200151 000

Analyses were made in 100‐day‐old mice. Data are means of results from three mice of each genotype.

The G93AGur and G93AGurdl mutant mice (Gurney et al., 1994) contained 35‐fold (not shown) and 20‐fold, respectively, more mutant protein in spinal cord than homozygous G127X mutant mice at 100 days (Fig. 4B). In the G93AGurdl mice the SOD1 activities were also markedly increased (Table 2).

Life‐long presence of detergent‐soluble G127X mutant SOD1 protoaggregates

CNS tissues are commonly extracted with buffers to which weak detergents, such as NP40 and deoxycholate, are added at ∼1% concentrations to solubilize myelin and various membranes (Johnston et al., 2000; Shinder et al., 2001; Wang et al., 2002b). To avoid potential artefactual influences on the mutant SOD1, we extracted murine CNS with PBS at the intracellular pH 7.0 in the absence of detergents. Despite disruption with the Ultraturrax followed by extensive sonication, regularly ∼15% of the wild‐type SOD1 of murine and human brain and spinal cord was sedimented by centrifugation at 20 000 g for 30 min. Higher speeds did not result in further sedimentation of enzyme. Less (5–10%) of the wild‐type SOD1 was pelleted from liver and kidney extracts from wild‐type mice. This phenomenon was further explored. When brain from a SOD1 knockout mouse was homogenized and sonicated in the presence of a haemolysate to yield the same SOD1 activity as in a homogenate of a wild‐type mouse, only ∼2% of the enzyme was present in the sediment. This shows that the pelleting is not caused by affinity of SOD1 for sedimenting components. When 20 000 g sediments from control mouse CNS were dissolved by sonication in buffer containing 0.1% NP40, the SOD1 enzymic activity released to the supernatant corresponded to ∼95% of the SOD1 protein that had been sedimented. This shows that the sedimented, apparently trapped, wild‐type SOD1 protein was enzymatically active. Likewise, ∼85% of the SOD2 activity was present in the first supernatant, and 15% was released from the NP40‐solubilized pellet. Since SOD2 is localized to the mitochondrial matrix, this shows that the mitochondria were efficiently disrupted by the primary homogenization. SOD1 that was localized to the mitochondrial intermembrane space should have been released. On the basis of these studies (detailed data not presented), we suggest that the sedimented SOD1 mainly derives from cytosol trapped in structures formed by disruption and closure of the abundant neuronal and glial processes that exist in the CNS. We refer to this sedimentable material as ‘processomes’.

First we homogenized spinal cords from mice of different ages (from day 2 until they were moribund) in pH 7.0 PBS. At all ages, a large proportion of the G127X mutant protein was sedimented by centrifugation at 20 000 g (Fig. 2E). The amount of processomes was determined in all experiments by analysis of wild‐type murine SOD1 in the pellet and the original homogenate. The processomes corresponded to ∼4% at day 2 and ∼15% of the wild‐type SOD1 of the homogenate at 50 days and later. After subtraction of these figures it was found that, at all ages examined, an additional 20% of the G127X SOD1 was sedimented (Fig. 4C). We propose that this additional material represents aggregated G127X SOD1. The true amounts of aggregated mutant enzyme were probably somewhat larger than indicated by the data in Fig. 4C, since the processomes were also likely to have trapped some aggregated enzyme. The SOD1‐immunoreactive complexes of ∼33 kDa (Fig. 5A) were ∼2‐fold more prone to form aggregates and accounted on average for 60% of the sedimented material (Fig. 2E). Similar data were found for the brain. In brain homogenates from four 100‐day‐old mice, on average 20% of the wild‐type mouse SOD1 was pelleted at 20 000 g, and an additional 17% of the G127X‐mutant SOD1. From liver and kidney, similarly homogenized and centrifuged, a minimal amount and 20% respectively more of the G127X SOD1 was sedimented compared with mouse SOD1.

The G127X SOD1 appeared to be very labile (see also the post‐mortem stability study below), which suggested that the aggregation could have been caused by the handling of the specimen. We therefore carried out a number of control experiments. Brains and spinal cords from 100‐day‐old G127X mice were cut into pieces, homogenized and analysed by centrifugation, either immediately after the animals had been killed or after being kept frozen at –80°C. No differences in amounts of sedimented mutant SOD1 or patterns on immunoblots could be discerned. Likewise, there were no differences between homogenates centrifuged immediately and homogenates kept for 2, 6, 12 and 24 h at 4°C before centrifugation. Freezing of a whole homogenate overnight at –80°C did not alter the distribution of G127X SOD1. Thus, following disruption of the tissue and dilution in the homogenization buffer, the physical and structural properties of the mutant protein appeared to be relatively stable.

Detergent‐resistant aggregates accumulate late in the disease in transgenic mice

Spinal cords and brains from mice of different ages were homogenized, centrifuged and analysed in detergent‐containing buffer. Here we chose the non‐ionic detergent NP40 at 0.1%, which was the lowest concentration found reliably to dissolve or prevent the formation of processomes. The sediments were washed five times in this buffer. For each washing step, some aggregated mutant SOD1 was dissolved and the final pelleted material thus showed relatively high detergent‐resistance. In this case, small but increasing amounts of G127X SOD1 were sedimented until 150 days (Figs 2D and 4C). For up to 150 days of age, the final sediment contained more of the complexes of ∼33 kDa than the 17‐kDa band. Near‐terminal disease, (i.e. at 200 days), and in moribund mice, there was a marked accumulation of the detergent‐resistant aggregates. As in the patient (Fig. 1B), there was a much greater amount of these aggregates in spinal cord than in brain (Fig. 2D). These terminally occurring aggregates were composed of multiple molecular forms and there was extensive high molecular weight smearing. There was no immunoreactive material at the point of application or in the stacking gel. No murine SOD1 could be demonstrated in the aggregates by blotting with the 131–153 antibody (not shown; detection limit ∼0.3% of the sedimented G127X mutant enzyme).

When spinal cords from G93AGur mice were analysed in the same way (i.e. extracted, centrifuged and washed five times in 0.1% NP40 and the sediment examined by immunoblotting), it was found that in these mice too there was a marked accumulation of detergent‐resistant aggregated mutant SOD1 in spinal cords in the last phase of the disease, but not in brain (not shown). Remarkably, the total amounts of such aggregates in moribund G127X (Fig. 2D) and G93AGur mice were similar: ∼15 µg/g wet weight.

Rapid post‐mortem aggregation of G127X mutant SOD1

The CNS from the ALS patient had a 20 h post‐mortem time, which may have altered the properties of the labile mutant enzyme. Pieces of brain from a G127X mouse were immediately frozen or kept before freezing for various times under conditions mimicking the temperature changes of a deceased human body (see Material and methods). The pieces were then homogenized in pH 7.0 PBS without detergent then centrifuged, and supernatants and pellets were examined by immunoblotting (Fig. 2F). While ∼50% of the mutant enzyme was sedimented from the homogenate of the snap‐frozen piece, nearly 100% was sedimented by centrifugation after 6, 12 and 24 h of storage. At none of these storage times was there any evidence for degradation or other structural alterations in immunoblots of the whole homogenates. The SOD1 in brain pieces from wild‐type mice similarly stored and analysed did not show any evidence for aggregation or other changes (not shown).

Early activation of tissue reaction markers in the mice and concordance with the patient with respect to terminal histopathology

Sections from the brain and cervical, thoracic and lumbar spinal cord were examined in 2‐, 50‐, 100‐ and 200‐day‐old and terminally ill mice. At least two animals (G127X mutant and C57BL/6J controls) were studied at each age. Spinal motor neurons showed diffuse immunoreactivity with the mutant‐specific Agaz antibody from 2 days of age onwards (Fig. 6A–C). In 100‐day‐old mice this immunoreactivity was also seen in small inclusions in motor neuron somata (Fig. 6B), and these inclusions had increased in size at 200 days of age and in terminally ill mice (Fig. 6C). Agaz‐positive inclusions were also seen in the neuropil and, in end‐stage mice, also in a few astrocytes. SOD1‐positive inclusions were also seen in terminal G93AGur mice (Fig. 6L, M), but in striking contrast the terminal G127X mice did not show inclusions and swellings in ventral roots and ventral funiculi (Fig. 6I, J). Possibly the very unstable G127X mutant SOD1 only rarely enters the axonal transport system. This suggests that axonal injury does not play a major role in motor neuron degeneration, at least in the G127X mouse model. Ventral root fibres had an unremarkable appearance in haematoxylin/eosin‐stained sections from the G127X mice.

Fig. 6 Histopathology of the spinal cord from transgenic mice. (AK) Material from G127X transgenic mice. (L, M) Material from an end‐stage G93AGur transgenic mouse. Scale bars = 10 µm. (AC) Staining with the Agaz antibody, showing staining primarily of motor neurons of increasingly particulate nature in (A) 50‐day‐old, (B) 100‐day‐old and (C) terminal G127X transgenic mice. (D, H) Haematoxylin/eosin staining, showing an LBHI (arrows) in (D) an astrocyte and (H) a motor neuron. (EG) GFAP staining, showing reactive astrocytes in (E) 50‐day‐old, (F) 100‐day‐old and (G) terminal spinal cord. (I, J, L, M) Staining with 3–20 anti‐SOD1 antibody to show ventral roots (I, L) and ventral horns (J, M) from (I, J) end‐stage G127X and (L, M) G93AGur mice. Note the heavy staining in motor neurons and neuropil. Only the G93AGur mouse shows aggregates and swellings in axons (L). K Motor neuron inclusions were labelled by an antibody directed against ubiquitin (red) and some were also labelled by the 3–20 anti‐SOD1 antibody (brown, arrows).

Weak GFAP immunopositivity was seen in the spinal cord in 50‐day‐old mice, and this increased in 100‐ and 200‐day‐old and terminally ill mice (Fig. 6E–G). Weak GFAP immunoreactivity was seen in cortical regions in 200‐day‐old and terminally ill mice, while no apparent differences from wild‐type mice were seen in younger animals.

Sections from 2‐day‐old mice were unremarkable, but in sections from mice aged ≥50 days αB‐crystallin‐immunoreactive astrocytes were seen throughout the spinal cord, including the white matter and dorsal horns (not shown). In the 200‐day‐old mice and the terminally ill mice this immunoreactivity was also seen in motor neurons. Intense immunoreactivity for an antibody directed against UCH‐L1 was seen in spinal motor neuron somata and to a lesser degree in the neuropil in mice aged ≥50 days (not shown).

Ubiquitin staining increased with time in 50‐, 100‐ and 200‐day‐old and terminal mice, more than in wild‐type mice of the same ages. Some of the ubiquitin staining appeared to colocalize with inclusions staining for human SOD1 (Fig. 6K).

Many similarities were thus noted between sections from terminally ill mice and the patient. Skein‐like and Lewy body‐like hyaline inclusions were seen in motor neurons and some astrocytes in both (Figs 3B, C, J and 6D, H) and some of these inclusions reacted with the mutant‐specific Inger or Agaz antibodies (Figs 3E–G and 6B, C, K). Astrocytes immunopositive for antibodies directed against αB‐crystallin (Fig. 3L) were also seen in both species, as was astrogliosis (Figs 3D and 6E–G).


The amounts of mutant protein found here in the spinal cord of G127X mice are about half those previously reported in G85R mice with similar survival length (Bruijn et al., 1997) and much lower than those of other transgenic mouse models (Gurney et al., 1994; Wong et al., 1995; Wang et al., 2002b). According to the northern blotting data, the synthesis rate of G127X mutant SOD1 should be similar to that of G93A SOD1 in G93AGurdl mice, which contain about 20‐fold more mutant enzyme in brain and spinal cord (Fig. 4B). This suggests that the G127X mutant protein is unstable and rapidly degraded. The instability is probably caused by the change in 5 and loss of 21 C‐terminal amino acids that are involved in multiple inter‐ and intra‐subunit interactions, including loss of the stabilizing Cys57–Cys146 disulphide bond (Tainer et al., 1982). Although all the ligands to the prosthetic metals are present, the G127X mutant protein lacked SOD activity, further supporting the notion of a misfolded structure. Moreover, two amino acid residues, Arg143 and Leu152, that interact with the copper chaperone for superoxide dismutase in heterodimer formation are absent (Lamb et al., 2001). CNS from the patient contained even less G127X mutant SOD1. No conclusions can be drawn from the absence of soluble G127X mutant SOD1 in CNS homogenates from the patient (Fig. 2A), since the post‐mortem storage study in the G127X mouse indicated that any soluble protein should have aggregated in the patient. There was most mutant protein in the ventral horns (Fig. 1B), but a large proportion of that probably existed in presumably innocuous sequestered inclusions. Such inclusions also occurred elsewhere in the CNS but were less abundant. The data shown in Table 1 and Fig. 4B suggest that the steady‐state level of non‐sequestered G127X mutant SOD1 should be similar in different areas of the CNS of the patient, and thus clearly less than the 0.4% of control SOD1 content found in the temporal lobe. Any proposed mechanism of injury must be compatible with these findings. Thus it is unlikely that prosthetic copper is involved in ALS pathogenesis, as also previously suggested (Subramaniam et al., 2002), and major noxious sequestering of relatively abundant components such as chaperones is also an unlikely explanation (Okado‐Matsumoto and Fridovich, 2002). The extremely high levels of mutant SOD1 found in some of the transgenic models may cause some alterations without relevance for human ALS pathogenesis.

Whereas the content of wild‐type SOD1 in the CNS is only intermediate (Marklund, 1984; Carlsson et al., 1995), the content of G127X mutant SOD1 was higher in the CNS than in other organs and there was little change from 2 to 200 days of age (Fig. 4B). The mutant protein aggregated rapidly (Fig. 2F), and ∼20% of the G127X mutant SOD1 in spinal cords homogenized without detergent appeared to occur as aggregates at 2, 50, 100 and 150 days (Fig. 4C). This indicates that the high levels of both the soluble misfolded and the aggregated G127X mutant SOD1 are not accumulated, but rather represent steady states between formation and degradation. Apparently there is a slow degradation of misfolded and aggregated (mutant) SOD1 in the CNS compared with other organs. These species could mislocalize and/or interfere with essential cellular functions and thereby induce the processes that finally lead to motor neuron degeneration. The early appearance of increased staining for UCH‐L1, αB‐crystallin and GFAP (Fig. 6E–G) in the present study, as well as findings in other transgenic models (Mourelatos et al., 1996; Williamson and Cleveland, 1999; Bendotti et al., 2001; Vukosavic et al., 2000), are in accordance with the notion of a long‐term noxious influence on spinal cord cells. Recent studies on proteins involved in other neurodegenerative diseases [amyloid‐β‐peptide (Walsh et al., 2002) and α‐synuclein (Volles and Lansbury, 2002)] as well as more artificial systems (Bucciantini et al., 2002) also suggest that early oligomeric protein aggregates are particularly noxious.

The NP40‐resistant aggregates appear to occur too late in the disease course to be causative (Figs 2D and 4C). The very high levels in end‐stage disease suggest a final dysfunction in SOD1 degradation, and the aggregates must occur in several cell types in the tissue. Notably, despite a 35‐fold difference in total mutant SOD1 content, the amounts of NP40‐resistant aggregates were similar in terminal stage G127X and G93AGur murine spinal cord. Similar late detergent‐resistant aggregates have previously been reported in the G93AGur (Johnston et al., 2000; Shinder et al., 2001; Wang et al., 2002a) and H46R/H48Q (Wang et al., 2002b) mouse models. The appearance in the spinal cord of first small granular inclusions and later larger, more dense inclusions of G127X mutant SOD1 (Fig. 6A–C) correlates with the course of appearance of the NP40‐resistant aggregates (Fig. 4C). This suggests that the inclusions seen in mouse spinal cord and in the anterior horns and the precentral gyrus of the patient are nearly congruent with the NP40‐resistant aggregates. We find no histological correlates of the protoaggregates, which may be amorphous. The inclusions may be so‐called aggresomes sequestered as a protective measure (Johnston et al., 1998). Interestingly, these terminal aggregates form a least common denominator between ALS patients carrying SOD1 mutations and the transgenic models. Possibly they should be seen as terminal markers of increased concentrations of noxious misfolded soluble SOD1 and protoaggregates.

Notable are SOD1‐immunoreactive complexes of ∼33 kDa, which were more abundant than the native band in both the detergent‐soluble and resistant aggregates in the mice (Fig. 2D, E). A band with the same relative abundance at 36 kDa occurred in afflicted areas in the human ALS case (Fig. 1B), whereas in the fibroblasts only the native subunit at 17 kDa was seen (Fig. 2B). In 100‐day‐old mice, there was no evidence that the high molecular weight complexes were ubiquitinated, whereas some ubiquitination of SOD1 was indicated in terminal mice (Fig. 5C). This suggests that at least part of the SOD1 is degraded in ubiquitin‐dependent pathways, and that these pathways are compromised in end‐stage disease. The potential involvement of the high molecular weight complexes in ALS pathogenesis deserves further exploration.

The primary cellular site(s) of noxious activity of mutant SOD1s is not known. Attempts to cause motor neuron disease by specific overexpression of mutant SOD1 in astrocytes (Gong et al., 2000), neurons in general (Pramatarova et al., 2001) or motor neurons only (Lino et al., 2002) have so far failed. This suggests that other cell types are primarily involved or that the simultaneous presence of mutant SOD1 in several cell types together causes a tissue reaction resulting in motor neuron injury. SOD1 levels differ at most by a factor of 2 between different areas in the human brain; the anterior horns and precentral gyrus show intermediate activities (Table 1). Thus, the normal human and murine promoter apparently directs abundant synthesis in many cell types in the CNS. The altered G127X mutant SOD1 material found in the present study should exist in several cell types in the spinal cord and brain.

Several studies have shown that neurons in the brain are also compromised in the transgenic mouse ALS models (Wong et al., 1995; Kostic et al., 1997; Lutz‐Bucher et al., 1999). Lesions outside the motor system have been described in ALS patients carrying SOD1 mutations (Hirano, 1998), and widespread CNS degeneration was found in a case in which survival was extended by placement on a respirator (Kato et al., 1996). Like the spinal cord, the brain contained high levels of G127X mutant SOD1 (Fig. 4B) as well as similar levels of protoaggregates. Possibly, processes similar to those in the spinal cord also occur in some areas in the brain, but at a somewhat slower rate.

Any theory advanced for the toxicity must be applicable to all of the >100 mutant SOD1s associated with ALS. The present data suggest that the G127X and G93A mutant SOD1s are similarly potent in causing ALS. Furthermore, the ALS‐linked mutant SOD1s cause dominantly inherited disease with rather similar phenotypes, suggesting similar degrees of toxicity. The mutant SOD1s show, however, widely different stabilities in the CNS and other organs. There is a 20‐fold difference in mutant SOD1 content in the spinal cord between the G93AGurdl and the G127X transgenic mice (Fig. 4B), and the difference between ALS patients carrying the D90A (content and activity equal to those of wild‐type enzyme in control individuals; K Ernhill, PA Jonsson, P Andersen, SL Marklund and T Brännström, unpublished data) and G127X mutations is >100‐fold. This suggests that noxious subfractions are causing the toxicity, rather than the bulk of the mutant SOD1s found in the spinal cord. We propose that the properties of such putative subfractions may be similar to those we here find in G127X mutant SOD1, and the question is how they could arise. Several studies have shown reduced structural stabilities in mutant SOD1s. A particularly distinguishing difference from the wild‐type enzyme is found for the metal‐free apo‐state, in which both stable mutants, such as D90A and G93A, and more structurally disrupted mutants, such as A4V and C6F, share pronounced destabilization (Lindberg et al., 2002). The apo‐forms are also more prone to aggregation under denaturing conditions (Stathopulos et al., 2003). Furthermore, the stabilizing intrasubunit disulphide bond should impose strict constraints on possible conformations of misfolded SOD1. In G127X mutant SOD1, as well as in several other ALS‐linked truncation mutations and C146R, this bond is absent, suggesting that it should be reduced in the putative noxious subfractions. The disulphide bond is more easily reduced in various ALS‐linked mutant SOD1s than in the wild‐type enzyme (Tiwari and Hayward, 2003). Metal‐free and disulphide‐reduced SOD1 exists following polypeptide synthesis, and probably also when the enzyme is finally degraded (Fig. 7). The synthesis and degradation must occur at equal rates in the patients for all mutant SOD1s, despite the widely different steady‐state levels in the CNS. Following synthesis of the polypeptide, however, their fates appear to differ widely. Some, such as D90A, mature almost completely to native active SOD1, whereas others, such as the extreme G127X, probably never assume a native‐like folding. The mutant SOD1s may, however, following loss of the metals and disulphide reduction during the degradation, share a propensity to attain a noxious folding state. The present finding—that misfolded mutant SOD1 is found in higher steady‐state levels in the CNS than elsewhere in the body—thus provides a partial explanation for the particular susceptibility of some areas of the CNS to the adverse effects of mutant forms of this ubiquitously expressed protein.

Fig. 7 Suggested mechanism of formation of noxious subfractions of mutant SOD1s. The steady‐state levels of mutant SOD1s in spinal cord of ALS patients differ more than 100‐fold between different mutations, but the rates of degradation should be equal. The present study shows that minute quantities of misfolded mutant SOD1 lacking the intrasubunit disulphide bond suffice to induce motor neuron degeneration. We suggest that the mutant SOD1s share a propensity to assume a noxious folding state, possibly following loss of metals and disulphide reduction during degradation. The present finding, that misfolded mutant SOD1 lacking the disulphide bond is found in higher steady‐state levels in the CNS than elsewhere in the body, provides a partial explanation for the particular susceptibility of some parts of the CNS to the adverse effects of mutant forms of this ubiquitously expressed protein.


We wish to thank Eva Bern, Ingalis Fransson, Karin Hjertkvist, Åsa Håkansson, Ann‐Charlott Nilsson, Karin Wallgren and Agneta Öberg for expert technical assistance and Mikael Oliveberg for stimulating discussions. The study was supported by the Swedish Science Council, the Swedish Medical Society/Björklunds Fund for ALS Research, the Swedish Association of Neurologically Disabled and the Council of Västerbotten County.


View Abstract