A novel SOD1-ALS mutation separates central and peripheral effects of mutant SOD1 toxicity

Transgenic mouse models expressing mutant superoxide dismutase 1 (SOD1) have been critical in furthering our understanding of amyotrophic lateral sclerosis (ALS). However, such models generally overexpress the mutant protein, which may give rise to phenotypes not directly relevant to the disorder. Here, we have analysed a novel mouse model that has a point mutation in the endogenous mouse Sod1 gene; this mutation is identical to a pathological change in human familial ALS (fALS) which results in a D83G change in SOD1 protein. Homozgous Sod1D83G/D83G mice develop progressive degeneration of lower (LMN) and upper motor neurons, likely due to the same unknown toxic gain of function as occurs in human fALS cases, but intriguingly LMN cell death appears to stop in early adulthood and the mice do not become paralyzed. The D83 residue coordinates zinc binding, and the D83G mutation results in loss of dismutase activity and SOD1 protein instability. As a result, Sod1D83G/D83G mice also phenocopy the distal axonopathy and hepatocellular carcinoma found in Sod1 null mice (Sod1−/−). These unique mice allow us to further our understanding of ALS by separating the central motor neuron body degeneration and the peripheral effects from a fALS mutation expressed at endogenous levels.

to further our understanding of ALS by separating the central motor neuron body degeneration and the peripheral effects from a fALS mutation expressed at endogenous levels.

Introduction
Amyotrophic lateral sclerosis (ALS) is a devastating neurodegenerative disease characterized by a loss of upper and lower motor neurons (LMNs), which causes muscle weakness, paralysis and ultimately death, typically within 3-5 years of disease onset. Approximately 10% of ALS cases have a clear family history (fALS), caused by mutations in specific genes, usually with a dominant pattern of inheritance (1)(2)(3). Mutations in superoxide dismutase 1 (SOD1) account for 10-20% of fALS cases (although pathogenicity has not be demonstrated for all these changes), and to date >155 mutations have been identified throughout all five exons of the SOD1 gene, >95% of which are dominant (4).
SOD1 is a 153 amino acid metalloenzyme (also called Cu/Zn SOD1) that forms a homodimer whose main known function is to remove superoxide radicals through creating molecular oxygen and hydrogen peroxide, although other functions are known (5). Mutant SOD1 takes on a toxic gain of unknown function in SOD1-fALS, causing many cellular abnormalities that ultimately result in death of motor neurons (6). Recent research has identified misfolded wild-type (WT) SOD1 in non-SOD1-fALS and in 'sporadic' ALS suggesting that it may also play a role in the pathogenesis of these ALS cases (7)(8)(9)(10).
SOD1 is highly conserved across species (11) and >12 different transgenic mouse models have been created that overexpress mutant forms of human SOD1 (6,12,13) and in one case mouse Sod1 (14). The majority of these mice recapitulate many characteristics of ALS, including progressive motor deficits, hindlimb paralysis, motor neuron degeneration and early death (6,12,13). A mouse strain carrying a spontaneous point mutation in mouse Sod1 has been previously described (15), although the equivalent mutation in humans has not been identified as pathogenic.
However, concerns remain regarding the translation of these models to human SOD1-fALS-particularly because SOD1 is generally overexpressed in transgenics and such raised expression levels affect the pathology of these animals (6,12). For example, the most widely used model of SOD1-fALS, the high-copy SOD1 G93A transgenic, carries ∼24 copies of the mutant human SOD1 gene, expresses SOD1 protein at ∼17-fold over the endogenous level, and has greatly accelerated disease compared with SOD1 G93Adl mice, a strain derived from the SOD1 G93A founder line but with lower levels of SOD1 protein because of a deletion in the transgene array (∼8-10 copies of SOD1 G93A gene, 8-fold SOD1 protein expression over WT) (16)(17)(18). As well as raised levels of mutant SOD1 affecting phenotype, increased levels of WT SOD1 also give rise to neurodegeneration-overexpression of WT human SOD1 at levels comparable with that found in SOD1 G93A transgenics results in an ALS-like syndrome with progressive loss of spinal motor neurons and premature death (19). Thus SOD1 'dose' is clearly important for determining phenotype-and as well as overexpression, reduced expression also gives rise to neuronal and non-neuronal phenotypes in heterozygous and homozygous SOD1 knockout mice (reviewed in 4).
Mutations in SOD1-ALS cause a toxic gain of function, which leads to motor neuron degeneration. However, curiously, the majority of studies that have analysed dismutase activity of SOD1-fALS patient samples show that SOD1 dismutase activity is reduced to an average of ∼58% of normal levels (reviewed in 4). SOD1 transgenic models overexpress mutant SOD1 and also express two copies of endogenous mouse Sod1, so dismutase activity is not reduced in these animals. Therefore, although SOD1 transgenics clearly model the SOD1 toxic gain-of-function leading to motor neuron degeneration, they do not generally model the possible effects on ALS pathogenesis of a reduction in dismutase activity, as observed in the majority of SOD1-fALS patient samples.
The effects of SOD1 loss of function on motor neurons have been recently readdressed through the study of Sod1 −/− mice (20,21), which suffer from a severe progressive denervation of hindlimb muscles, leading to striking motor phenotypes (20). Importantly however, several reports show that aged Sod1 −/− mice do not develop motor neuron degeneration at any age (20,22,23). Thus, SOD1 activity is critical in maintaining innervation of neuromuscular junctions, but its removal does not result in motor neuron cell body loss.
In order to create the genetically closest model of SOD1 ALS to date, and to investigate the effects of a SOD1 mutation at endogenous expression levels, we identified a mutant line that carries an N-ethyl-N-nitrosourea (ENU)-induced point mutation in the mouse Sod1 gene. Fortuitously, this mutation is identical to the nucleotide change found in human SOD1 D83G dominant fALS cases (24). In a D83G SOD1-fALS family, four of the five affected individuals had a rapid disease duration (range: 6-12 months), whilst one family member had a long disease duration (151 months). Two of the affected SOD1 D83G family members who were clinically examined in detail first presented with symptoms of LMN deficits, which were followed with upper motor neuron (UMN) symptoms (24).

Identification of an ENU-induced point mutation in the mouse Sod1 gene
To identify mouse lines carrying the equivalent of human ALS causative pathogenic mutations, we screened for mutations in Sod1 using genomic DNA from an ENU-induced mutagenesis archive containing over 10 000 mice (25,26). We identified a mouse mutant carrying an adenosine-to-guanine missense mutation resulting in a D83G substitution (Supplementary Material, Fig. S1). Importantly, the same point mutation (A-G) gives rise to dominant D83G SOD1-fALS (24).
Reactive gliosis of both astrocytes and microglia is observed in ALS patients and mouse models. In Sod1 D83G/D83G mice, lumbar spinal cord sections from 15-week-old mice showed striking astrogliosis (GFAP) and microgliosis (IBA1), which increased further at 52 weeks of age (Fig. 1C). Sod1 +/D83G mice did not show elevation in reactive gliosis compared with WT littermates (Fig. 1C).
To investigate UMNs in Sod1 D83G/D83G mice, we examined the survival of corticospinal motor neurons (CSMNs) at 15 and 29 weeks of age ( Fig. 1D and E). Although Nissl staining of the cerebral cortex did not reveal obvious abnormalities in Sod1 D83G/D83G mice at either age (Supplementary Material, Figure S2A), analysis of Cry-mu, a selective marker of CSMNs in layer V of the motor cortex (27), revealed signs of cellular degeneration in Sod1 D83G/D83G mice at 29 weeks (Supplementary Material, Figure S2B). This was supported by staining for the molecular marker chicken ovalbumin upstream promoter transcription factor-interacting protein 2 (CTIP2), a transcription factor expressed by CSMN in layer V of the motor cortex and a subset of layer VI neurons (27) (Fig. 1D). CTIP2 expression helps distinguish between a possible reduction in molecular marker expression and cellular degeneration, and is restricted to the nucleus, allowing for reliable quantitative analysis.
Analysis of functional motor units reveals a distal motor neuropathy in Sod1 D83G/D83G mice In view of the MN degeneration observed in Sod1 D83G/D83G mice, we next determined the number of functional motor neurons that innervated the extensor digitorum longus (EDL) hindlimb muscle by physiological analysis of motor unit survival (MUNE). We found no differences in EDL motor unit survival across genotypes at 15 weeks of age, suggesting that the loss of LMN cell bodies detected at this age is restricted to populations of motor neurons that innervate hindlimb muscles other than EDL. Indeed, EDL has been previously shown to be less vulnerable to disease in transgenic SOD1 models (28). However, by 52 weeks of age, we observed a significant reduction in the number of motor units in EDL muscles of Sod1 D83G/D83G mice compared with WT littermates (P < 0.001; Fig. 2A and B). Comparison between WT and Sod1 +/D83G littermates did not reveal any significant differences in EDL motor units at 15, 52 or 96 weeks of age ( Fig. 2A).
Morphological analysis of the innervation pattern of endplate neuromuscular junctions (NMJ) of the EDL muscle confirmed the distal progressive denervation occurring between 15 and 52 weeks of age in Sod1 D83G/D83G mice ( Fig. 2C-E). In agreement with the MUNE analysis, no significant differences in denervated EDL endplate NMJ were observed between any of the genotypes at 15 weeks of age (Fig. 2E). However, by 52 weeks of age, a significantly different proportion of endplate NMJ in Sod1 D83G/D83G EDL is denervated (WT 2.6 ± 0.9; Sod1 +/D83G 2.6 ± 1.1; Sod1 D83G/D83G 14.8 ± 2.8; P < 0.001) (Fig. 2E). Since no additional LMN body degeneration occurs between 15 and 52 weeks of age in Sod1 D83G/D83G mice, the progressive denervation and loss of EDL motor units are likely a peripheral neuropathy that is not the result of motor neuron death. Sod1 D83G/D83G mice display progressive motor and behavioural deficits We next examined whether the loss of motor neurons and progressive denervation in Sod1 D83G/D83G mice was reflected in deficits in motor function. Longitudinal phenotypic characterization of WT, Sod1 +/D83G and Sod1 D83G/D83G littermates showed motor function of Sod1 D83G/D83G mice deteriorates progressively with age, and these mice develop tremors, gait abnormalities and become severely kyphotic (Supplementary Material, Video S1).
We found a reduction in body weight of Sod1 D83G/D83G from 4 weeks of age compared with WT littermates (females, P = 0.001; males, P = 0.034) ( Fig. 3A; Supplementary Material, Fig. S3A). To evaluate the weight differences between littermates, the relative levels of lean and fat mass were determined using Echo MRI (29). Fifty-two-week-old male and female Sod1 D83G/D83G mice showed significantly less fat mass compared with sex-matched WT littermates, while there were no differences observed between WT and Sod1 +/D83G littermates at 6, 35, 52 or 88 weeks of age ( To assess motor function, we examined grip strength and performance on an accelerating rotarod, starting at 6 and 7 weeks of age, respectively. We observed reduced grip strength in female and male Sod1 D83G/D83G mice from 6 weeks of age (females, P ≤ 0.002; males, P < 0.05), but not in Sod1 +/D83G mice ( We also examined the performance of WT, Sod1 +/D83G and Sod1 D83G/D83G littermates using a modified SHIRPA analysis, which comprises of a battery of simple phenotypic tests with an emphasis on neurological function (30,31). Three traits differed between genotypes: the presence and progression of tremors, pelvic elevation and the ability to walk down a vertical wire grate (negative geotaxis) ( Table 1).
We found subtle behavioural deficits in heterozygous Sod1 +/D83G animals when examining the performance of mice on an in-cage wheel running system (32). A similar system has previously been used to identify presymptomatic motor abnormalities in SOD1 G93A transgenic mice (33). WT and Sod1 +/D83G littermates were assessed at 44 and 88 weeks of age by recording activity over 7 days. Sod1 +/D83G animals exhibited a significant deficit in nightly total distance travelled compared with WT littermates (P < 0.05), which declined further with age ( Fig. 3E). In addition, while WT mice at 88 weeks of age showed an increase in the duration for each running bout (duration per run) when compared with 44 weeks of age, 88-week-old Sod1 +/D83G mice showed a reduction in duration. Sod1 +/D83G mice also displayed a reduction in the duration per running bout at 44 and 88 weeks of age compared with WT littermates (Fig. 3F). Thus, Sod1 +/D83G mice present with subtle yet progressive deficits in motor function in the home cage, showing that the deleterious effects of the SOD1 D83G mutation are not restricted to homozygous mice.

Sod1 D83G/D83G mice lose muscle force with age
To gain a detailed understanding of the effect of the Sod1 D83G mutation on motor neuron and muscle function, we undertook a physiological analysis of the tibialis anterior (TA) and EDL hindlimb muscles of female WT, Sod1 +/D83G and Sod1 D83G/D83G littermates at 15 and 52 weeks age. We also assessed Sod1 +/D83G mice at 96 weeks of age.
TA muscle of 15-week-old Sod1 D83G/D83G mice was significantly weaker than in WT littermates (P = 0.02), and muscle force declined further at 52 weeks of age ( Fig. 4A and B). This may reflect Quantitative analysis of the percentage of denervated EDL endplate NMJ (denervated NMJ/total NMJ counted × 100) from all three genotypes at 15 weeks of age reveal no significant differences between any of the genotypes. By 52 weeks of age, a significant difference in the percentage of denervated endplate NMJ appear between Sod1 D83G/D83G and the other two genotypes (WT and Sod1 +/D83G ). At 15 weeks of age, at least 700 NMJ endplates were counted per genotype. At 52 weeks of age, at least 465 NMJ endplates were counted per genotype. Percentage of denervated endplate NMJ at 52 weeks: WT: 2.6 ± 0.9; Sod1 +/D83G : 2.6 ± 1.1; Sod1 D83G/D83G : 14.8 ± 2.8. ***P < 0.001. Scale bar is 20 μm.
the LMN degeneration detected in the spinal cord at this stage (Fig. 1B). However, as observed with motor unit survival and NMJ innervation, no deficit in EDL muscle force was detected in Sod1 D83G/D83G mice at 15 weeks, confirming that this muscle is less vulnerable to the effects of mutant SOD1 than other hindlimb fast twitch muscles. However, by 52 weeks, EDL muscle force in Sod1 D83G/D83G mice was significantly reduced by ∼35% compared with WT (P < 0.001; Fig. 4C and D). Therefore, TA and EDL muscles undergo a progressive loss in muscle strength in Sod1 D83G/D83G animals between 15 and 52 weeks of age, at a time when there is no progression in the death of LMN. In addition, as with SOD1 transgenic mouse models, the TA is affected earlier than EDL (28).
TA and EDL are fast twitch muscles that normally fatigue rapidly when repeatedly stimulated. Changes in the fatigue characteristics of fast twitch muscles are a typical feature of disease in  SOD1 transgenic mouse models (34) and so we undertook a fatigue test of EDL muscles in Sod1 D83G/D83G mice ( Fig. 4E and F). The force measured at the beginning of a 3 min period of stimulation was compared with that at the end of the test to produce a Fatigue Index (FI), a measure of the fatigability of the muscle. The results highlight the progressive deterioration of EDL in these mice, so that at 15 weeks, the FI of EDL in Sod1 D83G/D83G mice is no different from WT, but by 52 weeks of age, EDL FI is increased by 30% in Sod1 D83G/D83G mice compared with WT (P = 0.04) (Fig. 4E and F). Comparison between WT and Sod1 +/D83G littermate mice did not reveal any significant differences in TA or EDL muscle function at 15, 52 or 96 weeks of age (Fig. 4).  Fig. S4A). EDL muscles of 15-week-old Sod1 D83G/D83G mice, had a similar SDH staining pattern to WT littermates, but by 52 weeks, there was a marked increase in the proportion of intensely stained, likely type I muscle fibres (Supplementary Material, Fig. S4B). These results once again indicate that in mice expressing mutant SOD1, EDL muscles are affected later in the disease than other hindlimb muscles such as the TA.

Sod1 D83G/D83G mice have selective sensory deficits
Although ALS is predominantly a motor neuron disorder, a subset of ALS patients and SOD1 G93A transgenic mice develop deficits in the sensory system (35,36). Thus, we performed behavioural sensory phenotypic analyses with four sensory tests on female mice at 22 weeks of age (n ≥ 10 per group): (i) cold plate (noxious cold stimulus), (ii) Randall Sellito test of mechanosensation (withdrawal to a high-threshold ramp mechanical stimulus), (iii) Von Frey (low threshold mechanical stimulus), (iv) Hargreaves method (noxious heat stimulus) (Fig. 5A-D). Interestingly, Sod1 D83G/D83G mice have significant sensory deficits in response to both a low threshold mechanical stimulus (von Frey; P < 0.05) (Fig. 5C) and a noxious heat stimulus (Hargreaves method; P < 0.01) (Fig. 5D) but no deficit toward a noxious cold stimulus (cold plate) (Fig. 5A) or a high-threshold mechanical stimulus (Randall Sellito) (Fig. 5B). These results suggest that different primary afferent populations have selective vulnerabilities to the Sod1 D83G mutation, or they may be a consequence of central sensorimotor integration.
The reduction of mutant SOD1 D83G in the soluble fraction might be due to an unfolding and sequestration to the insoluble fraction. Therefore, we measured the level of SOD1 in the RIPAinsoluble fractions from spinal cords of WT, Sod1 +/D83G and Sod1 D83G/D83G littermates and found low levels of insoluble SOD1 in all genotypes, with a similar reduction in SOD1 D83G protein levels in Sod1 +/D83G and Sod1 D83G/D83G mice as that observed in the soluble fraction (Supplementary Material, Fig. S5A). Thus, SOD1 D83G protein does not accumulate in the insoluble fraction.
We were unable to identify misfolded SOD1 by immunoprecipitation using either the SEDI or USOD misfolded SOD1 antibodies from Sod1 +/D83G and Sod1 D83G/D83G tissue extracts or immunohistochemistry (Supplementary Material, Figs. S5B and S6A), or detect overt inclusion pathology using p62 or ubiquitin staining in Sod1 +/D83G and Sod1 D83G/D83G mice from spinal cord and brain (Supplementary Material, Figs. S5C and S6B). To examine whether Sod1 D83G mRNA is differentially regulated from WT Sod1, we performed quantitative PCR of Sod1 mRNA isolated from brains of 9-week-old WT, Sod1 +/D83G , Sod1 D83G/D83G littermates and did not observe a difference between genotypes (Supplementary Material, Fig. S7A). In addition, we directly compared the relative levels of Sod1 WT and D83G mRNA levels from the brains of 9-week-old Sod1 +/D83G mice using quantitative pyrosequencer analysis (see Materials and Methods) and did not see a difference between the relative expression levels of WT and mutant alleles ( Supplementary  Material, Fig. S7B).

Sod1 D83G/D83G have some phenotypes of Sod1 null mice
Since Sod1 D83G/D83G mice are dismutase inactive (Fig. 6A and B), they are likely to phenocopy at least some aspects of Sod1 null mice (Sod1 −/− ). Sod1 −/− mice develop a progressive peripheral motor neuropathy, in which motor axons retract from neuromuscular junctions resulting in muscle denervation (20,22). However, unlike Sod1 D83G/D83G mice, there is no reported motor neuron loss in Sod1 −/− mice at any age (20,22,23). To verify that LMN loss in Sod1 D83G/D83G mice is not due to a loss of SOD1 activity and to confirm that motor neuron survival is unaffected in Sod1 −/− mice, we compared LMN survival in female Sod1 D83G/D83G and Sod1 −/− mice at 15 weeks of age; both genotypes were backcrossed to a C57BL/ 6J background. In contrast to Sod1 D83G/D83G mice and in agreement with all previous reports, Sod1 −/− mice do not develop motor neuron cell body loss and the survival of motor neurons in the sciatic pool is similar in WT (484 ± 9 MN, n = 6) and Sod1 −/− (489 ± 6 MN, n = 5; P = 0.47) mice. Therefore, unlike Sod1 −/− mice, Sod1 D83G/D83G mice have an additional toxic gain of function that causes the degeneration of motor neuron cell bodies.
Lifespan is reduced in transgenic SOD1 mice that overexpress mutant SOD1 and in Sod1 −/− mice. We set the humane endpoint of life in Sod1 D83G/D83G mice as either the time at which a loss of 20% of maximum bodyweight occurs or the onset of piloerection (an involuntary erection of fur that is indicative of loss of health). The majority of WT, Sod1 +/D83G and Sod1 D83G/D83G littermates were culled due to a loss of 20% bodyweight (∼70% of Sod1 D83G/D83G mice), while the rest (∼30%) were culled for the presence of piloerection. The survival of Sod1 D83G/D83G mice is shortened compared with Sod1 +/D83G and WT mice ( Table 2 and Fig. 7A). We also found that male Sod1 D83G/D83G mice had a significantly reduced lifespan compared with female Sod1 D83G/D83G mice (495 ± 22 days versus 588 ± 24 days; P = 0.024).
At autopsy, Sod1 D83G/D83G mice had significantly more liver tumours than WT and Sod1 +/D83G littermates ( Table 2; P < 0.001). Interestingly, it has previously been shown that Sod1 −/− mice also develop liver tumours, most likely as a result of an increase in oxidative damage (38). Given that Sod1 D83G/D83G mice are dismutase inactive for SOD1 ( Fig. 6A and B), it is likely that the loss of SOD1 activity in these mice is the cause of liver tumours.
The identification of liver tumours in Sod1 D83G/D83G mice led us to examine end-stage male Sod1 D83G/D83G mouse livers. Livers from WT littermates had no pathology other than variable perivenular lymphocytic infiltration (Fig. 7B). In Sod1 D83G/D83G mice, tumours occurred in the presence of otherwise pathologically normal and abnormal livers. In Figure 7C, there is a single, well-circumscribed and well-differentiated hepatocellular carcinoma occurring in a liver that is otherwise morphologically  normal. In contrast, another Sod1 D83G/D83G liver demonstrated multiple ill-defined hepatocellular carcinomas and a well-circumscribed nodular lesion, composed of sheets of pleomorphic tumour cells with vesicular nuclei and basophilic cytoplasm compatible with a poorly differentiated hepatocellular carcinoma or cholangiocarcinoma (Fig. 7D). This liver also showed diffuse nuclear anisocytosis, moderate parenchymal inflammation and patchy fatty change (Fig. 7E). Hence, pathological analysis reveals that the hepatocellular carcinomas in Sod1 D83G/D83G mice occurred in the presence of both morphologically normal and abnormal livers.
Mitochondrial membrane potential defects in Sod1 +/D83G embryonic motor neurons Mitochondrial dysfunction has been identified as one of the earliest defects in motor neurons of transgenic SOD1 mice and it has been suggested that mitochondrial dysfunction may play a pivotal role in motor neuron degeneration in ALS (39). We therefore analysed the mitochondrial membrane potential (Δψ m ) of embryonic motor neurons. Δψ m is an indicator of mitochondrial energetic state and can be measured using tetramethylrhodamine methylester (TMRM) (40). Those from Sod1 D83G/D83G (as previously reported for Sod1 −/− ) did not survive for >72 h in culture and therefore Δψ m could only be examined in motor neurons from WT and Sod1 +/D83G littermates. We have previously shown that Δψ m in embryonic motor neurons from SOD1 G93A transgenic mice is disrupted (41). In embryonic motor neurons of Sod1 +/D83G mice, Δψ m was significantly elevated compared with WT littermates ( Fig. 8; P < 0.001).
Since the Sod1 +/D83G mice have an ∼45% reduction in SOD1 dismutase activity, it was unclear whether the elevation in Δψ m in Sod1 +/D83G motor neurons was due to a loss of SOD1 dismutase activity, or a gain of function of SOD1 D83G protein. We therefore analysed the Δψ m of embryonic motor neurons derived from WT and Sod1 +/− littermates and found that Sod1 +/− motor neurons also have elevated Δψ m (Fig. 8). This indicates that the mitochondrial defects present in Sod1 +/D83G motor neurons are likely caused by a partial loss of SOD1 function.

Discussion
We have identified and characterized the first mouse model carrying the equivalent of a human ALS pathogenic mutation in the endogenous mouse Sod1 gene, and this mutation is identical to a human fALS mutation (24). Since the mutation is within the endogenous Sod1 gene, mutant SOD1 is not overexpressed. Here we show that homozygous Sod1 D83G/D83G mice develop upper and LMN degeneration, unlike Sod1 null animals. This is therefore likely due to a toxic gain of function of mutant SOD1-presumably modelling that which causes motor neuron death in human ALS cases.
We found ∼20% loss of UMN by 29 weeks of age in homozygotes, but did not look at later time points and so do not know if this phenotype progresses. Given the importance of mutant SOD1 protein dose in accelerating the ALS-like phenotype (16), it is possible heterozygotes may develop UMN loss later in life. Alternatively, a copy of WT SOD1 might be protective; this remains to be determined.
We showed 23% loss of LMN in the lumbar spinal cord of homozygous Sod1 D83G/D83G mutants by 15 weeks of age that remained stable at 52 weeks of age. We speculate whether further ageing, or a 'second hit' might be required to cause a more dramatic loss of LMN in this mouse mutant; however, this remains to be determined. We note that a similar loss of MN in humans is unlikely to cause sufficient loss of motor function to prompt a visit to a physician. Sod1 +/D83G heterozygous mutants displayed  no LMN loss up to 52 weeks of age, either because the dose of mutant SOD1 was insufficient or because WT SOD1 might have a protective effect.
Our analyses of the Sod1 D83G mouse mutant show that mutation in the endogenous mouse Sod1 gene models critical features of human ALS (U and LMN cell death), albeit with a different profile than found in mice expressing human mutant SOD1 transgene arrays. This corroborates earlier findings in which a mutant genomic mouse Sod1 transgenic was used to model human SOD1 G85R-fALS-'Sod1 G86R' mice with 'high transgene expression' had both U and LMN degeneration by 3-4 months of age (14). Therefore, the endogenous mouse SOD1 D83G mutation is informative for determining why some U and LMN die in ALS but removes any possible confounding effects of overexpression observed in SOD1 transgenic mice. We note that we and others have not observed MN cell body degeneration in Sod1 −/− mice at any stage, and therefore MN loss in Sod1 D83G/D83G mice likely arises from a toxic gain of function of the mutant SOD1 protein, as in ALS (20)(21)(22).
The significant reduction in SOD1 D83G protein is not due to allele-specific differences at the transcriptional level, but the dose-dependent decreases in SOD1 protein in heterozygous and homozygous D83G mutant mice may result from instability of mutant SOD1 and its subsequent degradation (42). It has been previously reported that mutant SOD1 has decreased half-life compared with WT SOD1 (42), and potentially the inability of D83G SOD1 to coordinate zinc may contribute to its instability (37,43,44).
Given that SOD1 expression levels in different transgenic mouse models show dose-dependent toxicity (16,18), the low levels of SOD1 D83G protein may only be sufficient to cause moderate motor neuron degeneration. The low SOD1 mutant protein level (∼10% of WT in homozygotes) may be sufficient to elicit a degree of motor neuron degeneration but not the levels of loss seen in ALS, and may at least partially explain why loss of motor neurons within the sciatic motor pool appears not to progress between 15 and 52 weeks of age.
With respect to loss-of-function effects, which are of interest because of the loss of dismutase activity in the majority of human pathogenic SOD1-fALS mutations (average dismutase activity in SOD1-fALS is ∼58% of normal (4)), we found mitochondrial membrane potentials (Δψ m ) of both Sod1 +/D83G and Sod1 +/− embryonic MNs are hyperpolarized. Thus, SOD1 loss of dismutase function to just 56% of WT (Sod1 +/D83G level) contributes to mitochondrial abnormalities. Distal axonopathy in Sod1 −/− mice can be rescued by expressing mitochondrial-targeted WT SOD1 (20); the relevance of this to ALS remains to be determined. Sod1 D83G/D83G mice develop a severe peripheral neuropathy similar to that arising from the loss of function in Sod1 −/− mice. This peripheral neuropathy is likely driving the phenotypic deterioration from 15 weeks of age in Sod1 D83G/D83G mice (Supplementary Material, video S1), as no further LMN loss occurs after that. As dismutase activity is only 1% of WT animals in the SOD1 D83G homozygotes, this peripheral neuropathy may be due to a loss of dismutase activity, potentially leading to increased vulnerability of peripheral motor axons to oxidative stress (45). It is also possible that the neuropathy develops because of a loss-of-unknown SOD1 function.
In at least one patient with diagnosed ALS who unexpectedly died from other causes, NMJ degeneration was found to precede motor neuron cell body death (45). Similarly, in transgenic human mutant SOD1 mouse models (with normal dismutase levels), NMJ degeneration leading to synaptic dysfunction precedes motor neuron cell body death and behavioural motor deficits (45,46). These findings led to the concept of ALS as a "dying back" disorder in which muscle denervation precedes the death of the motor neuron cell body, and suggests that SOD1 gain of function can also lead to NMJ degeneration-and potentially also contributing to the distal neuropathy seen in Sod1 D83G/D83G mice.
Here our data, and that from other SOD1 mouse models, suggests that the axonal and neuronal cell body degeneration may be separate events that could be modulated by different sets of genes and/or environmental factors. In further support of this hypothesis, ablation of Bax in the human G93A SOD1 transgenic mouse model leads to complete dissociation between motor neuron soma death and motor dysfunction via distal denervation (47). While cell bodies are protected in Bax deficient G93A SOD1 transgenic mice, the degree of NMJ denervation and overall survival is not altered compared with G93A transgenic mice alone. Thus, Sod1 D83G/D83G mice provide a novel model to test therapeutics aimed at preserving NMJ as well as ameliorating the early stages of motor neuron cell body degeneration.
Overall, these results suggest that Sod1 D83G/D83G mice not only model the early stages of human ALS but are also able to separate the effects of central neuronal death from the peripheral distal neuropathy in a system in which the mutant gene is expressed at endogenous levels.
The Sod1 D83G model is the first of its kind carrying a known pathogenic point mutation in the mouse endogenous Sod1 gene, which is identical to a human fALS mutation. Homozygous Sod1 D83G/D83G mice develop a degree of UMN and LMN degeneration, as well as progressive motor dysfunction due to a distal neuropathy, and hence model key aspects of the early stages of ALS. The Sod1 D83G model therefore provides a unique mammalian system in which to assess the contribution of both central neuronal loss and peripheral axonal dysfunction, which will further our understanding of ALS.

Mice
Ethics statement: all experiments were performed under licence from the UK Home Office. WT, Sod1 +/D83G and Sod1 D83G/D83G mice were initially on a C57BL/6J-C3H background and backcrossed at least four generations to C57BL/6J. For all experiments, littermates for all genotypes were used, produced by intercrossing Sod1 +/D83G mice. Experiments were performed blind to genotype and lifespan was defined as a loss of 20% of maximum body weight or the presence of piloerection. Animals were assessed daily and weighed at least biweekly.

Identification of Sod1 D83G mutation
The Harwell ENU-DNA archive (http://www.har.mrc.ac.uk/ services/dna_archive/) was screened with high-resolution melting analysis using the LightScanner platform (Idaho Technology Inc., Salt Lake City, Utah, USA) (48). Coding sequence from exon 4 of Sod1 including flanking (∼70 bp) splice sites was tested in DNA derived from ∼10 000 F1 ENU mutagenized animals. Sod1 exon 4 was amplified with the following primers: SOD1_4F CATCCACT CATACGTATTTGAC, SOD1_4R ACCATAAAGTCATGGGAAGG using the lightscanner mastermix plus LC green (Idaho Technology Inc.) and possible mutations analysed with Sanger sequencing (GATC, Germany). We identified the chromosome 16 A90,224, 530G mutation, corresponding to the D83G amino acid change in SOD1. The mouse line was rederived using sperm from the (C57BL/6J × C3H) F1 founder by in vitro fertilization. Genomic DNA from the F1 founder carrying the Sod1 D83G mutation was deep sequenced (see Supplementary Material, Materials and Methods for details) to identify any other possible ENU-induced mutations linked to Sod1 on mouse chromosome 16. The closest ENU-induced coding mutation was ∼46 Mb proximal to the Sod1 gene and did not segregate with lifespan or weight abnormalities in Sod1 D83G/D83G mice.

Behavioural analysis
See Supplementary Material, Materials and Methods for a description of behavioural and Echo MRI analysis.
Physiological assessment of hindlimb muscle force, motor units and fatigue index Muscle force, motor unit number and muscle fatigue characteristics for female WT, Sod1 +/D83G and Sod1 D83G/D83G littermates were examined at 15 and 52 weeks of age, and female WT and Sod1 +/D83G littermates also at 88 weeks of age, as described previously (49). Cohort sizes per time point were at 15 weeks: 5 WT, 5 Sod1 +/D83G and 5 Sod1 D83G/D83G animals; 52-weeks: 5 WT, 7 Sod1 +/D83G and 6 Sod1 D83G/D83G animals; 88 weeks: 5 WT, 7 Sod1 +/D83G . Briefly, mice were anaesthetized (4.5% chlorohydrate), the distal tendons of the TA and EDL muscles in both hindlimbs cut and attached to isometric force transducers (Dynamometer UFI Devices, UK), and the sciatic nerves exposed and sectioned. Isometric contractions were elicited by stimulating the nerve to the TA or EDL muscles using squarewave pulses of 0.02 ms duration at supra-maximal intensity. The number of motor units innervating the EDL muscle was determined by stimulating the motor nerve with stimuli of increasing intensity, resulting in incremental increases in twitch tension due to successive recruitment of motor axons with increasing threshold. The number of increments in twitch force was counted giving an estimate of the number of functional motor units present in muscle. The fatigue characteristics of EDL were examined by undertaking a fatigue test in which the EDL muscle was repeatedly stimulated at 40 Hz for 250 ms every second for 180 s. The resulting contractions were recorded on a pen recorder (Lectromed Multitrace 2) producing a fatigue trace from which an FI, a measure of muscle fatigability, can be determined by expressing the force at the end of the test as a ratio of the force at the start: FI = F t180 /F t0 . A muscle that is completely fatigue resistant has an FI approaching 1.0.

Morphological assessment of muscle and neuronal tissue
For assessment of survival of motor neurons in the sciatic pool of the lumbar spinal cord (16,18), animals were perfused with saline followed with 4% paraformaldehyde (PFA) fixation. Spinal cords were removed, postfixed in 4% PFA and cryopreserved in 30% sucrose. Transverse sections (20 μm) of the lower L2-lower L6 lumbar region of fixed spinal cords cut on a cryostat serially onto glass slides and stained for Nissl (gallocyanin) (50). Nissl-stained motor neurons in the sciatic pool in every third section of the L3-L6 lumbar region, over 40 sections in total, were counted. Only large, polygonal neurons with a distinguishable nucleus and nucleolus and clearly identifiable Nissl structure were included in the counts. This method avoids the possibility of counting the same motor neuron in consecutive sections. At least five mice were analysed from each experimental group per time point.
Muscle histochemistry was performed on TA and EDL muscle sections by staining for the oxidative enzyme, as previously described (53). Briefly, serial cross-sections of fresh frozen TA and EDL were cut on a cryostat at 12 μm, collected on glass slides and stained for SDH activity to determine the oxidative capacity of the muscle fibres.

Neuromuscular junction quantification
EDL muscles were dissected from mice of all three genotypes at 15 and 52 weeks of age. The numbers of whole EDL muscles analysed at 15 weeks of age: 7 WT, 4 Sod1 +/D83G and 8 Sod1 D83G/D83G . At 52 weeks: 11 WT, 8 Sod1 +/D83G and 8 Sod1 D83G/D83G . Mice were perfused trans-cardial with saline solution and 4% PFA. Muscles were postfixed in 4% PFA and cryopreserved in sucrose. Whole EDL muscles were then embedded in OCT media (Tissue-Tek), frozen on dry-ice and stored at −80°C. A Bright™ cryostat was used to cut 20 µm longitudinal section of frozen EDL muscles which were subsequently collected onto poly-lysine-coated slides. Primary antibodies used: mouse monoclonal anti-synaptic vesicle (Developmental Studies Hybridoma Bank) Mouse monoclonal anti-neurofilament (165 kDa) (Developmental Studies Hybridoma Bank). M.O.M. biotinylated secondary antibody (Vector Labs) was used following manufacturer instructions. α-Bungarotxin-rhodamine was used to label postsynapsis and sections mounted. Images were acquired on a confocal microscope (Leica DFC 420C) and NMJ were manually counted. qPCR, western blotting, dismutase activity and immunoprecipitations qPCR, western blotting, dismutase activity and immunoprecipitation protocols are available in Supplementary Material, Materials and Methods.
Embryonic motor neuron culture and mitochondrial membrane potential Motor neuron culture and measurement of mitochondrial membrane potential are described in Supplementary Material, Materials and Methods.

Statistical analysis
An ANOVA test was used to compare between WT, Sod1 +/D83G and Sod1 D83G/D83G genotypes per time point followed by Bonferroni's multiple comparisons testing correction for weight, dismutase activity, protein quantification, Echo MRI analysis, grip strength, rotarod, startle response and sensory analysis. Wheel running behaviour was analysed using a repeated measures ANOVA. Cox survival analysis was used to analyse SHIRPA and survival data. The Mann-Whitney test was used to compare between genotypes per time point or Sod1 D83G/D83G cohorts across time points for TA and EDL maximum muscle force, surviving motor units, FI and motor neuron survival. Two-tailed tests were used in all instances and significance level was set at P < 0.05.

Supplementary Material
Supplementary Material is available at HMG online.