Using the set of rabbit antibodies raised against peptides in the SOD1 sequence, we found small round inclusions in spinal cord motoneurons of all the 29 sporadic and 8 familial ALS patients lacking mutations in the SOD1 gene and in the 2 SBMA patients (Figures 1A, B, D, E and N) and (Supporting Information Figures S1B, E, H and K, and S2A]. These inclusions were somal and in many cells they were particularly abundant in the axon hillock (Figures 1B and S2A). They were relatively homogenous in size and measured 0.5–3 µm. The proportion of motoneurons exhibiting such inclusions varied from only a few to most of the motoneurons, and they were seen in all spinal cord cross-sections investigated (Table 1). Small granular SOD1-immunoreactive inclusions were also seen in neurons of Clarke's nucleus in 22 of 23 SALS and in 3 of the 5 FALS patients in which the nucleus of Clarke was available for study, and in both SBMA patients. In hematoxylin/eosin-stained sections, some Lewy body-like hyaline inclusions were seen in motoneurons, and in a few patients a small proportion of such structures were immunoreactive for SOD1. No correlation between the disease duration and amount of inclusions could be seen in those patients from whom reliable data on duration were available. Other pathomorphological observations in the patients are reported as in Supporting Information (Results).

In most of the control patients, the motoneurons lacked granular SOD1 staining (Figure 1J and K). In 3 out of the 28 control patients with other neurodegenerative diseases (1 with Huntington's disease [HD]; 1 with Parkinson's disease [PD]; 1 with Alzheimer's disease [AD]), and in 2 of the 19 control patients without neurological disease, a few granular SOD1-immunoreactive inclusions were seen in some motoneurons (Figure 1C and L). The abundance of inclusions was, however, smaller than in any of the ALS patients (Table 1). Also, two of these control patients showed a few small granular SOD1-immunoreactive inclusions in the neurons of Clarke's nucleus. In control patients, no granular SOD1 inclusions were seen in cortex or hippocampus (AD patients), striatum (HD patient) or mesencephalon (PD patients).

To gain insight into the subcellular localization of the granular inclusions, we used double-labeling immunohistochemistry with the 57–72Ra-ab anti-SOD1 antibody and antibodies directed against marker proteins for lysosomes, mitochondria and endoplasmatic reticulum. Using confocal laser microscopy, we could show a partial colocalization of SOD1 inclusions with the lysosomal marker cathepsin D in all patients studied (Figure 1O). Colocalization was found in about 25% of all lysosomes, while the proportion of all granular inclusions colocalizing with the lysosomal marker was less than 25%. We could not detect any colocalization of the SOD1 inclusions and the endoplasmic reticulum markers KDEL and GRP-78, the mitochondrial markers mitochondrial Hsp70 and mitochondrial marker, clone SPM198 or in inclusions containing TAR DNA binding protein-43 (TDP-43), the autophagosomes marker MAP1LC3A or ubiquitin (Figures. S1 and S2).

In most previous studies, antibodies raised against whole SOD1 have been used [22], [23], [24], [25], [26], [27], [28]. We compared a set of 4 such antibodies with our rabbit anti-SOD1 peptide antibodies for reactivity against denatured and native SOD1 in an ELISA (Figures 2 and S3). As seen in Figure 2A, the 4–20Ra-ab, 57–72Ra-ab and 131–153Ra-ab antibodies reacted strongly with denatured SOD1, but hardly at all with native SOD1. Similar results were found for the other rabbit anti-SOD1 peptide antibodies (Figure S3). Among the antibodies raised against whole SOD1, our in-house rabbit antibody (Rabbit-1) reacted equally to denatured and native SOD1 in the ELISA. The sheep antibodies from the Binding Site and Calbiochem were a little and clearly more, respectively, active versus denatured than against native SOD1 (Figure 2B). The monoclonal mouse antibody was almost completely specific for native SOD1. Most of the antibodies were also immobilized on Sepharose and tested for their affinities for native and denatured SOD1 in solution (Figures S4 and S5). The outcomes were in principle identical; the antipeptide antibodies bound denatured SOD1 but lacked affinity for native, whereas antibodies raised against whole SOD1 bound both native and denatured SOD1.

For comparison purposes we tested dilution series of the set of 4 antibodies raised against whole human SOD1 on sections from 4 patients with abundant small granular inclusions. Using the antibody with the highest relative reactivity for denatured SOD1 (the Calbiochem antibody) (Figure 2B), granular inclusions could sometimes be discerned against the background staining of cytosolic SOD1 (Figure 1F). Thus, the failure in previous studies to detect the granular SOD1 inclusions might be explained the masking effect of staining of the abundant native SOD1 in the motoneurons.

To demonstrate the specificity for SOD1 in the histopathological studies, the antibodies were preincubated with denatured SOD1 (not shown) or the immunizing peptides in increasing concentrations. This resulted in gradual and finally full disappearance of immunoreactivity (Figure S2D, E and F). An antibody raised against keyhole limpet hemocyanin, the carrier protein to which the peptides were coupled for immunization, gave no specific staining. The antibodies stained inclusions of human SOD1 in transgenic mice (Figure S2G), but did not stain tissues from normal control or SOD1 knockout mice (Figure S2H and I). The rabbit peptide antibodies showed very high specificities for SOD1 as analyzed by western immunoblots of human spinal cord extracts (Figure 3). In western blots of extracts of murine spinal cords, there was strong staining for human SOD1 in wild-type human SOD1 transgenic mice, variable weak staining of murine SOD1 (cross-reaction) and no discernable staining in SOD1 knockout mice (Figure S6).

To further validate the finding of SOD1-staining inclusions in ALS, antibodies were raised in chicken against three of the SOD1 peptides. The antibodies showed high specificities for SOD1 in western blots (Figure S7). Immobilized on Sepharose they efficiently bound denatured but not native SOD1 (Figure S4). Histopathologically, they all stained granular inclusions in motoneurons of ALS cases in a fashion indistinguishable from that of the rabbit antibodies, and there were no immunopositive inclusions in controls (Figure S7).

Inclusions containing SOD1 are considered hallmarks of ALS caused by mutations in the enzyme [22], [28], [29]. Using the present anti-peptide antibodies, SOD1-immunoreactive inclusions have previously been demonstrated in patients carrying the G72C and A4V mutations [30], and in a case carrying the G127X C-terminal truncation mutation [9]. With the latter, the 4–20Ra-ab anti-SOD1 antibody allowed visualization of the inclusions, but not the 131–153Ra-ab anti-SOD1 antibody, which is directed against a sequence absent from the mutant protein. In most patients carrying SOD1 mutations, the inclusions tend to be larger and both skein- and Lewy body-like [9], [30], [31] (e.g. Figure 1H), but small granular SOD1-immunoreactive inclusions were also found in the patient with the G127X mutation (Figure 1I). Importantly, in patients homozygous for the wild type-like D90A mutant SOD1, the majority of the SOD1-immunoreactive inclusions were of the small granular type (Figure 1G).

Homogenates of tissues from the temporal lobe, the precentral gyrus and from the cervical and lumbar spinal cord ventral horns of 5 SALS patients, 4 FALS patients and 5 control patients were examined by western blot using anti-SOD1 peptide antibodies. The denatured SOD1 monomer was seen as a single band of the appropriate molecular weight in all samples (Figure 3). There was no evidence of smearing or accumulation of SOD1 protein in the loading wells, nor were there any other alterations detected with the three antibodies used that could correlate with disease.

The activites of SOD isoenzymes were analysed in 15 different parts of the CNS from controls, SALS and FALS patients (Tables S1 and S2). SOD1 showed throughout the highest activity, and there were no significant differences in this enzyme between the groups. The SOD1 activities in the areas mainly afflicted by ALS, the spinal cord ventral horns and the precentral gyrus, were intermediate among gray matter areas. Using the specific activity of human SOD1 [32], the concentration of the enzyme was calculated to be 2 µM. Ventral horns were found to contain about 60 mg protein per g wet weight. SOD1 thus accounts for about 0.1% of the tissue protein.

The anti-SOD1 antibodies used were polyclonal rabbit antibodies raised against keyhole limpet hemocyanin-coupled peptides corresponding to amino acids 4–20, 24–39, 43–57, 57–72, 80–96, 100–115 and 131–153 in the human SOD1 sequence or corresponding chicken antibodies raised against peptides corresponding to amino acids 24–37, 57–72 and 131–153. They were purified on a Sulfolink coupling gel with the corresponding peptide coupled [12]. An antibody against a neopeptide in G127X mutant SOD1 was similarily prepared [9]. The other antibodies used are presented in Table S3.

ELISA plates with 96 wells (Maxisorp, Nunc A/S, Roskilde, Denmark) were coated overnight at 4°C with 100 µl of 2 µg/ml native human SOD1 (Sigma, Schelldorf, Germany) or human SOD1 that had been denatured with 4 M guanidinium chloride and 5 mM of the chelator diethylenetriaminepentaacetic acid (DTPA). This resulted in a complete denaturation of the SOD1 (Figure S7). After washing and blocking, primary antibodies were added and incubated at 23°C for 90 min. Finally, peroxidase-labeled secondary antibodies were added followed by development in substrate (1,2-phenylenediamine and H₂O₂) for 6 min, and the absorbance was read at 490 nm. Complete reactions in wells without SOD1 coating were used as blanks and the ELISA readings were subtracted. For the peptide antibodies, the subtractions amounted to an A₄₉₀ of 0.015–0.020. Secondary antibodies alone gave negligible reactions in wells coated with SOD1.

Tissues were collected at autopsy from patients prospectively enrolled at the Department of Neurology, Umeå University Hospital. All patients fulfilled the El Escorial criteria for ALS [43]. This group consisted of 16 patients with SALS (mean age 70±9 [49–83] years) and 6 patients with FALS (mean age 63±4 [55–68] years). All 22 ALS patients were genotyped for SOD1 mutations as described previously [4], and none were found.

Furthermore, a retrospective search for clinicopathologically confirmed cases of ALS from the hospital records yielded 2 FALS patients (aged 33 and 49 years) and 13 SALS patients (mean age 70±10 [55–88] years), from whom material was subjected to histological study. No statistically significant differences between prospectively and retrospectively collected patients with the same type of ALS were found, and the materials were combined in the final analysis, giving one group of 8 FALS patients (mean age 58±12 [33–68]; years) and one group of 29 SALS patients (mean age 70±9 [49–88] years). From the hospital records two cases with SBMA (aged 52 and 76 years, respectively) were found. The diagnosis was verified by the presence of the trinucleotide repeats in the androgen receptor. For comparison, spinal cord tissue from two patients with SOD1 gene mutations (D90A and G127insTGGG [G127X]), respectively, was also subjected to histological study.

Control tissue from 28 control patients with other neurodegenerative diseases (mean age 74±19 [2–92]; 15 with Alzheimer's disease; 7 with Parkinson's disease; 3 with multiple sclerosis; 1 with frontotemporal lobar degeneration; 1 with tuberous sclerosis; 1 with Huntington's disease) and from 19 control patients without neurological disease (mean age 69±17 [37–91] years; the patients died from heart conditions or pneumonia) were used. Of these, 7 patients with neurodegenerative diseases and 3 without neurodegenerative diseases were prospectively enrolled and the rest collected from archival material.

The study adhered to the tenets of the Helsinki Declaration. Information about the study was given orally and in written format to next of kin, and in most cases also to the patient. Informed consent was obtained from next of kin from all patients prospectively enrolled in the study, the oral consent was then noted in the hospital files. The procedure was approved by the Ethical Committee of Umeå University which also approved the use of retrospective archival material from the hospital laboratory of clinical pathology.

For the prospectively collected material, tissue samples were taken from the cervical, thoracic and lumbar regions, and for the retrospective material the tissues available mostly consisted only of blocks from the cervical and/or thoracic parts of the spinal cord. Tissue samples were also taken from hippocampus, ventral cingulate gyrus and frontal cortex, middle and superior temporal gyrus, striatum and mesencephalon. Tissues for histopathological studies were immersion-fixed in 4% paraformaldehyde in 0.1 M Na phosphate, pH 7.4. The immunohistochemical and fluorescent microscopy procedures are described in detail in Supporting Information (Material and Methods).

Tissues were homogenized as described in Supporting Information (Material and Methods) and were without prior centrifugation diluted 1:1 in SDS-PAGE loading buffer, heated for 5 min at 95°C and separated on 12% Ready Gels (BioRad, Hercules, CA, USA). Probing and detection were done as previously described [42], [44].

Statistical analysis was done using the STATISTICA data analysis software system (version 7.1; StatSoft Inc., Tulsa, OK, USA) or SPSS (version 15; SPSS Inc., Chicago, IL, USA). The nonparametric Kruskall-Wallis statistics (when comparing three groups) and Mann-Whitney statistics (when comparing two groups) were used for the morphological comparisons. In estimation of staining, the numbers given are median and range. When comparing SOD activities, analysis of variance (ANOVA) was used.