Amyotrophic lateral sclerosis (ALS) is a common, fatal motor neuron disorder with no effective treatment. Approximately 10% of cases are familial ALS (FALS), and the most common genetic abnormality is superoxide dismutase-1 (SOD1) mutations. Most ALS research in the past decade has focused on the neurotoxicity of mutant SOD1, and this knowledge has directed therapeutic strategies. We recently identified TDP-43 as the major pathological protein in sporadic ALS. In this study, we investigated TDP-43 in a larger series of ALS cases (n = 111), including familial cases with and without SOD1 mutations.
Ubiquitin and TDP-43 immunohistochemistry was performed on postmortem tissue from sporadic ALS (n = 59), ALS with SOD1 mutations (n = 15), SOD-1–negative FALS (n = 11), and ALS with dementia (n = 26). Biochemical analysis was performed on representative cases from each group.
All cases of sporadic ALS, ALS with dementia, and SOD1-negative FALS had neuronal and glial inclusions that were immunoreactive for both ubiquitin and TDP-43. Cases with SOD1 mutations had ubiquitin-positive neuronal inclusions; however, no cases were immunoreactive for TDP-43. Biochemical analysis of postmortem tissue from sporadic ALS and SOD1-negative FALS demonstrated pathological forms of TDP-43 that were absent in cases with SOD1 mutations.
These findings implicate pathological TDP-43 in the pathogenesis of sporadic ALS. In contrast, the absence of pathological TDP-43 in cases with SOD1 mutations implies that motor neuron degeneration in these cases may result from a different mechanism, and that cases with SOD1 mutations may not be the familial counterpart of sporadic ALS. Ann Neurol 2007;61:427–434
Amyotrophic lateral sclerosis (ALS) is a common neurodegenerative disorder in which the relentless destruction of motor neurons causes progressive weakness, typically leading to death within a few years.1 The cause is unknown, and no effective treatment currently exists. Although most cases are sporadic (SALS), 10% of cases have a family history (FALS).2 In 1993, mutations in the gene encoding the antioxidant enzyme Cu/Zn superoxide dismutase-1 (SOD1) were identified in some FALS families.3 Since then, more than 120 different SOD1 mutations have been reported, accounting for approximately 20% of FALS cases.1, 2 A number of other genetic loci and genes subsequently have been identified; however, SOD1 remains the most common and the only one to cause autosomal dominant, classic ALS.1, 2 Although it is widely believed that SOD1 mutations cause ALS because of a toxic gain of function, this remains unproved.1, 4, 5 Similarities in the clinical course and neuropathology of SALS and FALS have lead to the hypothesis that studying the familial forms will elucidate the pathophysiological mechanism(s) underlying SALS.1, 2 As a result, most ALS research in the past decade has focused on the biological basis of mutant SOD1 neurotoxicity, and this knowledge has directed therapeutic strategies.5–8 Moreover, transgenic mice, overexpressing human mutant SOD1, are widely accepted as the most informative animal models for basic and translational ALS research.1, 5, 7–9 However, the appropriateness of this approach is dependent on the assumption that the disease mechanism(s) in SALS is similar to that caused by SOD1 mutations.
The neuropathology of ALS is characterized by the abnormal accumulation of insoluble proteins in the cytoplasm of degenerating motor neurons.10, 11 Until recently, little was known about the specific biochemical composition of these neuronal cytoplasmic inclusions (NCIs), except that the abnormal protein was ubiquitinated. These ubiquitin-immunoreactive (ub-ir) NCIs are most common in lower motor neurons and most often appear as either filamentous skeins or compact, round bodies.10, 11 A number of studies have confirmed these NCIs to be a highly sensitive and specific marker for ALS.11 Recently, we identified the TAR DNA-binding protein, TDP-43, as a major component of the NCI in SALS, as well as in the most common pathological subtype of frontotemporal dementia (frontotemporal lobar dementia with ubiquitinated inclusions [FTLD-U]).12 These findings have subsequently been confirmed.13–15 Evidence that TDP-43 is the major pathological protein in these conditions includes: (1) colocalization of TDP-43 and ubiquitin in the pathological inclusions; (2) specificity of TDP-43 immunoreactivity for the NCI in these conditions; (3) demonstration of abnormal Mr fragments of TDP-43 in postmortem tissue; (4) demonstration that the pathological forms of TDP-43 are ubiquitinated and hyperphosphorylated; and (5) loss of the normal nuclear TDP-43 staining in cells harboring NCI, suggesting loss of normal TDP-43 function.12, 13 These findings, combined with recent studies showing both clinical16, 17 and pathological18 overlap between ALS and frontotemporal dementia, support the view that SALS and FTLD-U represent a spectrum of neurodegenerative disease linked mechanistically to pathological TDP-43.
In this study, we extend our investigation of pathological TDP-43 to a much larger series of ALS (n = 111) that includes SALS, FALS with and without SOD1 mutations, and one of the most common clinical variants, ALS with dementia (ALS-d) (Table). Our findings demonstrate that pathological TDP-43 is a consistent feature of SALS, ALS-d, and SOD1-negative FALS, but most remarkably, TDP-43 pathology is not present in cases with SOD1 mutations, suggesting that mutant SOD1 may cause ALS through disease mechanisms that are distinct from those that underlie SALS.
Table . Immunohistochemistry in Amyotrophic Lateral Sclerosis with and without SOD1 Mutations
Patients and Methods
The cases used in this study were obtained from brain banks affiliated with the University of British Columbia, University of Pennsylvania, Washington University in St. Louis, University of Munich, Northwestern University, Sheffield University, and Virginia Mason Medical Center, Seattle. Consent for autopsy was obtained from the legal representative in accordance with local institutional review boards. Cases included patients with SALS (n = 59), FALS with confirmed SOD1 mutations (n = 15), FALS with SOD1 mutations excluded (n = 11), and ALS-d (n = 26; 15 with SALS and 11 with FALS-d or dementia) (see the Table). In cases with dementia, previous workup had excluded other causes (such as Alzheimer's, Lewy body, and Pick's diseases) and had demonstrated the ub-ir pathology in extramotor neocortex and hippocampus (FTLD-U) that is now accepted as being characteristic of ALS-d, as well as a subset of cases of frontotemporal dementia without motor features.18 In most cases of SALS and SALS-d, the SOD1 gene was not analyzed. SOD1 mutations had been excluded in 4 of 11 cases of FALS-d. It is unlikely that any of the remaining FALS-d cases had SOD1 mutations because dementia has not been previously reported in patients with SOD1 mutations. In cases of SOD1-negative FALS, the specific genetic defect was unknown. Cases with SOD1 mutations were obtained from the University of British Columbia, University of Munich, Northwestern University, Sheffield University, and Virginia Mason Medical Center. Sequencing of SOD1 had previously been performed at each of these centers using methods described elsewhere.3, 19–21 The 15 cases with SOD1 mutations were unrelated and included 7 different mutations (see the Table). No cases of FALS were included in which the SOD1 gene status was unknown.
Immunohistochemistry was performed at each of the participating centers, using previously published methods.12, 15, 18 In brief, 5μm-thick sections of formalin-fixed, paraffin-embedded tissue, from spinal cord or medulla, were obtained from each case. Sections of frontal neocortex and hippocampus were also obtained where available. Sections were pretreated with either formic acid or microwaving to enhance immunoreactivity. Primary antibodies included polyclonal anti-ubiquitin (1:1,000; DAKO, Glostrup, Denmark), monoclonal anti-ubiquitin (monoclonal 1510; 1:40,000; Chemicon, Temecula, CA), and polyclonal anti–TDP-43 (1:3,000; ProteinTech Group, Chicago, IL). In Vancouver, immunohistochemistry was performed using the Ventana BenchMark XT automated staining system (Ventana, Tucson, AZ), whereas other centers used a manual method using an avidin-biotin complex detection system (Vector Laboratories, Burlingame, CA). Sections were developed using either 3,3′-diaminobenzidine or aminoethylcarbizole and counterstained with hematoxylin. Double-labeling immunofluorescence was performed on formalin-fixed, paraffin-embedded sections of spinal cord using primary antibodies against ubiquitin (monoclonal 1510; Chemicon; 1:20,000) and TDP-43 (1:2,000; ProteinTech Group) and Alexa Fluor 488– and 594-conjugated secondary antibodies (Molecular Probes, Eugene, OR), as described previously.12
Quantitation of Pathology
Semiquantitative analysis of ub-ir and TDP-43–immunoreactive pathology was performed on sections of spinal cord or medulla and extramotor cortex with maximal numbers of NCI. Data were obtained for each type of inclusion (fibrillar NCI, round NCI, conglomerate NCI, and glial inclusions) using the following scoring system: 0 = none; + = 1–3 inclusions/section; ++ = 4–10 inclusions/section; +++ = >10 inclusions/section. Extramotor cortical pathology was graded in a similar semiquantitative fashion: 0 = none; + = mild; ++ = moderate; +++ = severe.
The presence of pathological TDP-43 was analyzed in urea solubilized spinal cord extracts from cases with SOD1 mutations (n = 3), SOD1-negative FALS (n = 2), and SALS (n = 2), using previously described methods.12 In brief, postmortem spinal cord tissue was dissected and weighed. Tissue was homogenized in 5ml/gm low-salt (LS) buffer (10mM tris[hydroxymethyl]aminomethane [Tris], pH 7.5, 5mM EDTA, 1mM dithiothreitol, 10% sucrose, and a cocktail of protease inhibitors) and sedimented at 25,000g for 30 minutes at 4°C. Pellets were then washed by reextraction in LS buffer and sedimentation. The resulting pellets were subjected to 2 sequential extractions in 5ml/gm Triton-X buffer (LS + 1% Triton X-100 [Sigma, St. Louis, MO] + 0.5M NaCl) and sedimented at 180,000g for 30 minutes at 4°C. Myelin was removed from pellets by homogenization in Triton-X buffer containing 30% sucrose, followed by centrifugation. The resulting pellets were then homogenized in 5ml/gm sarkosyl buffer (LS + 1% N-lauroyl-sarcosine + 0.5M NaCl) and incubated at 22°C on a shaker for 1 hour before sedimentation at 180,000g for 30 minutes at 22°C. The remaining pellets were extracted in 0.25ml/gm urea buffer (7M urea, 2M thiourea, 4% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate [CHAPS], 30mM Tris-HCl, pH 8.5) before centrifugation at 25,000g for 30 minutes at 22°C. Supernatants were saved as the urea fractions. Proteins were resolved in Tris-glycine 5 to 20% gradient sodium dodecyl sulfate polyacrylamide gel electrophoresis, transferred to nitrocellulose, and probed with mouse monoclonal anti–TDP-43 (Abnova, Taipei City, Taiwan), and secondary antibodies (horseradish peroxidase–conjugated anti–mouse IgG; Jackson ImmunoReasearch, West Grove, PA). Blots were developed with Renaissance Enhanced Luminol Reagents (NEN Life Science Product, Boston, MA), and digital images were acquired using a Fujifilm Intelligent Darkbox II (Fuji Systems USA, Stamford, CT).
Immunohistochemistry on sections of postmortem spinal cord or brainstem always demonstrated ub-ir NCI in lower motor neurons in SALS cases (n = 59; see the Table). Most had a filamentous morphology (Fig 1A), but compact, round NCI were also common (see Fig 1B). Similar numbers of NCI with the same morphology were always detected with TDP-43 antibodies (see Figs 1C, D). Neurons with cytoplasmic inclusions consistently demonstrated a loss of the normal nuclear TDP-43 staining (see Figs 1C, D). Double-labeling immunofluorescence confirmed colocalization of ubiquitin and TDP-43 in these NCIs (Fig 2). Similar NCIs, positive for both ubiquitin and TDP-43, were also always found in cases of SALS-d and FALS-d (n = 26). Of the 15 cases with SOD1 mutations, 11 (73%) had ub-ir NCIs. Although some of the SOD1 cases had small numbers of filamentous or round NCI, the most common type consisted of irregular, ill-defined, multilobulated masses (see Fig 1E). These “conglomerate” NCIs demonstrated ubiquitin immunoreactivity that was inhomogeneous and variable in intensity. These likely correspond to the “hyaline conglomerates” that have previously been reported in cases with SOD1 mutations but that are generally not found in SALS.19 Importantly, none of the cases with SOD1 mutations demonstrated any NCIs (filamentous, round, or conglomerate) that were TDP-43–immunoreactive. In some sections stained for TDP-43, unstained conglomerate NCIs were faintly visible (see Fig 1F). Double immunofluorescence confirmed that the ub-ir NCIs in SOD1 mutation cases were not immunoreactive for TDP-43 (see Fig 2). In contrast, 10 of 11 (91%) cases of SOD1-negative FALS had filamentous or round NCI that were immunoreactive for both ubiquitin and TDP-43. None of the SALS or SOD1-negative FALS cases had any conglomerate NCIs.
In addition to lower motor neuron pathology, cases of ALS-d also always showed ub-ir NCI and neurites in superficial layers of the frontal neocortex and dentate granule cell layer of the hippocampus (FTLD-U), characteristic of this condition (see the Table).18 Similar extramotor pathology was also found in 15 of 55 (27%) SALS cases with no history of dementia. This prevalence is similar to previous studies and suggests that classic SALS and ALS-d represent a spectrum of disease with a common underlying pathology.18 In every case where it was present, this extramotor pathology was also immunoreactive for TDP-43 (see Fig 1G). In contrast, none of the cases with SOD1 mutations had any ub-ir extramotor pathology.
In addition to NCIs, TDP-43 immunohistochemistry also demonstrated cytoplasmic inclusions in small cells with glial morphology in spinal cord and brainstem (see Fig 1H). Most of the affected cells had small, round nuclei, suggestive of oligodendroglia. This finding is consistent with a recent study describing oligodendroglial TDP-43 pathology in FTLD-U.15 Glial inclusions were present in the lower motor nuclei and adjacent white matter in 57 in 59 (97%) of SALS and all cases of ALS-d. Similar glial TDP-43 pathology was always present in SOD1-negative FALS cases, but not in any cases with SOD1 mutations.
We also examined TDP-43 biochemically, in the insoluble urea extracted fraction from postmortem spinal cord tissue. In addition to the normal 43kDa protein, both SALS and SOD1-negative FALS also showed a distinct pathological band at approximately 25kDa, representing C-terminal fragments, and a high Mr smear (Fig 3). Immunoblots from three cases, each with a different SOD1 mutation, were devoid of this pathological TDP-43 signature. With shorter exposure time, an approximately 45kDa band representing phosphorylated TDP-43 was observed in all SALS and SOD1-negative FALS cases, as well as in one of the three cases with a SOD1 mutation (not shown).
TDP-43 was first cloned as a human protein capable of binding to the transactive response DNA of human immunodeficiency virus type 1 (HIV-1),22 and later identified as part of a complex involved in splicing the cystic fibrosis transmembrane conductance regulator gene.23 TDP-43 contains two RNA recognition motifs and a glycine-rich C-terminal region.24 It is a highly conserved, widely expressed nuclear protein with presumed functions in transcription regulation and exon skipping.25, 26 It also acts as a scaffold for nuclear bodies through an interaction with survival motor neuron protein.27 The physiological function of TDP-43 in the brain is currently unknown; however, it is normally localized to the nucleus of neurons and some glial cells.12, 15
Although the specific role of TDP-43 in neurodegeneration remains speculative, a number of findings, in this and previous studies, suggest that this protein is directly involved in the pathogenesis of SALS, ALS-d, and FTLD-U.12–15 Accumulation of abnormal TDP-43 is a consistent and specific feature of this group of conditions. Immunohistochemistry shows TDP-43 to be a component of the protein inclusions that form specifically within degenerating neuronal populations. Biochemical analysis demonstrates that the accumulated TDP-43 is abnormally processed and does not simply represent entrapped normal protein. This sort of accumulation of insoluble, misfolded proteins in neurons is a common feature of many neurodegenerative disorders and has been speculated to cause cellular dysfunction by interfering with essential processes such as intracellular transport.28 Furthermore, cells that contain TDP-43–positive inclusions consistently show an absence of the normal nuclear staining pattern, raising the possibility that some essential normal function of TDP-43 may be lost. Finally, the presence of TDP-43–positive inclusions in glia suggests that disruption of some protective or supportive role of these cells may contribute to pathogenesis.4 However, even if the accumulation of pathological TDP-43 is secondary to some more central cause of neurodegeneration, it nonetheless appears to be a highly sensitive and specific marker of the disease process underlying ALS, ALS-d, and FTLD-U.
In contrast, we found no evidence of pathological TDP-43 in cases with SOD1 mutations. Because the cases we examined represent only a fraction of the different pathogenic SOD1 mutations reported, we cannot exclude the possibility that some other mutations could be associated with abnormal TDP-43, but it is certainly not a consistent feature. It is worth noting that our cases included several examples with the SOD1 mutations that are most common in North America (A4V) and Europe (I113T), as well as the mutation expressed in a widely used transgenic mouse model of ALS (G85R).2 Importantly, SOD1 mutations account for only approximately 20% of FALS, and several other FALS genes or genetic loci have been identified and others undoubtedly exist.1, 2 Although we do not know the specific genetic abnormality in our 11 SOD1-negative FALS cases, the presence of pathological TDP-43 in all suggests that some FALS cases may be caused by mechanisms that are more similar to SALS.
One issue that remains unresolved is the specific protein composition of the NCIs in cases with SOD1 mutations. A number of studies have shown these to label with antibodies against normal or misfolded SOD1.20, 29 Although such results are not consistent,30 this may simply reflect differences in antibody sensitivity for various species of mutant SOD1. Other studies have found that the NCIs in SOD1 mutation cases are immunoreactive for neurofilament proteins.19, 31 However, this finding is not absolutely sensitive or specific for SOD1 mutation cases. It is also unclear whether this reflects a primary abnormality of neurofilament processing or some secondary phenomena. Although fully resolving the biochemical composition of the NCIs in SOD1 mutation cases may be important for understanding the mechanism of SOD1 toxicity, the crucial finding in the context of this study is that these do not appear to contain TDP-43 and are therefore fundamentally different from the NCIs that characterize SALS.
In summary, our data have significant implications for elucidating pathogenic mechanism(s) in ALS. The finding of pathological TDP-43 in all cases of the largest series of SALS, examined to date, strengthens the role of this protein in disease pathogenesis. The presence of similar TDP-43 pathology in all cases of ALS-d cases supports the hypothesis that classic ALS and ALS-d are part of a disease spectrum that also includes FTLD-U.18 Furthermore, the recognition of TDP-43–positive glial inclusions is consistent with the hypothesis that ALS results from dysfunction of both neurons and their supporting cells.4 However, our most striking finding is the absence of pathological TDP-43 in any of our cases with SOD1 mutations. The significance of this result is the implication that the pathological processes underlying motor neuron degeneration in SALS are different from those associated with SOD1 mutations. Most ALS research and many of the therapeutic strategies considered in the past decade have been based on the assumption that SALS and SOD1-linked FALS are pathogenically similar. However, the finding of an absolute biochemical distinction between these two groups suggests that this view should be reexamined and raises questions about the role of SOD-1 models in ALS research. This may also partially explain why therapeutic strategies, shown to be effective in SOD1 mouse models, have generally not been effective in clinical trials of SALS patients.8, 32
After the submission of this manuscript, a related study was published that shows similar results.33
This research was funded by the Canadian Institutes of Health Research (74580, I.R.A.M.), the NIH (National Institute on Aging, P30 AG-10124, P01 AG-17586, V.M-Y.L., J.Q.T., M.S.F.; P50 P50-AG05681, N.C., U01 AG-16976, N.J.C.; AG-13854, E.H.B.), the German Federal Ministry of Education and Research (01G10505, H.A.K., M.N.), the Wellcome Trust (P.J.S.), and UK Medical Research Council (G0301152, P.J.S., P.G.I.).
We thank M. Luk, T. Schuck, A. Truax, J. Robinson, M. Getahun, Y. Xu, I. Pigur, and D. Carter for their expert technical assistance. We thank Drs D. Seilhean and F. Salachas for contributing tissue and genetic information.