Amyotrophic lateral sclerosis (ALS) is a fatal adult-onset neurodegenerative disease that causes degeneration of motor neurons and paralysis. Approximately 20% of familial ALS cases have been linked to mutations in the copper/zinc superoxide dismutase (SOD1) gene, but it is unclear how mutations in the protein result in motor neuron degeneration. Transgenic (tg) mice expressing mutated forms of human SOD1 (hSOD1) develop clinical and pathological features similar to those of ALS. We used tg mice expressing hSOD1-G93A, hSOD1-G37R, and hSOD1-wild-type to investigate a new subcellular pathology involving mutant hSOD1 protein prominently localizing to the nuclear compartment and disruption of the architecture of nuclear gems. We developed methods for extracting relatively pure cell nucleus fractions from mouse CNS tissues and demonstrate a low nuclear presence of endogenous SOD1 in mouse brain and spinal cord, but prominent nuclear accumulation of hSOD1-G93A, -G37R, and -wild-type in tg mice. The hSOD1 concentrated in the nuclei of spinal cord cells, particularly motor neurons, at a young age. The survival motor neuron protein (SMN) complex is disrupted in motor neuron nuclei before disease onset in hSOD1-G93A and -G37R mice; age-matched hSOD1-wild-type mice did not show SMN disruption despite a nuclear presence. Our data suggest new mechanisms involving hSOD1 accumulation in the cell nucleus and mutant hSOD1-specific perturbations in SMN localization with disruption of the nuclear SMN complex in ALS mice and suggest an overlap of pathogenic mechanisms with spinal muscular atrophy.
Amyotrophic lateral sclerosis (ALS) is a human adult-onset neurodegenerative disease characterized by progressive weakness, muscle atrophy, spasticity, paralysis, and death generally within 3 to 5 years after onset of symptoms. The symptoms are related to degeneration and loss of upper motor neurons in the cerebral cortex and lower motor neurons in the brainstem and spinal cord (1, 2). Amyotrophic lateral sclerosis occurs in sporadic and familial forms (2, 3). Approximately 90% of cases are sporadic, with no known genetic component aside from those reported to be linked to mutations in the TAR-DNA–binding protein (4). Familial forms of ALS (fALS), comprising approximately 10% of cases, can be caused by either autosomal dominant or recessive mutations in a variety of proteins (5, 6). Approximately 20% of fALS cases are linked to mutations in the superoxide dismutase 1 (SOD1) gene (7, 8), which encodes a metalloenzyme that converts the highly toxic superoxide anion to molecular oxygen and hydrogen peroxide (9). More than 100 different mutations in the human SOD1 (hSOD1) protein have been linked to ALS (3, 5). These mutations appear to confer a toxic gain of function to the protein rather than a loss of dismutase activity (7, 10, 11). One SOD1 mutation causes a glycine to alanine substitution in the protein at position 93 (G93A) (3, 5, 12). Transgenic (tg) mice expressing high levels of hSOD1-G93A mutant ubiquitously in tissues throughout the body develop disease at 6 to 8 weeks of age and become completely paralyzed and die at approximately 16 to 18 weeks of age (13), with severe pathological alterations of motor neurons (14–16). Transgenic mice expressing lower levels of hSOD1-G93A and other mutant variants of hSOD1 develop a later onset disease (3, 5, 14).
Neither the underlying causes of the motor neuron degeneration nor the extraordinary vulnerability of motor neurons to the disease in ALS or animal models of ALS is understood. Abnormal protein aggregation and cytoplasmic inclusion formation, excitotoxicity, mitochondrial dysfunction, axon abnormalities, target deprivation with axonal dying-back, and oxidative stress from reactive oxygen species or reactive nitrogen species have been implicated in causing motor neuron degeneration in ALS (3, 12, 17, 18). Whether mutant protein acquires a toxic property or function through misfolding and aggregation or an aberrant oxidative chemistry in models of mutant hSOD1–linked fALS is highly controversial (18, 19). Nitric oxide and peroxynitrite have been implicated in the process of motor neuron death in cell culture models (20, 21) and in the process of motor degeneration in vivo in tg mice expressing hSOD1-G93A (16, 22). Mitochondria have garnered much attention in ALS (3, 12, 16, 23), but the nucleus is a cellular compartment that is rarely studied in human ALS (24) and animal models of ALS (25).
Several lines of evidence suggest precedence for nuclear pathobiology in ALS. Abnormal nuclear inclusions positive for ubiquitin, promyelocytic leukemia protein, and ataxin-3 can be found in motor neurons in human ALS (26). Mutation in the fused in sarcoma (FUS) gene is associated with some fALS cases. Interestingly, FUS protein undergoes nucleocytoplasmic shuttling, and some FUS mutants have impaired nuclear localization with accumulation in the cytoplasm (27, 28). Moreover, sporadic ALS cases have elevated levels of 8-hydroxy-2-deoxyguanosine, a robust marker of oxidatively damaged DNA (29), in vulnerable CNS regions (30), and specifically in upper and lower motor neurons with activated p53 (31). Human ALS motor neurons also accumulate DNA double-strand breaks (32). The DNA base excision repair enzyme apurinic-apyrimidinic endonuclease-1 is increased and active in human ALS–vulnerable CNS regions (33). The hSOD1-G93A tg mice also accumulate 8-hydroxy-2-deoxyguanosine in the spinal cord (34), and motor neurons show elevated levels of DNA single- and double-strand breaks (16). Although these data are provocative regarding possible cell nucleus–based mechanisms of disease in ALS, it is not clear whether they reflect causal mechanisms or disease consequences.
To substantiate further a possible role for cell nucleus abnormalities in the pathobiology in ALS, we tested the hypothesis that hSOD1 mutants gain access to the nucleus and compromise the organization of nuclear gems (35). We found that hSOD1 tg mouse lines have differential abnormalities in the cell nucleus. Accumulations of hSOD1-G93A, -G37R, and -wild-type proteins are found in tissue nuclear fractions of mouse spinal cord at a young age and are localized in the nucleus of spinal motor neurons. Moreover, there is hSOD1 mutant–specific decrease in motor neuron nuclei containing survival motor neuron protein (SMN, gemin 1) complexes. The number of motor neuron nuclei lacking the complexes increases with age and disease progression. These results suggest a possible new mechanism for mutant hSOD1 cytotoxicity. Disruption of nuclear SMN complexes could be related to motor neuron degeneration in SOD1-linked fALS in yet to be clarified mechanisms overlapping with spinal muscular atrophy (35, 36).
Materials and Methods
The ALS mice used were tg mice expressing hSOD1. Four different tg mouse lines were studied. One mouse line (B6SJL-TgN[SOD1-G93A]1Gur/J, G1H line, stock 002726; The Jackson Laboratory, Bar Harbor, ME) has a high copy number of hSOD1-G93A mutant allele (∼20 copies) and a rapid disease onset (13). We also used hSOD1-G93A mice with reduced transgene copy number and a much slower disease progression (B6SJL-Tg[SOD1-G93A]dl1Gur/J mice, stock 002300; The Jackson Laboratory). Another tg mouse line studied, clinically similar to the G93A low–expressing tg mice, was hemizygous for a low copy number of hSOD1-G37R allele (line 29, some [n = 6]; these mouse tissue samples were provided by Dr. David Borchelt, McKnight Brain Institute, University of Florida, and living mice were derived from a founder B6.Cg-Tg [SOD1-G37R 29Dpr/J], stock 008229; The Jackson Laboratory). Nontransgenic littermate control mice studied at the same ages as the mutants and tg mice expressing the normal wild-type human SOD1 gene (B6SJL-TgN[SOD1] 2Gur, stock 002297; The Jackson Laboratory) at levels comparable to the mutant protein in hSOD1-G93Ahigh mice were used as controls. The institutional Animal Care and Use Committee approved the animal protocols.
Preparation of Nuclear-Enriched Tissue Fractions
The mice were killed by anesthetic overdose and were quickly decapitated. The brains and spinal cords were removed rapidly, weighed, and placed in ice-cold PBS. The tissue was hand dounce-homogenized in 5× buffer (320 mmol/L sucrose, 5 mmol/L CaCl2, 3 mmol/L magnesium acetate, 100 mmol/L EDTA, 10 mmol/L Tris-HCl, 0.1% NP-40, pH 8.0). The homogenate was incubated with 5 mg/mL of bead-coupled proteinase K for 30 minutes at room temperature (RT) with gentle mixing. The coupling of proteinase K to agarose beads was used to prohibit permeation of enzyme into the nucleus. The nuclei were then separated by discontinuous sucrose gradient ultracentrifugation as follows: 1.5 mL of 0.9 mol/L sucrose was layered on top of 1.8 mol/L sucrose and the homogenate was layered on to the top, this was then centrifuged at 78,800g for 120 minutes, and the off-white pellet (nuclei) was resuspended in PBS.
Another protocol was developed to isolate nucleus-pure fractions of spinal cord using a modification of a previously reported method that is not reliant on sucrose gradient ultracentrifugation but used trypsin digestion and differential centrifugation (37). Fresh spinal cords were washed and minced in ice-cold Ca2+-free and Mg2+-free Hank’s balanced salt solution and then digested in 0.25% trypsin-EDTA at 37°C for 20 minutes with gentle trituration. The volume of the tissue digest was adjusted with Tris-NP40-EDTA and then centrifuged at 20g for 5 minutes at RT. The P1 pellet, containing undigested tissue, white matter, and vessels, was discarded, and the supernatant (S1) was collected and centrifuged at 400 rpm (∼50 gav) for 5 minutes at RT. The P2 pellet was saved and resuspended in Tris-NP40-EDTA, and the S2 supernatant was diluted in Tris-NP40-EDTA. These fractions were centrifuged at 1000 rpm at RT; the pellets (containing cell nuclei) were combined and resuspended in Tris-NP40-EDTA and centrifuged again at 1000 rpm to wash away trypsin-EDTA. The pellet was resuspended in Tris-NP40-EDTA and digested for 15 minutes at 37°C with 5 mg/mL proteinase K–agarose (Sigma-Aldrich, St. Louis, MO) and then centrifuged at 20g to remove the proteinase K beads. The supernatant containing cell nuclei (confirmed by microscopy) was collected and centrifuged at 14,000g to collect nuclei, which were then resuspended in Tris-20% glycerol with protease inhibitors. Protein concentrations were measured by a protein assay (BioRad, Hercules, CA) with bovine serum albumin as standard.
The nucleus-enriched fractions were assayed for relative purity by immunoblotting, bright-field microscopy, and immunofluorescence. For immunofluorescence, a small aliquot was allowed to dry on a gelatin-coated slide before staining.
Tissue fractions were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis on 4% to 12% NuPAGE gels (Invitrogen, Eugene, OR) or on 16% nonprecasted gels under denaturing and reducing conditions. The proteins were transferred to a nitrocellulose membrane using the Xcell Surelock system (Invitrogen) according to the manufacturer’s protocol. The membranes were stained with Ponceau S (Sigma-Aldrich) to determine transfer efficiency, destained, blocked with 1% bovine serum albumin/0.05% Tween 20 in Tris-buffered saline for 1 hour, and then incubated with a primary antibody in blocking solution. The following antibodies were used: mouse monoclonal antibody to light neurofilament (Millipore, Billerica, MA), 1:1000; rabbit polyclonal antibody to lamin-associated polypeptide 2 (Lap2) (provided by Dr. Kathy Wilson, Johns Hopkins University), 1:2000; goat polyclonal antibody to Lap2 (Santa Cruz Biotechnology, Santa Cruz, CA), 1:1000; rabbit polyclonal antibody to methyl-CpG–binding protein 2 (Imgenex, San Diego, CA), 1:500; mouse monoclonal to neuronal nucleus protein NeuN (Millipore), 1:1000; mouse monoclonal antibody to hSOD1 (MBL, Woburn, MA), 1:1000; rabbit monoclonal antibody to hSOD1 (Epitomics, Burlingame, CA), 1:5000; rabbit polyclonal antibody to SOD1 (Millipore), 1:500. After primary antibody incubation, blots were washed in blocking buffer, incubated with species-appropriate horseradish peroxidase–conjugated secondary antibody (BioRad) diluted at 1:10,000–50,000 in blocking solution, washed in blocking buffer, and developed with ECL Supersignal West Pico reagent (Thermo Scientific, Rockford, IL). Immunoreactivities were visualized on the membranes using a CCD camera and BioRad Quantity One software or x-ray film. Western blots were done in triplicate or more. The fold difference in immunoreactivity was determined by dividing the optical density of the nuclear hSOD1 immunoreactivity by the endogenous SOD1 immunoreactivity optical density. Results are presented as mean ± SD. Statistical significance was determined by the Student t test. The level of significance was set at p < 0.05.
Immunofluorescence and Immunohistochemistry
Mice were killed by anesthetic overdose and were fixed in situ by intracardial perfusion of 4% paraformaldehyde for 20 minutes. The tissues were then removed and postfixed overnight and then cryoprotected in 20% glycerol for 48 hours at 4°C. Brain and spinal sections were cut at 40 μm using a sliding microtome. The sections were stored in antifreeze buffer at -20°C until used.
Sections were washed overnight in PBS at 4°C to remove the antifreeze buffer. They were then permeabilized in 0.4% Triton X-100 (Tx) in PBS for 30 minutes, blocked in 10% normal donkey serum/0.1% Tx in PBS for 60 minutes, and incubated in a primary antibody in blocking solution for 48 hours at 4°C. The primary antibodies used were mouse monoclonal anti-NeuN (Millipore) 1:200; rabbit polyclonal anti-Lap2, 1:500; goat polyclonal anti–choline acetyltransferase (Millipore), 1:100; mouse monoclonal anti-hSOD1 (MBL), 1:50; rabbit polyclonal anti-SOD1 (Millipore), 1:50; mouse monoclonal anti-snurportin 1 (Santa Cruz Biotechnology), 1:50; and mouse monoclonal anti-SMN (BD Biosciences, San Jose, CA), 1:100. The sections were then washed in PBS, incubated in species-appropriate secondary antibody (1:400) conjugated to Alexa-Fluor-488, -594, or -647 (Molecular Probes, Eugene, OR), washed in PBS containing Hoechst-33342 dye (Molecular Probes) for nuclear staining, washed again, mounted with Vectashield (Vector Laboratories, Burlingame, CA), and coverslipped. Stained sections were viewed on a Zeiss LSM 510 Meta inverted confocal microscope with the appropriate lasers and filters, and images were captured and analyzed using the LSM image browser software (Zeiss). Image processing was done using the LSM software, with a slight alteration of brightness and contrast without changing the content and actual result.
For immunohistochemistry, the tissue sections were rinsed to remove the antifreeze buffer and permeabilized in 0.4% Tx in PBS for 30 minutes, blocked in 4% normal goat serum/0.1% Tx in PBS for 60 minutes, and incubated in a primary antibody in blocking solution for 48 hours at 4°C, as follows: mouse monoclonal anti-hSOD1 (MBL), 1:100; rabbit monoclonal anti-hSOD1 (Epitomics), 1:500; mouse monoclonal anti-SMN (BD Biosciences), 1:100; mouse monoclonal anti-TrkA (BD Biosciences), 1:100; mouse monoclonal anti-TrkB (BD Biosciences), 1:100; mouse monoclonal anti–estrogen receptor α (Leica Microsystems–Novocastra, Buffalo Grove, IL), 1:100; mouse monoclonal anti–phospho-CREB (Cell Signaling Technology, Beverly, MA), 1:100; and mouse monoclonal anti-Ogg1 (Novus Biologicals, Littleton, CO), 1:100. The sections were washed in PBS and then developed using the Vectastain Elite ABC kit (Vector Laboratories) according to the manufacturer’s protocol with diaminobenzidine as chromogen. Sections were mounted, counterstained with cresyl violet, and coverslipped. Stained sections were viewed on an Olympus microscope at the indicated magnification, and digital images were captured with a Nikon camera and ACT-1 software. Counts of SMN complexes in the nucleus were done at 1000× magnification. The SMN-positive discrete round bodies were observed by careful focusing throughout the z-axis of the nucleus of spinal motor neurons. Statistical analysis of the data was done using a heterogeneous χ2 test.
Preparation of a Nuclear-Enriched Fraction From CNS Tissue Homogenate
To demonstrate that hSOD1 is present in the nucleus of cells within tg mouse CNS using Western blotting, we developed and characterized new protocols to enrich CNS tissue homogenates for the cell nucleus. In one approach, the tissue was gently homogenized, digested with proteinase K beads, and subjected to sucrose gradient ultracentrifugation to obtain a highly enriched nuclear fraction (Fig. 1A). When examined by immunoblot, both the brain and spinal cord nuclear fractions were highly enriched for the nuclear envelope marker LAP2, with the exclusion of the cytosolic marker neurofilament light chain (Fig. 1A). Pretreatment with proteinase K efficiently degraded cytosolic contamination of the nuclear fraction but did not appear to affect the total nuclear protein stability, as determined by a comparison of predigestion and postdigestion nuclear and cytosolic markers (Fig. 1A). The purity of the nuclear fractions from the brain and spinal cord was further corroborated by microscopic assessment of morphology and immunodetection of nuclear markers and DNA staining (Fig. 1B). The fractions showed a relatively high purity for nuclei, as shown morphologically and by Hoechst staining for DNA and Lap2 (nuclear membrane) immunostaining (Fig. 1B). The fractions from the brain consisted almost exclusively of neuronal nuclei, as determined by dual staining of NeuN (neuronal nuclear marker) and DNA (Fig. 1B). Although the spinal cord preparation was also highly enriched in neuronal nuclei, there was some debris contamination, probably attributable to myelin (Fig. 1B), but this contamination was completely eliminated using the alternative approach (data not shown). In addition, the nuclei from the spinal cord were more heterogeneous than that of the brain. Many nuclei were NeuN-negative nonneuronal cell nuclei, and only a small percentage of the nuclei were likely from motor neurons (as determined by morphology, DNA staining patterns, and size). A low percentage of motor neuron nuclei was anticipated because motor neurons constitute a minority of total cells within the spinal cord.
Human SOD Accumulates in the Cell Nuclear Compartment of Nervous Tissue From Tg Mice
Using nuclear-enriched CNS tissue fractions, we determined by Western blotting the nuclear presence of endogenous mouse SOD1 in non-tg mice and hSOD1 in tg mice expressing hSOD1-G93A, hSOD1-G37R, and hSOD1-wild-type (Fig. 2). In both brain and spinal cord of non-tg mice, endogenous mouse SOD1 was present in nuclear fractions (Fig. 2A). Non-tg mouse brain nuclear fractions had higher levels of endogenous SOD1 than spinal cord nuclear fractions (Fig. 2A). In spinal cord cell nucleus-pure fractions prepared by trypsin digestion and differential centrifugation and confirmed by the high enrichment of the nuclear marker methyl-CpG–binding protein 2 and exclusion of neurofilament, the level of mouse SOD1 in non-tg mice was very low compared with those in brain nuclear fractions (Fig. 2A, B). The presence of some endogenous SOD1 in the nuclear compartment of CNS tissue agrees with work on liver homogenates (38). The hSOD1 was present in the nuclear fractions of the brain and spinal cord of tg mice (Fig. 2A). In tg mouse spinal cord, hSOD1-G93A and hSOD1-G37R were both detected in the nuclear fractions at presymptomatic (6 weeks of age) stages of disease (Fig. 2A, B). At this age, the levels of hSOD1-G93A were about 30% to 40% endogenous, whereas the levels of hSOD1-G37R were approximately 120% of endogenous (Fig. 2C). The hSOD1-wild-type was also detected in the nuclear fractions of spinal cord in 6-week-old mice (Fig. 2B), and the levels were greater than those detected for the mutants (Fig. 2C). By late-stage disease (16 weeks old), the hSOD1-G93A present in the nuclear fraction of spinal cord had significantly increased (Fig. 2A). This finding was in sharp contrast to brain tissue that showed similar levels of hSOD1-G93A in the nuclear fractions at both stages of the disease (Fig. 2A). These findings were further corroborated using immunofluorescence for nuclear markers and hSOD1 on nuclear fraction aliquots (Fig. 2D). In both the brain and spinal cord, hSOD1 was seen to localize within the nuclei (Fig. 2D).
Human SOD Is Localized in the Nucleus of Motor Neurons Early in the Disease in Tg Mice and the Motor Neuron Nucleus Architecture Becomes Disrupted
The immunoblotting findings demonstrating the hSOD1 presence in whole–spinal cord nuclear fractions prompted us to determine whether hSOD1 was specifically present in motor neurons of tg mice. Because the morphological data on nuclear fractions from the spinal cord indicated a heterogeneous population of cell nuclei with only a small percentage being the motor neuron nuclei (Figs. 1C, 2D), we used immunofluorescence and immunohistochemistry on tg mouse spinal cord sections to directly show the hSOD1 nuclear localization in cells (Fig. 3). As expected based on Western blots (Fig. 2A, B), non-tg mice did not show any hSOD1 immunoreactivity, demonstrating that the antibody was specific for hSOD1 and did not detect the endogenous mouse SOD1 (Fig. 3H). The hSOD1-wild-type tg mouse spinal cord sections showed a strong immunoreactivity for hSOD1, further demonstrating the specificity of the hSOD1 antibody, and showed cytoplasmic and nuclear immunoreactivity in neurons of the ventral horn (Fig. 3A) and the dorsal horn (Fig. 3B). Presymptomatic hSOD1-G93A tg mice had intense hSOD1 immunoreactivity localized to the nucleus of motor neurons in the spinal cord ventral horn (Fig. 3C). As judged by signal intensity, there appeared to be more hSOD1-G93A in the nucleus than in the cytoplasm (Fig. 3C, C′). To corroborate the hSOD1 immunofluorescence and confocal microscope images, immunoperoxidase immunohistochemistry was used to detect hSOD1 (Fig. 3G, H). Spinal cord sections from non-tg mice showed no immunoreactivity, even after long developing periods (Fig. 3G), whereas those from presymptomatic hSOD1-G93A mice displayed robust immunoreactivity in the nucleus of motor neurons before cytoplasmic vacuolation, yet the presence of the disease was evident by the neuropil vacuolation (Fig. 3H). The nuclear localization of hSOD1-G93A was ubiquitous within the pools of motor neurons (Fig. 3C, H). Prominent hSOD1-positive aggregates were seen in the nucleus of motor neurons in presymptomatic hSOD1-G93A mice (Fig. 3C, C′, H). No other cell type in the spinal cord, including dorsal horn neurons, demonstrated nuclear localization of hSOD1 in presymptomatic mice (Fig. 3D, D′), suggesting that the hSOD1-G93A immunoreactivity observed by immunoblotting of spinal nuclear fractions of 6-week-old mice was caused largely by the motor neuron nuclei (Fig. 2A). However, by late-stage disease, spinal cord motor neuron nuclei showed scant hSOD1-G93A immunoreactivity (Fig. 3E, E′), whereas there was a significant increase in immunoreactivity in dorsal horn neuron nuclei (Fig. 3F, F′). This histological observation agrees with the immunoblotting observation showing a major increase in hSOD1-G93A immunoreactivity in spinal cord nuclear fractions from late-stage disease mice (Fig. 2A).
The loss of nuclear hSOD1-G93A immunostaining and nuclear DNA staining in motor neurons of hSOD1-G93Ahigh at end-stage disease (Fig. 3E, E′) and several other morphological observations in these mice (16) prompted a confocal microscope investigation of motor neuron nuclear architecture in the mice. We found that the nuclear architecture was severely disrupted in spinal motor neurons of hSOD1-G93Ahigh mice at late-stage disease (Fig. 4). The nucleus of each motor neuron in non-tg mice and presymptomatic tg mice (Fig. 4) showed a very structured architecture consisting of a nucleolus, Cajal bodies, and areas of interchromatin, heterochromatin, and euchromatin (Fig. 4). In contrast, the nucleus of remaining motor neurons in severely symptomatic mice had indiscernible nucleoli and associated structures and significant remodeling of the chromatin (Fig. 4).
Human SOD Is Localized in the Nucleus of Motor Neurons in Tg Mice With a Slow Disease Onset
Western blotting experiments indicated that hSOD1 also accumulates very early in spinal cord nuclear fractions in hSOD1-G37R tg mice that have a much slower disease progression (Fig. 2B). To determine if the nuclear localization of mutant SOD1 in motor neurons was specific to high-expressing hSOD1-G93A tg mice, we used in situ immunofluorescence on the low-expressing hSOD1-G93A tg mice (Fig. 5A, B) and on the hSOD1-G37R tg mice (Fig. 5C, D). These 2 tg mouse lines develop disease similar to that seen in the G93Ahigh expresser model, although they have increased time to disease onset and longer life spans (39). The hSOD1-G93Alow mice showed prominent hSOD1 localization in the nucleus of motor neurons (Fig. 5B). As with hSOD1-G93Ahigh mice, the nuclear localization appeared equivalent to or greater than the cytoplasmic content and specific to motor neurons in presymptomatic mice. Interestingly, unlike the hSOD1-G93Ahigh mice, hSOD1-G93Alow mice did not appear to have the prominent morphological aggregation in either the cytoplasm or the nucleus of spinal motor neurons. The hSOD1 nuclear localization was also observed in motor neurons in hSOD1-G37R tg mice (Fig. 5D); however, not all motor neurons displayed nuclear hSOD1-G37R (Fig. 5D). Furthermore, these mice did not appear to form the morphologically detectable hSOD1 aggregates (Fig. 5D) that were prominent in hSOD1-G93Ahigh mice (Fig. 3C).
Nuclear Protein Import Markers and SMN Complex in Motor Neurons of hSOD1 Tg Mice
In presymptomatic hSOD1-G93A tg mice, many of the hSOD1 aggregates appeared to be localized in proximity to the nuclear envelope (Fig. 3C′, H inset). Based on this observation, we sought evidence for mutant hSOD1 interference with nuclear import. To morphologically study the possibility of a general import problem, we localized snurportin-1. Snurportin is an importin β-binding protein involved in the nuclear import of spliceosomal small nuclear ribonucleoproteins (40). Snurportin localizations in motor neurons were similar in non-tg and hSOD1-G93A tg mice (Figure, Supplemental Digital Content 1, ). The protein was localized to the cytoplasm, consistent with the observation that nuclear imported snurportin is shuttled to the cytosol (40). This observation suggests that the nuclear hSOD1 aggregates are not sequestering snurportin-1. However, other nuclear import abnormalities could remain unrevealed by mere snurportin localization, such as inhibition of specific importins independent of the nuclear pore. To determine if the nuclear localized hSOD1 might be affecting the import of nuclear-localized proteins, we screened, by immunohistochemistry, several proteins involved in various nuclear pathways (Figure, Supplemental Digital Content 2, ). In hSOD1-G93A tg mice at different stages of disease and non-tg same-age littermates, we compared the nuclear localizations of receptor tyrosine kinases, estrogen receptor, CERB, and the DNA base excision repair enzyme Ogg1. There were little apparent differences in the protein localizations in presymptomatic and early symptomatic mice when compared with age-matched non-tg littermates (Figure, Supplemental Digital Content 2, ). However, by late-stage disease, nearly all of the proteins examined were excluded from the nucleus of the remaining spinal motor neurons, supporting our observation that nuclear architecture is disrupted at end-stage disease (Fig. 4) and confirming an earlier study showing that p53 and p73 are excluded from the nucleus of motor neurons (16). This would indicate that in advanced disease, there is a general disruption of nuclear organization and function in motor neurons. An exception was Ogg1, which was retained within the nuclei of motor neurons throughout the disease course (Figure, Supplemental Digital Content 2, ), consistent with previous observations demonstrating a marked accumulation of DNA damage in motor neurons in hSOD1-G93A mice (16).
In contrast, we found that SMN showed prominent differences in nuclear immunolocalization in presymptomatic hSOD1-G93Ahigh mice versus non-tg littermates (Fig. 6). Nuclear localized SMN was abnormal in motor neurons in tg mice, as evidenced by the dissipation of nuclear SMN complexes. This abnormality appeared restricted to the nuclei of motor neurons, and the changes appeared asynchronous in motor neurons (Fig. 6B–D; Table). Lumbar spinal cord motor neurons in tg mice could be classified as having nuclei with normal SMN complexes (Fig. 6B), nuclei devoid of SMN complexes (Fig. 6D), and nuclei in an intermediate state of SMN complex dissolution (Fig. 6C). Most of the motor neuron nuclei in presymptomatic hSOD1-G93Ahigh tg mice showed some disruption of the SMN complex (58.9% ± 7.2%), with most of those having no nuclear SMN complex (35.7% ± 4.5%). By late-stage disease, many motor neurons in hSOD1-G93Ahigh tg mice had degenerated, but of those that still had intact nuclei, there was no observable SMN complex (Table).
To determine if the SMN abnormality was specific to mice expressing high levels of hSOD1-G93A mutant protein, we examined, by immunofluorescence and confocal microscopy, SMN complexes in spinal cord motor neurons in G93A and G37R tg mice (Fig. 7); moreover, we used immunohistochemistry to quantitatively assess SMN complexes in hSOD1-G93Alow, -G37R, and -wild-type tg mice and their age-matched non-tg littermates (Table). Immunohistochemistry and immunofluorescence coupled to confocal microscopy were complementary but not identical approaches because the latter offers greater resolution in cellular detail. By immunofluorescence and confocal microscopy, spinal cord motor neurons in non-tg mice showed uniformly prominent nuclear SMN complexes (Fig. 7A), whereas age-matched littermate G93Ahigh tg mice had SMN abnormalities in motor neuron nuclei (Fig. 7C), similar to findings made using immunohistochemistry (Fig. 6; Table). In these tg mice, there were normal motor neurons with a prominent nuclear SMN complex (Fig. 7B), similar to that seen in non-tg control mice (Fig. 7A), and motor neurons completely lacking the prominent nuclear SMN complex (Fig. 7C). Interestingly, in motor neurons that did not have the SMN complex, the SMN appeared to localize to the cytoplasmic side of the nuclear envelope, forming a distinct annulus around the nucleus (Fig. 7C, arrows). Age-matched non-tg controls for G37R tg mice had normal spinal cord motor neurons with prominent nuclear SMN complexes (Fig. 7D). Both immunofluorescence and immunohistochemistry revealed that motor neurons in G37R tg mice also showed nuclear SMN complex abnormalities (Fig. 7E, F; Table), similar to those seen in hSOD1-G93Ahigh mice, despite the G37R mice having less ubiquitous hSOD1 nuclear localization. However, unlike the hSOD1-G93Ahigh mice, the SMN in these motor neurons did not localize to the nuclear envelope, rather it was seen in prominent spherical structures in the cytoplasm (Fig. 7F, arrows). The SMN nuclear complex formation was also abnormal in motor neurons in hSOD1-G93Alow mice. These mice at early symptomatic stages (24 weeks) displayed abnormal SMN profiles in approximately 40% of the motor neurons (Table). The disruption of the nuclear SMN complex was seen in the mutant hSOD1 tg mice and not in the age-matched hSOD1-wild-type tg mice. However, 1-year-old hSOD1-wild-type tg mice showed a minor loss of SMN complexes (Table) and the gems appeared small (Fig. 6E).
We describe a new approach to interrogate the cell nucleus in mouse models of neurodegenerative disease and, by doing so, reveal a novel pathobiology in motor neurons of hSOD1 tg mouse models of ALS. We show that hSOD1 accumulates in the nuclear fractions of hSOD1 tg mouse spinal cord and is present in the nucleus of motor neurons in the ventral horn of these mice. Mutant (G93A and G37R) and wild-type variants of hSOD1 were detected in cell nuclei at a young age (6 weeks). In mutant tg mice, the presence of hSOD1 in the nucleus was detected irrespective of the level of transgene expression. Interestingly, mutant and wild-type tg mice significantly differed in SMN immunostaining patterns in the nucleus. Mutant hSOD1 mice displayed a similar motor neuron nuclear pathology seen as the loss of SMN-positive gems, whereas age-matched wild-type hSOD1 tg mice had normal patterns. Thus, the nuclear accumulation of hSOD1 per se appears not to cause cytotoxic effects because wild-type hSOD1 expression in the CNS of tg mice does not cause major overt ALS-like disease but rather mutant hSOD1 in the nucleus appears to exert specific toxic effects on the SMN complex. Disruption of nuclear gems in motor neurons could be a common mechanism of disease in motor neuron disorders (35, 36).
Accurate information on the precise subcellular localization of a protein is critical for understanding its function and pathobiology. A potential confounder in determining the subcellular location of SOD1, both endogenous and transgenically expressed, is its high level (41) or nonphysiological level (42) of expression, which complicates obtaining pure subcellular fractions because of a ubiquitous presence. We demonstrate the feasibility of obtaining CNS tissue subcellular fractions that are highly enriched for cell nuclei and have nearly undetectable contamination of cytosolic proteins. Our initial efforts to obtain a nuclear fraction containing no detectable cytoplasmic components proved to be problematic. Using the sucrose ultracentrifugation protocol by Jiang et al (43), the spinal cord nuclear yield was very low, with significant particulate contamination. By modifying the protocol to include an additional lower percentage sucrose layer and a predigestion step with proteinase K–coupled beads, we obtained a highly enriched nuclear fraction as determined by immunoblotting and fluorescent microscopy. Another protocol, developed and characterized previously (37) and modified and further validated here, was also used for corroboration. The latter approach can be used without ultracentrifugation. The nuclear fractions were fluorescent for DNA and highly immunoreactive for nuclear proteins, whereas nearly no immunoreactivity for abundant cytoplasmic proteins was detected. The relative purity of the fractions, coupled with the spinal cord tissue section confocal microscopy, conclusively shows that normal endogenous SOD1 gains access to the nucleus at low levels, consistent with previous studies (38, 44), whereas hSOD1 mutant and wild-type variants accumulated in the nuclei of spinal cord cells to a greater extent in tg mice. The nuclear presence of hSOD1-G93A was not an artifact of expression of high levels of mutant protein because tg mice expressing much lower levels of hSOD1 as G39A and G37R mutants also showed accumulation in the nucleus. The localizations and subcellular distributions of hSOD1 in hSOD1 tg mice have been studied infrequently using in situ histological methods because antibodies that preferentially or specifically bind hSOD1 were lacking. The antibodies we used here were either highly specific for hSOD1 (Epitomics) or preferentially detected hSOD1 with only a slight detection of mouse SOD1 when a large amount of proteins was loaded. With the recent development of these antibodies, a nuclear presence has been suggested from immunohistochemistry (16, 45), but our study is the first to conclusively show the accumulation of hSOD1 in the nuclear fraction of CNS tissues at levels greater than endogenous and, specifically, in motor neurons in vivo.
The presence of endogenous and mutant SOD1 in the nucleus begs the questions of how it gains access to the nucleus and its function in the nucleus. Liver cell nuclear SOD1 can be depleted by dialysis, demonstrating that it is trafficked freely by the nucleus possibly through nuclear pores (38). The function of nuclear SOD1 could be to catalyze the dismutation of superoxide radicals produced by nuclear membrane cytochromes P450 and b5 (38) and NADPH oxidase (46). Importantly, nuclear SOD1 is known to mitigate the accumulation of abasic site DNA lesions formed directly or indirectly by superoxide radical generated in or near the nucleus (47). However, cell culture studies of human neuroblastoma cells transfected with hSOD1-G93A plasmid suggest that the mutant form gains access to the nucleus, where it associates with chromatin and has enhanced peroxidase activity that causes DNA damage (48). The hSOD1-G93A mouse motor neurons accumulate considerable DNA lesions at presymptomatic stages of disease (16) and thus may induce SOD1 expression in the nucleus as a compensatory mechanism, but our evidence suggests that endogenous mouse SOD1 was not upregulated in tg mice.
We show that the magnitude of hSOD1-G93A accumulation in the nuclear fraction changes at different stages of disease in tg mice. There are specific temporal and spatial patterns in the hSOD1-G93A localization. It is present in the nucleus of spinal motor neurons before disease onset. In the lumbar spinal cord, motor neuron nuclei were unique in this feature because no other spinal cord cell nuclei, even neurons of the dorsal horn, showed hSOD1 localization in the presymptomatic stage. This is in sharp contrast to end-stage disease where the motor neuron nuclei lose hSOD1-G93A immunoreactivity, whereas numerous dorsal horn neuron nuclei accumulate hSOD1-G93A. Our histological finding is consistent with the immunoblot results, where hSOD1 immunoreactivity was low in presymptomatic mouse spinal cord nuclear fractions, but high levels were detected by late-stage disease.
Immunoreactivity for hSOD1-G93A and -G37R was found in the nucleus as large aggregates that appeared in the matrix or close to the nuclear envelope or was concentrated in a cytoplasmic perinuclear location. We reasoned that this finding might have pathophysiological significance stemming from possible effects on nuclear import and/or export. We hypothesized that the submembrane or perimembrane aggregates could interfere with nuclear import by disrupting nuclear pore transportation through association, hence the distinct membrane staining. However, immunolocalization analysis of snurportin, a nuclear pore import protein (40), did not reveal obvious differences in subcellular localization between tg and non-tg mice. We then thought that, rather than a nuclear pore abnormality, the hSOD1-G93A protein associates with importins, sequestering them within the nucleoplasm and thus inhibiting protein import. Several nuclear localized proteins were analyzed, but they did not show any obvious differences in localization between the tg and non-tg mice early in disease. However, late in the disease, degenerating motor neurons have prominent morphological abnormalities in the chromatin and the nucleolus and appear to acquire nuclear import/export abnormalities based on the exclusion of several nuclear proteins. Of the proteins we examined, constituting neurotrophin signaling, nuclear signaling, transcription factors, and repair enzymes, we observed a loss of motor neuron nuclear immunoreactivity for all of these proteins except for the DNA repair enzyme Ogg1, which appeared to be sustained or increased. A previous study has shown loss in nuclear immunoreactivity for importin α and β in motor neurons of late symptomatic G93A mice (25). Nevertheless, from these experiments, it was still not evident if the nuclear localization of hSOD1-G93A could mechanistically contribute to motor neuron degeneration in ALS mouse models. Our observations in mice have particular relevance because aggregates of misfolded hSOD1 are often found in spinal motor neuron and glial cell nuclei in human cases of SOD1-linked fALS and sporadic ALS (49).
An important finding that provides new insight into the mechanisms of motor neuron degeneration in ALS was that motor neurons in tg mice expressing mutant hSOD1, but not wild-type hSOD1, lost nuclear SMN-positive complexes. The SMN is a ubiquitously expressed protein involved in the formation of the spliceosomal small nuclear ribonucleoproteins (35). The SMN functions as one component of a multiprotein complex consisting of 7 other proteins, gemins 2 to 8 (50). The SMN and the other gemins form complexes in the nucleus, called gems, and in the cytoplasm (35). The formation of the spliceosome is crucial to cells because they are responsible for the pre-mRNA splicing necessary for correct mRNA translation. The critical importance of SMN in motor neuron biology is well established by the fatal genetic disease spinal muscular atrophy, which is the second most common genetic cause of childhood mortality and is characterized by progressive symmetrical limb and trunk paralysis, muscular atrophy, and motor neuron degeneration (36). Spinal muscular atrophy is caused by mutations in the Smn gene (36). We find here that mutant hSOD1 tg mice fail to form these SMN complexes in motor neurons. The abnormality is a presymptomatic event, and by late-stage disease, motor neurons with nuclear SMN complexes are sparse. This is not surprising given that, by late-stage disease, most motor neurons have degenerated or are in the process of degenerating. Although both G93A and G37R mutants had disruption to motor neuron nuclei SMN complexes, their respective SMN localization was distinct. In G93A mice, the SMN protein appears to localize in a ring around the nuclear envelope, whereas, in G37R mice, the SMN can be found in cytosolic aggregates. This could indicate that whereas the overall cell toxicity of mutant hSOD1 is caused by SMN complex disruption, the mediation of the disruption might be different in each type of hSOD1 mutant causing perturbation in RNA processing. Molecular genetic studies identifying ALS-linked mutations in genes encoding proteins involved in RNA processing support this idea (51). Further investigations of this possibility are needed.
Our findings suggest a new role for the nucleus in ALS pathobiology and a possible new mechanism for mutant hSOD1 cytotoxicity. Disruption of the formation and organization of nuclear SMN complexes could be related to motor neuron degeneration in ALS, reminiscent of disease events occurring in spinal muscular atrophy (52). Turner et al (53) have shown in cell culture that the expression of G93A- and G37R-hSOD1 in transfected mouse NSC34 cells causes SMN protein depletion, but wild-type hSOD1 expression does not; furthermore, SMN protein levels are reduced in presymptomatic G93A-hSOD1 tg mouse spinal cord. In further support of our findings, studies show that the SMN1 gene copy number is a risk factor for human sporadic ALS (54), SMN overexpression protects against mutant SOD1 toxicity in cultured cells (55), and genetic knockdown of SMN in G93A-hSOD1 tg mice worsens the disease (53).
The authors thank Antoinette Price and Yan Pan for technical assistance and advice. Dr. Kathy Wilson, Johns Hopkins University, provided a sample of Lap2 antibody.