Motor neuron degeneration and skeletal muscle denervation are hallmarks of amyotrophic lateral sclerosis (ALS), but other neuron populations and glial cells are also involved in ALS pathogenesis. We examined changes in inhibitory interneurons in spinal cords of the ALS model low-copy Gurney G93A-SOD1 (G1del) mice and found reduced expression of markers of glycinergic and GABAergic neurons, that is, glycine transporter 2 (GlyT2) and glutamic acid decarboxylase (GAD65/67), specifically in the ventral horns of clinically affected mice. There was also loss of GlyT2 and GAD67 messenger RNA-labeled neurons in the intermediate zone. Ubiquitinated inclusions appeared in interneurons before 20 weeks of age, that is, after their development in motor neurons but before the onset of clinicalsigns and major motor neuron degeneration, which starts from 25weeks of age. Because mutant superoxide dismutase 1 (SOD1) in glia might contribute to the pathogenesis, we also examined neuron-specific G93A-SOD1 mice; they also had loss of inhibitory interneuron markers in ventral horns and ubiquitinated interneuron inclusions. These data suggest that, in mutant SOD1-associated ALS, pathological changes may spread from motor neurons to interneuronsin a relatively early phase of the disease, independent of the presence of mutant SOD1 in glia. The degeneration of spinal inhibitory interneurons may in turn facilitate degeneration of motor neurons and contribute to disease progression.
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease, characterized by late-onset progressive degeneration of motor neurons resulting in paralysis of limb, facial, and respiratory muscles and usually leading to death within 5 years after diagnosis. In most patients, the disease is sporadic, whereas approximately 10% of cases have a monogenetic inherited form (1). Mutations associated with ALS have been identified in the gene for superoxide dismutase 1 (SOD1) (2) and, more recently, in genes encoding the DNA/RNA binding proteins, TAR-DNA-binding protein-43 (TDP-43) and FUS/TLS (1, 3). In addition, mutations in optineurin, VAPB, angiogenin, DCTN1, FIG4, senataxin, and alsin are associated with some ALS cases and ALS-like disorders (1, 4). Nuclear-to-cytoplasmic redistribution, aberrant processing and aggregation of TDP-43 are prominent pathological features in most ALS patients, independent of the presence of TDP-43 mutations (3, 5, 6). SOD1-associated ALS (SOD1-ALS) and FUS/TLS-ALS patients instead develop aggregates of SOD1 (2, 7, 8) and FUS/TLS (3, 9, 10), respectively. These and additional data point to an important role for protein aggregation in the pathogenesis of ALS, but the relationships between protein aggregation, neuronal degeneration, and clinical manifestations in different types of ALS are poorly understood (3, 7, 8, 11).
Hypotheses on ALS pathogenesis usually focus on toxic mechanisms that primarily affect motor neurons; their vulnerability has been linked to their large size, long axons, high metabolic activity, specialized physiological properties, and their dependence on skeletal muscle- and glia-derived factors (12, 13). Although motor neuron degeneration and skeletal muscle denervation represent hallmarks of ALS, cell loss and degenerative changes also occur in other neuron populations throughout the nervous system, particularly in the motor cortex and in glia (14-21). The involvement of multiple cell types favors disease models in which abnormalities in non-motor neurons or non-neuronal cells occur concurrently or sequentially with motor neuron degeneration and raises questions about the relationship between the degeneration of different cell types. For example, multiple clinicopathological studies have investigated the relationship between degeneration of spinal motor neurons and cortical (upper) motor neurons that underlie lower and upper motor neuron manifestations, respectively. These studies have provided variable results that suggest independent degeneration of cortical motor neurons and spinal motor neurons, anterograde propagation of disease from motor cortex to spinal cord, or retrograde progression of disease from spinal cord to motor cortex (14, 22-26). A linkage between upper and lower motor signs in initial stages of disease has been suggested, that is, when there are focal manifestations; this linkage may be lost in later stages of disease as neurodegeneration and clinical manifestations spread independently in the cortex and spinal cord (27). Taken together, clinicopathological studies indicate that neurons interconnected or contiguous to affected regions develop degenerative changes more readily than other cells and point to the occurrence of disease mechanisms that mediate or facilitate spreading of disease to neighboring or interconnected neurons (14).
Considerable attention has been devoted to the role of glia in SOD1-ALS (28). For example, SOD1 aggregates are present in astrocytes in the spinal cords of SOD1-ALS patients and transgenic mouse models (29). Other evidence pointing to a role of glial abnormalities in SOD1-ALS comes from studies with mutant SOD1 transgenic mice that express no or reduced levels of mutant SOD1 in glia. These mice show slower disease progression and prolonged survival compared with ubiquitous mutant SOD1-expressing mice, although disease onset is unaltered (30-33). On the other hand, expression of mutant SOD1 in neurons determines disease onset and is sufficient to cause an ALS-like disease in mice, indicating that mutant SOD1 in glia does not contribute to disease initiation but rather to disease propagation (28, 34). How mutant SOD1 in glia contributes to disease progression remains to be defined.
The intermediate zone of the spinal cord (Rexed laminae IV-VIII) contains several neuron types that play important roles in controlling the activities of motor neurons, for example, in reflexive and patterned movements and the maintenance of muscle tone (35, 36). Abnormalities in 1 or more populations of interneurons may cause overexcitation or overinhibition of motor neurons and possibly causing them excitotoxic stress (37). There are reports of loss of intermediate zone interneurons in the spinal cord of ALS patients (17, 38) and mutant-SOD1 transgenic mice (39, 40). Furthermore, abnormalities in inhibitory interneurons or inhibitory synapses may occur before motor neuron degeneration in mutant-SOD1 transgenic mice (41-43). We previously showed that spinal interneurons in a low-copy line of G93A-mutant SOD1 mice (G1del mice) (44) start to express the stress transcription factors c-Jun and ATF3 from 20 weeks of age (45, 46). This is several weeks later than the onset of c-Jun and ATF3 activation in motor neurons (at 14-15 weeks), but before the onset of motor manifestations, which starts after the age of 25 weeks in these mice (13, 45). To explore the role of non-motor neuron abnormalities in ALS further, we studied degeneration of spinal inhibitory interneurons in mutant SOD1-ALS mice.
Materials and Methods
Ubiquitous G93A-SOD1 mice were originally derived from Gurney G1 mice, but because of a reduction in the transgene copy number (8 instead of ∼20 transgene copy numbers per haploid genome), they show a delayed disease onset and are termed G1del mice (34, 44). The G1del mice were maintained under standard housing conditions in a FVB/N background by mating hemizygote males with nontransgenic females. Nontransgenic offspring served as controls (34). Neuron-specific G93A-hSOD1 mice carrying the complementary DNA (cDNA) of G93A-mutant hSOD1 cloned into the Thy1.2 expression cassette were generated as described (34). Data from this study were obtained from homozygotes of the T3 line (T3T3 mice) generated by intercrossing T3 hemizygotes and from T3hSOD1 double transgenic mice generated by crossing T3 mice with line N29 wild-type hSOD1-overexpressing mice (34). Onset of clinical disease was determined on the basis of weight, the ability to extend the hind limbs, and the ability to hang upside down on a grid for 1 minute (34). Mice reached end-stage disease when they could not right themselves within 5 seconds when placed on their back, when they lost more than 30% of their maximal weight, or when they developed infection of 1 eye. End-stage mice also were unable to hang in the hanging grid test, which predominantly depends on performance of forelimb muscles.
GlyT2-GFP transgenic mice bred in the C57BL6/J background (47) were obtained from Dr Hanns Ulrich Zeilhofer (Institute of Pharmacology Toxicology, University of Zurich). All animal experiments were approved by the Erasmus University animal care committee and performed in accordance with the guidelines of the Principles of Laboratory Animal Care (National Institutes of Health Publication No. 86-23) and the European Community Council Directive (86/609/EEC).
Mice were anesthetized with pentobarbital and perfused transcardially with 4% paraformaldehyde. The lumbar and cervical spinal cord were carefully dissected, embedded in gelatin blocks, and sectioned at 40 μm with a freezing microtome, as described (48). Free-floating sections were processed for immunohistochemistry using a standard avidin-biotin-immunoperoxidase complex method (ABC; Vector Laboratories, Burlingame, CA) with diaminobenzidine (0.05%) as the chromogen (48). Guinea pig anti-glycine transporter 2 (GlyT2; Millipore, Billerica, MA; 1:10000) and rabbit anti-glutamic acid decarboxylase (GAD65/67; Millipore; 1:4000) were used as primary antibodies to label glycinergic and GABAergic nerve terminals, respectively. Other primary antibodies were rabbit anti-ATF3 (Santa Cruz Biotechnology, Santa Cruz, CA; 1:1000), rabbit anticalbindin (Swant, Marly, Switzerland; 1:10000), goat anti-choline acetyl transferase (ChAT; Millipore; 1:500), rat anti-Mac2 (Cedarlane, Burlington, Ontario, Canada; 1:2000), rabbit anti-phospho-c-Jun (Ser63; Cell Signaling, Beverly, MA; 1:1000), mouse antiparvalbumin (Swant; 1:10000), mouse antiubiquitin (clone FK2; Affinity BioReagents, Golden, CO; 1:2000), and rabbit antiubiquitin (Dako, Glostrup, Denmark; 1:2000). Immunoperoxidase-stained sections were analyzed and photographed using a Leica DM-RB microscope and a Leica DC300 digital camera.
To examine the relative intensity of glycine transporter 2 (GlyT2) and GAD65/67 immunoperoxidase staining, lumbar L3-L5 sections were photographed using a 5× objective, and optical densities were determined from TIFF files using MetaMorph image analysis software (Molecular Devices, Sunnyvale, CA). Optical densities were determined in rectangular areas of 200 × 250 μm and 100 × 400 μm for the ventral and dorsal horns, respectively. To minimize variability resulting from the sectioning and staining procedures, these analyses were performed with spinal cord specimens embedded in a single gelatin block and processed in the same run. Gelatin blocks typically contained 12 cervical C5-C8 or lumbar L3-L5 specimens from G1del mice of 3 different ages or disease stages (e.g. 20 weeks, 30 weeks, end-stage disease, n = 3 per group) and nontransgenic littermates aged 20 to 30 weeks (n = 3). Other blocks contained specimens from 40-week-old T3T3 mice (n = 3), symptomatic (70-100 weeks old, n = 3), and 2-year-old nontransgenic or T3 hemizygote littermates (34).
In Situ Hybridization
In situ hybridization (ISH) was performed on free-floating 4% paraformaldehyde-fixed 30-μm-thick frozen sections using digoxigenin-labeled RNA probes, as described (49). Sense and antisense probes were transcribed from linearized plasmid constructs containing the partial GlyT2 or Gad67 cDNA sequences using a digoxigenin labeling kit (Roche, Indianapolis, IN). Sections were incubated overnight at 65°C with the probes diluted at 200 ng/mL. After hybridization, the digoxigenin-labeled RNA-RNA complex was detected by using alkaline phosphatase-conjugated sheep-antidigoxigenin (Roche; diluted 1:4000, incubated 48 hours at 4°C), with 5-bromo-4-chloro-3-indolyl phosphate and nitro-blue tetrazolium as substrate and chromogen, respectively.
Alkaline phosphatase-stained sections were analyzed and photographed using a Leica DM-RB microscope and a Leica DC300 digital camera. For quantitative analysis of the number of GlyT2 and Gad67 messenger RNA (mRNA)-labeled cells, lumbar L3-L5 sections were examined with an Olympus microscope fitted with a Lucivid miniature monitor coupled to StereoInvestigator software (version 4.37; MicroBrightField, Colchester, VT). The sections were systematically sampled across the spinal cord (each 10th section). The counting fields, 350 μm apart, were 0.0225 mm2; cells in contact with the left and lower boundaries of the counting fields were excluded. The area of the gray matter multiplied by the number of cells per squared millimeter gave the number of cells per section. Sections used for this analysis were produced in a single staining run to avoid variability in staining intensities and numbers of labeled cells resulting from differences in staining conditions between runs.
To combine GlyT2 or GAD67 ISH with immunofluorescence, hybridized digoxigenin complementary RNA was visualized using a tyramide amplification method with fluorescein isothiocyanate-labeled tyramide (50). After hybridization, sections were incubated with biotinylated-sheep antidigoxigenin antibody (Roche) (1:500; 48 hours at 4°C in phosphate-buffered saline [PBS], 2% milk powder, and 0.5% Triton X-100), followed by incubation with avidin-biotin-peroxidase complex (Vector Laboratories) and subsequently reacted with fluorescein isothiocyanate-tyramide conjugate (4 μg/mL) in the presence of H2O2 (0.001%) in PBS containing 0.1 mol/L imidazole, pH 7.6 (51). Thereafter, the sections were washed in PBS and processed for immunofluorescence with rabbit anti-ATF3 and rabbit antiubiquitin as the primary antibodies and Cy3-labeled donkey-anti-rabbit (1:200; Jackson ImmunoResearch, Bar Harbor, ME) as the secondary antibody. Double-labeled sections were analyzed with a Zeiss LSM 510 confocal laser scanning microscope using 40×/1.3 and 63×/1.4 oil-immersion objectives.
Spinal cord specimens were homogenized and sonicated in 20 volumes of PBS containing 0.5% Nonidet P-40 (NP-40) and protease inhibitors cocktail (Sigma) and centrifuged at 800 × g for 5 minutes at 4°C, and protein concentrations of the supernatants (S1) were determined. Samples containing 2 to 10 μg protein were electrophoresed on 8% or 10% SDS-PAGE gels and blotted on polyvinylidene fluoride membranes (Millipore). The membranes were blocked with 5% nonfat dry milk in PBS with 0.05% Tween 20, incubated in primary antibody, diluted in PBS with 0.05% Tween 20 with 1% dry milk followed by an incubation in peroxidase-conjugated secondary antibody, incubated in chemiluminescence reagent (Amersham, Piscataway, NJ), and exposed to film or a Kodak Image station. Primary antibodies used for Western blot included mouse antiactin (Millipore; 1:4000), guinea pig anti-GlyT2 (Millipore; 1:5000), rabbit anti-GAD65/67 (1:5000), and rabbit antimurine SOD1 (SOD101; Stressgen, Victoria, British Columbia, Canada; 1:8000).
Reverse Transcription-Polymerase Chain Reaction
Semiquantitative reverse transcription-polymerase chain reaction (PCR) was performed as previously described (52). Total RNA was extracted from spinal cord tissue using TRIzol and treated with DNAse. The RNA (5 μg) was converted into cDNA using oligo(dT) primer and reverse transcriptase in a total reaction volume of 20 μl. Polymerase chain reaction was performed with 0.1 μL of the reverse transcriptase reaction mixture in a reaction volume of 25 μl. Primers used were as follows: ChAT, 5′-GCGAATCGTTGGTATGACAAGTC-3′ (forward) and 5′-TTGAAGTTTCTCTGCCGAGGAG-3′ (backward); G6dph, 5′-CTTTGGACCCATCTGGAATCG-3′ and 5′-CACTTTGACCTTCTCATCACGGAC-3′; Gad67, 5′-TACGGGGTTCGCACAGGTC-3′ and 5′-CCCCAGGCAGCATCCACAT-3′; GlyT2, 5′-TACCGCTACCCTAACTGGTCCATGG-3′ and 5′-ATCCACACGACTGGACTAGCACTGA; GlyT2A, 5′-ACTCTACGGTTCAATCTGTTGTCC-3′ and 5′-GGTCCTAGGTGCACGAGGACTATCCCGG-3′. The number of PCR cycles and quantity of cDNA were determined to be within the linear range of the reactions. For quantification, the PCR products were electrophoresed on a 2% agarose gel, stained with ethidium bromide, and scanned on a Molecular Dynamics Typhoon instrument.
Statistical analyses were done using GraphPad Prism software (San Diego, CA). Means from different age groups and different transgenic mouse lines were compared using 1-way analysis of variance and the Tukey post test.
Reduced GlyT2 and GAD65/67 Protein and mRNA Levels in the Spinal Cord of Symptomatic G1del Mice
G1del (also termed G1slow or low-copy G1) mice ubiquitously express G93A-mutant SOD1; they develop weakness in 1 or more limbs from age 24 to 30 weeks and die of fatal paralysis between 32 and 40 weeks of age (48, 53). Early pathological and molecular changes in motor neurons include swelling and vacuolization of a subset of mitochondria starting before 3 weeks (53), the appearance of ubiquitinated pathology, Golgi fragmentation, endoplasmic reticulum stress, and ATF3 activation starting from 13 to 15 weeks (13, 45). These changes precede neuromuscular denervation, loss of motor neurons, and astrogliosis and microgliosis. Abnormalities in interneurons (ATF3 and phospho-c-Jun expression) occur after 20 weeks (45). In this study, we examined G1del mice at ages 10 to 15 weeks (early presymptomatic), 20 to 25 weeks (late presymptomatic), and 30 weeks (symptomatic) and mice that had reached disease end stage.
To identify inhibitory interneurons, we used the expression of GlyT2 and GAD as markers for glycinergic and GABAergic neurons, respectively (49, 54, 55). Immunohistochemistry showed that GlyT2 was expressed throughout the spinal cord gray matter except for the most superficial part of the dorsal horn (Figs. 1A-D). At high magnification, labeling was distributed in punctae, consistent with a predominant localization in nerve terminals (Figs. 1A-D) and in accord with previous studies (54). The distribution of GlyT2 immunoreactivity in spinal cords of 10- and 20-week-old G1del mice was indistinguishable from that in nontransgenic mice (Figs. 1B, C), whereas in symptomatic and end-stage G1del mice, there was a marked reduction of GlyT2 immunoreactivity, particularly in the ventral horn motor neuronal cell groups (Figs. 1D, H). GlyT2 immunoreactivity was not altered in the dorsal horn (Figs. 1D, H). Reduced GlyT2-immunoreactivity in the ventral horn occurred in both cervical and lumbar cord but, in general, was more prominent in the latter.
GABAergic interneurons were labeled with an antibody that binds both GAD67, the predominant GAD isoform in the ventral horn, and GAD65 that is expressed in a more restricted set of GABAergic interneurons and terminals (55, 56). GAD65/67 immunoreactivity was present throughout the spinal cord gray matter with more prominent labeling in the dorsal horn compared with that of GlyT2 (Fig. 1E). As with GlyT2, no change in overall staining occurred in spinal cord sections from 10- and 20-week-old G1del mice (Figs. 1F, I). There was reduced GAD65/67 immunoreactivity in the ventral horn of symptomatic G1del mice (Figs. 1G, I).
Western blot showed reduced GlyT2 and GAD65/67 immunoreactivity in spinal cord homogenates of symptomatic and end-stage G1del mice (Figs. 2A, B). In addition, we performed semiquantitative reverse transcription-PCR analysis to examine changes at the mRNA level. Consistent with the immunohistochemistry and Western blotting, GlyT2 and GAD67 mRNA levels in the spinal cord of 20-week-old G1del mice were the same as in the spinal cord of nontransgenic mice, whereas GlyT2 and GAD67 mRNA levels were reduced in symptomatic and end-stage G1del mice (Figs. 2C, D). In the end-stage disease mice, GlyT2 and GAD67 mRNA levels were reduced by approximately 50%, which was comparable to the loss of ChAT mRNA, which is predominantly produced by motor neurons.
We studied GlyT2 and GAD67 mRNA expression using ISH. Consistent with previous reports (49, 54), GlyT2 mRNA-labeled cells were predominantly localized in the deep dorsal horn and the intermediate zone (Rexed laminae V-VIII), whereas they were detected at a low frequency in the motor columns in the ventral horn (Rexed laminae IX) (Fig. 3A) and the superficial dorsal horn. Labeled cells were small- to medium-sized (15-25 μm in diameter) and showed varying labeling intensities. The distribution of GlyT2 mRNA was the same as in nontransgenic mice in 10- and 20-week-old G1del mice, whereas 30-week-old symptomatic G1del mice showed a reduction of GlyT2 mRNA, which was due to both reduced numbers of GlyT2 mRNA-labeled cells and reduced staining intensities in remaining cells (Figs. 3A-C). Quantitative analysis indicated a 50% loss of GlyT2 mRNA-labeled cells in lumbar L3-L5 sections of 30-week-old G1del mice compared with controls and 20-week-old G1del mice (Fig. 3G).
GAD67 mRNA labeling was most prominent in dorsal lamina of the spinal cord with a high density of small intensely stained cells in Rexed laminae II-III (Fig. 3D). Labeled cells in other lamina usually were less intensely detected, whereas a very low number of GAD67 mRNA-positive cells were present in the motor columns. As with GlyT2 mRNA, there was reduced GAD67 mRNA in 30-week-old and end-stage G1del mice (Fig. 3F). Neurons in superficial dorsal horn seemed relatively spared. Quantitative analysis indicated 30% to 40% loss of labeled cells in lumbar L3-L5 sections of 30-week-old G1del mice versus control and 20-week-old G1del mice (Fig. 3G).
ATF3 Expression and Ubiquitinated Aggregates in Inhibitory Spinal Interneurons
To determine whether ATF3 was associated with inhibitory interneurons, we combined fluorescent ISH of GlyT2 or GAD67 mRNA with ATF3 immunofluorescence. Double labeling with ATF3 revealed multiple GlyT2 mRNA/ATF3- and GAD67 mRNA/ATF3 double-labeled neurons in the intermediate zone and deep dorsal horn of 30-week-old and end-stage G1del mice (Figs. 4A-D). Systematic analysis of lumbar L3-L5 sections of 30-week-old G1del mice indicated that 27% ± 5% (mean ± SE, n = 3 animals, 30-40 ATF3 cells analyzed per mouse) of ATF3-positive cells were GlyT2 mRNA-positive, whereas 33% ± 5% were Gad67 mRNA-positive. A subset of double-labeled neurons showed eccentric nuclei (Figs. 4A, C, D) (suggesting pathologic alteration), consistent with previous observations (45).
To determine whether spinal inhibitory interneurons also develop ubiquitinated SOD1 aggregates, we double-stained for GlyT2 or GAD67 mRNA and polyubiquitinated epitopes. Multiple GlyT2 or GAD67 mRNA-stained neurons containing ubiquitin-immunoreactive structures were identified in the spinal cords of symptomatic G1del mice (Figs. 4E-G). The mRNA staining did not overlap with that of ubiquitin. In lumbar L3-L5 sections (2 mice, 3 sections per mouse), approximately 50% (14/26) of intermediate zone neurons with ubiquitin immunopositivity also stained for GlyT2 mRNA. Similarly, in GAD67 mRNA/ubiquitin-stained sections, approximately 50% (12/27) of ubiquitinated neurons stained for GAD67 mRNA. These data indicate that G1del mice show loss of spinal inhibitory interneurons, which is preceded by ATF3 expression and the appearance of ubiquitinated aggregates, 2 key pathological hallmarks that also precede motor neuron loss.
Early Dendritic Ubiquitin Pathology in Inhibitory Interneurons in G1del Spinal Cord
Ubiquitinated aggregates in motor neurons of G1del mice occur initially and more frequently in dendrites than in the cell soma (34, 45). These aggregates are strongly immunoreactive for human SOD1 and first appear at 13 to 15 weeks and only in the motor columns (45). Double labeling with ChAT showed that at 15 weeks, essentially all ubiquitinated dendrites were immunoreactive for ChAT (Figs. 5A, B). However, a proportion of ubiquitinated dendrites was ChAT-negative and also occurred in the intermediate zone in 20-week-old and older mice (Fig. 5A). Thus, ubiquitin pathology also occurred in dendrites of non-motor neurons in older mice.
To determine whether non-motor neuron dendritic ubiquitin pathology also occurs in the dendrites of inhibitory interneurons, we double stained for ubiquitin and parvalbumin (Figs. 5A, C). Parvalbumin, a small calcium binding protein, was used as a substitute marker instead of GlyT2 or GAD65 because staining for GlyT2 and GAD67 protein or mRNA does not outline the dendritic compartment of interneurons. Parvalbumin is reported to be present in the somatodendritic compartment of a subset of inhibitory interneurons that project to motor neurons (57). Accordingly, we obtained evidence in GlyT2-GFP transgenic mice that more than 90% of parvalbumin-positive neurons in the intermediate zone and ventral horn of spinal cord are glycinergic (47) (Figure, Supplemental Digital Content 1, ). Double labeling for ubiquitin and parvalbumin showed that at 20 weeks, but not at 15 weeks, a subset of ubiquitinated neurites were parvalbumin positive (Figs. 5A, C). Parvalbumin-positive ubiquitinated dendrites were preferentially localized in the intermediate zone, although sometimes they occurred in the motor columns. Triple staining for ubiquitin, ChAT, and parvalbumin showed that parvalbumin was only present in a subset of ChAT-negative ubiquitinated dendrites, indicating the presence of ubiquitin in dendrites of other populations of spinal interneurons. Of note, we never observed ChAT and parvalbumin double-stained neurons and neurites, thereby excluding the possibility that parvalbumin-positive dendrites were from motor neurons.
Early Microglial Activation in the Intermediate Zone in G1del Mice
There was a low density of activated microglia, identified using an antibody against Mac2 (58), in the motor columns of 15-week-old G1del mice, whereas higher densities of Mac2-positive microglia were seen in the motor columns at older ages (Figs. 5E-H). Activated microglia appeared in the intermediate zone between 20 and 25 weeks. Activated microglia in motor columns were usually in close apposition with motor neurons. Sections from 25-week-old G1del mice stained for both Mac2 and parvalbumin revealed multiple examples of Mac2-positive microglia contacting parvalbumin-positive interneurons (Fig. 5G). Together, these data further support the relatively early pathological alterations of spinal interneurons, including a subclass of parvalbumin-positive cells.
Interneuron Abnormalities in Neuron-Specific G93A-SOD1 Mice
To determine whether neuron-specific mutant SOD1 expression also is sufficient to trigger interneuron abnormalities, we examined homozygous T3T3 mice that express G93A-mutant SOD1 in neurons throughout the CNS, including spinal motor neurons and interneurons (34). Hemizygote T3 mice do not develop clinical and pathological signs of motor abnormalities up to 2 years, whereas homozygous T3T3 mice develop a motor neuron disease strongly resembling that in G1del mice. The age of onset in T3T3 mice is higher (>54 weeks) and considerably more variable (54 to >104 weeks) than in G1del mice, likely because of lower mutant SOD1 expression levels in motor neurons (34). ATF3 immunohistochemistry revealed ATF3-immunoreactive motor neurons and interneurons in spinal cord of symptomatic (n = 2) and end-stage (n = 5) T3T3 mice (Figs. 5A, E, I). ATF3 staining was also observed in the spinal cord of 2-year-old presymptomatic T3T3 mice (n = 2) but not in 40-week-old T3T3 mice (n = 4). Hemizygote T3 mice and nontransgenic mice showed ATF3 staining in some motor neurons at 2 years (Fig. 6) but not at 40 and 70 weeks (not shown). Old T3 and nontransgenic mice did not show ATF3-positive interneurons (Fig. 6).
Immunostaining with an anti-phospho (ser63)-c-Jun antibody resulted in motor neuron and interneuron staining patterns that strongly resembled those obtained with anti-ATF3 antibody. Thus, phospho-c-Jun-positive motor neurons and interneurons occurred in spinal cord of 2-year-old, symptomatic, and end-stage T3T3 mice, whereas no or sporadic-labeled motor neurons occurred in spinal cord of nontransgenic, T3, and 40-week-old T3T3 mice (Figs. 6B, F). Immunohistochemistry showed the presence of intensely ubiquitin-immunoreactive dendrites and cell bodies in the spinal cord of symptomatic T3T3 mice and 2-year-old but not 40-week-old T3T3 presymptomatic mice. Ubiquitinated dendrites were in both the ventral horn and the intermediate zone, indicating that the ubiquitinated aggregates were in interneurons in addition to motor neurons (Fig. 6G). Direct evidence for death of interneurons in T3T3 mice was obtained by reanalyzing sections stained with a silver degeneration staining method that outlines dying neurons and their processes (34). This analysis revealed multiple argyrophilic interneurons (Figure, Supplemental Digital Content 2, ).
Symptomatic T3T3 mice showed reduced levels of GlyT2 immunoreactivity in the ventral horn but not in the dorsal horn (Fig. 6H). No changes in GlyT2 immunoreactivity occurred in spinal cord of 40-week-old T3T3 mice or T3 mice up to 2 years, indicating that reduced GlyT2 immunoreactivity correlated with the occurrence of other degenerative changes and clinical manifestations.
As in G1del mice, all symptomatic T3T3 mice showed ubiquitin-parvalbumin double-labeled dendrites and cell bodies (Figs. 7A, B), as well as parvalbumin-immunoreactive neurons with nuclear ATF3 staining (Fig. 7D). We also combined ubiquitin with GlyT2 staining, which enabled us to examine whether motor neurons with ubiquitinated aggregates were contacted by glycinergic boutons (Figs. 7B, C). Most ubiquitin-positive motor neurons (15/17 identified in lumbar sections from symptomatic T3T3 mice) were surrounded by GlyT2-immunoreactive boutons (Fig. 7C), whereas ubiquitinated motor neurons in lumbar sections from 20- to 25-week-old G1del mice generally showed a normal pattern of GlyT2-immunoreactive boutons (not shown). Quantitative analyses to identify subtle losses of GlyT2-immunoreactive boutons (as previously demonstrated for high copy G1 mice [41) were beyond the scope of this study.
Dendritic Ubiquitin Pathology and Microglia Activation in the Intermediate Zone After the Ventral Horn in Neuron-Specific G93A Mice
As indicated above, no ubiquitin pathology was observed in spinal cord of 40-week-old T3T3 mice, whereas in presymptomatic and symptomatic T3T3 mice, ubiquitinated aggregates were always in both motor neurons and interneurons. Thus, it was not possible to determine whether ubiquitin pathology appeared at an earlier time point in motor neurons than in interneurons, as in G1del mice (Fig. 5A). To address this question, we used T3hSOD1 mice, which develop a considerably more predictable disease phenotype (34). The T3hSOD1 mice take advantage of the fact that coexpression of high levels of wild-type SOD1 (via a yet poorly understood mechanism) facilitates onset an progression of disease, and lowers the threshold of the concentration of mutant SOD1 required to cause disease within the normal lifespan of mice (48, 59, 60). T3hSOD1 mice develop signs of muscle weakness starting from 1 year of age, but dendrites with ubiquitinated aggregates appear before the age of 20 weeks, long before clinical onset (34). Double labeling of ChAT and ubiquitin showed that the large majority of these ubiquitinated dendrites were cholinergic (Figs. 7E, G), whereas only a few dendrites were ChAT-negative (Figs. 7F, G). A higher level of ChAT-negative ubiquitinated dendrites occurred in a later presymptomatic stage at 35 weeks (Fig. 7G). These data indicate that in T3 neuron-specific G93A mice (as in G1del mice), dendritic ubiquitin pathology occurs first only in motor neurons, and later, albeit still in a presymptomatic stage, also in spinal interneurons. Similarly, as in G1del mice, neuron-specific G93A mice show microglia activation first in the motor columns and subsequently in the intermediate zone (Figs. 7H-K). Taken together, these data indicate that neuron-specific and ubiquitous G93A-SOD1 mice develop similar interneuronal pathological features, specifically after the onset of the disease in motor neurons.
We have shown that SOD1-ALS transgenic mice expressing G93A mutant SOD1 lose the markers of glycinergic and GABAergic inhibitory neurons GlyT2 and GAD65/67 mRNA and protein in the spinal cord, indicating their involvement in the disease. These interneurons develop 2 key pathological hallmarks that also characterize degenerating motor neurons: ubiquitinated inclusions and expression of the stress transcription factor ATF3. These findings complement previous studies showing the degeneration of spinal interneurons in sporadic ALS patients and in transgenic SOD1-ALS mice (17, 38, 39, 61). We also show that ubiquitin pathology appears first in motor neurons (before 15 weeks of age), and subsequently in interneurons (onset between 15 and 20 weeks of age). This is consistent with the sequential expression of ATF3 in motor neurons and interneurons in G93A-SOD1 mice (45).
Because ubiquitin aggregates occur first and more frequently in dendrites rather than in the cell soma of neurons (34, 45) and because GlyT2 and GAD67 mRNA and protein are not present in dendrites, we used parvalbumin as a substitute marker for the dendrites of inhibitory interneurons. Although there is compelling evidence that parvalbumin in spinal cord intermediate zone is predominantly associated with inhibitory neurons (Figure, Supplemental Digital Content 1, ) (57), it cannot be excluded that some parvalbumin-immunoreactive dendrites in the intermediate zone belong to excitatory neurons. Nevertheless, our data show that in 15-week-old G1del mice, ubiquitin staining is only associated with motor neuronal somata and dendrites (identified by ChAT staining),while at 20 weeks, approximately half of the ubiquitin-stained dendrites are ChAT-negative and are located in the intermediate zone, indicating that they are not of motor neurons. In view of our data that at least 50% of the ubiquitin-immunoreactive intermediate zone neurons express GlyT2 or GAD67 mRNA in a later phase of disease, it is likely that a significant proportion of the non-motor neuron ubiquitinated dendrites at 20 weeks belong to inhibitory interneurons.
The onset of degenerative changes in interneurons precedes the onset of behavioral motor manifestations and most of the motor neuron degeneration (45), raising the question as to whether interneuron degeneration contributes to further degeneration and motor signs. Our data indicate that the overall level of GlyT2 and GAD immunoreactivity in the ventral horn is unaltered at 20 weeks, but subtle changes at inhibitory synapses on individual motor neurons cannot be excluded. Several studies suggest that changes in synaptic inputs to motor neurons contribute to motor neuron degeneration and motor manifestations of ALS (42, 62-65). Two recent studies also indicate that there are subtle losses of inhibitory synapses innervating motor neurons in presymptomatic high-copy G1 mice (41, 43), and compensatory sprouting of inhibitory glycinergic axons may also occur (41). These and our present data support the notion that early degenerative changes in inhibitory interneurons may contribute to the degeneration of motor neurons in SOD1-ALS mice, for example, by facilitating excitotoxic stress, one of the potential factors contributing to motor neuron degeneration in SOD1-ALS (28, 37, 42).
Ubiquitin pathology and ATF3 expression appeared in interneurons several weeks after their appearance in motor neurons. This raises the possibility that interneuron degenerative changes somehow are a consequence of motor neuron degeneration. A linkage between interneuron and motor neuron degeneration is also suggested by the fact that GlyT2 and GAD67 immunoreactivity is specifically reduced in the ventral horn, indicating the selective involvement of inhibitory interneurons innervating motor neurons. Longitudinal analyses of muscle denervation in SOD1-ALS mice have shown that large motor neurons that innervate type IIB muscle fibers and form the forceful fast-fatigable (FF) units degenerate before clinical onset (66, 67). The time of appearance of ubiquitin pathology and the onset of ATF3 expression (at 13-15 weeks) coincide with the onset of early molecular changes in FF motor neurons, as identified in gene profiling experiments of motor neuron subtypes (13). Instead, fast fatigue-resistant and slow motor neurons become involved in later phase of disease (13, 66, 67). Together, these data indicate that the disease in SOD1-ALS mice starts in FF motor neurons and subsequently involves other motor neurons and at the same time spinal interneurons.
The appearance of interneuron abnormalities after the onset of motor neuron degeneration, combined with evidence that mutant SOD1 in glia contributes to disease progression after initiation of disease in motor neurons, may indicate that mutant SOD1 in glia plays an important role of the spreading of disease to interneurons. However, we found that neuron-specific G93A-SOD1 mice develop the same interneuronal abnormalities as in ubiquitous mutant SOD1-expressing G1del mice, indicating that glial mutant SOD1 expression is dispensable for triggering spinal interneuron degeneration. One proposed glial mechanism is that microglial activation triggered by motor neuron degeneration could cause or facilitate the degeneration of other neurons (28, 68). Some evidence suggests that toxic microglial actions requires microglial mutant SOD1 expression (31, 69), but mutant SOD1-independent actions of activated microglia may also occur; indeed, deleterious actions of microglia have been proposed for multiple neurodegenerative conditions (70). However, a major role of activated microglia in triggering early interneuron alterations in SOD1-ALS mice seems unlikely in view of our data that microglial activation is still limited at the time of the first appearance of interneuron dendritic ubiquitin pathology (i.e. between 15 and 20 weeks in ubiquitous G1del mice and 20 and 35 weeks in neuron-specific T3hSOD1 mice).
An alternative mechanism that could explain the appearance of interneuronal pathology subsequent to motor neuronal pathology is transcellular transmission of protein aggregation by seeds of aggregated species (71). Recent evidence indicates that such a mechanism may contribute to the spreading of pathology in neuronal protein aggregation disorders, including synucleinopathies and tauopathies (71). Furthermore, seeding-like properties have been demonstrated for mutant SOD1 (72). Hence, a possible scenario that is compatible with our data would be that mutant SOD1 aggregates released by degenerating motor neurons are taken up by interneuronal nerve endings and retrogradely transported to their parent cell to trigger or facilitate further SOD1 aggregation leading to the ubiquitinated inclusions and, eventually, cell death.
The availability of SOD1 transgenic mice models has enabled the precise dissection of the pathological progression of disease, indicating the involvement of different cell types at different disease stages as well as different cell autonomous and non-cell autonomous disease mechanisms. A central question is whether similar mechanisms also operate in sporadic ALS, which is predominantly characterized by TDP43 aggregates. The recent availability of new mouse models, such as TDP43 transgenic mice that reproduce aspects of this TDP43 pathology (73-76), may help resolve this question and further unravel the disease mechanisms underlying ALS.
The authors thank Drs N. Nelson and N.J.K. Tillakaratne for GlyT2 and GAD67 cDNA constructs and Dr H.U. Zeilhofer for providing GlyT2-GFP transgenic mice.