Growing evidence suggests that amyotrophic lateral sclerosis (ALS) is a multisystem neurodegenerative disease that primarily affects motor neurons and, though less evidently, other neuronal systems. About 75% of sporadic and familial ALS patients show a subclinical degeneration of small-diameter fibers, as measured by loss of intraepidermal nerve fibers (IENFs), but the underlying biological causes are unknown. Small-diameter fibers are derived from small-diameter sensory neurons, located in dorsal root ganglia (DRG), whose biochemical hallmark is the expression of type III intermediate filament peripherin. We tested here the hypothesis that small-diameter DRG neurons of ALS mouse model SOD1G93A suffer from axonal stress and investigated the underlying molecular mechanism. We found that SOD1G93A mice display small fiber pathology, as measured by IENF loss, which precedes the onset of the disease. In vitro small-diameter DRG neurons of SOD1G93A mice show axonal stress features and accumulation of a peripherin splice variant, named peripherin56, which causes axonal stress through disassembling light and medium neurofilament subunits (NFL and NFM, respectively). Our findings first demonstrate that small-diameter DRG neurons of the ALS mouse model SOD1G93A display axonal stress in vitro and in vivo, thus sustaining the hypothesis that the effects of ALS disease spread beyond motor neurons. These results suggest a molecular mechanism for the small fiber pathology found in ALS patients. Finally, our data agree with previous findings, suggesting a key role of peripherin in the ALS pathogenesis, thus highlighting that DRG neurons mirror some dysfunctions found in motor neurons.
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized by the loss of motor neurons in the spinal cord, brainstem and cerebral cortex. It has long been believed that mainly motor neurons undergo dysfunction and degeneration in ALS, whereas sensory neurons localized in dorsal root ganglia (DRG) are spared. However, accruing evidences have highlighted the abnormalities of sensory neurons in ALS mouse models and subclinical abnormalities of sensory nerve fibers in ALS patients (1–6).
On the basis of morphology, sensory nerve fibers are classified into small fibers (axonal diameters 0.5–4 μm) and large fibers (12–20 μm). Small fibers include unmyelinated C fibers and thinly myelinated Aδ fibers, derived from small-diameter DRG neurons (<30 µm), and terminate as free nerve endings in muscle, skin and viscera. Large myelinated fibers derive from large-diameter DRG neurons (>30 µm) and have peripheral branches encapsulated by accessory structures that modify sensory transduction (7).
Studies on ALS mouse models have disclosed swollen axons in the dorsal roots deriving from DRG neurons (8), and accumulation of misfolded SOD1 in DRG neurons (9). Clinical studies on sporadic and familial ALS patients have detected rare impairment of large sensory fibers (1–3) versus a wide degeneration of small fibers, as showed by the loss of intraepidermal nerve fibers (IENFs) in up to 75% of ALS patients including patients carrying SOD1 gene mutations (4,6). These findings suggest that, besides motor neurons, ALS primarily affect small-diameter DRG neurons. However, the underlying molecular mechanisms have never been investigated yet.
We tested here the hypothesis that SOD1 gene mutations in mouse model SOD1G93A cause small nerve fiber pathology by inducing axonal stress in small-diameter DRG neurons. Because axonal stress stems from abnormalities of neurofilaments (NFs) (10) and NF abnormalities are widely detected in spinal motor neurons of ALS mouse models and in familial and sporadic ALS patients (11–17), we hypothesized that small-diameter DRG neurons of SOD1G93A mice display NF abnormalities.
Because the NF subunit peripherin is selectively expressed in spinal motor neurons and in small-diameter DRG neurons (18–21) and because mutant SOD1 modifies peripherin isoform levels and assembly in the NFs of motor neurons thus inducing neurotoxic effects (11,17,22–24), we hypothesized that a similar mechanism occurs in small-diameter DRG neurons. We tested the hypothesis that peripherin levels and assembly are dysregulated in small-diameter DRG neurons of SOD1G93A mouse model and that peripherin dysregulation may underlie axonal stress through NF disassembly.
Decreased IENF density in SOD1G93A mouse models
To test the hypothesis that SOD1 gene mutations in mouse model SOD1G93A cause small nerve fiber pathology, we analyzed intraepidermal nerve fiber density (IENFD) in skin biopsies obtained at pre-symptomatic and symptomatic stages of two ALS mouse models expressing SOD1G93A in two different strains (B6.SJL.SOD1G93A and B6.Cg.SOD1G93A). IENFD did not statistically differ between early pre-symptomatic non-transgenic (NTg) and SOD1G93A mice (Supplementary Material, Fig. S1), but significantly decreased in both the SOD1G93A mouse models at pre-symptomatic and symptomatic stages (IENFD mean ± SEM at 56 days: B6.SJL.NTg 25.27 ± 1.94, n = 8; B6.SJL.SOD1G93A 15.11 ± 1.18, n = 8, two-tailed unpaired t-test, t = 4.469, df = 14, ***P = 0.0005; IENFD mean ± SEM at 126 days: B6.SJL.NTg 30.50 ± 1.94, n = 7; B6.SJL.SOD1G93A 18.82 ± 1.70, n = 6, two-tailed unpaired t-test, t = 4.446, df = 11, ***P = 0.0010; IENF density mean ± SEM at 67 days: B6.Cg.NTg 28.15 ± 2.43, n = 4; B6.Cg.SOD1G93A 16.30 ± 0.94, n = 4, two-tailed unpaired t-test, t = 4.552, df = 6, **P = 0.0039; IENFD mean ± SEM at 115 days: B6.Cg.NTg 35.25 ± 1.05, n = 6; B6.Cg.SOD1G93A 27.85 ± 1.58, n = 6, two-tailed unpaired t-test, t = 3.908, df = 10, **P = 0.0029, Fig. 1).
These results first show a loss of small sensory fibers in SOD1G93A mouse models. These results also show that the loss of small sensory fibers precedes the disease onset.
DRG neurons of SOD1G93A mouse models display axonal stress features
IENF are the terminal ending of small-diameter DRG neuron neurites. Because SOD1 protein is expressed in DRG neurons (25), we hypothesized that mutant SOD1 expression damages DRG neurons and that IENF loss in SOD1G93A mice results from either loss of small DRG neurons or early axonal stress associated to neurite length decrease. We, first, confirmed that DRG neurons of SOD1G93A mice express hSOD1G93A protein (Supplementary Material, Fig. S2a and b) and then performed small DRG neuron count. The number of small DRG neurons did not differ between symptomatic SOD1G93A mice and NTg littermates (Supplementary Material, Fig. S3); these data showed that SOD1G93A expression does not elicit a loss of small DRG neurons. We then tested the axonal stress hypothesis. We analyzed the levels of activating transcription factor 3 (ATF3), a transcription factors whose levels rapidly increase upon axonal stress (26–28), in in vitro DRG neurons prepared from SOD1G93A mice at pre-symptomatic and symptomatic stages. ATF3 levels were higher in the DRG neurons of SOD1G93A compared with age-matched NTg at both pre-symptomatic and symptomatic stages (normalized ATF3 intensity ± SEM at 56 days: B6.SJL.NTg 1.00 ± 0.11, B6.SJL.SOD1G93A 1.25 ± 0.04, two-tailed unpaired t-test *P = 0.0183, t = 2.487, df = 32; at 126 days, B6.SJL.NTg 1.00 ± 0.08, B6.SJL.SOD1G93A 1.54 ± 0.09, two-tailed unpaired t-test ***P = 0.0003, t = 4.272, df = 21; Fig. 2A). ATF3 increase in DRG tissues of SOD1G93A mice was also confirmed by gene expression analysis (symptomatic mice at 84–126 days, B6.SJL.NTg 1.00 ± 0.19, B6.SJL.SOD1G93A 2.97 ± 0.63, two-tailed unpaired t-test *P = 0.0391, t = 2.463, df = 8; Supplementary Material, Fig. S4). To confirm that SOD1G93A DRG neurons display an early axonal damage, we performed a time-course analysis of neurite length from pre-symptomatic to symptomatic stage. At all ages, neurite lengths of in vitro SOD1G93A DRG neurons were significantly lower compared with neurite length of NTg neurons (normalized neurite length ± SEM at 56 days: B6.SJL.NTg 1.00 ± 0.06, B6.SJL.SOD1G93A 0.86 ± 0.03, two-tailed unpaired t-test *P = 0.0287, t = 2.190, df = 1217; at 70 days: B6.SJL.NTg 1.00 ± 0.05, B6.SJL.SOD1G93A 0.71 ± 0.06, two-tailed unpaired t-test ***P = 0.0003, t = 3.651, df = 623; at 84 days: B6.SJL.NTg 1.00 ± 0.03, B6.SJL.SOD1G93A 0.81 ± 0.03, two-tailed unpaired t-test ***P < 0.0001, t = 4.150, df = 2100; at 105 days, B6.SJL.NTg 1.00 ± 0.05, B6.SJL.SOD1G93A 0.72 ± 0.03, two-tailed unpaired t-test ***P < 0.0001, t = 4.696, df = 1386; at 126 days, B6.SJL.NTg 1.00 ± 0.05, B6.SJL.SOD1G93A 0.79 ± 0.04, two-tailed unpaired t-test **P = 0.0018, t = 3.135, df = 791; Fig. 2B). To confirm that axonal stress features are a result of mutant SOD1 actions and not due to high-SOD1 reductase activity, we transfected in vitro DRG neurons with plasmids encoding wild-type SOD1 or SOD1G93A and analyzed ATF3 levels and neurite lengths. We found that SOD1G93A increased ATF3 levels and decreased neurite length while wild-type SOD1 left the two parameters unchanged (Supplementary Material, Fig. S5). Overall, these results show that DRG neurons of SOD1G93A mice display early axonal stress features and suggest that this phenomenon is the result of mutant SOD1 expression.
Small-diameter DRG neurons of SOD1G93A mouse models selectively display axonal stress features
Clinical studies on sporadic and familial ALS patients have detected a wide degeneration of small fibers versus rare impairment of large fibers (4–6). To test whether these clinical observations derive from a selective dysfunction of small-diameter DRG neurons, we analyzed axonal stress features in in vitro small and large DRG neurons obtained from SOD1G93A mice. Adult small-diameter DRG neurons (diameter <30 µm; ∼70% of total neuron population) are characterized by peripherin expression whereas large-diameter DRG neurons (diameter >30 µm; ∼30% of total neuron population) are characterized by NFH expression (29,30). We in vitro distinguished the two different neuron subpopulation according to cell diameter and found that SOD1G93A small DRG neurons displayed increased ATF3 labeling whereas large SOD1G93A DRG neurons did not (normalized ATF3 intensity in neurons with diameter <30 µm ± SEM: small B6.SJL.NTg neurons 1.00 ± 0.20 versus small B6.SJL.SOD1G93A neurons 2.00 ± 0.10, two-tailed unpaired t-test ***P = 0.0007, t = 4.662 df = 11; small B6.Cg.NTg neurons 1.00 ± 0.06 versus small B6.Cg.SOD1G93A neurons 1.23 ± 0.08, two-tailed unpaired t-test *P = 0.0260, t = 2.245 df = 171; normalized ATF3 intensity neuron with diameter >30 µm ± SEM: large B6.SJL.NTg neurons 1.00 ± 0.09 versus large B6.SJL.SOD1G93A neurons 1.40 ± 0.14 two-tailed unpaired t-test P > 0.05; large B6.Cg.NTg neurons 1.00 ± 0.15 versus large B6.Cg.SOD1G93A neurons 0.84 ± 0.12, two-tailed unpaired t-test P > 0.05; Fig. 3A).
SOD1G93A small DRG neurons identified according to cell size or peripherin immunoreactivity had reduced neurite length compared with small NTg neurons (normalized neurite length of neurons with diameter <30 µm ± SEM at 126 days, B6.SJL.NTg 1.00 ± 0.06 versus B6.SJL.SOD1G93A 0.78 ± 0.06, two-tailed unpaired t-test *P = 0.012, t = 2.541, df = 147; at 115 days, B6.Cg.NTg 1.00 ± 0.08 versus B6.Cg.SOD1G93A 0.59 ± 0.06, two-tailed unpaired t-test ***P = 0.0002, t = 3.944, df = 95, Fig. 3B; normalized neurite length of peripherin-positive neurons ± SEM at 126 days, B6.SJL.NTg 1.00 ± 0.10 versus B6.SJL.SOD1G93A 0.67 ± 0.05, two-tailed unpaired t-test **P = 0.0045, t = 2.928, df = 73, Fig. 3C). These results were further confirmed by neurite length analyses of DRG neurons positive to isolectin IB4 that labels a subclass of small DRG neurons (31) (Supplementary Material, Fig. S6). Conversely, large SOD1G93A DRG neurons had equal neurite length compared with large NTg DRG neurons (Fig. 3D).
Hence, our experiments highlight that only small-diameter DRG neurons of SOD1G93A mouse models display increased ATF3 labeling and neurite shortening. Although our data cannot rule out other abnormalities in SOD1G93A large-diameter DRG neurons, these results suggest that small DRG neurons are damaged more severely by SOD1G93A expression than large DRG neurons.
Peripherin56 accumulation in small-diameter DRG neurons of SOD1G93A mouse model
Above described results suggested that SOD1G93A induces axonal stress in DRG neurons through a molecular mechanism that is specifically activated in small-diameter DRG neurons. The biochemical hallmark of small-diameter DRG neurons is the expression of the type III intermediate filament peripherin that is conversely not expressed in large-diameter DRG neurons (29,32).
Because previous studies showed that mutant SOD1 modifies peripherin levels and assembly in the NFs of motor neurons (17,22,23) we hypothesized that a similar mechanism occurs in small-diameter DRG neurons. Mouse peripherin gene encodes three mRNA alternative splice variants that lead to three translation products: peripherin58, the most expressed isoform (475 amino acids, predicted 58 KDa), peripherin61 (507 amino acids, predicted 61 KDa) and peripherin56 (461 amino acids, predicted 56 KDa) (33,34) (Fig. 4A). We analyzed peripherin isoform levels in the assembled fraction of NFs (triton X-100 insoluble fraction) and in non-assembled fraction of NFs (triton X-100 soluble fraction) (35) of DRG tissues of SOD1G93A and littermate NTg by using two previously described antibodies (i.e. per56, per61) specific for the peripherin56 and 61 isoforms (23) and one antibody recognizing all the peripherin isoforms (i.e. per56/58/61). Peripherin61 was not expressed in DRG tissues (Fig. 4A). Peripherin56 was present at low levels and only in assembled NF fraction of NTg mice whereas in SOD1G93A mice peripherin56 was clearly detectable at higher levels in the non-assembled NFs (ratio of intensity × area normalized to actin mean ± SEM in insoluble fraction: B6.SJL.NTg 1.00 ± 0.50, n = 3 versus B6.SJL.SOD1G93A 0.92 ± 0.21, n = 3, two-tailed unpaired t-test, P > 0.05; mean ± SEM in soluble fraction: B6.SJL.NTg 1.00 ± 0.53, n = 3 versus B6.SJL.SOD1G93A 13.77 ± 1.79, n = 3, two-tailed unpaired t-test, **P = 0.0024; Fig. 4A). Peripherin56 antibody specificity was verified on lysates of HEK293T cells transfected with plasmids encoding peripherin isoforms (Supplementary Material, Fig. S7). This result suggested that SOD1G93A expression in DRG neurons induces peripherin56 to accumulate in the non-assembled fraction of NFs. To confirm these data, we analyzed peripherin56 levels in in vitro small-diameter DRG neurons (diameter <30 µm). Peripherin56 levels were higher in small-diameter neurons of SOD1G93A DRG mice than NTg DRG littermate (peripherin56 fluorescence intensity mean ± SEM: B6.SJL.NTg 1.00 ± 0.07 versus B6.SJL.SOD1G93A 2.51 ± 0.16, two-tailed unpaired t-test, ***P < 0.0001, t = 8.139 df = 105; Fig. 4B). To confirm this result, we analyzed peripherin56 labeling in DRG tissues of symptomatic SOD1G93Amice. The percentage of peripherin56-positive neurons in SOD1G93A mice was significantly higher than in NTg littermate (mean ± SEM ratio peripherin56 positive/ peripherin positive cells: B6.Cg.NTg 24.32 ± 3.47, n = 16 image analyzed from n = 3 mice versus B6.Cg.SOD1G93A 41.90 ± 3.47, n = 21 image analyzed from n = 3 mice, two-tailed unpaired t-test **P = 0.0012, t = 3.516, df = 35; Fig. 4C). Peripherin56 was expressed in small-diameter DRG neurons and not expressed in large-diameter neurons (Supplementary Material, Fig. S8). Higher magnification images suggested that while antibody recognizing all the peripherin isoforms labeled extensive filament networks, per56 labeling was associated with non-filamentous structures in the soma of SOD1G93A DRG (Fig. 4D). Hence, these results showed that small-diameter SOD1G93A DRG neurons display peripherin56 accumulation in the non-assembled fraction of NFs.
Peripherin56 causes axonal stress by disassembling NFs
Based on previous observations that peripherin alternative splice variants are neurotoxic to motor neurons (23), we hypothesized that peripherin56 damages axons of DRG neurons. We transfected DRG neurons with plasmids encoding the splice variants peripherin58 or peripherin56 (23) and found that peripherin56 significantly decreased neurite length (absolute neurite length ± SEM: green fluorescence protein (GFP): 183.40 ± 12.53 µm, peripherin58: 238.00 ± 23.44 µm, peripherin56: 134.80 ± 13.10 µm, Kruskal–Wallis test and Dunn's multiple comparison test, ***P < 0.001 per58 versus per56, *P < 0.05 per56 versus GFP; Fig. 5A). Neurite length maintenance requires a proper assembly of NFs (36–38) that depends on the interactions between the C-terminal tails of the NF subunits (39,40). In small-diameter DRG neurons, NFs are formed by the subunits peripherin, NFM and NFL (38,41). Because peripherin56 differs from the main isoform peripherin58 in the C-terminal tails (33,34) (Fig. 4A), and C-terminal-deleted peripherin mutants can disrupt the endogenous NF network (42), we hypothesized that peripherin56 decreases DRG neurite length by interfering with the normal NF assembly. To test this hypothesis, we transfected F11 cells, a DRG-derived line, with plasmids encoding peripherin56 or 58 and analyzed NF assembly by quantifying NFL and NFM levels in assembled NFs (triton X-100 insoluble fractions). We found that peripherin56 overexpression significantly decreased NFM and NFL levels in the assembled NFs fraction (ratio of intensity × area NFM/actin in insoluble fraction: GFP 1.00 ± 0.09 versus per56 0.62 ± 0.05 versus per58 1.27 ± 0.05; ratio of intensity × area NFL/actin in insoluble fraction: GFP 1.00 ± 0.09 versus per56 0.44 ± 0.06 versus per58 0.80 ± 0.06, one-way analysis of variance (ANOVA) and Tukey test, *P < 0.05; **P < 0.01; Fig. 5B). To test whether the same phenomenon occurs in DRG neurons of SOD1G93A mice, we analyzed NFM and NFL levels in assembled NFs (triton-insoluble fractions) and non-assembled NFs (triton X-100 soluble fractions) fractions prepared from DRG tissues of SOD1G93A mice and NTg littermate. NFM and NFL levels were significantly decreased in the assembled NF fractions of SOD1G93A mice compared with NTg littermate (ratio of intensity × area NFM/actin in insoluble fraction: NTg 1.00 ± 0.25 versus SOD1G93A 0.38 ± 0.09, two-tailed unpaired t-test, *P = 0.0283, t = 2.523 df = 11; ratio of intensity × area NFL/actin in insoluble fraction: NTg 1.00 ± 0.05 versus SOD1G93A 0.58 ± 0.09, two-tailed unpaired t-test, **P = 0.0013; Fig. 5C). These results confirmed that NFM and NFL subunit assembly in NFs is impaired in SOD1G93A DRG neurons.
This paper provides a new knowledge in the field of ALS through the following three novelties: (i) SOD1G93A mouse models display a loss of IENF that precedes the disease onset (Fig. 1); (ii) small-diameter DRG neurons of SOD1G93A mouse models display early axonal stress (Figs. 2 and 3) and (iii) small-diameter DRG neurons of SOD1G93A mouse models display accumulation of peripherin56 and altered assembly of NFs (Figs. 4 and 5).
Showing a loss of IENF in SOD1G93A mouse models, our results agree with the previous data describing a loss of IENF in sporadic and familial ALS patients including patient carrying SOD1 mutations (4,6). The evidence of IENF loss before disease onset in SOD1G93A mouse models suggests that small fiber pathology is a feature of ALS pathological process rather than the result of an unspecific tissue degeneration.
Because IENFs are the terminal endings of somatic small-diameter DRG neurons (43), in vivo IENF loss is consistent with the axonal stress features detected in in vitro small-diameter DRG neurons of SOD1G93A mouse models. These results suggest that mutated SOD1 causes axonal degeneration in small-diameter DRG neurons and add to previous evidence suggesting that the degeneration in ALS is not specifically directed toward motor neurons, but involves other neurons as well (44,45).
We found that axonal stress features in SOD1G93A small-diameter DRG neurons are associated with the accumulation of the splice variant peripherin56. This result agrees with previous data showing that SOD1G93A expression modifies peripherin splicing (23). However, different from DRG neurons, in mouse motor neurons mutant SOD1 expression induces the expression of splice variant peripherin61 (23), whereas motor neurons of ALS patients up-regulate the human-specific isoform peripherin28 (17). Overall, this evidence shows that peripherin splicing is altered in ALS, and also suggests that the resultant peripherin splicing pattern may be modulated by, still unknown, species and cell-type-specific factors.
Similar to peripherin61 that assembles incompetent and is neurotoxic to motor neurons (23), peripherin56 accumulates in non-assembled fraction and damages DRG neurites. Where do the potential neurotoxic features of peripherin56 derive from? Compared with the main isoform peripherin58, peripherin56 is 14 amino acids shorter since the 21 C-terminal amino acids are replaced by a unique C-terminal end of eight amino acids (33,34) (Fig. 4A). It has been shown that C-terminal-deleted peripherin mutants disrupt the endogenous NF network (42). This evidence suggests that an intact peripherin tail is required to form a normal filamentous network and suggests that peripherin56 toxicity stems from C-terminal end truncation. This hypothesis is also supported by the evidence that a conserved motif within the tails of type III intermediate filaments accounts for the tails' self-association during late stages of filament assembly, maintains intermediate filament network architecture and prevents filament bundling (46).
In small-sized DRG neurons, peripherin co-assembles with NFL and NFM into a single-NFs network (41) whose assembly and network organization strongly depend on the interactions between the subunit C-terminal tails (39). Accordingly, peripherin56 overexpression may alter NF homeostasis by decreasing NFM and NFL assembly in DRG neurons. This hypothesis is supported by our experimental data showing that peripherin56 expression decreases neurite length and impairs NFM and NFL assembly in DRG-like cells and by our results showing decreased NFM and NFL levels in the assembled fraction of NFs prepared from DRG tissue of SOD1G93A mice. The altered NFL assembly observed in DRG-like cells mirrors the altered NFL assembly previously observed in spinal cord and sciatic nerve of SOD1 mouse models (47). Overall, the altered NF homeostasis may underlie the neurites shortening we observed in in vitro SOD1G93A DRG neurons and in skin biopsies of SOD1G93A DRG neurons.
Besides being a subunit of peripheral nerve NFs, peripherin is also expressed in the neurons of the CNS that have projections to the periphery, including spinal motor neurons (18–20,48). Because several lines of evidence support a role for peripherin in the etiology of ALS, we can hypothesize that missplicing of peripherin leads to axonal damage in both DRG neurons and motor neurons of ALS mouse model and ALS patients. Future experiments aimed at preventing axon degeneration by modulating peripherin splicing/expression are needed to provide direct proof for this hypothesis.
Hence, our results show that small-diameter DRG neurons are damaged in SOD1G93A mouse models and suggest that the underlying molecular mechanisms involve peripherin missplicing and altered homeostasis of NFs. These findings provide the biological bases for small fiber pathology detected in ALS patients and strengthen the hypothesis that ALS is a multisystem neurodegenerative disease.
Materials and Methods
We used two different ALS transgenic mouse strains: B6.SJL.SOD1G93A and B6.Cg.SOD1G93A. Both express ∼20 copies of the human gene SOD1 with a glycine to alanine substitution in position 93 (SOD1G93A) either on B6.SJL strain [B6.SJL-Tg(SOD1*G93A) 1Gur, Jackson Laboratories, here indicated as B6.SJL.SOD1G93A] or on C57BL/6J strain [B6.Cg-Tg(SOD1*G93A) 1Gur/J, Jackson Laboratories, indicated as B6.Cg.SOD1G93A]. Mice were maintained and bred at the animal house of the ‘Carlo Besta’ Neurological Institute (B6.SJL strain) and Mario Negri Institute (C57BL/6J strain) in compliance with institutional guidelines and international law (EU Directive 2010/63/EU EEC Council Directive 86/609, OJL 358, 1, December 12, 1987, NIH Guide for the Care and Use of Laboratory Animals, US National Research Council, 1996). The experiments were planned and conducted with the aim of minimizing the number of sacrificed animals. Genotyping was performed according to previous published protocols (49,50). All the experiments have been conducted on male mice.
Measure of IENF density
Hind paw footpad biopsies were collected at sacrifice using a 3 mm round punch and IENFD was quantified as previously described (51). Briefly, specimens were immediately fixed by immersion in 2% paraformaldehyde–lysine–periodate for 24 h at 4°C, cryoprotected overnight and serially cut with a standard cryostat (Bio-Optica) to obtain 20 μm sections for mice. Three sections from each sample were randomly selected and immunostained with rabbit polyclonal anti-protein gene product 9.5 antibody (PGP9.5, 1:1000, AbD Serotec, Kidlington, Oxfordshire, UK) followed with secondary antibody incubation (1:200, anti-rabbit Alexa Fluor 488 coniugated, Molecular Probes) using a free-floating protocol. In each section, the total number of PGP9.5-positive IENFs was counted under a fluorescence microscope at high magnification. Individual fibers were counted as they crossed the dermal–epidermal junction whereas secondary branching within the epidermis was excluded. The length of the epidermis was measured using a computerized system (Image-Pro Plus; Media Cybernetics, Inc., Silver Spring, MD, USA), and the linear density of IENFs was obtained. IENFs were counted by an examiner blinded to the experimental conditions.
DRG neuron counts
Three mice of either strain were used; four to six DRGs from each animal were excised. DRG were embedded in OCT and immediately frozen in liquid nitrogen. Samples were sectioned (8 µm) using a standard cryostat (Bio-Optica). Immunofluorescence was performed following standard protocols by using rabbit anti-peripherin and mouse NFH antibody as primary antibodies (peripherin ab4666 Abcam; NF-H 2836, Cell Signaling). Fluorescence images were acquired with the D-Eclipse C1 confocal microscope (Nikon, Tokyo, Japan) mounted on a light microscope Eclipse TE2000-E (Nikon) with sequential acquisition setting. Four DRG sections of each animal were randomly chosen for counting of the labeled neurons, and the numbers of peripherin-positive neurons were expressed as the percentage of the total number of neurons per section. Image acquisition and neuron counts were performed by investigators blinded to the experimental condition.
Spinal DRG were collected from NTg mice and SOD1G93A mice at pre-symptomatic and symptomatic stages, and processed using a modified Choi's method (52). DRG were enzymatically dissociated with 0.25% trypsin and collagenase 1% in the Leibovitz's L-15 medium (Leibovitz, Sigma-Aldrich, St Louis, MO, USA). Dissociated cells (mainly sensitive neurons and glial cells) were plated in 24-well plates on collagen-coated glass coverslips, pre-treated with poly-d-lysine (Sigma-Aldrich) with Dulbecco's modified Eagle's medium (DMEM) (Invitrogen), 10% fetal bovine serum, 10% F-12 (Invitrogen). Cells were incubated for 24 h with 5% CO2 at 37°C. After 24 h, the DRG plating media was removed and replaced with the DRG feeding media containing DMEM, 1 × B27 supplement (50× stock, Gibco Life Technologies, Monza, Italy), 2 mm GlutaMAX (Invitrogen) and penicillin–streptomycin mixture (1 U/l) (Invitrogen). Experiments were conducted on DRG neurons maintained in culture 4 days (DIV4).
Plasmids and transient transfections
Plasmids encoding peripherin56, 58 and 61 were previously described (23). DRG neurons were transfected by electroporation according to a previous described protocol (53). Briefly, freshly dissociated neurons, added with 1 μg of peripherin56 or 58 plus 1 μg of pGFP plasmid were electroporated using 4D Nucleofector X Unit (Lonza), pulse CU133. HEK293T cells and F11 cells were transfected using lipofectamine 3000 according to manufacturers' protocol.
Measurement of neurite lengths
Neurons were stained with antibody against neuronal class III β-tubulin (TUJ1) (1:1000, Covance) and neurite length analyzed as previously described (54,55). The longest identifiable neurites of at least 100 neurons in each condition were measured using an image analysis system (Image-Pro Plus, Media Cybernetics, Silver Spring, MD, USA) on fluorescence microscope-obtained images (Axiophot-Zeiss, Oberkochen, Germany). Experiments were repeated at least three times using cultures prepared on separate days.
Immunofluorescence assay on in vitro DRG and DRG tissues was performed using the following antibodies: class III β-tubulin (TUJ1) (1:1000, mms-435P Covance), ATF3 (1:200, 180842 Abcam), peripherin (1:1000, ab4666 Abcam), hSOD1 (1:500, 2770 Cell Signaling), NF-L (1:250, 2837 and 2835, Cell Signaling), NF-H (1:250, 2836, Cell Signaling), peripherin56 and 61 (rabbit polyclonal peripherin isoform-specific antiserum, 1:500 (23) and isolectin GS-IB4 (1:100, I21411, alexafluor 488 coniugate, Molecular Probes). Secondary antibodies were anti-mouse/anti-rabbit IgG conjugated to Alexa Fluor 488 (green) or Alexa Fluor 594 (red) (1:200; Molecular Probes). The following antibodies were used for western blot experiments: peripherin (1:10000, Ab4666 Abcam), NF-L (1:1000, 2837 Cell Signaling); NF-M (1:1000, 2838 Cell Signaling), peripherin56 and 61 (1:500), hSOD1 (1:1000, 2770 Cell Signaling), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (1:500, sc-25778 Santa Cruz,), β-actin (1:20000, A5316, Sigma) or tubulin (1:1500, T9026, Sigma).
Immunocytochemistry of cultured cells and immunohistochemistry on DRG tissues
In vitro DIV4 DRG neurons were fixed with 4% paraformaldehyde, permeabilized with 0.5% triton X-100, thoroughly washed and finally blocked with 1% bovine serum albumin (BSA) in phosphate-buffered saline. Cells were incubated first with primary antibody overnight at 4°C, thoroughly washed and then incubated with secondary antibody for 1 h. Coverslips were mounted with FluoroSave Reagent (Millipore, 345789). For immunohistochemistry assays, DRG were harvested at sacrifice, embedded in OCT and immediately frozen in liquid nitrogen. Samples were sectioned (8 µm) using a standard cryostat (Bio-Optica). Immunofluorescence was performed following standard protocols. Fluorescent images were acquired and quantified following a published protocol (56). Labeled neurons were chosen randomly for quantification from four coverslips from three independent experiments. Fluorescence images were acquired with D-Eclipse C1 confocal microscope (Nikon, Tokyo, Japan) mounted on a light microscope Eclipse TE2000-E (Nikon) with sequential acquisition setting. Morphometric and fluorescence intensity measurements were made with the Image-Pro Plus image analysis software. Image acquisition and morphometric quantification were performed by investigators blinded to the experimental condition.
Protein fractioning and immunoblotting
Cells or tissue samples were lysed in low salt triton X-100 extraction buffer [20 mm Tris–HCl pH 7.5, 150 mm NaCl, 1 mm ethylenediaminetetraacetic acid, 50 μm MG132, 1% (v/v) triton X-100] added with protease inhibitor cocktail (Roche, Switzerland) for 30 min on ice and sonicated. The triton X-100 soluble fractions were recovered as supernatant following centrifugation for 30 min at 16 000 × g. The triton X-100 insoluble pellets were solubilized in 2% (w/v) SDS with sonication and made to the equivalent volume of the soluble fraction with extraction buffer (57). Protein quantification was performed using bicinchoninic acid protein assay by comparison with BSA standards. Western blots were performed with Novex® NuPAGE® sodium dodecyl sulfate–polyacrylamide gel electrophoresis gels (Invitrogen).
Data were subjected to the normality test (Kolmogorov–Smirnov test) and equal variance test (Bartlett's test). Unpaired Student's t-test was used to compare two groups and one-way ANOVA followed by appropriate post hoc tests was used to compare multiple groups. Data that did not satisfy the assumptions of the parametric tests were analyzed using the Kruskal–Wallis test followed by Dunn's multiple comparisons test. Data are presented as mean ± standard error of the mean (SEM). Analyses were performed using the GraphPad Prism4 software. Figures were prepared using the GraphPad Prism4 and GraphPad Prism6 software.
This work has been partly supported by the Fondazione Italiana di Ricerca per la SLA – Sclerosi Laterale Amiotrofica (AriSLA) and the Italian Ministry of Health (RF-INN-2007-644440).
The authors thank Janice Robertson (Department of Laboratory Medicine and Pathobiology, University of Toronto) for providing antibodies against peripherin56 and 61 and plasmids encoding peripherin isoforms (23), and Bradley Turner (Florey Institute of Neuroscience and Mental Health, University of Melbourne for providing plasmids pEGFP-N1, pEGFP-N1-hSOD1 and pEGFP-N1-hSOD1G93A (58).
Conflict of Interest statement. None declared.