All animal procedures were carried out following the European guidelines for the use of animals in research (2010/63/EU) and the requirements of Italian laws (D.L. 26/2014); they were approved by the Italian Ministry of Health (protocol number 293/2021-PR).

Transgenic hemizygous SOD1-G93A male mice (B6.Cg-Tg [SOD1 ^G93A]1Gur/J) were obtained from The Jackson Laboratory (Bar Harbor, ME, United States of America; RRID:MGI:4835776) and then crossbred with C57BL/6 female mice to gain offspring. Generated animals were genotyped by PCR through hSOD1 oligos using interleukin-2 (IL-2) as a PCR internal control (sequences in Table 1). At the end of the training protocols, trained and untrained mice were anesthetized using Rompum (xylazine, 20 mg mL⁻¹, 0.5 mL kg⁻¹, Bayer, Milan, Italy) in combination with Zoletil (tiletamine and zolazepam, 100 mg mL⁻¹, 0.5 mL kg⁻¹, Virbac, Milan, Italy). After the pedal-withdrawal reflex was no longer present, the mice were euthanized for tissue dissection.

Female SOD1-G93A and WT littermate mice were randomly categorized into an intense endurance training group and a sedentary control group (n = 12/group). At 35 days of age, mice started running on a motorized treadmill (Ugo Basile, Rodent Treadmill NG cat. 47300) five times a week for 8 weeks, increasing speed, uphill slope, and time as shown in Table 2; we modified the protocol from Alizadeh et al. (2015) to obtain an intense endurance exercise. The treadmill was equipped with a mild shock to encourage the animal to keep running.

To reduce animal stress and experimental variability, the training was carried out in the morning between 8:00 a.m. and 1:00 p.m., and it included, on top of the training, 1 hour of acclimatization in the experimental room, 10 min of warm-up, and 10 min of cool-down.

The day after the last training section, trained and sedentary mice were anesthetized and euthanized, and the tissues were collected.

The disease progression was evaluated starting from 35 days of age (i.e., the start of the training) until the end of the training period (91 days of age) by measuring weight and grip strength and scoring the motor skills.

Total RNA was obtained from the tibialis anterior (TA) of 91-day-old trained and sedentary SOD1-G93A and WT mice (n ≥ 5 for each experimental group) with TRIzol^® Reagent (Ambion, Life Technologies) and reverse-transcribed using the ImProm-II™ Reverse Transcription System (Promega). Real-time qPCR was used to evaluate the levels of mRNA expression with a LightCycler 480 SYBR Green System (Roche, ETC). We used the “second derivative max” algorithm of the LightCycler software to calculate the crossing point (Cp) values. We chose the TATA box-binding protein as the housekeeping gene for normalization. Primer sequences are listed in Table 1.

Protein samples, obtained from the TA of sedentary and trained 91-day-old SOD1-G93A and WT littermate mice (n ≥ 5 for each experimental group), were separated by SDS–PAGE and transferred onto nitrocellulose membranes (PerkinElmer, Cat# NBA085B). The membranes were blocked with 5% milk, incubated with the appropriate primary antibodies (Table 4), incubated with the appropriate peroxidase-conjugated secondary antibodies (Table 4), and developed using enhanced chemiluminescence (Bio-Rad Clarity^TM Western ECL Substrate, Cat# 170–5061. Densitometric analyses were performed using ImageJ software (U.S. National Institutes of Health, Bethesda, Maryland, United States of America, https://imagej.nih.gov/ij/), and we analyzed five samples per group, divided into two different blots. The expression of proteins of interest was normalized to GAPDH expression levels.

The TA of 91-day-old trained and sedentary SOD1-G93A and WT mice (n ≥ 5 for each experimental group) was embedded into the Tissue-Tek Optimal Cutting Temperature (O.C.T.) Compound and sectioned at 10 μm using a Leica Cryostat.

For nicotinamide adenine dinucleotide dehydrogenase (NADH)-tetrazolium reductase activity staining, TA cryosections were hydrated in 0.1 M Tris-HCl, pH 7.5, for 10 min and incubated in NADH-tetrazolium reductase activity solution (0.4 mg/mL NADH, 0.8 mg/mL NTB, and 1 M Tris-HCl, pH 7.5) for 40 min at 37°C in a humidified chamber. The slides were dehydrated in increasing concentrations of ethanol, cleared in xylene, and mounted in a Eukitt mounting medium. Images of the whole muscle section were acquired using an Olympus BX51 Microscope. The percentage of NADH-positive areas was measured using ImageJ software (U.S. National Institutes of Health, Bethesda, Maryland, United States of America, https://imagej.nih.gov/ij/).

For sirius red collagen staining, TA sections were fixed 1 h with Bouin solution at 56°C and stained for 1 h with picrosirius red 0.1% (0.1% Direct Red in picric acid), dehydrated in increasing concentrations of ethanol, cleared in xylene, and mounted in a Eukitt mounting medium. Images of the whole muscle section were acquired using an Olympus BX51 Microscope. The percentage of the red area was measured using Fiji software.

For immunofluorescence (IF) staining, TA sections were blocked in PBS containing 0.3% Triton X-100 and 10% NDS for 1 h at room temperature (RT) and incubated overnight at 4°C with the primary antibody (Table 4) in PBS containing 0.3% Triton X-100 and 2% NDS, followed by incubation with the appropriate secondary antibody for 3 h at room temperature in the same solution. Immunofluorescence on the TA was visualized at 10X or ×20 magnification using a Zeiss LSM 800 Confocal Laser Scanning Microscope and processed by ZEN 2.6 (blue edition) (Carl Zeiss, Milan, Italy), taking at least four photos per sample.

The cross-sectional area (CSA) was analyzed using ImageJ Plugin “Macro_seg_5_modif.ijm.txt.” To evaluate the percentage of positive fibers for MyHC IIa and MyHC IIb, we used Plugin “Weka_Segmentation.”

The co-localization area of α-bungarotoxin and synaptophysin in NMJs was analyzed using ImageJ (U.S. National Institutes of Health, Bethesda, Maryland, United States of America, https://imagej.nih.gov/ij/).

We used ATPB staining to analyze the mitochondria. The area of the mitochondria in the entire TA was measured using Fiji, while the footprint of mitochondria in the single fibers was processed using Mitochondrial Network Analysis (MiNA), an ImageJ plugin, following the toolset in the work of Valente et al. (2017). We chose the fibers with a diameter <1000 μm (at least three fibers per image); then, the analysis consisted of steps that preserved the quality image, converted it into a binary image, and created a mitochondrial network for the quantitative analysis.

The spinal cords and sciatic nerves of sedentary and trained 91-day-old SOD1-G93A and WT mice (n ≥ 5 for each experimental group) were post-fixed in 4% paraformaldehyde, cryoprotected in 30% sucrose in PB, embedded in the O.C.T. compound, and cut using a cryostat at 30 μm thickness for the spinal cord (L3–L5) and 15 μm thickness for the sciatic nerve.

To count the motor neurons, we performed cresyl violet staining (Nissl staining). In brief, the sections were stained with 1% cresyl violet, dehydrated in 50%–100% alcohol, cleared in xylene, and mounted in a Eukitt mounting medium (Sigma-Aldrich). The images of the ventral horns were obtained at ×10 magnification using a Zeiss Axioskop 2 Microscope. The neurons with a cell body area ≥200 μm² and a definable cytoplasm with a nucleus and nucleolus were counted using Neurolucida software (MBF Bioscience, United States of America). For immunofluorescence on the spinal cord and sciatic nerve, the sections were blocked for 1 h at RT in PBS containing 0.3% Triton X-100 and 10% NDS and incubated for 48 h at 4°C with the primary antibody (listed in Table 4) in PBS containing 0.3% Triton X-100 and 2% NDS, followed by incubation with the appropriate secondary antibody (listed in Table 4) for 3 h at RT in the same solution, as previously described (Scaricamazza et al., 2022).

Immunofluorescence on the spinal cord and sciatic nerve sections was visualized at ×20 magnification using a Zeiss LSM 800 Confocal Laser Scanning Microscope, processed using ZEN 2.6 (blue edition) (Carl Zeiss, Milan, Italy), and analyzed using ImageJ (U. S. National Institutes of Health, Bethesda, Maryland, United States of America, https://imagej.nih.gov/ij/, 1997–2016).

Data are presented as the mean ± SEM. For the evaluation of the onset, data were analyzed using the Kaplan–Meier curve with a log-rank (Mantel–Cox) test and a log-rank test for trend. Then, the non-parametric Mann–Whitney test was used to evaluate the difference in the onset between groups (SD G93A and TR G93A). For the analysis of weight, grip strength tests, motor skill tests, histology, and protein and RNA expression, data were evaluated using a two-way ANOVA with the Tukey test as a post hoc analysis. Differences between groups were considered significant when the p-value was less than 0.05. The normality of the data was assessed using the Shapiro–Wilk test. Statistical analysis was carried out using GraphPad Prism 5 software.

SOD1-G93A and wild-type (WT) female mice underwent an intense endurance exercise protocol on a rodent treadmill for 8 weeks, 5 days a week, starting at 35 days of age, which is well before any ALS symptoms (Scaricamazza et al., 2022). To understand the effect of intense endurance exercise on ALS, we analyzed several parameters indicating functional and structural hallmarks of the disease throughout the training. Specifically, we monitored weight, muscle strength, and motor skills and compared them with the sedentary groups.

As shown in Figure 1A, unlike the WT mice that gradually gained weight, the weight of SOD1-G93A mice increased only in the first training weeks. Interestingly, trained SOD1-G93A mice stopped gaining weight earlier than the sedentary mice, at 70 and 77 days of age, respectively (Figure 1A).

The training improved the muscular strength of WT mice but did not affect the strength of SOD1-G93A mice (Figure 1B). We noticed that the trained transgenic mice tended to be weaker than the sedentary group, even though the difference was not statistically significant (Figure 1B).

The training significantly worsened the motor abilities of SOD1-G93A mice. Following the protocol of Milanese et al. (2021), we used two different tests to evaluate the motor skills: the hind limb extension reflex and the gait. As shown in Figure 1C, the score of the hind limb extension reflex of trained SOD1-G93A mice started decreasing at 52 days of age, 2 weeks before the sedentary transgenic mice. The gait test showed a similar pattern, with the trained transgenic mice starting to lower their score at 59 days of age and the sedentary transgenic mice at 70 days of age (Figure 1D).

By taking together the weight, grip strength, and score of the motor skill tests, we defined the onset of the disease (see Section 2 for details). As shown in Figure 1E, the training accelerated the onset of the disease by approximately 1 week, starting in trained mice at 63 days of age and sedentary mice at 70 days of age.

Skeletal muscle is a plastic tissue able to change and model its fibers in response to functional and metabolic needs. Intense endurance exercise pushes muscle fibers toward a more oxidative phenotype, promoting the shift from glycolytic to oxidative fibers (Ausoni et al., 1990; Schiaffino and Reggiani, 2011; Mukund, 2019). We have recently demonstrated that the same shift occurs early in the glycolytic skeletal muscles of sedentary SOD1-G93A mice before any denervation events (Scaricamazza et al., 2020). Thus, we wondered if the oxidative status that the training induced caused the early deficit of motor skills in trained SOD1-G93A mice by hastening the damage to skeletal muscles. In detail, we chose to analyze the tibialis anterior because it is a glycolytic muscle that we previously characterized and demonstrated to be affected in our model (Scaricamazza et al., 2020; 2022).

To define the oxidative status of the muscle fibers, we performed, at the end of the eighth-week training protocol, an in situ NADH-tetrazolium reductase activity assay on the TA muscle of sedentary and trained WT and SOD1-G93A mice (91 days of age). Since the NADH-positive area in the TA of the sedentary SOD1-G93A mice covered almost the entire sample, we could not see any difference between trained and sedentary mice (Supplementary Figures S1A, B). However, we observed an upregulation in the expression levels of sarcolipin (Sln) and neuron-derived orphan nuclear receptor (Nor1) and a slight downregulation, although not statistically significant, in the expression of glucose transporter type 4 (Glut4) in trained SOD1-G93A mice (Figure 2A), indicating that training shifted muscle fibers toward oxidative metabolism.

Consistent with that, immunofluorescence analysis revealed that training led to a significant increase in MyHC IIa/MYH2-positive fibers (Figures 2B, C) and a strong decrease in MyHC IIb/MYH4-positive fibers in SOD1-G93A mice (Figures 2B, C), while it did not affect the MyHC IIx/MYH1-positive area, which was assessed by considering the unstained fibers (Figures 2B, C). According to the literature, MyHC IIb-positive fibers decrease in WT mice after training (Figures 2B, C) (Yan et al., 2011; Qaisar et al., 2016). In agreement, we observed the same trend when analyzing the mRNA expression of Myh2 and Myh4 in the TA of the same experimental groups (Figure 2D).

The CSA measure corroborated the fiber shift: although the total CSA did not change (Figures 3A, B), by dividing for a specific dimension range and giving 100% to the total number of fibers, we observed that the percentage of small oxidative fibers (<1,000 μm²) increased, while the percentage of large glycolytic fibers (2000–3,000 μm²) decreased if compared to sedentary SOD1-G93A mice (Figure 3C). In agreement with these results, immunofluorescence on the same muscles revealed that training increased the mitochondrial marker ATPB, both in WT and SOD1-G93A mice (Figures 3A, D). Moreover, we detected with the MiNA 2.0 toolset that the oxidative fibers of trained SOD1-G93A mice had larger mitochondrial areas than sedentary WT and SOD1-G93A mice (Figures 3E, F), further confirming oxidative metabolism.

Overall, these data suggest that intense endurance training exacerbates the oxidative shift occurring in the glycolytic muscles of SOD1-G93A mice.

Considering that the skeletal muscle of SOD1-G93A mice had a worse oxidative status after training, we next assessed whether intense endurance exercise also affected muscle integrity, particularly fibrosis and denervation.

Picrosirius red staining (Supplementary Figures S2A, B) and collagen type III alpha 1 chain (Col3a1) mRNA expression analysis (Supplementary Figure S2C) on the TA of 91-day-old mice showed that training did not affect fibrosis as the percentage of red staining and Col3a1 mRNA expression did not change in sedentary and trained transgenic mice (Supplementary Figure S2B).

On the contrary, denervation strongly advanced in trained SOD1-G93A mice. Specifically, the expression levels of atrophy and denervation markers, such as F-box-only protein 32/atrogin-1, acetylcholine receptor α subunit (AChRα), and voltage-gated sodium channel 1.5 (Na_v1.5), increased in SOD1-G93A mice after training compared to the sedentary mice (Figure 4A). To further investigate the denervation process, we evaluated the neuromuscular junction (NMJ) dismantling in the TA of trained transgenic mice. Comparing trained and sedentary transgenic mice, immunofluorescence analysis showed a reduced co-localization of nicotinic acetylcholine receptor (nAChR) clusters (defined with α-bungarotoxin, α-BTXN) and synaptophysin (SYN) in trained mice, indicating increased NMJ instability (Figures 4B, C). Dystrophin (Dys) and low-density lipoprotein receptor-related protein 4 (LRP4) are essential for NMJ stability and postsynaptic differentiation, and, as we previously reported, they increase in SOD1-G93A mice starting at 120 days of age (Scaricamazza et al., 2022), probably to counter the damage of denervation. As shown in Figures 4D–F, training accelerated the upregulation of LRP4 and Dys in SOD1-G93A mice, further confirming the reduced NMJ stability.

We assessed whether intense endurance exercise had an impact on demyelination, axonal loss, and inflammation in the peripheral nervous system. To this end, cross sections of the sciatic nerves of 91-day-old sedentary and trained WT and SOD1-G93A mice were probed with antibodies recognizing myelin basic protein (MBP), neurofilament heavy chain (NFH), and pro-inflammatory immune cells (CD68). As shown in Figure 5, training enhanced Wallerian degeneration in SOD1-G93A mice by prompting myelin degradation and loss of the axonal cytoskeleton, as indicated by the reduction in MBP and NFH expression levels, respectively (Figures 5A–C). The damage in the sciatic nerve came with immune cell infiltration, as the stronger staining of CD68 in the sections of trained SOD1-G93A mice revealed in Figures 5A, D.

Further confirming demyelination, sciatic nerve cross sections of trained transgenic mice showed increased staining of glial fibrillary acidic protein (GFAP) (Figures 5E, F), usually expressed in non-myelinating Schwann cells (Yang and Wang, 2015).

Motor neuron loss is the final stage and the most characteristic hallmark of ALS. To determine if intense endurance exercise affected motor neurons, we performed Nissl staining on cross sections of the lumbar spinal cord (L3–L5) of sedentary and trained WT and SOD1-G93A mice. As shown in Figure 6, 91-day-old sedentary SOD1-G93A mice had a reduction of approximately 20% in the percentage of ventral horn motor neurons compared with the sedentary WT mice (Figures 6A, B). Interestingly, after training, SOD1-G93A mice reached a 50% reduction (Figures 6A, B).

In ALS, the death of motor neurons comes with a neuroinflammatory response characterized by the activation of reactive microglia and astrocytes (Volonté et al., 2019). Here, the identification of microglial and astrocytic markers by immunofluorescence (CD68 and GFAP, respectively) showed that these two cell populations were more active in trained SOD1-G93A mice than in sedentary controls (Figures 6C–E).