Gamma motor neurons survive and exacerbate alpha motor neuron degeneration in ALS

Abstract

The molecular and cellular basis of selective motor neuron (MN) vulnerability in amyotrophic lateral sclerosis (ALS) is not known. In genetically distinct mouse models of familial ALS expressing mutant superoxide dismutase-1 (SOD1), TAR DNA-binding protein 43 (TDP-43), and fused in sarcoma (FUS), we demonstrate selective degeneration of alpha MNs (α-MNs) and complete sparing of gamma MNs (γ-MNs), which selectively innervate muscle spindles. Resistant γ-MNs are distinct from vulnerable α-MNs in that they lack synaptic contacts from primary afferent (I_A) fibers. Elimination of these synapses protects α-MNs in the SOD1 mutant, implicating this excitatory input in MN degeneration. Moreover, reduced I_A activation by targeted reduction of γ-MNs in SOD1^G93A mutants delays symptom onset and prolongs lifespan, demonstrating a pathogenic role of surviving γ-MNs in ALS. This study establishes the resistance of γ-MNs as a general feature of ALS mouse models and demonstrates that synaptic excitation of MNs within a complex circuit is an important determinant of relative vulnerability in ALS.

Keywords: ALS; fusimotor; gamma motor neuron; motor neuron disease.

Fig. 1.

Small γ fusimotor neurons are selectively spared at the end stage of disease in SOD1^G93A mice. (A and B) MNs in the L5 segment of WT (A) and SOD1^G93A (B) animals at the end stage of disease (∼ day 150). Animals are heterozygous for the Hb9::GFP transgene in which GFP is selectively expressed in large α-MNs. (A) α-MNs thus are identified by positive immunoreactivity for ChAT (red), GFP (green), and NeuN (blue), and γ-MNs are identified by positive immunoreactivity for ChAT and negative immunoreactivity for NeuN and GFP. (B) γ-MNs (arrowheads) are preserved in SOD1^G93A animals at end stage. (Scale bars: 200 μm.) (C) Size distribution of all ChAT⁺ MNs in WT (gray bars; 50-μm² bins; n = 5 mice) and SOD1^G93A animals (white bars, n = 5 mice). ChAT⁺ MN cell body sizes showed a bimodal distribution best fit by two Gaussian curves (correlation = 0.93) representing small WT (solid red line) and small SOD1 (dashed red line) and large WT (solid blue line) and large SOD1 (dashed blue line) populations. These measurements were used to determine the cutoff of the small ChAT⁺ γ-MNs. In WT animals, the small-size ChAT⁺ population had a mean cross-sectional area (± SD) of 310 ± 67 μm². The large population had a wider distribution with a mean cross-sectional area of 687 ± 211 μm². The size cutoff distinguishing small and large ChAT⁺ MNs was 440 μm². Small MNs represent 36.2 ± 1.7% of the total ChAT⁺ MNs. In the SOD1^G93A mice, the small ChAT⁺ MNs had a mean cross-sectional area (± SD) of 300 ± 96 μm², and the large ChAT⁺ MNs had a mean cross-sectional area (± SD) of 526 ± 238 μm². The subpopulation of ChAT⁺ α-MNs (HB9::GFP/NeuN⁺) is represented by two different Gaussian curves: WT, 715 ± 186 μm², correlation 0.76 (solid green line), and SOD1^G93A, 563 ± 173 μm², correlation 0.80 (dashed green line). Error bars represent the 95% confidence interval. (D) Quantification of α-MNs and γ-MNs in the L4/L5 segment of the spinal cord. At end stage (day 150), there is a significant (78%, ***P < 0.0001) loss of the large-size α-MN population in animals carrying the SOD1^G93A transgene as compared with WT controls. No significant difference is observed in the small-sized Hb9::GFP⁻, NeuN⁻ γ-MN population. The loss of α-MNs becomes significant at day 90 (60 d: P = 0.26; 90 d: *P = 0.0267; 120 d: **P = 0.0009). (E) MNs in the L5 segment of end-stage WT (E₁–E₃) and SOD1^G93A (E₄–E₆) animals demonstrate that small-sized ChAT⁺ cells (white arrowheads) express high levels of the Gfrα1-TLZ compared with large-sized ChAT⁺ MNs (white asterisks). (Scale bar: 100 μm.) (F) There is no significant difference in the number of Gfrα1^TLZ+ MNs in SOD1^G93A (gray) and WT (white) animals at P90 and at end stage (∼P150). Error bars represent the 95% confidence interval.

Fig. 2.

Innervation of intrafusal muscle fibers is preserved in end-stage SOD1^G93A mice. (A–D) Representative images of the TA muscle from WT (A and B) and SOD1^G93A (C and D) animals showing NMJs on extrafusal (A and C) and intrafusal (B and D) muscle fibers. Innervation of muscle was determined by colocalization of markers for motor axon terminals (anti-PGP 9.5 antibody) and the acetylcholine receptors of the postsynaptic surface of the NMJ (fluorophore-conjugated α-BTX). Marked denervation of extrafusal NMJs is observed in the SOD1^G93A mouse at end stage (C). Annulospiral primary sensory endings (asterisk in B1 and D1), motor axons in intrafusal fibers, and intrafusal muscle also show immunoreactivity for PGP 9.5, allowing identification of the intrafusal motor endings in the juxtaequatorial and polar regions of the muscle spindle. In both WT (B1) and SOD1^G93A (D1) animals, intrafusal NMJs are innervated by PGP9.5⁺ motor axons (arrowheads). (Scale bars: 50 μm.) (E, Upper) The majority (66 ± 6.3%; ***P < 0.0001) of NMJs in the extrafusal fibers of the TA are vacant at P150 in the SOD1G93A mice. (Lower) The great majority (92 ± 3.5%) of the intrafusal NMJs remained innervated. The limited amount of denervation presumably is caused by the selective loss of the fusimotor collaterals of β-MNs. Error bars represent the 95% confidence interval.

Fig. 3.

Gamma fusimotor neurons are also spared in the TDP-43^A315T and hFUS^P525L mutants. Shown is the shared selectivity of MN degeneration across different models of familial ALS. (A and B) Representative images of the L5 segment from WT (A) and TDP-43^A315T (B) mice at the end stage of disease (∼P150) showing immunoreactivity for ChAT (A₂ and B₂, red), GFP (A₃ and B₃, green), and NeuN (A₄ and B₄, blue). Small-sized ChAT⁺, GFP⁻, NeuN⁻ γ-MNs (white arrowheads in A and B) are preserved in the ventral horn of the spinal cord of the TDP-43^A315T mice at end stage. (Scale bar: 100 μm.) (C) Bimodal distribution of cell-body size of all ChAT⁺ MNs in WT mice (gray bars; 50-μm² bins; n = 7) and TDP-43^A315T mice (white bars; n = 7) fit by two Gaussian curves (correlation = 0.71) representing small WT (solid red line) and TDP43 (dashed red line) and large WT (solid blue line) and TDP-43 (dashed blue line) populations. In the WT mice, the small ChAT⁺ MNs had a mean (± SD) cross-sectional area of 318 ± 75 μm². Large ChAT⁺ MNs showed a wider size distribution around a mean (± SD) of 688 ± 204 μm². We used an area of 465 μm² as the cutoff point to distinguish between small and large MNs. In the TDP43^A315T mice, the small-sized ChAT⁺ population had a mean cross-sectional area (± SD) of 311 ± 77 μm², and the large-size MNs had a mean area (± SD) 652 ± 271 μm². All α-MNs (Hb9::GFP/NeuN⁺) are represented by two different Gaussian curves: WT (green solid line): 718 ± 180 μm², correlation 0.70, and TDP-43^A315T (green dashed line): 724 ± 289 μm², correlation 0.72. Error bars represent the 95% confidence interval. (D) Quantification of ChAT⁺ MNs shows an 18.6% reduction (*P = 0.02) in in the total number of L5 MNs [WT (gray): 510 ± 24 MNs; TDP-43^A315T (white): MNs 416 ± 25 MNs. This reduction could be accounted for entirely by the 27.4% reduction in the number of α-MNs (WT: 339 ± 17 MNs; TDP-43^A315T: 246 ± 9 MNs; **P = 0.003). No difference in the total number of γ-MN (ChAT⁺, NeuN⁻; <465 μm²) cells was observed. Error bars represent the 95% confidence interval. (E) Distributions of cell body size of all ChAT⁺ MNs in τ^ONhFUS^WT animals at P360 (gray bars; 50-μm² bins; n = 4 animals) and from age-matched τ^ONhFUS^P525L animals (red bars). Body sizes of ChAT⁺ MN cells showed a bimodal distribution best fit by two Gaussian curves (correlation = 0.91) representing small (τ^ONhFUS^WT) (solid gray line), small τ^ONhFUS^P525L (solid red line), large (τ^ONhFUS^WT) (dashed gray line), and large τ^ONhFUS^P525L (dashed red line) populations. These measurements were used to determine the cutoff of the small, ChAT⁺ γ-MNs. n = 4 for all genotypes. Error bars represent the 95% confidence interval. (F) Expression of myc-hFUS (red) in the γ-MN population. The γ-MN population was defined by size (<440 μm²), the presence of ChAT (gray), and the absence of NeuN (green) as indicated by the dotted white lines. (Scale bar: 30 μm.) (G) Number of ChAT⁺ MNs in the L5 segment in τ^ONhFUS^WT (gray; n = 4) and τ^ONhFUS^P525L (red; n = 4) animals. At P360, the large (>440 μm²) α-MN population is significantly reduced (41%; P = 0.0001) in τ^ONhFUS^P525L animals compared with τ^ONhFUS^WT animals and WT littermate controls (τ^ONhFUS^WT: 366 ± 10.2 MNs vs. τ^ONhFUS^P525L 215 ± 13 MNs). No difference in the total number of γ-MNs (<440 μm²) was observed (τ^ONhFUS^WT: 159.7 ± 10.8 MNs vs. τ^ONhFUS^P525L 160 ± 9 MNs; P = 0.821). n = 4 for each genotype. **P < 0.01 using one-way ANOVA at each time point with Bonferroni’s post hoc test. Error bars represent the 95% confidence interval.

Fig. 4.

Loss of sensory inputs increases α-MN survival in SOD1^G93A mutant mice. (A) Schematic representation of the spinal reflex circuit (A₁). This circuit is perturbed in the GDNF^FLOX mice in which reduced γ-MN survival results in a reduction in proprioceptive feedback (A₂). In Egr3^KO mice, the circuit is disrupted further, in that degeneration of the muscle spindles results in a deficit in the strength of direct (I_A) sensory-motor connections (A₃). (B) Survival analysis of SOD1^G93A; Egr3^KO double-mutant mice and SOD1^G93A mutant controls revealed no significant difference in lifespan (median survival 159 d for Egr3^WT; SOD1^G93A mice vs. 157 d for Egr3^KO; SOD1^G93A mice; P = 0.67). (C) Quantification of the number of ChAT⁺ MNs at end stage (∼P157) in SOD1^G93A; Egr3^KO (138 ± 8; n = 4), WT (455 ± 38; n = 4; ***P = 0.001), and SOD1^G93A (198 ± 12; n = 4; *P = 0.01) animals. As anticipated, a significant decrease in the number of γ-MNs was observed in the Egr3^KO; SOD1^G93A mice compared with WT and SOD1^G93A mutant mice (**P = 0.01). In contrast, there was a 10% increase in the number of α-MNs in the double mutant compared with SOD1^G93A mice. Error bars represent the 95% confidence interval. (D) A significant 14% increase in innervation was observed in the TA muscle of Egr3^KO; SOD1^G93A animals compared with SOD1^G93A animals (*P = 0.02). Error bars represent the 95% confidence interval. (E) The hind limb reflex score declines earlier in the course of the disease in the SOD1^G93A; GDNF^FLOX/FLOX;Egr3^NOCRE animals than in SOD1^G93A; GDNF^FLOX/FLOX;Egr3 ^CRE/CRE animals from P90 to P140 (P < 0.05; two-way ANOVA). (F) Lifespan is extended by 10 d in SOD1^G93A; GDNF^FLOX/FLOX; Egr3^CRE/CRE animals compared with controls (median survival: 157.5 d in SOD1^G93A; GDNF^FLOX/FLOX; Egr3^NOCRE animals vs. 167 d for SOD1^G93A; GDNF^FLOX/FLOX; Egr3^CRE/CRE animals; ***P ≤ 0.0001). (G) Quantification of ChAT⁺ MNs in the L5 segment demonstrates a significant reduction in the total number of MNs in SOD1^G93A; GDNF^FLOX/FLOX; Egr3 CRE/CRE (dark blue) animals when compared with controls [GDNF^FLOX/FLOX (no CRE, white); SOD1^G93A; GDNF^FLOX/FLOX (no CRE, light blue); **P = 0.0002]. This difference could be accounted for by the elimination of the γ-MN population in the SOD1^G93A; GDNF^FLOX/FLOX; Egr3^CRE/CRE animals (SOD1G93A; GDNFFLOX/FLOX; Egr3^NOCRE vs. SOD1^G93A; GDNF^FLOX/FLOX; Egr3^CRE/CRE: **P = 0.003). There was no significant difference in the number of α-MNs in SOD1^G93A; GDNF^FLOX/FLOX; Egr3^CRE/CRE when compared with the SOD1^G93A; GDNF^FLOX/FLOX; Egr3^NOCRE animals (P = 0.13). A decrease in the number of α-MNs is observed in animals carrying the SOD1^G93A transgene when compared with WT control animals (***P = 0.0001). Error bars represent 95% confidence interval. (H) Quantification of TA muscles showed no significant increase in innervation at disease end stage in SOD1^G93A; GDNF^FLOX/FLOX; Egr3^CRE/CRE (dark blue) compared with SOD1^G93A; GDNF^FLOX/FLOX; Egr3^CRE− (light blue) animals (P = 0.10). Error bars represent the 95% confidence interval. (I) MNs in segment L6 visualized in transgenic mice carrying the tdTomato reporter (red). Excision of the loxP-flanked STOP cassette by Cre-mediated recombination (ChAT Cre) results in the expression of tdTomato in cholinergic cells. MNs located in Onuf’s nucleus (white dotted square shown in higher magnification in I₂) are immunonegative for VGluT1 inputs. As previously described (42), Onuf’s nucleus was identified by its characteristic location at the lateral border of the ventral horn. (Scale bars: 250 μm in I₁ and 25 μm in I₂.) Image courtesy of Francisco Alvarez (Emory University School of Medicine, Atlanta, GA).