Charcot-Marie-Tooth disease (CMT) is an inherited peripheral neuropathy typified by degeneration of motor and sensory neurons, which triggers progressive muscle wasting and sensory deficits mainly in the feet and hands1. When genetic neuropathy results from demyelination and presents with slowed nerve conduction speeds, it is categorised as CMT1, whereas CMT2 results from axon degeneration unrelated to myelin disruption. Mutations in >100 genes cause CMT, the majority of which lead to greater dysfunction in lower limbs2. There is thus a length-dependency to CMT, such that peripheral nerves with the longest axons are generally more affected; this suggests that cellular processes most impacted by the extreme morphology of neurons, e.g. axonal transport, may play an important role in pathogenesis3,4,5.

Caused by dominantly inherited, missense mutations in the widely and constitutively expressed gene, GARS1 (ref. 6), the 2D subtype of CMT (CMT2D) usually manifests during adolescence and, unlike most forms of CMT, frequently displays upper limb predominance7,8. The GARS1-encoded protein, glycyl-tRNA synthetase (GlyRS), charges glycine to tRNA for protein translation; yet, it is a toxic gain-of-function that likely drives neurodegeneration9,10,11. Contrasting with most other aminoacyl-tRNA synthetases12, cytoplasmic and mitochondrial GlyRS isoforms are encoded by a single gene (GARS1) rather than two separate loci13. All CMT2D-causing mutations occur downstream of the mitochondrial targeting sequence11, while heterozygous mutant Gars mice modelling CMT2D show no disturbance in levels of proteins involved in oxidative phosphorylation14, together suggesting that mitochondrial dysfunction is unlikely to play a causative role in disease. The mechanisms underlying neuronal selectivity in CMT2D remain unresolved; however, aberrant protein–protein interactions caused by relaxation of the GlyRS structure appear to underlie neomorphic function15,16. Indeed, neuropathy-causing mutant GlyRS mis-interacts with the extracellular domains of neuronal transmembrane receptors, neuropilin 1 and tropomyosin receptor kinases (Trks) A–C16,17. Permitting this, GlyRS is secreted from several different cell types and circulates in mammalian serum, likely for a non-canonical and only partially understood function16,18,19.

The neuromuscular junction (NMJ) is the specialised synapse connecting lower motor neurons to muscle fibres, and is dysfunctional in several CMT models20,21,22,23,24,25. Indeed, CMT2D mice display loss of NMJ integrity in multiple hindlimb muscles without spinal cord motor neuron degeneration9,26,27,28,29,30. By directly comparing a proximal and a distal muscle, we have previously shown that the neuromuscular synapse is an important site of selective and early pathology28, replicating the muscle weakness pattern of patients. Also contributing to reduced strength and independently from denervation, these neuropathic mice display pre-synaptic disruption of neurotransmission that correlates with disease severity and worsens with age29. Neuromuscular degeneration is replicated in a Drosophila melanogaster model for GARS1 neuropathy and is dependent on toxic accumulation of mutant GlyRS at the NMJ19. Suggestive of a non-cell autonomous mechanism, this pathological build-up requires muscle-secreted, but not neuron-derived, GlyRS, and appears to be mediated by deviant interaction with Plexin B31.

Here, we extend our neuromuscular analyses in CMT2D, through a comprehensive assessment of developmental and degenerative processes at the mouse NMJ across several anatomically and functionally distinct muscles. By correlating varied neuropathic NMJ phenotypes with extensive morphological data on developing and mature wild-type synapses32, we have begun to identify key features underlying the selective vulnerability of neuromuscular connections in GARS1 neuropathy.

Materials and methods


Mouse work was carried out under license from the UK Home Office in accordance with the Animals (Scientific Procedures) Act 1986 and approved by the UCL Queen Square Institute of Neurology Ethical Review Committee. GarsC201R/+ mice (RRID: MGI 3849420) were maintained as heterozygote breeding pairs on a C57BL/6J background and genotyped as previously described27. Animals sacrificed for 1- and 3-month timepoints were 27–35 and 84–95 days old, respectively, and both sexes were used unless otherwise stated.

Grip strength testing

Grip strength was assessed in forelimbs as previously described17.

Muscle dissection and immunohistochemistry

Muscles were dissected and stained as wholemount preparations as outlined formerly33,34. The following antibodies were used to co-stain motor nerve terminals and axons: 1/25 mouse pan anti-synaptic vesicle 2 (SV2, Developmental Studies Hybridoma Bank [DSHB], Iowa City, IA, supernatant); 1/250 mouse anti-neurofilament (2H3, DSHB, supernatant). 1/1000 Alexa Fluor 555 α-BTX (Thermo Fisher Scientific, Waltham, MA, B35451, RRID:AB_2617152) was used to identify AChRs.

NMJ imaging and analysis

NMJs were imaged using a LSM780 laser scanning microscope (Zeiss, Oberkochen, Germany). Denervation, polyinnervation, and perforation analyses were performed as detailed elsewhere28,33.

Statistical analysis

Data were assumed to be normally distributed unless evidence to the contrary could be provided by the D’Agostino and Pearson omnibus normality test, while equal variance between groups was assumed. GraphPad Prism 8 (version 8.4.0, La Jolla, CA) was used for all statistical tests, which were all two-sided. Datasets were statistically compared using a one- or two-way analysis of variance (ANOVA) followed by either Sidak’s or Tukey’s multiple comparisons test. Correlation was assessed using Pearson’s product moment correlation. To maintain an α of 0.05, Bonferroni correction was applied to all correlation analyses when performing multiple tests. Sample sizes, which were pre-determined using previous experience28 and power calculations, are reported in all figure legends and represent biological replicates (i.e. individual animals). Once imaged, no muscles were excluded from analyses. As no treatment groups were involved, randomisation was not performed. All reported error bars depict standard error of the mean (SEM). All analyses were performed blinded to genotype.


NMJ denervation underlies greater hindlimb weakness in CMT2D mice

Mutant Gars mice display a developmental perturbation of sensory neuron fate in hindlimb-innervating lumbar dorsal root ganglia (DRG)17, which is not present in forelimb-innervating cervical DRG35, suggestive of length-dependent phenotypes in CMT2D mice. To determine whether the neuromuscular system displays a similar pathological pattern, we performed grip strength testing of mild GarsC201R/+ mice27. Previous data indicate that muscle function is impaired at both 1 and 3 months in male and female GarsC201R/+ mice when all limbs are simultaneously tested17. Assessing just forelimb muscle function of female wild-type and GarsC201R/+ mice at the same timepoints, we observed a similar early and persistent deficit in CMT2D strength (Supplementary Fig. S1). Hindlimbs alone cannot be tested because of the manner in which mice grasp the grip strength meter used. Nevertheless, to compare grip strength from all limbs17 with forelimb data generated here, we normalised to wild-type at each timepoint in each limb category. The relative forelimb strength was significantly greater at both ages compared with all limbs (72% vs. 57% and 69% vs. 59%, Fig. 1a), suggesting that CMT2D mice may display greater weakness in hindlimbs.

Fig. 1: NMJ denervation underlies the greater weakness in CMT2D mouse hindlimbs.
figure 1

a Grip strength relative to wild-type [WT] is significantly impaired in all four limbs and just the forelimbs of GarsC201R/+ mice at 1 (left, genotype P < 0.001, limbs P = 0.049, interaction P = 0.049; two-way ANOVA) and 3 months (right, genotype P < 0.001, limbs P = 0.125, interaction P = 0.125; two-way ANOVA). The weakness is more pronounced when all limbs are tested than when just the forelimbs are assessed. n = 4 (GarsC201R/+ forelimbs at 3 months), 6 (1 month and all limbs at 3 months) and 8 (wild-type forelimbs at 3 months). ##P < 0.01; unpaired t-test. b Representative collapsed z-stack confocal images of NMJs in hindpaw [Lumb. (hind)] and forepaw [Lumb. (fore)] lumbrical muscles dissected from 3-month-old wild-type and GarsC201R/+ mice. Lower motor neurons are visualised using a combination of SV2/2H3 (green) and post-synaptic AChRs with α-BTX (magenta). Asterisks identify partially denervated synapses. Scale bars = 25 µm. c The hindpaw, but not forepaw, lumbrical muscles show significant denervation at 1 month. By 3 months, loss of innervation is worse in both muscles, with significant denervation now also observed in forepaw lumbricals. P < 0.001 for all one-way ANOVAs comparing the percentage of fully innervated, partially denervated and vacant NMJs in both muscles at both timepoints (six tests, three per age). N.b., forepaw and hindpaw lumbricals were dissected from the same mice. n = 6. Means ± SEM are plotted in all graphs. ***P < 0.001, *P < 0.05; Sidak’s multiple comparisons test. See also Supplementary Fig. S1.

Lumbrical muscles involved in hindpaw clasping and grip strength, show a progressive loss of lower motor neuron connectivity in mutant Gars mice, correlating with overall model severity28. To compare NMJ innervation between a hindlimb and a forelimb muscle of similar function and relative limb location, we dissected both hindpaw and forepaw lumbrical muscles and stained them with antibodies against SV2/2H3 to visualise motor neurons, and α-bungarotoxin (α-BTX) to identify post-synaptic acetylcholine receptors (AChRs) (Fig. 1b). At 1 month, hindlimb lumbricals displayed significantly more partially and fully (i.e. vacant) denervated NMJs than wild-type mice, becoming worse by 3 months (Fig. 1c). In contrast, forelimb lumbricals displayed no significant loss of innervation at 1 month, with comparatively mild denervation manifesting by 3 months. When statistically compared, loss of NMJ integrity was consistently greater in hindpaw lumbricals, indicating that neuromuscular denervation contributes to the greater hindlimb weakness of CMT2D mice.

Impaired NMJ maturation precedes denervation

Prior to denervation, hindpaw lumbricals in CMT2D mice display impairments in key maturation events, such as synapse elimination and post-synaptic plaque-to-pretzel migration28; however, these developmental delays were not seen in a postural, anterior abdominal muscle called the transversus abdominis (TVA)36, which also lacked denervation at 3 months in mutant Gars mice28. To test whether this link between perturbed NMJ maturation and degeneration is consistent across CMT2D muscles, we assessed synaptic development in hindpaw and forepaw lumbricals (Fig. 2). At birth, post-synaptic endplates are contacted by several different motor neurons (see Fig. 2a for an example) and, through a process called synapse elimination, go from being polyinnervated to monoinnervated by 2 weeks in mice37,38. In tandem, AChRs migrate and become concentrated in close apposition to motor nerve terminals, resulting in their conversion from simple, circular plaques at birth to complex pretzel-like structures with increasing numbers of perforations39. Delayed synapse elimination was replicated in GarsC201R/+ hindpaw lumbrical NMJs, which showed significant polyinnervation at 1 month and 3 months (Fig. 2b). Contrastingly, forepaw lumbricals showed no such defect. To assess post-synaptic development, perforations per endplate were counted, with higher numbers indicative of advanced maturation. Once again, hindpaw lumbrical NMJs showed impaired AChR development, which was also present in forepaw lumbricals, albeit not to the same extent (Fig. 2c). These data indicate that defective pre-synaptic maturation is associated with subsequent neuromuscular degeneration, whereas post-synaptic disruption may represent a systemic effect, as it is also seen in the non-denervated TVA28.

Fig. 2: Impaired CMT2D NMJ maturation is linked to denervation.
figure 2

a Representative collapsed z-stack confocal images of NMJs in hindpaw [Lumb. (hind)] and forepaw [Lumb. (fore)] lumbrical muscles dissected from 1-month-old GarsC201R/+ mice. Lower motor neurons are visualised using a combination of SV2/2H3 (green) and post-synaptic AChRs with α-BTX (magenta). Asterisks identify partially denervated synapses, the arrowhead a vacant synapse, and the arrow a polyinnervated NMJ. Scale bars = 25 µm. b Hindpaw, but not forepaw, lumbrical muscles of GarsC201R/+ mice display significantly more polyinnervated NMJs at 1 month compared with wild-type (left, genotype P < 0.001, muscle P < 0.001, interaction P < 0.001; two-way repeated measures ANOVA). This pattern is also present at 3 months (right, genotype P = 0.001, muscle P = 0.191, interaction P = 0.029; two-way repeated measures ANOVA), although there are fewer polyinnervated NMJs in the hindpaw lumbricals at this stage, indicative of a delay rather than cessation of synapse elimination. c Both hindpaw and forepaw lumbrical NMJs show post-synaptic maturation deficiency as assessed by counting the perforations per endplate. This defect was significantly worse in hindpaw lumbricals. P < 0.001 for all one-way ANOVAs comparing the percentage of NMJs with none, one, two, and more than two perforations in both muscles (i.e. four tests). N.b., hindpaw and forepaw muscles are the same as those analysed in Fig. 1. n = 6. Means ± SEM are plotted in all graphs. ***P < 0.001, **P < 0.01, *P < 0.05; Sidak’s multiple comparisons test.

Hindlimb FDB muscles also display delayed NMJ development and degeneration

To extend our analyses into additional hindlimb and forelimb muscles, we stained NMJs of hindpaw flexor digitorum brevis (FDB) and forelimb epitrochleoanconeus (ETA) muscles dissected from 3-month-old mice (Fig. 3a). FDB muscles aid hindpaw opening and the ETA contributes to forearm supination40,41, and both can be wholemount stained to assess innervation32. The FDB displayed defective synapse elimination associated with severe denervation, whereas the ETA was unaffected (Fig. 3b, c). These data uphold that CMT2D hindlimbs display greater pathology, and that impaired NMJ development consistently accompanies degeneration.

Fig. 3: Hindlimb muscles behave similarly and show marked differences with forelimb muscles.
figure 3

a Representative collapsed z-stack confocal images of NMJs in FDB and ETA muscles dissected from 3-month-old wild-type and GarsC201R/+ mice. Lower motor neurons are visualised using a combination of SV2/2H3 (green) and post-synaptic AChRs with α-BTX (magenta). The asterisk identifies a partially denervated NMJ and the arrowhead a fully denervated synapse. Scale bars = 25 µm. b The FDB, but not ETA, muscle of GarsC201R/+ mice displays significantly more polyinnervated NMJs at 3 months (genotype P < 0.001, muscle P = 0.005, interaction P < 0.001; two-way repeated measures ANOVA). c At 3 months, denervation in the FDB muscle is severe, whereas the ETA shows little degeneration. P < 0.001 for all one-way ANOVAs comparing the percentage of fully innervated, partially denervated and vacant NMJs in both muscles (i.e. three tests). n = 6. Means ± SEM are plotted in all graphs. ***P < 0.001; Sidak’s multiple comparisons test.

CMT2D polyinnervation correlates with degeneration and is independent of synapse elimination

Including previous data from the TVA28, we have now assessed NMJ polyinnervation and denervation at 3 months in five wholemount muscles (Fig. 4a). We detected a broad spectrum of denervation (partial and full combined) in CMT2D mice, ranging from 2% in the TVA to 63% in the FDB (Fig. 4b, Supplementary Table S1). A similar phenotypic continuum was identified for disturbed synapse elimination, with 0.7% of TVA NMJs remaining polyinnervated and 7.3% of FDB NMJs (Fig. 4c, Supplementary Table S1). We tested the relationship between these two phenotypes and found a highly significant positive correlation between percentage of vacant and polyinnervated synapses occurring at 3 months (Fig. 4d). A similar correlation was present when vacant and partially denervated synapses were combined (% denervation, Fig. 4e). To determine whether perturbed synapse elimination is simply a delay of the physiological process occurring in wild-type animals, we correlated the CMT2D motor input counts with previously generated32 P7 wild-type polyinnervation data from the five muscles (Fig. 4f). No correlation was identified, suggesting that impaired synapse elimination in mutant Gars mice may be caused by selective, muscle-specific perturbation of pathways relevant to removal of supernumerary motor axons, as opposed to a systemic slow-down in the process.

Fig. 4: CMT2D NMJ polyinnervation correlates with denervation and is not simply a delay of wild-type synapse elimination.
figure 4

a Five thin and flat muscles were analysed in this study: transversus abdominis (TVA, yellow), epitrochleoanconeus (ETA, purple), forelimb lumbricals (green, Lumb. [fore]), hindlimb lumbricals (cyan, Lumb. [hind]) and flexor digitorum brevis (FDB, red). The schematic was created with BioRender ( b The five muscles display a spectrum of vulnerability to denervation in Gars201R/+ mice at 3 months; the TVA remains unaffected, while the FDB muscles show stark degeneration. c Defective synapse elimination at 3 months is more pronounced in muscles with greater denervation. d, e There is a significant correlation between the percentage of polyinnervation observed in each muscle and the percentage of vacant (d, ***P < 0.001, r = 0.993) or denervated (e, **P = 0.006, r = 0.971) NMJs at 3 months in GarsC201R/+ mice. f There is no correlation between the percentage of wild-type NMJ polyinnervation at P7 and the polyinnervation percentage at 3 months in GarsC201R/+ mice (P = 0.690, r = −0.246), suggesting that the mutant phenotype is not due to a delay of wild-type synapse elimination. See Supplementary Table S1 for pairwise statistical testing between muscles in panels (b) and (c). Data in panels (e) and (f) were analysed by Pearson’s product moment correlation. The colour-coding of muscles in panels (e) and (f) is maintained from panels (ac). TVA data28 and wild-type P7 polyinnervation data32 were generated for previous studies. n = 6 (all muscles except TVA) and 8 (TVA). Means ± SEM are plotted in all graphs.

Fibre type does not associate with CMT2D denervation

Motor neurons innervating hindpaw muscles (FDB and lumbricals) are more impacted in CMT2D mice than those targeting forepaws (lumbricals) followed by more proximal muscles (ETA and TVA) (Fig. 4). This indicates a general distal-to-proximal vulnerability axis consistent with length-dependent axon degeneration. However, adjacent muscles innervated by neurons of similar length show significant distinctions in denervation (Supplementary Table S1), suggesting that additional factors contribute to NMJ vulnerability. Fast-fatiguable motor neurons innervating fast twitch muscle fibres are more susceptible to degeneration in amyotrophic lateral sclerosis (ALS) than motor nerves contacting slow twitch fibres42. To assess whether this is also observed in CMT2D, we correlated previously published fast twitch fibre type percentages32,40,41,43,44,45 with mutant Gars mouse denervation (Supplementary Fig. S2). No relationship was observed indicating that fibre type is unlikely to be driving inter-muscle disparities in neuropathy.

CMT2D denervation correlates with demand for post-natal NMJ growth

We then evaluated NMJ structure as a possible determinant in neuropathology, since NMJs with smaller synaptic volumes have less neuronal input to lose to become fully denervated. Using a robust and standardised semi-automated workflow called NMJ-morph46, we have previously generated detailed data on morphology of developing (P7) and mature (P31-32) wild-type NMJs in the same five muscles assessed here32. We found that developing synapses show greater inter-muscle variability and that post-natal NMJ growth occurs at different rates across muscles32—two phenotypes that may provide a substrate for differential NMJ degeneration in CMT2D. We therefore first correlated all 41 morphological variables from P7 and P31-32 wild-type NMJs with the percentage of vacant synapses in 3-month mutant Gars mice (Supplementary Table S2); however, no significant relationships were identified. We then tested for correlations between morphology and combined percentages of partially and fully denervated NMJs (Supplementary Table S3). From 41 variables, only AChR perimeter of developing NMJs showed a significant interaction, indicating that a smaller AChR perimeter associates with greater denervation levels (Supplementary Fig. S3). Given that many of the assessed NMJ variables are interrelated (e.g. AChR perimeter/area), the finding that only one out of 82 tests showed a significant correlation suggests that synaptic architecture has little impact on neuropathy.

Finally, we assessed the role of post-natal NMJ growth and development by correlating the percentage change in each morphological variable with the extent of mutant Gars degeneration. Several different pre-synaptic (terminal area, number of branches and branch length), post-synaptic (AChR area) and overlapping (synaptic contact area) variables significantly correlated with the percentage of vacant CMT2D NMJs (Fig. 5a, Supplementary Table S4). Two further pre-synaptic variables (branch point number and complexity) significantly associated with the percentage of denervated (partial and full) mutant NMJs (Fig. 5b, Supplementary Table S5). All seven correlations were positive, indicating that NMJs undergoing greater post-natal change in morphology are more prone to degeneration in CMT2D mice.

Fig. 5: NMJs that undergo greater post-natal morphological change are more vulnerable to degeneration in CMT2D mice.
figure 5

The percentages of vacant (a) and denervated (b) NMJs in 3-month GarsC201R/+ mice were correlated with post-natal NMJ growth in 20 different pre-synaptic (i), post-synaptic (ii) and overlapping (iii) morphological variables. Wild-type P7 to P31-32 morphological data32 and TVA vacancy data28 were previously published. Correlation was assessed by calculating Pearson’s product moment correlation coefficient (r), the results of which are presented in Supplementary Table S4 and Supplementary Table S5 along with associated P values. Only significant correlations are presented, i.e. Bonferroni-corrected P < 0.00256. n = 6 (all muscles except TVA) and 8 (TVA). Lumb. (fore), forepaw lumbricals; Lumb. (hind), hindpaw lumbricals.


CMT patients usually present with weakness and atrophy in the feet and then hands, indicating that peripheral neurons with longer axons are more impacted by the disease. However, mutations in a few genes, including GARS1, can cause an upper limb predominance7,8,47. This suggests that while axon length is important, additional characteristics contribute to pathogenesis and the spectrum of motor neuron involvement. Identification of such factors may elucidate pathomechanisms and provide potential targetable pathways. While we did not observe upper limb weakness in mutant Gars mice (possible reasons for which are discussed elsewhere35), we do observe significant distinctions in NMJ pathology between adjacent muscles innervated by motor neurons of similar length. This suggests that, like in human patients, CMT2D mouse motor neurons possess features that modify their predisposition to neuropathology.

We first assessed the impact of muscle fibre type on neuropathy, because large, fast-fatiguable motor neurons targeting fast twitch fibres are more vulnerable in ALS42, a motor neuron disease with parallels to CMT. There was no correlation between fibre type and CMT2D degeneration, suggesting that in contrast to ALS, muscle fibre types have little impact on motor neuropathy. Perhaps impacting accuracy, fibre type percentages, including from rat TVA, were obtained from several studies; thus, in the future fibre typing should be performed in parallel across all five muscles.

Since pathology in mutant Gars mice and patients is restricted to distal segments of peripheral nerves, we then probed NMJ morphology. Again, we found little evidence to indicate that distinctions in pre- or post-synaptic architecture cause the spectrum of synapse loss. However, we did find that CMT2D denervation correlated closely with wild-type NMJ change and development. This suggests that motor terminals under the greatest pressures to grow and mature during early post-natal life, perhaps due to different functions of their associated muscles32, are more likely to degenerate. Possibly contributing to this, severe CMT2D mice possess fewer mitochondria in pre-synaptic motor terminals in the proximal levator auris longus muscle29. Accordingly, motor neurons with higher bioenergetic capacity and thus energy supply are more resistant to disease in spinal muscular atrophy mice48, while computational modelling indicates energetic demand also contributes to selective vulnerability in ALS49.

While developmental demand may contribute to early synaptic disruption, GARS1-associated neuropathy is progressive and denervation continues in CMT2D mice up to 3 months and beyond. Additional mechanisms are therefore likely driving the life-long neurodegeneration caused by mutant GlyRS.

Synapse elimination is mediated by reciprocal nerve-muscle signalling and electrical activity to refine nervous system architecture38,50. The strong correlation across CMT2D muscles between polyinnervated and denervated synapses, and the lack of association with wild-type polyinnervation, indicate that synapse elimination is linked to pathology and not caused by systemic interruption.

Synapse elimination delay is unlikely to be driving denervation for two reasons; firstly, having multiple motor neurons innervating an endplate is probably beneficial from an innervation standpoint as several motor nerve terminals must degenerate for full denervation to occur. Corroborating this, the percentage of denervation (partial and full) across all FDB NMJs is 63 ± 4% at 3 months, whereas at polyinnervated synapses, only 6 ± 4% of NMJs show signs of degeneration (data not shown). Secondly, mechanisms mediating removal of supernumerary motor inputs share many features with neurodegenerative pathways51, and hence a delay in elimination would likely also postpone neuropathy. Alternatively, the phenotype could represent a compensatory response to degeneration, similar in essence to Schwann cell-mediated neuronal sprouting52 or pre-synaptic homeostatic plasticity53. However, if this were the case, it would probably be observed throughout the lifespan of CMT2D mice and in different neuromuscular disease models, but this does not appear to occur. We therefore suggest that the polyinnervation phenotype is an epiphenomenon resulting from aberrant interaction of mutant GlyRS with one or several key trans-synaptic proteins. There is precedent for this in a CMT2D Drosophila model; muscle-secreted mutant GlyRS aberrantly accumulates at NMJs prior to degeneration19 and coincides with neurodevelopmental wiring defects31. This phenocopies ectopic motor neuron branching caused by loss-of-function mutations in plexA and plexB, which encode neuronal transmembrane proteins that bind secreted semaphorins to facilitate axonal guidance, retrograde signalling and synaptic plasticity54,55. This observation led to the finding that plexin B serves as an erroneous binding partner for mutant GlyRS at the NMJ; however, while mutant GlyRS interferes with the fidelity of axon guidance pathways causing wiring defects, this was de-coupled from NMJ pathologies, i.e. was an epiphenomenon31.

We have previously shown that mutant GlyRS aberrantly interacts with neurotrophin receptors (TrkA-C)17, of which TrkB is the main one present at the mammalian NMJ56,57. TrkB and its ligand, brain-derived neurotrophic factor (BDNF), play an important role in NMJ synapse elimination57,58,59. This may, in part, be driven by BDNF-mediated clustering in motor nerve terminals of voltage-gated calcium channels, which facilitate calcium influx, synaptic neurotransmission, and NMJ maturation60,61,62,63. Mutant GlyRS mis-interacting with TrkB at the NMJ could delay synapse elimination through impairing calcium channel insertion into pre-synaptic motor terminals, thereby hindering neurotransmission and neuronal activity. Consistent with this idea, it has been shown that CMT2D mice display pre-synaptic transmission defects, including reduced frequency of spontaneous neurotransmitter release, which is calcium transient-dependent29. A primary mechanism causing the weakened synaptic transmission was ultimately not identified, instead it was suggested that several different release processes function sub-optimally29. It is therefore conceivable that accumulation of mutant GlyRS at NMJs through association with TrkB, akin to the synaptic build-up of pathological protein observed in the fly model19,31, may be contributing to the impaired synaptic function prior to NMJ degeneration.

Upon BDNF binding, TrkB is internalised into neuron terminals. Activated receptor complexes can then precipitate local signalling events such as those just discussed or become sorted into signalling endosomes for long-range retrograde axonal transport critical to pro-survival gene transcription64,65. The erroneous association of mutant GlyRS with TrkB at motor nerve terminals may thus diminish retrograde neurotrophic signalling, contributing to neuropathy. Correspondingly, signalling endosome transport is impaired in primary sensory neurons cultured from severe mutant Gars mice66, while mutations in essential signalling endosome retrograde effectors, Rab7 (ref. 67) and dynactin, cause peripheral neuropathy68,69. In addition, activation of IGF1R, a neurotrophic factor receptor found at the NMJ, is involved in regulating signalling endosome dynamics in motor neurons70, which may be a common function of similar receptors like TrkB71. Furthermore, high levels of phosphorylated IGF1R at NMJs may confer resistance to motor degeneration in ALS mice72, suggesting that levels of neurotrophic factors and/or their receptors (e.g. BDNF/TrkB) could indeed play a role in selective vulnerability to motor terminal loss in peripheral neuropathy. Alternatively, mutant GlyRS sequestration of TrkB could divert BDNF to bind to the pan-neurotrophin death receptor, p75NTR, resulting in over-activation of degenerative signals73. Consistent with these views and the importance of TrkB signalling to peripheral synapse integrity, reduced TrkB activity is known to impair NMJ neurotransmission, maturation and innervation, without impacting motor neuron number56,74, phenotypes very similar to those found in CMT2D mice. It is therefore conceivable that different levels of TrkB and/or BDNF at NMJs across muscles contribute to both the differential denervation and subverted synaptic maturation.

In summary, by assessing developmental and degenerative NMJ phenotypes across five different wholemount muscles from CMT2D mice, we have identified a spectrum of vulnerability to neurodegeneration, which is fully supported by findings in two muscles from the more severe mutant Gars mouse28. Through correlation analyses with muscle fibre type, NMJ architecture, and post-natal synaptic growth, we identified that the magnitude of developmental demand likely contributes to neuromuscular synapse loss in mutant Gars mice. We believe that these results point the way forward for an improved understanding of the molecular mechanisms driving differences in synaptic vulnerability to neuropathy.