Disease mutation in gigaxonin-E3 ligase recapitulates giant axonal neuropathy in mice

Abstract

The nervous system evolved a variety of connections and neuron types to sustain diverse functions. While challenging, unlocking the universal mechanisms that support neuron integrity can be addressed in giant axonal neuropathy (GAN), a rare and fatal disease with broad deterioration of the nervous system. Here, we describe a new mouse strain that recapitulates key aspects of the GAN pathology following the introduction of a disease-causing mutation in GAN. Unlike previous GAN knock-out mice which show no overt phenotype, GAN^A49E/A49E mice exhibit early sensory-motor deficits and ataxia, giant axons and demyelination which, together with increased abundance, dramatic compaction and disorganization of neurofilaments across the nervous system, mimics the human disease. Using this model, we uncover novel alterations within neuromuscular junctions and muscles that might contribute to GAN pathogenesis. Interestingly, we pinpoint a sex bias whereby females show more severe histopathological damage and disease severity. Altogether, the GAN^A49E strain provides the first robust rodent model for GAN, recapitulating the symptoms and histological hallmarks of the human pathology. This model will be invaluable when investigating the cellular and molecular mechanisms that uphold neuron integrity along with effective therapies for GAN.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40478-025-02138-1.

Introduction

Amongst neurodegenerative diseases, giant axonal neuropathy (GAN) is unique for its widespread deterioration of the entire nervous system [2, 5]. Rare and fatal, GAN is extremely severe, starting during infancy with walking defects as the first clinical signs and evolving to complete loss of sensitivity and ambulation in the adolescence [28, 31]. In its classical form, the disease subsequently spreads to the central nervous system (CNS) and causes a wide range of symptoms, encompassing dysarthria, nystagmus, intellectual disabilities, ataxia and epilepsy [5, 28, 31, 35]. The outcome is usually fatal in the third decade of life. A key feature of GAN is the generalized disorganization of the cytoskeleton, which entails the bundling of Intermediate Filaments (IFs) [10, 35]. Neurofilaments (NFs) are not the only IF types densely packed within giant axons [2]—such defects also extend to non-neuronal tissues [52], including desmin (muscle) [44, 55], keratin (curled hair) [24], and vimentin in various cell types [14, 51]. As a neuropathy of axonal type, GAN is characterized by giant axons that can extend up to 50 µm in diameter [2, 21, 48, 50, 52] and by axonal degeneration associated with demyelination [2, 52]. Together, the broad deterioration of the entire nervous system and wide aggregation of IFs identify GAN as a unique disease to study the pivotal mechanisms of neuron and cytoskeleton homeostasis.

GAN is a recessive disease [15] for which we identified the defective protein, gigaxonin [13], an adaptor of Cul3-E3 ubiquitin ligase that is ubiquitously expressed [16] and particularly prominent within the nervous system [25]. To date, ~ 100 mutations have been found in 80 families [35]. Mutations are scattered across the entire GAN gene and are of various types (missense, nonsense, deletion….) with no obvious correlation to disease severity. Despite this broad array of mutations, patients share the common feature of a substantial decrease in the abundance of gigaxonin. Combining structural and in vitro analysis of gigaxonin mRNA and protein, we previously uncovered a general mechanism of instability of disease mutants in patients, regardless of the type or position of the mutation [8, 35].

The first attempts to model the disease focused on the creation of knock-out (KO) mouse strains by depleting early exons in the GAN gene, e.g., exon1 (GAN^Δex1)[18] and exons 3–5 (GAN^Δex3−5)[20, 25]. These early efforts introduced premature STOP codons into the mRNA but failed to recapitulate the severity and the histopathological characteristics of the human disease [18, 25]. Indeed, the homozygous KO mice exhibit mild symptoms with late onset (> 12 months of age), no overt degeneration, absence of giant axons and compaction of NFs, which is seen in patients [35]. Still, defects were observed, including an increased abundance of NF subunits throughout the nervous system [18, 25]. A phenomenon of genetic compensation was invoked to account for the weak phenotype in the GAN-KO mice, and is indeed supported by two facts. First, there are differences in the penetrance of symptoms depending on the genetic background given exclusively mild motor deficits in pure 129/SvJ-genetic background and sensory deficits in pure C57BL/6 animals [25]. Second, exacerbating the NF unbalance by crossing the GAN^Δex1 with peripherin (TgPer) transgenic mice slightly increased the phenotype (motor deficits and ataxia at 12 months of age), but without producing giant axons and NF compaction as they occur severely in patients [46].

In spite of their incompletely penetrant and relatively mild phenotype, the GAN KO mice have been instrumental, together with patient’s cells and other animal models, in revealing the cellular functions of gigaxonin. The first signature of GAN, i.e. the wide aggregation of the IF cytoskeleton [10, 52], has been widely studied in primary fibroblasts of patients. Vimentin, which forms ovoid bundles in patient’s cells [14, 16, 51] was shown to interact with gigaxonin through its rod domain [42]. Conserved between IF types, the rod domain enables the targeting by the gigaxonin-E3 ligase of the entire family towards degradation [10, 35]. This has been demonstrated by the complete destruction of a wide range of IFs in WT and GAN-KO cells upon excessive expression of gigaxonin [27, 42], and the demonstrated gigaxonin-dependent ubiquitination of peripherin and NF-L [27, 49]. Interestingly, the study of GAN cortical neurons from GAN KO mice revealed an unexpected function of gigaxonin, in the formation of the autophagosomes through the ubiquitin-dependent degradation of the ATG16L1 protein [11, 38, 57]. Thus, absence of gigaxonin was shown to diminish the autophagy flux [57], which could further enhance the aggregation of IF proteins and other substrates prone to autophagy degradation.

A key feature of GAN is the wide deterioration of the entire nervous system. Using zebrafish as a model, we uncovered a role for gigaxonin in sustaining motility through the control of motoneuron specification mediated by the Sonic Hedgehog pathway and degradation of the Patched receptor [1, 35]. While it would be interesting to investigate the gigaxonin-dependent specification of other neuronal populations, the control of fundamental developmental pathways by the gigaxonin-E3 ligase opens the intriguing question of a possible developmental origin or contribution to disease. While this remains to be established, creating a robust model for GAN in rodents is crucial, to both scrutinize the mechanisms underlying the wide panel of symptoms in patients and conduct robust preclinical studies for therapy. In this regard, the efficacy of the gene therapy clinical trial developed for GAN (ClinicalTrials.gov ID: NCT02362438) could not be adequately and fully addressed in the GAN-KO animals, with potential benefits only assessed in aged mice [3].

Armed with the knowledge of a possible genetic compensation arising from nonsense-induced transcription of related genes, as a result of premature STOP codon onto the mRNA in the KO strategy [23, 40], we generated a first knock-in (KI) mouse model for GAN. We selected a missense allele (A49E) which, when homozygous in patients [8] results in normal mRNA level but decreases gigaxonin abundance and vimentin aggregation in fibroblasts, in an effort to produce the first robust rodent model for GAN. Combining behavioral and histopathological analysis over a 24 month-period, we demonstrate that GAN^A49E homozygous mice exhibit early-onset and severe symptoms from 3-months onward. Strikingly, the GAN^A49E mice reproduce the hallmarks of the pathology as severely as seen in patients, including giant axons, demyelination, abnormal compaction and increased abundance of NFs. Interestingly, we also document a sex bias in the expression of the phenotype, whereby the degree of muscle denervation and fragmentation of neuromuscular junctions correlates with the severity of the symptoms.

Materials and methods

Generation of the GAN^A49E mutant mouse

The constitutive Knock-In GAN^A49E mice, created by the CIPHE (Centre d’immunophénomique, Aix-Marseille, Université, France), was obtained by site-directed mutagenesis to introduce the c.146C > A substitution (Fig. 1C) in the GAN gene. The targeted point mutation was introduced into the first exon of the GAN gene, substituting a cytosine (C) with an adenine (A) (c.146C > A) at position 146 using a targeting vector containing a NeoR cassette. The targeting vector was generated using BAC clones from a C57BL/6J RPCI-23 BAC library and were electroporated into C57BL/6 embryonic stem (ES) cell line. Correctly targeted clones were detected by Southern blot (EcoRV—BAMHI digest) with two probes that hybridized upstream of the 5′ homology arm or downstream of the 3′ homology arm. The targeted GAN^A49E ES cells were injected into C57BL/6 blastocysts to generate chimeras. Serial backcrossing of the F1 heterozygous with commercially available wildtype C57BL/6 mice (Janvier Labs) generated C57BL/6 GAN mice. Heterozygous breeding allowed us to obtain GAN^+/+ (WT), GAN^A49E/+ (Het) and GAN^A49E/A49E (KI) littermates. Genotyping was performed by PCR using two GAN-specific oligonucleotide primers: the 5′-GAG GAG ATC CCG GTG CAG-3′ (forward) and 5′-ACA GGG TTC TCC TTC TGC-3′ (reverse) (Fig. 1C). Animals were treated in accordance with the European Union guide for the care and use of animals in research (directive 2010/63/UE) and the project was authorized by the French Ministry of Research (APAFIS#9338-2017031712486270 v4). All mice were housed in a 12 h light/dark cycle with food and water ad libitum.

Fig. 1.

Generation of the GAN^A49E mouse model. a Instability of gigaxonin in the GAN patient carrying the A49E mutation (F26 in [8]), revealed by immunoblotting. CosT and controls correspond to ectopic Flag-tagged gigaxonin expressed in COS cells and to 3 unrelated control individuals, respectively. b Immunofluorescence revealing vimentin aggregation (white arrow) in human primary fibroblast derived from the F26 GAN patient. Scale bar 10 µm. c Targeting strategy to introduce the 146C > A point mutation in the first exon of the murine GAN gene. d Detection of a portion of the GAN gene in wild-Type (WT), heterozygote (Het) and homozygous GAN Knock-In (KI) mouse, by PCR: WT and KI alleles generate a 425- and 510-bp PCR product, respectively. e Abondance of gigaxonin in neuronal and non-neuronal tissues from WT and GAN-KI mice at 3 months of age, revealed by immunoblotting. CosT correspond to ectopic Flag-tagged gigaxonin expressed in COS cells; Cereb. = cerebellum; Sc(C) = cervical section of spinal cord; Sc(L) = lumbar section of spinal cord; SN = sciatic nerve.

Reverse transcription and PCR (RT-PCR)

Total RNA from brain samples of 24-month-old WT and KI mice was isolated using the Nucleospin® kit (Macherey–Nagel #7404404) according to the manufacturer’s recommendations. For each sample, 1 μg of RNA was used for reverse transcription with MultiScribe reverse transcriptase from High-Capacity RNA-to-cDNA Kit (Applied Biosystems #4387406). PCR was performed on 50 ng of cDNA, using GAN-exon9-10 specific primers (GGGTAGCGAGATGGTAACTTG and CGGATGGAAGGAGTGGTTTAG) and GAPDH-exon2-3 specific primers (ATGGTGAAGGTCGGTGTG and CATTCTCGGCCTTGACTG), with Go Taq Flexi kit (Promega #M7808).

Behavioral phenotyping

Behavioral studies were performed every 3 months, over a 24 month-period on littermate WT and KI mice.

Rotarod test

Mice were trained at day 0 by letting them to walk freely on the rotarod (Bioseb, France) while gradually increasing the speed of rotation from four to nine rotations per minute (rpm) with acceleration of 1 rpm per minute. They were expected to adapt 1 min per rotation speed tested. Tests were then carried out on days 1–3. Mice were placed on the rotating rod with preset acceleration from 4 to 40 rpm for 5 min. At each testing day, latencies to fall were recorded in three successive trials with 15 min rest intervals between trials. Mice were scored though the average of the last three testing sessions performed on the last day.

Openfield test

Individual mice were placed in the middle of a 45 × 45 cm arena and move freely for 10 min (Bioseb, France). Mouse displacements were continuously recorded by infra-red movement sensors present in the system. Recordings were analyzed with the Openfield Actitrack v2.7 software.

Grid test

Mice were allowed to acclimate to the grid for 5 min before the experiment. Then, mice were tested for their ability to cling to grids harboring different weights (40, 30, 20 or 10 g) and suspended by their tails for 30 s or until they drop the grid. Three such measurements were made per grid and the best time performance was selected, with maximum performance being defined as the ability to stand > 30 s (cut-off limit). Mouse performance was scored for each grid weight included in the design. The strength score was calculated as: .

Hotplate test

Mice were placed on a 53 °C metal surface until they show sign of discomfort (shaking or licking paws, jumping or shouting) (Bioseb, France). The response latencies were recorded three times with 15 min intervals (cut-off 30 s).

Von-Frey test

Mice were isolated for 30 min in wire-mesh floor cages. The plantar surface of each hind limb was stimulated using a metal rod (Dynamic plantar aesthesiometer, Ugo Basile). The intensity of the mechanical stimuli was gradually increased until the mouse withdraws its paw (cut-off at 15 g). The amount of force that generated withdrawal was recorded (five measurements, 15 min apart).

Catwalk™ test

Gait was analyzed using the Catwalk™ system (Noldus Information Technology, Netherlands) in a room where all lights were turned off. Briefly, mice were allowed to voluntarily cross a 100-cm-long, 5-cm-wide walkway with a glass platform illuminated by green fluorescent light. In this device, the green light is scattered when the mouse paw touches the glass floor, producing an illuminated image. The light intensity correlates with the force exerted by the paw. Footprints captured by a high-speed camera placed under the glass floor were analyzed using the CatWalk™ XT 10.1 software. Mice were allowed to spontaneously cross the walkway as often as needed to obtain four compliant runs (minimum of five consecutive step cycles without stop). Data are reported as the average of four runs per mouse.

Clasping reflex

Hindlimb-clasping reflex was carried out three times a week from 6 months of age, by lifting the mice by the tail and scoring their hindlimb positions. When hind limbs were completely extended and spread outwards, a score of 1 was given; if the mice briefly tightened their hind limbs onto their abdomen for a few seconds, it was scored as 0.5; if they manifested such behavior for prolonged time or and in case of no reaction, a score of 0 was applied.

Immunoblotting

Proteins were extracted from tissues homogenized in 5:1 (v/w) of lysis buffer (150 mM NaCl; 50 mM Tris pH 7.5; 0.1 mM DTT, Halt protease inhibitor (Thermo Fisher 78442) and 1 mM PMSF). A total of 20–80 µg proteins were resolved on 8 or 10% SDS–polyacrylamide gels and transferred onto nitrocellulose membranes. The membranes were incubated for 60 min at room temperature (RT) in blocking solution [5% non-fat milk in PBSTw (PBS with 0.05% Tween 20)], overnight at 4°C with primary antibody, washed with PBSTw, incubated for 60 min at RT with the secondary antibody and washed again with PBSTw before revelation. Antibodies were diluted in blocking solution. Primary antibodies used included: mouse anti-gigaxonin [16] (N12, 1:150), mouse anti-NFL (Millipore MAB1615, 1:250), mouse anti-NFH (Millipore MAB5266, 1:200), mouse anti-NFM (Millipore MAB5254, 1:500), mouse anti-GAPDH (Ambion AM4300, 1:4000); they were followed by horseradish peroxidase-conjugated secondary antibodies (Jackson Immunoresearch, 1:5000) that were then visualized using regular ECL substrate (Sigma, WBKLS0100). Quantifications of proteins levels were performed by measuring their band intensities on the blots with the Image J software (https://imagej.net/ij/) and normalized to GAPDH signal intensity.

Morphological analyses

To preserve tissue structures, mice were sacrificed with pentobarbital intraperitoneal injection and perfused transcardially with PBS and then 4% paraformaldehyde (PFA). After dissection, tissues were post-fixed for 4 h in 4% PFA at 4 °C and then kept in PBS with 30% sucrose for 2 days at 4 °C before embedding in Optimal Cutting Temperature medium (OCT) and cryopreserved at − 80 °C.

For electron microscopy analysis, tissues were first kept overnight in modified Karnovsky’s fixative (2.5% glutaraldehyde, 2.0% paraformaldehyde, 0.1% tannic acid in 0.1 M cacodylate buffer pH 7.3). Then, tissues were post-fixed for 60 min in 1% osmium tetroxide (aqueous), dehydrated in graded ethanols and finally embedded in EmBed 812 using an Automated Microwave Tissue Processor for Electronic Microscopy (Leica EM AMW). Axons were manually counted on Semi-thin sections of sciatic nerve and L4-roots (1 μm; Leica-Reichert Ultracut E). Semi-thin sections (1 µm) were collected and stained with 1% toluidine blue and normalized with the area of the nerve determined by the ImageJ software. Myelin thickness, axonal calibers and g-ratio analysis were performed using the g-ratio plug-in (http://gratio.efil.de) [26] in Image J (https://imagej.net/ij). Ultrathin sections (70 nm; Leica-Reichert Ultracut E) were collected at different levels of the block and contrasted with uranyl acetate 1.5% in 70% ethanol, and then lead citrate. Images were acquired on a Tecnai F20 transmission electron microscope at 120kV (Institut des Neurosciences de Montpellier: Electronic Microscopy facilities, INSERM U1298, Université Montpellier, Montpellier France).

Immunohistochemistry

For neurofilaments immunostaining, 20 µm sections of lumbar spinal cord and brain were cut using a NX50 cryostat and stored in AMARAL (Glycerol 30%, Ethylene glycol 30%, phosphate buffer 0.1M) at − 20 °C. Five tissues per animal for the spinal cord and three tissues per animal for the brain were analyzed. Floating tissue sections were washed three times with PBS before blocking for 120 min at RT with the blocking buffer: PBS, 0.3% Triton X-100 (BP15-500, Fisher Scientific), 3% Horse Serum (Sigma H0146) and 4% bovine serum albumin (BSA, Sigma A9418). Tissues were then incubated overnight at 4 °C with the following primary antibodies: mouse anti-NFL (Millipore MAB1615, 1:200), mouse anti-NFH (Millipore MAB5266, 1:100), mouse anti-NFM (Millipore MAB5254, 1:100). On the next day, tissues were washed again with PBS with 0.1% Triton X-100 (PBST 0.1%) and incubated with the secondary antibodies (from Jackson Immunoresearch) anti-mouse AlexaFluor 594 (1:500) and anti-rabbit Alexafluor-594 (1:500) for 90 min at RT. After several washes with PBST 0.1%, spinal cord tissues were incubated with NeuroTrace Stain (ThermoFisher Scientific, N21480, 1:500) for 20 min at RT, and then washed again with PBST 0.1% and PBS. Spinal cord tissues were mounted with Fluoromount G (00-4958-02, ThermoFisher Scientific), and brain tissues with Fluoromont G-DAPI (00-4959-52, ThermoFisher Scientific). Images were acquired on a Zeiss 880 confocal microscope and analyzed using the Image J software.

For neuromuscular junctions, teased fibers from the gastrocnemius muscle were kept in AMARAL at − 20 °C and rinsed 3 times before blocking for 60 min at RT with blocking buffer (PBST 0.5%, 4% BSA, 5% normal goat serum). Primary antibodies anti-SV2-c and anti-2H3-c (DSHB, 1:200 diluted in the blocking buffer) were applied for two consecutive nights at 4 °C. Fibers were washed three times before incubating with alpha-bungarotoxin 488 (Invitrogen, B13422, 1:500) and secondary antibodies anti-mouse AlexaFluor 594 (Jackson Immunoresearch, 1:500) in the blocking buffer for 60 min at RT. Three washes were performed before mounting using Fluoromount G (00-4958-02, ThermoFisher Scientific). Images were acquired on a Zeiss LSM880 confocal microscope with a 63X/1.4 NA oil immersion objective for the NMJ structure and on a confocal spinning disk microscope high content screening CQ1 (Yokogawa) with a 40X/0.95 NA objective for the number of MNJ per muscle fiber. Images were treated and analyzed with ImageJ software, with an ImageJ NMJ-morph macro [29, 43].

For studies focused on muscle, tissues were stained on 10µm cryosections or on teased fibers. For transversal analysis, muscle sections, directly on slices, were permeabilized for 10 min with PBST 0.5%, washed three times with PBS, blocked for 60 min using PBS with 2% BSA, and incubated with the primary antibody rabbit anti-laminin (Merck, L9393, 1:200) overnight at 4 °C or for 2 h at 37 °C in a moist chamber. Samples were then washed three times with PBS and incubated with a secondary anti-rabbit AlexaFluor 488 antibody (Jackson Immuno-research, 1:500) for 45 min at 37 °C in a moist chamber or for 60 min at RT. After five washes, Hoechst counterstaining (1:1000) was applied for 5 min and quickly washed once with PBS. Processed sections were then mounted using Fluoromount G (00-4958-02, ThermoFisher Scientific). Images were acquired on a Zeiss Axio Observer Z1 microscope using MetaMorph Software. Muscle fiber areas were quantified using the semi-automated ImageJ macro Open-CSAM [19].

Mouse embryonic fibroblast culture and immunostaining

Animals were treated in accordance with the European Union guide for the care and the use of animals in research (2010/63/UE). MEF were obtained by dissection and dissociation of E14.5 disemboweled embryos in 0.05% trypsin–EDTA. After centrifugation at 150 g for 5 min at RT, cells were resuspended and plated in Dulbecco’s modification of Eagle medium (DMEM) medium (Thermo Fischer) containing 15% fetal bovine serum, 1% non-essential amino acids, 1% sodium pyruvate, 1% L-glutamin and 1% Penicillin/Streptomicin. Genotyping of embryos was performed after MEF plating, as outlined above. MEFs were fixed in PFA 4% at RT for 20 min and washed 3 × 10 min with PBS 1X. Blocking and permeabilization were performed in PBS/0.1%Triton, 4% Bovine Serum Albumine (Sigma), 4% Donkey Serum (Sigma) for 1 h at RT. Subsequently, preparations were incubated overnight with anti-vimentin antibody (V2258 Sigma) diluted in blocking buffer at 4 °C. Following washes with PBS/0.1%Triton, preparations were incubated with secondary antibodies were incubated for 1 h at RT, and mounted with Fluoromount.

Statistical analyses

Statistical tests applied are described within the figure legends for each graph. Graphs and statistical calculations were made using GraphPad Prism version 6.0 for Windows, (GraphPad Software, San Diego, California USA, www.graphpad.com) or R version 3.6 [53]. For behavioral phenotyping, data were analyzed by two-way analysis of variance (ANOVA) at 3, 18, 21 and 24 months, or using three-way ANOVA on repeated measures from the age of 6 months to the age of 15 months. Pairwise comparisons were calculated with pairwise t-tests or pairwise Wilcoxon rank sum tests when the data did not follow a Gaussian distribution. Distributions of axonal calibers were compared using Chi-squared test. A regression analysis was run to examine the relationship between the g-ratio and the axon diameter in sciatic nerve, dorsal and ventral roots. t tests and Mann–Whitney test for nonparametric data were also performed when analyzing only two groups. Comparisons of percentages for NMJs and muscle area distribution were performed with Chi-squared tests.

Results

Generation of the GAN^A49E knock-in mouse

To overcome potential genetic compensation arising when the GAN gene is inactivated, as suspected in mildly affected GAN-KO mouse models [18, 20, 25], we created a knock-in model with a mutation resulting in expression of the mutated form of the GAN mRNA. Among the numerous missense mutations identified in patients across the GAN gene [35], we selected the A49E variant, which fulfills the criteria of normal mRNA expression and disease severity. The A49E mutation was identified in a severely affected patient (F26 in the original publication [8]), who developed early deficits at 2 years of age, a severe sensorimotor decline with loss of ambulation at age 8, and widespread symptoms in the CNS at age 10. Importantly, we showed that while the A49E mutation leads to gigaxonin instability ([8], Fig. 1a and Supplementary file 1), presumably due to impaired homodimerization [8], its mRNA level is not compromised in patient [8]. Moreover, aggregation of several IFs has been evidenced in patient samples, including NF compaction in giant axons within peripheral nerves [8] and vimentin bundles in skin-derived fibroblasts (Fig. 1b).

The homozygous GAN^A49E mouse was obtained using a gene targeting strategy (Fig. 1c), by substituting the nucleotide C at position 146 with A in the first exon of the GAN gene (c.146C > A). The presence of the inserted 146C > A mutation, associated with a NeoR residue was further confirmed by PCR (Fig. 1d). Immunoblotting showed no detectable gigaxonin in KI tissues, including neural tissues in which gigaxonin is enriched in control mice (Fig. 1e and Supplementary file 1). As is the case in the F26 patient, we show that the A49E mutation in the murine GAN gene does not alter mRNA levels (supplementary Fig. 1a) and induces vimentin aggregates in primary fibroblasts of mutant mice (supplementary Fig. 1b). These data confirm the creation of the first GAN knock-in model, exhibiting instability of the A49E-mutated gigaxonin and collapse of the vimentin IF network, as observed in patients.

GAN^A49E mice develop early onset, severe sensory and motor phenotypes and ataxia

Considering the severity of the sensory-motor symptoms in GAN patients, we conducted a behavioral study on the GAN^A49E mice over a 24-month-period, analyzing both sensory and motor performances (Fig. 2). As we noticed a possible variation between sexes, females and males were analyzed separately.

Fig. 2.

GAN^A49E mice develop early onset and severe sensory and motor phenotypes. Longitudinal analysis of the behavior of WT and KI mice (black and red, respectively), separated by sex (females in circles and males in triangles) over a 3–24-month period. Please note that connected points represent the same cohorts of animals. 18, 21 and 24 time points are represented separately due to the inequal number of animals in different groups, as a result of occasional death of animals during aging. The 3-month data set is separated because it corresponds to another cohort, generated afterwards. a Body weight measurement reveals an overall normal growth curve in KI mice. b Sensory deficits in KI mice are evidenced upon thermal (Hotplate test, left panel) and mechanical (Von-Frey test, right panel) stimulation. c The decreased motor performances of KI mice are evidenced in the Rotarod (left panel), the Openfield (middle panel) and the Grid test (right panel). d Ataxic behavior is evidenced in KI mice, with decreased distance between right and left front paws and increased spacing between hind paws using the Catwalk (BOS: Base of Support). All data are presented as mean ± standard error (SE) or median ± SE depending on the normality of the distribution of data. For longitudinal tests (Hotplate, Von-Frey, Rotarod, Openfield), two-way ANOVA (at 3, 18, 21 and 24 months) or three-way ANOVA (from 6 to 15 months of age) are performed to account for sexe, genotype and time. For the Hotplate test, n = 8–15/group at 3 months of age, 9 between 6 and 15 months and 4–7 at 18 months. For the other tests, n = 11–17 mice/group at 3 months; 13 mice/group between 6 and 15 months excepted for the catwalk (n = 12) for which we were unable to perform the test for one mouse at least at one time point; 9–13 mice at 18 months; 7–12 mice at 21 months; and 5–9 mice at 24 months of age. The grid test is performed at 24 months of age, n = 7–13 animals/group. P-values are calculated using pairwise comparison tests to evaluate statistical significance between WT and KI mice for both sexes. e Qualitative analysis of the gait reveals abnormalities in 50% of KI mice in the 6–24-month-period, most of them being severely affected with paralysis. To note, among 27 control mice, only one male aged mouse (23-months of age) exhibited a mild phenotype. Statistical analysis is performed using a Chi-square test, n = 27 mice WT (13 males and 14 females) and 32 mice KI (17 males and 15 females). Bottom: Representative picture of a KI mouse with paralysis of hind limbs. f Scoring of the clasping reflex (in arbitrary units – a.u.) of hindlimbs reveals defects in KI mice over the 6–24-month-period. Statistical analysis is performed using Mann–Whitney test, each point is a mouse score, data are presented with median, n = 35 mice WT (17 males and 18 females) and 43 mice KI (23 males and 20 females). Right: Representative postures of WT and KI mice. For all tests, statistical significance is represented by asterisk:.p = 0.06, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. See also Supplementary videos 1, 2 and 3.

Overall, we did not observe any weight difference between WT and KI mice, with only variations at 6 and 15 months of age for females and males, respectively (Fig. 2a). Strikingly, GAN^A49E mice develop strong sensory deficits, with insensitivity to both thermal and mechanical stimuli that are more pronounced in females (Fig. 2b). We observed a significant and progressive increase in the latency to exhibit discomfort signs in the Hot plate test for KI mice between 9 and 15 months (Fig. 2b left panel). This effect appears first in females, from 9 months onwards (17.73 ± 1.18 and 14.67 ± 1.02 s, respectively for KI and WT mice), and at 12 months for males (17.63 ± 0.83 and 14.82 ± 0.49 s, for KI and WT mice, respectively). Regarding mechanical responses, a persistent decrease in sensitivity was detected from 6 months in KI female (Fig. 2b right panel), with retraction of paws upon filament inducing a pressure of 6.29 ± 0.37 g for KI versus 5.43 ± 0.32 g for WT mice.

To evaluate motor performances of GAN^A49E mice, we associated three behavioral tests, using assisted (Rotarod) and free (Openfield) locomotor assays, and measurement of the grip strength (Grid test) (Fig. 2c). Overall, we confirmed an early and severe impairment of motor functions in both females and males KI mice. Indeed, the Rotarod test revealed a robust and persistent impairment of motor coordination and/or fatigue and muscle weakness in both sexes, starting at 3 months of age (Fig. 2c left panel). Of note, the locomotion defects of KI mice at 12 and 15 months are masked by the decline of the capacities of WT mice at these ages. In addition, the spontaneous motor activity, assessed with the Openfield test, revealed a persistent reduction from 6 months of age in female KI mice (Fig. 2c middle panel). We also conducted a Grid test, to measure the muscle strength at 24 months of age. Regardless of sex, KI mice exhibit a decreased ability to hang on to the grids of increasing weight (Fig. 2c right panel).

As GAN patients present ataxic cerebellar signs, we measured various features of the gait in GAN^A49E mice, using the Catwalk system. The longitudinal analysis revealed persistent alterations of gait parameters from 3 months of age, that are more pronounced in females (Fig. 2d). In particular, average distances between the front and hind paws significantly differ between KI and WT mice. For front paws (Fig. 2d left panel), KI mice have a smaller paw spacing than their WT littermates (1.43 ± 0.03 vs. 1.52 ± 0.02 cm at 9 months and 1.39 ± 0.06 vs. 1.58 ± 0.03 cm at 24 months). The opposite effect was observed for the hind paws, which are positioned with a larger distance in KI female (2.86 ± 0.06 cm at 9 months) compared to WT (2.47 ± 0.06 cm) (Fig. 2d right panel). These results suggest an alteration of the balance in GAN^A49E mice, which may attempt to compensate with a stable triangle shape of their limbs.

To complement this quantitative analysis of symptoms, we monitored the spontaneous mobility of mice at regular intervals (three times a week, from 6-months to 24-months of age) (Fig. 2e, Supplementary Video 1, 2 and 3). We categorized behavior into three groups: absence of abnormalities (Supplementary Video 1), mild troubles encompassing impairment of the balance or tactile sensitivity, abnormal positioning of the paws on a grid leading to falling (Supplementary Video 2) and severe defects with paralysis of at least one hind limb (Fig. 2e bottom and Supplementary Video 3). Strikingly, our analysis reveals that 50% of KI mice exhibit symptoms, amongst which 62,5% have a severe phenotype. Sex-dependent analyses revealed alterations in both sexes, but with a greater severity in females (80% of abnormalities are paralysis, versus 54,5% in males) (Fig. 2e upper panel). Concomitantly, we performed a hindlimb clasping test, a measure of reflex that is frequently used in genetic models of neuromuscular disorders. Thus, we scored the ability of mice to extend their hindlimbs (score of 1), as seen normally upon suspension by the tail (Fig. 2f). To reduce the variability in responses that is occasionally seen in controls, mice were tested individually three times a week (scores 1, 0 or intermediate 0,5), and mean reflex scores were obtained for individual mice over the 6–24 month-period. Robustly, the clasping test revealed a reduction of the reflex in KI mice in both sexes, hence providing another demonstration of their altered neuronal capacity.

Giant axons and demyelination in GAN^A49E nerves

To determine whether the severe sensory-motor defects seen in GAN^A49E mice may result from neuronal loss, we carried out neuronal and axonal counting on cross sections of the spinal cord and proximal/distal nerves from 24-month-old mice. We first quantified the number of motoneurons in the ventral horn of lumbar spinal cord and this revealed no overt difference between WT and KI mice (Supplementary Fig. 2). Secondly, we quantified both the overall surface and the total axonal count in proximal (ventral roots, VR, composed of motor fibers, dorsal roots, DR composed of sensory fibers) and distal (sciatic nerve, SN, with mixed fibers) regions of the nerve. Measurement of nerve surface area on toluidine blue-stained semithin sections revealed no difference between WT and KI proximal nerves but evidenced a significant larger surface in the distal nerves of KI mice (Supplementary Fig. 3a and b). The axon counting showed a constant total number of axons between WT and KI mice, in both proximal (roots) and distal (SN) regions of the nerve (Supplementary Fig. 3c), suggesting that the increased surface of the SN in KI mice may result from increased axon diameter. Accordingly, the distribution of axon calibers in the SN evidenced enlargement of axons, while roots display a dual shift towards both smaller (≤ 6 µm) and larger calibers (≥ 13 µm) in KI mice (Fig. 3a and b). Importantly, these analyses revealed the presence of extremely enlarged axons (up to 34 µm in diameter) both in proximal and distal portions of the nerve (Fig. 3a and Supplementary Fig. 3a). To our knowledge, this represents the first demonstration of the presence of giant axons, the main pathological hallmark of the human disease [2, 5], in a mouse model for GAN (see [18, 20, 25]).

Fig. 3.

GAN^A49E mice exhibit giant axons and demyelination in peripheral nerves. a Semithin cross-sections of sciatic nerve (SN), L4 dorsal (DR) and ventral (VR) roots of 24-month-old WT and KI mice, stained with toluidine blue. White asterisks indicate giant axons. Scale bars 20 µm. b Histograms (means ± standard deviation) of the distributions of axonal calibers in KI mice and their control littermates. In DR and VR, KI mice present more smaller and larger axons, as evidenced by an exact binomial test. Note that the statistical significance of the difference seen in the sciatic nerve could not be obtained (n.a, for non-applicable) due to the absence of axons with ≥ 13 µm in controls. c Scatter plots of g-ratios, as function of axonal diameters in SN, DR and VR, showing increased values in peripheral nerves of KI mice. P values calculated using t tests evidence differences in the regression slopes between KI and WT mice. In all nerve sections in panel c, representations and statistical relevance are provided separately for males and females. N = 200 axons per animal, selected randomly throughout the sections, n = 3–4 animals per genotype and sexe. p = 0.09,*p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

Consistent with data obtained from patient biopsies [2, 21, 48, 50, 52], giant axons from GAN^A49E mice have almost no myelin sheaths, as a result of excessive axonal size or secondary demyelination (Fig. 3a). As myelination is correlated with the size of the axon, we further quantified the myelination of axons in all nerve portions. This was achieved by measuring the g-ratio, which corresponds to the ratio between inner and outer (myelinated) axonal diameters. First, we showed that overall, the average g-ratios do not change between individual KI and WT mice in the SN and DR but it is higher in VR from KI mice, suggesting a massive demyelination of motor axons (Supplementary Fig. 3d). Secondly, to determine more precisely how myelin thickness varies as a function of axon diameter, we performed a regression analysis of the g-ratios and axon diameters. In both WT and KI mice, g-ratios exhibit a normal increase with axonal caliber as shown by the trend lines (Fig. 3c). Moreover, we observed a strong effect of the genotype on the interaction between the g-ratio and axonal diameter in proximal portions of the nerves, specifically for females. Importantly, the distribution of data revealed that the increased g-ratios (decreased myelination) in KI females target large caliber axons in DR, and small axons in VR. Considering the high proportion of small-middle range axon in sections, this may explain why the alteration of g-ratio was only evidenced in VR in the global population (Supplementary Fig. 3d).

Altogether, these results represent the first evidence of the histological hallmarks of the human disease in a GAN mouse model, with the presence of giant axons and myelin defects throughout peripheral nerves.

Disorganization, compaction and increased abundance of neurofilaments in GAN^A49E mice

In GAN patients, nerve tissue biopsies feature giant axons filled with disorganized NFs and clustering/exclusion of microtubules and organelles at the axoplasm periphery [2, 22]. Therefore, we analyzed the ultrastructure organization of the axoplasm in transverse sections of peripheral nerves of 24-month-old WT and KI mice (Fig. 4a for DR and Supplementary Fig. 4 for VR and SN). Strikingly, in both proximal and distal regions of the nerve, KI mice exhibit an abnormal distribution and compaction of NFs, in giant but also in axons of normal caliber. In KI nerves, NFs lose their parallel orientation along the axon and accumulate in dense bundles, to the same degree than what is found in patients (Fig. 4a).

Fig. 4.

Disorganization and compaction of NFs, and misdistribution of damaged organelles in GAN^A49E mice. Electron micrographs of the axons in a the dorsal roots and b the sciatic nerves of WT and KI mice (KI#1 and #2), and in a GAN patient biopsy (from Asbury et al. [22] Copyright with permission from Springer link). a White arrows point dense electron materials common between KI and patient samples. Higher magnifications of the axoplasm are taken form regions marked with white stars. Similar to patients, NFs are densely packed and disorganized in KI mice, with rare microtubules (back arrows). b In KI mice, various organelles are damaged, segregate and accumulate at the periphery of the axoplasm: mitochondria (arrowheads), autophagic structures (thin arrows), lysosomal vesicles (thick arrows) and dense material (long arrows). Scale bar (1 µm) and their higher magnification (200nm). See also Supplementary Fig. 4.

Similar to patient biopsy, densely packed NF regions in KI axons are devoid of microtubules, which are normally interspaced with NFs in control mice (Fig. 4a and Supplementary Fig. 4). Conversely, some axons that are devoid of NFs are completely occupied by microtubules, a feature that has been reported in GAN patients (Supplementary Fig. 4c).

GAN^A49E mice exhibit additional similarities with patient biopsies. In particular, our analyses revealed the presence of multiple electron-dense material across the axoplasm (Fig. 4a) which could correspond to NF aggregates, and organelles which are often pushed at the periphery of the axoplasm (Fig. 4b). In addition to their mislocalization, several organelles, including mitochondria, lysosomes and double-membrane autophagic vesicles, can be found accumulated and damaged in KI samples (Fig. 4b).

In light of the dramatic compaction of NFs in the peripheral nerves of KI mice, we assessed the distribution (Fig. 5) and abundance (Fig. 6) of the NF-L, NF-M and NF-H subunits within the CNS. Immunostaining of sections from 24-month-old KI mice revealed both peri-nuclear aggregation in the cortex and enhanced axonal signals in the white matter for NF-L and NF-H (Fig. 5, left, middle), with only modest effect observed for NF-M. This was confirmed by immunoblotting, which show increased abundance of NF-L and NF-H (and also NF-M) in the brain of KI mice (Fig. 6a and b and Supplementary file 1). Interestingly, this enrichment affects both soluble and insoluble (polymerized) forms of the NF subunits, with, respectively, a 4.05-, 3.14-, 2.7-fold increase for NF-L, NF-M and NF-H in the supernatant fractions, and 2.51-, 2.2-, 3.19-fold in the pellets. Overall, NFs are less compromised in the spinal cord of KI mice, with possible modest alterations of NF-L and NF-H (Fig. 5, right) and no significant difference in the abundance of NF subunits in KI tissues (Fig. 6a and b).

Fig. 5.

Aggregation of NFs in the central nervous system of GAN^A49E mice. Transversal sections of brain cortex, white matter and ventral horn of lumbar spinal cord from 24-month-old mice are immunostained for NF-L, -M and -H and DAPI or Nissl. KI samples exhibit increased abundance and aggregation (white arrows) of all neurofilament subunits in CNS, particularly NF-L and NF-H in the brain cortex and the white matter. Higher magnifications are taken form regions marked with white stars. Scale bar 20 µm.

Fig. 6.

Increased levels of NF subunits in the brain of GAN^A49E mice. a Immunodetection of NF-L, NF-M and NF-H subunits from soluble (supernatants, SN) and insoluble (pellets) fractions from brain and lumbar spinal cord (Sc(L)) of 24-month-old mice. b Histograms (medians with interquartile ranges) representing the quantification of the relative NF levels in KI tissues over control (WT) samples. The mean values of three experiments per animal are normalized to GAPDH and to WT means, n = 4 animals per group (2 females and 2 males). Statistical differences are evidenced by Mann–Whitney tests *p ≤ 0.05.

Alterations of neuromuscular junctions and muscle atrophy in GAN^A49E mice

Considering the numerous macroscopic and microscopic (ultrastructural) alterations seen in GAN^A49E axons, we next set out to assess whether this could alter the integrity of the synapse and/or muscle, possibly underlying the severe locomotor deficits seen at the behavioral level.

We first examined the neuromuscular junction (NMJ) in the gastrocnemius muscles of KI mice (Fig. 7). To preserve the three-dimensional structure of the NMJ, we teased muscle fibers from 24-month-old mice and immunostained the preparations with NF-M and synaptic vesicles (SV2) to identify the presynaptic structures, and α-bungarotoxin to label acetylcholine receptors in the post-synaptic region (Fig. 7a). This analysis revealed numerous alterations of the NMJ in GAN^A49E mice which, interestingly, differ between sexes. We observed that NMJ of KI females are severely damaged, showing gross denervation (Fig. 7a and b) or fragmentation (Fig. 7a and c). More detailed examination of the NMJ structure in female preparations revealed specific alterations of the presynaptic compartment, with a decreased of axon terminal diameter (Fig. 7d) and a reduction of the contact sites with muscle fibers (Fig. 7e). By comparison, NMJ preparations from KI males do not show overt denervation (Fig. 7b) or fragmentation (Fig. 7c), but are overall smaller with reduced pre- and post-synaptic contact sites (Fig. 7e and f), together with a decreased axonal diameter (Fig. 7d).

Fig. 7.

Severe alterations of the neuromuscular junction in GAN^A49E mice. a NMJs from gastrocnemius muscles of 24-month-old female and male, WT and KI mice are labeled with NF-M (2H3) and SV2 for the pre-synaptic terminals and fluorescently-tagged α-bungarotoxin for the post synaptic areas. Scale bar 20µm. Images were treated and analyzed with ImageJ software, with an ImageJ NMJ-morph macro [29, 43]. b The overlap between pre- and post-synaptic regions of the NMJ is reduced in females KI mice, reflecting a reduction of innervation. c NMJs from female KI mice are fragmented compared to WT (20% VS 5,7%) and to male KI animals. d Decreased axon diameter at the terminal segment in KI mice, as seen by NF-M/SV2 staining. e KI mice exhibit a decreased in the pre-synaptic area. f Males KI have a significant decrease of the post-synaptic acetylcholine receptor area. g, h Muscle fibers are mostly mono-innervated in control mice, while denervation and poly-innervation increase, respectively in female KI and male KI mice. Scale bar 20µm. For b–f, n = 49–57 NMJs from 2 animal per group, b, d–f are represented as medians ± interquartiles and analysis are performed using a Mann–Whitney test, c is analyzed using a Fisher’s exact test. For h, n = 230–539 muscle fibers from 2 animal per group, and analysis is performed using a Chi-square test. For all, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

Other interesting features in KI mice came to the fore when examining innervation at the tissue level. In normal conditions, muscle fibers are mostly innervated by single axons (mono-innervated), with appearance of poly- and non-innervation during aging (see 24-month-old WT mice in Fig. 7g and h). We observed that muscle fibers from KI female have increased denervation (18.7% in WT to 41.3%) (Fig. 7h in agreement with Fig. 7b), whereas, on the contrary, poly-innervation is enhanced in fibers from KI males (30.4% in WT to 46.4%) (Fig. 7g and h). Altogether, these results evidence not only a more pronounced alteration of the NMJs in KI females, but also an induction of poly-innervation in males. This phenomenon is reported as a recovery mechanism upon nerve injury [4, 6, 41] and, while evidenced at later stage (24-month-old mice), it might explain the less penetrant locomotor dysfunctions seen in males with GAN (Fig. 2).

Finally, we analyzed the morphology of muscles in GAN^A49E mice (Fig. 8a). We performed immunostaining for laminin on transversal sections of gastrocnemius muscles and evidenced a global reduction of the surface of the muscle fibers in KI mice (Fig. 8b). This was accompanied by a shift in the distribution of muscle fiber size towards smaller fibers (Fig. 8c), with a maximum of fibers measuring 1000 µm² in KI muscle compared to 1400 µm² in WT. Closer examination of individual fibers revealed a fourfold increase of central nuclei in GAN-KI mice, a characteristic feature of degeneration and impaired regeneration of the muscle.

Fig. 8.

GAN^A49E mice exhibit muscle atrophy with increased central myonuclei. a Longitudinal sections from gastrocnemius muscles (dashed lines) of 24-month-old WT and KI mice, labeled with laminin and DAPI. Scale bar 1000 µm. b Decrease of muscle fiber area in KI mice. Histograms represent medians with interquartile ranges. Statistical difference is calculated using the Mann–Whitney test. c The distribution of muscle fiber area significatively differs between KI and WT mice, with a shift towards smaller fibers in KI muscles. Statistical difference is calculated by a Chi-square test. The central positioning of nuclei in myofibers, seen using laminin and DAPI staining (d) is increased in proportion in KI tissues. Scale bar 100 µm. Statistical significance obtained with a Fisher’s exact test: ****p ≤ 0.0001, n = 3410 and 3941 muscles fibers for WT and KI, respectively with 4 animal per group (2 females and 2 males).

Altogether, these findings demonstrate that in parallel to neuronal dysfunction, KI mice exhibit alterations of the NMJ, muscle atrophy, and pericentral nuclei.

Discussion

Previous attempts to model GAN in mice failed to reproduce the severity of the pathology [35]. Indeed, inactivating the GAN gene (GAN-KO) impacts NFs but to a lesser extent than what is seen in patients, and does not reproduce the giant axons, which are hallmarks of the pathology. In addition, depleting the GAN gene generates only mild sensory and motor symptoms, with late onset (> 12 months). In line with the discovery of genetic compensatory responses arising from the degradation of truncated mRNA in KO designs [23, 40], we demonstrate here that introducing, in the GAN locus of mice, a patient mutation that does not alter mRNA stability produces an early-onset and severe phenotype, and recapitulates the histological and structural hallmarks of the human pathology.

The GAN-KI mouse was generated by reproducing the A49E missense mutation, selected for the severe and early-onset expression of symptoms in patient, the decreased abundance of mutated gigaxonin in lymphoblasts cell lines but with normal mRNA level, and aggregation of IFs in the nerve biopsy and skin-derived fibroblasts [8] (Fig. 1a and b, Supplementary Fig. 1). In agreement with the predicted instability of the A49E protein from impaired dimerization, we could not detect gigaxonin in all tissues from the GAN^A49E mice (Fig. 1e).

Our longitudinal behavioral analyses revealed robust and persistent symptoms in each of the eight tests, involving both the peripheral and central nervous systems (Fig. 2). Motor functions were impacted first, with a decline as early as 3 months of age, followed by sensory deficits starting at 6 months of age. Anomalies observed include decreased performances on assisted (Rotarod) and spontaneous (Openfield) motility, reduced grip strength (Grid) and diminished sensitivity to heat (Hot plate) and touch (Von Frey). The GAN-KI mice also display the ataxic gait of the human pathology, starting at 3 months with abnormal positioning of the hind paws and subsequently front paws. From 6 months of age, an impaired balance is seen in 50% of the GAN^A49E mice, evolving towards paralysis in 62,5% of the mice from 15 months onwards. An additional feature is the persistent reduction of the clasping reflex, as reported in rodent models of neuromuscular diseases. Interestingly, we evidenced a more penetrant phenotype in females in most of the behavioral tests, a feature that will be further discussed below.

For the first time, the histological and structural hallmarks of the human pathology are expressed with the same severity in the GAN-KI model. Indeed, as in nerve biopsies from patients, we evidenced giant axons that can reach up to 34µm in diameter in both proximal and distal portions of the nerves (Fig. 3 and Supplementary Fig. 3). Similarly to patients [2, 21, 48, 50, 52], these enlarged axons are demyelinated (Fig. 3a), as are the large caliber axons in the dorsal root and smaller axons in the ventral root (Fig. 3c). Our results are consistent with the imaging and electrophysiological analysis performed in patients, revealing a secondary sensory-motor demyelination [17, 45]. In addition, the ultrastructural analysis of giant axons revealed both an alteration of orientation and compaction of the NF array, as dramatic as in patient biopsies [2, 22], and this in both proximal and distal parts of the nerves (Fig. 4 and Supplementary Fig. 4). Interestingly, NF alteration can also be seen in normal-sized axons, indicating that enlargement of axon is not solely caused by the local compaction of NFs. In the CNS, NFs and in particular the NF-L and NF-H subunits form dense perinuclear bundles in the cortex, associated with massive accumulation in axons for NF-L and bundling for NF-M and NF-H (Fig. 5). Measurements of protein content in the brain revealed an overall increase of all NF subunits in both soluble and insoluble fractions (Fig. 6). The latter demonstrates that NF aggregation involves not only a reorganization of the NF array but also a defective turnover of the soluble and polymerized forms of NFs, as suggested by the E3 ligase activity of gigaxonin and its role on NF-L ubiquitination [49]. In the spinal cord, NFs only show modest alterations and no overt changes in protein abundance in the GAN^A49E mice (Fig. 5 and 6). Those results are in agreement with our previous GAN-KO model, which also exhibits preferential alteration in the brain [25], yet to a lower extend than in the KI mice. Overall, GAN^A49E mice exhibit a stronger phenotype with increased abundance of NF subunits, greater compaction of NFs and giant axons, hence supporting the hypothesis of a genetic compensation in the KO model. Recently, another group showed that crossing the mild GAN-KO mice with TgPer mice enhances NF unbalance and induces neurodegeneration at 12 months of age but, overall, motor deficits are late onset and neither giant axons nor densely packed NFs could be evidenced [46]. Also, various characteristics of the GAN^−/−;TgPer mice do not significantly differ from TgPer mice or GAN^−/−.

Interestingly, the GAN^A49E model exhibits additional features of the human disease. First, microtubules are no longer interspaced with NF in nerves, but excluded from NF-dense area (Fig. 4 and Supplementary Fig. 4). Occasionally, as seen in patients [22], axons are missing NFs and appear entirely filled with microtubules (Supplementary Fig. 4), a phenomenon explained by the presence of a tubulin-binding site in the head of NFs that acts as a negative regulator of microtubules polymerization [7]. Second, electron-dense material was found across the axoplasm of GAN^A49E mice (Fig. 4a), which may represent NF aggregates [22, 58]. Third, again as seen in patients [22, 60], organelles (mitochondria, lysosomes and autophagic vesicles) are damaged, mislocalized, and pushed at the periphery of the axoplasm (Fig. 4b). This may be caused by an alteration of their transport within the axons. Indeed, IFs have been shown to control the axonal transport of organelles [12, 30, 61], and alteration of NF content, as seen upon NF accumulation or NF loss due to, respectively dominant and recessive mutations in the Nefl gene in Charcot-Marie-Tooth disease is sufficient to alter the motility of organelles [30]. Therefore, the defective transport of mitochondria seen in patient fibroblasts and GAN DRG neurons [27, 39, 47] and the mislocalized organelles seen in the GAN KI model may result from the bundling of the IF network.

Overall, the GAN^A49E mouse currently represents the first phenotypic mouse model for GAN, which exhibit early-onset, severe sensorimotor deficits and the histological hallmarks of the human disease, with giant axons filled with densely-packed NFs.

The only feature that is not reproduced in our model is the neuronal degeneration that could be observed in the rare autopsies of patients [32, 59]. Indeed, we could not identify overt degeneration of motoneurons in the spinal cord in KI mice (Supplementary Fig. 2), nor axonal loss in nerves (Supplementary Fig. 3). Assessing whether this aspect preserves KI animals from the premature death reported in patients is something that remains to be elucidated. Alternatively, considering that our histopathological analyses were conducted at 24 months of age, general neurodegeneration might occur in older mice, thereby masking the specific effect of the GAN^A49E allele in KI mice, as already seen for the assessment of motor performances in the Rotarod test (Fig. 2). If not the case, our data would demonstrate that without neurodegeneration, perturbations of axonal integrity is sufficient to generate severe physiological symptoms.

Still, GAN^A49E axons exhibit a significative increase in diameter (including giant axons) along the nerves, and an additional shift toward smaller axons in roots (Fig. 3). This impact on large and small axons may be explained by the local abundance and distribution of NFs along the axon. On the one hand, giant axons have been consistently evidenced in genetic conditions leading to NF aggregation (dominant forms of CMT1F and CMT2E) [30]. On the other hand, reduced axon caliber has been featured in conditions reducing NF content (as seen in recessive CMT2E forms and NF-genetically modified mice) [33, 34, 56]. This is in agreement with the local aggregation of NFs along the nerve in the GAN^A49E mice (feature that has also been described in patients [2, 9, 48]), and the presence of axons depleted in NFs (as seen in electron microscopy images in supplementary Fig. 4).

In this study, we analyzed parameters that are not typically examined in patients, thus yielding new insight into the pathogenicity of the disease. First, the analysis of NMJ revealed a sex-effect in the GAN phenotype, with increased severity in females. More specifically, female KI mice exhibit denervation of the muscles, with the remaining NMJ being fragmented and altered in the presynaptic compartment. Less severely affected, NMJ of male KI mice are smaller (both pre and post synaptic compartments), and exhibit a pattern of poly-innervation indicating a process of degeneration-reinnervation. This adaptive response has been described following denervation upon nerve injury in frog, rat and mice, where the level of polyneuronal innervation at reinnervated NMJs was found to be increased [4, 41, 62]. Considering that females have significant more pronounced sensory-motor deficits than males, it would be interesting to determine whether the severe damages at the synapse, seen at later stage (24 months of age) could be contributing to the physiological deficits in females. Thus, our study may indicate a correlation between the degree of deterioration of the NMJ and behavior, and might suggest the existence of either compensatory mechanisms that are preferentially activated in males or, conversely, impaired regenerative capacity occurring preferentially in females. Identifying the sex-dependent factors (such as hormones) that may act as modifiers of the pathology now represents an issue of great interest. Second, we found that, regardless of sex, gastrocnemius muscles of GAN^A49E mice are atrophied, with a shift of fibers towards smaller size and presence of central nuclei (Fig. 8), which are also indicative of degeneration and failed regeneration of muscle fibers. Determining whether the muscle phenotype is a cell- or non-cell autonomous effect is of interest – in the meantime, the presence of desmin aggregates in muscle biopsies of patients [44, 55] may indicate an additional role of gigaxonin in muscle homeostasis.

Additional open questions are the age of onset of the pathology and the functions of gigaxonin that contribute to the behavioral deficits seen in the GAN^A49E mouse model. As we evidenced motor dysfunctions and ataxia at 3 months of age, it would be important to determine whether symptoms appear earlier at post- but also pre-natal stages. Indeed, depletion of gigaxonin during development in zebrafish revealed impaired locomotion [1], due to disrupted NMJ, axonal defects and Shh-dependent defective genesis of motoneurons. Another important question relates to the function(s) of gigaxonin that drives the pathogenicity. In other words, which gigaxonin substrate(s) contribute(s) to the symptoms in GAN? Substantial attention is given to NFs, due to the dramatic alterations found in the nerve biopsy of patients and the similar severity in the GAN^A49E mouse model. The implication of NFs in the GAN phenotype is legitimate, considering their direct implication in the related neuromuscular diseases called Charcot-Marie-Tooth disease [30], but other substrates may also be implicated. In particular, the gigaxonin-E3 ligase has been shown to regulate the autophagy pathway, through the degradation of the ATG16L1 protein [11, 57]. Considering that NF proteins are targeted to autophagy degradation in vitro and in vivo [54], one could suggest that the reduction of autophagy capacities in absence of gigaxonin may exacerbate the aggregation of NFs and therefore contribute to the symptoms in GAN^A49E mice and patients. Until we evaluate the possible degeneration of neurons at early time points (before 24 months), a role of the Shh pathway is still to be considered for GAN, with the Ptch receptor as another target of gigaxonin [1].

To our knowledge, our study presents the first phenotypic model for GAN, that can now be used to scrutinize novel functions of the gigaxonin-E3 ligase and address key questions. In particular, this novel model will help us assess the possible developmental origin of the disease, and investigate muscle cell-autonomous and sex biases along with the molecular pathways underlying the observed symptoms in mouse. Most importantly for patients, this GAN^A49E model presents with the unique opportunity to evaluate the efficacy of therapeutic strategies, including gene therapy and pharmacological approaches, with small molecules already shown to rescue the motor deficits in the gan zebrafish model [36, 37].

Conclusion

In this study, we generated and characterized the first robust model for GAN, reproducing the early onset and severity of the disease. The GAN^A49E disease-causing mutation induces early and progressive defects of both motor and sensory performances. In addition to the aggregation of NFs, we observe for the first time the main characteristic of the pathology, i.e. giant axons, in peripheral nerves of GAN mice. Our histopathological study also reveals a sex-effect in the muscle denervation and fragmentation of NMJs in GAN females, which correlates with the severity of symptoms. In conclusion, the GAN^A49E mouse model represents a unique opportunity to unveil the pathogenesis and develop effective therapies for the fatal yet incurable GAN disease.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

DR

Dorsal root

GAN

Giant axonal neuropathy

IF

Intermediate filament

KI

Knock-in

KO

Knock-out

NF

Neurofilament

NMJ

Neuromuscular junction

SN

Sciatic nerve

VR

Ventral root

WT

Wild-type

Author contributions

Conceptualization, P.B.; Methodology, F.F. and P.B.; Software: N.P. and L.A; Investigation, N.P., C.L., E.C., L.A., M.C.-C., A.P. AL.M-B and C.C.; Formal Analysis, C.H.-K. and C.L.; Writing—original draft, C.L., C.H.-K. and P.B.; Writing—Review & Editing, C.L. and P.B. with corrections from all; Ressources and Funding Acquisition, P.B.; Supervision, P.B. All authors read and approved the final manuscript.

Funding

This work is supported by the Institut National de la Santé et de la Recherche Médicale (INSERM), ATIP-Avenir program (INSERM), the Association Française contre les Myopathies (AFM-22225) and the Fondation Maladies Rares (FMR-R15041FF) to PB. CL is the recipient of a fellowship from the Centre Hospitalo-Universitaire of Montpellier.

Data availability

Upon reasonable request, the data described in this study are available from the corresponding author.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.