GDAP1 Involvement in Mitochondrial Function and Oxidative Stress, Investigated in a Charcot-Marie-Tooth Model of hiPSCs-Derived Motor Neurons

Mutations in the ganglioside-induced differentiation associated protein 1 (GDAP1) gene have been associated with demyelinating and axonal forms of Charcot-Marie-Tooth (CMT) disease, the most frequent hereditary peripheral neuropathy in humans. Previous studies reported the prevalent GDAP1 expression in neural tissues and cells, from animal models. Here, we described the first GDAP1 functional study on human induced-pluripotent stem cells (hiPSCs)-derived motor neurons, obtained from normal subjects and from a CMT2H patient, carrying the GDAP1 homozygous c.581C>G (p.Ser194*) mutation. At mRNA level, we observed that, in normal subjects, GDAP1 is mainly expressed in motor neurons, while it is drastically reduced in the patient’s cells containing a premature termination codon (PTC), probably degraded by the nonsense-mediated mRNA decay (NMD) system. Morphological and functional investigations revealed in the CMT patient’s motor neurons a decrease of cell viability associated to lipid dysfunction and oxidative stress development. Mitochondrion is a key organelle in oxidative stress generation, but it is also mainly involved in energetic metabolism. Thus, in the CMT patient’s motor neurons, mitochondrial cristae defects were observed, even if no deficit in ATP production emerged. This cellular model of hiPSCs-derived motor neurons underlines the role of mitochondrion and oxidative stress in CMT disease and paves the way for new treatment evaluation.


Introduction
Charcot-Marie-Tooth (CMT) disease is a heterogeneous group of sensory-motor disorders belonging to the larger class of genetic neuropathies. With an estimated prevalence of 1:2500, it is considered as the most frequent inherited pathology of the peripheral nervous system. It indifferently affects both sexes, of any geographical origin and age, and it is clinically defined by muscular weakness and atrophy, foot deformities such as pes cavus, and sometimes sensory loss and balance issues [1]. Traditionally, based on electrophysiological studies, demyelinating forms characterized by reduced nerve conduction velocity (NCV)

Subjects
Ethics approval was obtained from the ethic committee of Limoges University Hospital (n • 384-2020-40, 10/07/2020), as well as the informed consent of all participants. The study was conducted in accordance with the Declaration of Helsinki. The family of the CMTpatient presented two cases ( Figure 1). The propositus, here reported as "patient", showed first gait disturbances when he was 18 months old. This young boy was characterized by a severe axonal neuropathy, with subacute progression and polyvisceral disorders, leading him to an early death at the age of three. His 5-year-old younger brother developed motor impairment in feet, distal atrophy and abolished deep tendon reflexes associated to mental retardation. Parents, with a first degree of consanguinity, were asymptomatic, as well as the 13-year-old elder brother. In the GDAP1 gene genetic analyses detected the c.581C>G (p.Ser194*) mutation, homozygous in the patient and his affected brother, and heterozygous in the other family members. No other mutation in CMT-and peripheral neuropathiesassociated genes was detected by targeted Next Generation Sequencing (NGS) (see [34] for the detailed protocol). Two control subjects, without any clinical neurological signs, were enrolled in this study: Ctrl-1, a 24-year-old man, and Ctrl-2, a 28-year old woman. Sanger sequencing excluded any GDAP1 mutation in these controls. Ethics approval was obtained from the ethic committee of Limoges University Hos-101 pital (n°384-2020-40, 10/07/2020), as well as the informed consent of all participants. The 102 study was conducted in accordance with the Declaration of Helsinki. The family of the 103 CMT-patient presented two cases ( Figure 1). The propositus, here reported as "patient", 104 showed first gait disturbances when he was 18 months old. This young boy was charac-105 terized by a severe axonal neuropathy, with subacute progression and polyvisceral disor-106 ders, leading him to an early death at the age of three. His 5-year-old younger brother 107 developed motor impairment in feet, distal atrophy and abolished deep tendon reflexes 108 associated to mental retardation. Parents, with a first degree of consanguinity, were 109 asymptomatic, as well as the 13-year-old elder brother. In the GDAP1 gene genetic anal-110 yses detected the c.581C>G (p.Ser194*) mutation, homozygous in the patient and his af-111 fected brother, and heterozygous in the other family members. No other mutation in 112 CMT-and peripheral neuropathies-associated genes was detected by targeted Next Gen-113 eration Sequencing (NGS) (see [34] for the detailed protocol). Two control subjects, with-114 out any clinical neurological signs, were enrolled in this study: Ctrl-1, a 24-year-old man, 115 and Ctrl-2, a 28-year old woman. Sanger sequencing excluded any GDAP1 mutation in 116 these controls.   121 Skin biopsies were obtained from patient, Ctrl-1, and Ctrl-2, and incubated in 122 CHANG Medium ® D (Irvine Scientific, Santa Ana, CA., USA), with 10% Fetal Bovine Se-123 rum (FBS) (Gibco, Thermo Fisher SCIENTIFIC, Waltham, MA, USA). After two weeks, 124 once fibroblasts (FBs) have migrated from the skin fragment and grew in the culture dish, 125 they were isolated using trypsin. In the first three days, fibroblasts were cultivated com-126 bining the Chang Medium ® D (25%) with the RPMI 1640 medium (75%) (Gibco, Thermo 127 Fisher SCIENTIFIC), supplemented with 10% FBS. Then, Chang medium D was com-128 pletely replaced by RPMI medium and FBS. 129

Skin Biopsies and Fibroblasts Cell Culture
Skin biopsies were obtained from patient, Ctrl-1, and Ctrl-2, and incubated in CHANG Medium ® D (Irvine Scientific, Santa Ana, CA., USA), with 10% Fetal Bovine Serum (FBS) (Gibco, Thermo Fisher SCIENTIFIC, Waltham, MA, USA). After two weeks, once fibroblasts (FBs) have migrated from the skin fragment and grew in the culture dish, they were isolated using trypsin. In the first three days, fibroblasts were cultivated combining the Chang Medium ® D (25%) with the RPMI 1640 medium (75%) (Gibco, Thermo Fisher SCIENTIFIC), supplemented with 10% FBS. Then, Chang medium D was completely replaced by RPMI medium and FBS.

RNA Analysis
Total RNA was extracted from fibroblasts, hiPSCs, NPs and MNs of Ctrl-1 and patient, using the miRNeasy Mini kit (QIAGEN ® , Venlo, The Netherlands). After verifying RNA integrity with the Bioanalyzer 2100 system (Agilent Technologies), it was converted in cDNA with the QuantiTect ® Reverse Transcription kit (QIAGEN ® ). For the quantitative PCR (qPCR, or Real-Time PCR), primers were designed between the fifth and the sixth exon of GDAP1, and between the fifth and the sixth exon of TBP (TATA-Box Binding Protein), chosen as reference gene. Reactions were prepared with the Rotor-Gene SYBR-Green PCR Kit (400) (©QIAGEN) and performed on the Corbett Rotor-Gene 6000 Machine (© QIAGEN). All qPCR reactions were performed four times.
The chromogenic 3,3 -Diaminobenzidine (DAB) staining was used as complementary ICC method to immunofluorescence, by reason of its high sensitivity. For the DAB staining, MNs were fixed, permeabilized, and incubated with the GDAP1 primary antibody overnight. Next day, the VECTASTAIN ® Elite ABC kit (Vector Laboratories, Burlingame, CA, USA) was used for the avidin-biotin/peroxidase detection. The DAB+ chromogen, i.e., the peroxidase substrate solution, was added to induce the formation of the brown precipitate, visualized with a light microscope.

Electron Microscopy
All manipulations for the electron microscopy were performed in Neurology and Anatomic Pathology departments at University Hospital of Limoges. Cells were fixed in 2.5% glutaraldehyde, then incubated 30 min, at RT, in 2% OsO4 (Euromedex, Souffelweyersheim, France). After washing them with distilled water, they were dehydrated 10 min in a series of ethanol dilutions (30%, 50%, 70%, 95%) and three times in 100% ethanol. At the end, they were embedded overnight in Epon 812. Thin blocks were selected and stained with uranyl acetate and lead citrate and examined using a Jeol 1011 electron microscope.

Adenosine Triphosphate (ATP) Quantification
ATP was dosed using CellTiter-Glo ® Luminescent Cell Viability Assay kit (Promega, Madison, WI, USA), and the luminescent signal was recorded with the Fluoroskan Ascent ® FL (Thermo Fisher SCIENTIFIC,) following manufacturer instructions. DAPI staining was used to normalize luminescence's values to the number of cells. Reactions were performed in triplicate, and experiments were repeated three times.

Succinate Dehydrogenase (Complex II) Activity
Succinate dehydrogenase activity was measured using the Cell Proliferation Kit I (Roche, Basel, Switzerland), following manufacturer conditions. Absorbance of formazan crystals, at 595 nm, was recorded with the Multiskan™ FC Microplate Photometer (Thermo Fisher SCIENTIFIC), and normalized to the number of cells, measured with the DAPI staining. Reactions were performed in triplicate, and experiments were repeated three times.

Mitochondrial Superoxide Quantification
Fibroblasts and MNs were analyzed in basal conditions as well as in stressed conditions. Stressed wells were treated two hours with 1 mM H 2 O 2 solution, prepared in culture medium. After the treatment, 5 µM MitoSOX™ Red mitochondrial superoxide marker (Molecular Probes, Thermo Fisher SCIENTIFIC) was added to the whole of the plate and incubated for 10 min at 37 • C. Fluorescent signal was detected using the Leica DM IRB microscope and normalized to the number of cells, measured with the DAPI staining. Reactions were performed in triplicate, and experiments were repeated three times.

Statistical Analysis
All statistical analyses were performed using the GraphPad Prism 5 software (Graph-Pad Software, Inc., San Diego, CA, USA). Data were expressed as mean ± SEM (Standard Error of the Mean). They were compared using the nonparametric Mann-Whitney U test; p < 0.05 was considered significant.

Control and CMT2H hiPSCs Efficiently Differentiate into MNs
Fibroblasts of Ctrl-1, Ctrl-2, and the CMT2H patient were reprogramed in hiPSCs. After validating hiPSCs of the three subjects for all quality controls ( Figure S2), our differentiation protocol was applied in order to generate NPs first, then MNs (all differentiation steps for Ctrl-1, Ctrl-2, and patient are reported in the Figure S3). Figure 2A shows that all cells were PGP9.5-positive (red) and Tuj-1-positive (green), validating their neuronal profile. Cells expressed also the cholyne acetyltransferase (ChAT) enzyme (green), confirming the cholinergic function of these hiPSCs-derived motor neurons ( Figure 2B).
After validating hiPSCs of the three subjects for all quality controls ( Figure S2), our differentiation protocol was applied in order to generate NPs first, then MNs (all differentiation steps for Ctrl-1, Ctrl-2, and patient are reported in the Figure S3). Figure 2A shows that all cells were PGP9.5-positive (red) and Tuj-1-positive (green), validating their neuronal profile. Cells expressed also the cholyne acetyltransferase (ChAT) enzyme (green), confirming the cholinergic function of these hiPSCs-derived motor neurons ( Figure 2B).

GDAP1 Protein Is Expressed in MNs of Controls and Absent in p.Ser194* MNs
To complete the expression study, we evaluated GDAP1 protein expression on MNs, the cellular type known to express GDAP1, in comparison with fibroblasts. The experiment was performed for the two control subjects and the CMT patient ( Figure 4).
For the immunofluorescence analysis, MNs were stained with GDAP1 antibody (red), and Tuj-1 antibody (green), specific of neural cells. As shown in Figure 4A, GDAP1 protein was detected in MNs of Ctrl-1 and Ctrl-2, located in neurons' cell body. In contrast, no fluorescent red signal was observed in patient's MNs, suggesting the weak expression of GDAP1 protein ( Figure 4A).
These results were supported by the DAB staining showing the higher expression of GDAP1 protein in MNs of Ctrl-1 compared to patient ( Figure 5).

311
Mitochondrial Morphology 312 HiPSCs-derived motor neurons of controls and patient were analyzed by electron 313 microscopy. Surprisingly, in the cytoplasm of multiple patient's MNs, we observed the 314 emergence of several round structures, of various sizes, suspected to be lipid droplets 315 (LDs) ( Figure 8F). These structures appeared electron-dense, with a homogeneous content 316 surrounded by a more electron-dense line, presumably a lipid monolayer, supporting 317 their identity as LDs. However, we cannot exclude the possibility of a bilayer, and a dif-318 ferent nature of these structures. They were not observed in controls' MNs, nor in fibro-319 blasts of the three subjects ( Figures 8A,B,C,D,E). Given the mitochondrial localization of GDAP1 protein, we investigated mitochon-324 drial morphology and structure. Looking at MNs' ultrastructure, any difference in mito-325 chondrial size and shape was remarked between controls and patient. Moreover, MNs of 326

In MNs, GDAP1 Mutation Is Associated with Cytosolic Lipid Droplets and Perturbed Mitochondrial Morphology
HiPSCs-derived motor neurons of controls and patient were analyzed by electron microscopy. Surprisingly, in the cytoplasm of multiple patient's MNs, we observed the emergence of several round structures, of various sizes, suspected to be lipid droplets (LDs) ( Figure 8F). These structures appeared electron-dense, with a homogeneous content surrounded by a more electron-dense line, presumably a lipid monolayer, supporting their identity as LDs. However, we cannot exclude the possibility of a bilayer, and a different nature of these structures. They were not observed in controls' MNs, nor in fibroblasts of the three subjects ( Figure 8A-E).

In MNs, GDAP1 Mutation Is Associated with Cytosolic Lipid Droplets and Perturbed Mitochondrial Morphology
HiPSCs-derived motor neurons of controls and patient were analyzed by electron microscopy. Surprisingly, in the cytoplasm of multiple patient's MNs, we observed the emergence of several round structures, of various sizes, suspected to be lipid droplets (LDs) ( Figure 8F). These structures appeared electron-dense, with a homogeneous content surrounded by a more electron-dense line, presumably a lipid monolayer, supporting their identity as LDs. However, we cannot exclude the possibility of a bilayer, and a different nature of these structures. They were not observed in controls' MNs, nor in fibroblasts of the three subjects ( Figures 8A,B,C,D,E). Given the mitochondrial localization of GDAP1 protein, we investigated mitochondrial morphology and structure. Looking at MNs' ultrastructure, any difference in mitochondrial size and shape was remarked between controls and patient. Moreover, MNs of both subjects presented elongated and fragmented mitochondria. However, focusing on mitochondrial cristae, we observed that their organization was altered in mitochondria of Given the mitochondrial localization of GDAP1 protein, we investigated mitochondrial morphology and structure. Looking at MNs' ultrastructure, any difference in mitochondrial size and shape was remarked between controls and patient. Moreover, MNs of both subjects presented elongated and fragmented mitochondria. However, focusing on mitochondrial cristae, we observed that their organization was altered in mitochondria of patient's MNs. In particular, cristae's regular distribution and thickness were perturbed, preventing to discriminate their structure in the internal mitochondrial compartment. Swollen cristae were also observed ( Figure 9F). This disorganization of mitochondrial cristae was not present in Ctrl-1 and Ctrl-2 MNs, as well as in fibroblasts of the three subjects ( Figure 9A-E).
Biomedicines 2021, 9, x FOR PEER REVIEW 11 of 19 Swollen cristae were also observed ( Figure 9F). This disorganization of mitochondrial cristae was not present in Ctrl-1 and Ctrl-2 MNs, as well as in fibroblasts of the three subjects ( Figures 9A,B,C,D,E).

GDAP1 Mutation does Not Strongly Alter Oxidative Phosphorylation
The alteration of cristae organization in mitochondria of patient's MNs, led us to investigate the oxidative phosphorylation through the activity of the electron transport chain (ETC) complexes and the ATP production. Given the limited availability of hiPSCsderived MNs, we performed the MTT test to evaluate the activity of the succinate dehydrogenase (complex II). In both fibroblasts and motor neurons, succinate dehydrogenase activity seemed to be slightly increased in patient's cells, compared to Ctrl-1 and Ctrl-2, and it reached significant difference in fibroblasts (Ctrl-2 FB 0.97 ± 0.03 vs. patient FB 1.099 ± 0.03; p < 0.05) (Figure 10). In contrast, ATP levels were not significantly different between patient's and controls' fibroblasts and motor neurons ( Figure 11).

GDAP1 Mutation Does Not Strongly Alter Oxidative Phosphorylation
The alteration of cristae organization in mitochondria of patient's MNs, led us to investigate the oxidative phosphorylation through the activity of the electron transport chain (ETC) complexes and the ATP production. Given the limited availability of hiPSCsderived MNs, we performed the MTT test to evaluate the activity of the succinate dehydrogenase (complex II). In both fibroblasts and motor neurons, succinate dehydrogenase activity seemed to be slightly increased in patient's cells, compared to Ctrl-1 and Ctrl-2, and it reached significant difference in fibroblasts (Ctrl-2 FB 0.97 ± 0.03 vs. patient FB 1.099 ± 0.03; p < 0.05) (Figure 10). In contrast, ATP levels were not significantly different between patient's and controls' fibroblasts and motor neurons ( Figure 11).

GDAP1 Mutation Could Promote Mitochondrial Oxidative Stress
Mitochondria are also the main producers of reactive oxygen species (ROS), such as superoxide anion, inducing oxidative stress. A perturbation of mitochondrial cristae could promote redox imbalance. As expected, and shown in Figure 12, in patient's MNs, superoxide anion levels were significantly higher than in Ctrl-1 (Ctrl-1 MN 0.92 ± 0.17 vs. patient MN 1.37 ± 0.15; p < 0.05). This significant difference was even observed in fibroblasts (Ctrl-1 FB 0.92 ± 0.17 vs. patient FB 2.61 ± 0.68; p < 0.05). The same trend emerged in the comparison with Ctrl-2 cells, but significant difference was not reached. Mitochondria are also the main producers of reactive oxygen species (ROS), such as 354 superoxide anion, inducing oxidative stress. A perturbation of mitochondrial cristae could 355 promote redox imbalance. As expected, and shown in Figure 12, in patient's MNs, super-356 oxide anion levels were significantly higher than in Ctrl-1 (Ctrl-1 MN 0.92 ± 0.17 vs. patient 357 MN 1.37 ± 0.15; p < 0.05). This significant difference was even observed in fibroblasts (Ctrl-358 1 FB 0.92 ± 0.17 vs. patient FB 2.61 ± 0.68; p < 0.05). The same trend emerged in the com-359 parison with Ctrl-2 cells, but significant difference was not reached.

365
More than 80 mutations in GDAP1 gene have already been reported to be responsible 366 for demyelinating, axonal, and intermediate forms of Charcot-Marie-Tooth disease, and 367 associated to heterogeneous phenotypic manifestations [37]. However, GDAP1 role in cel-368 lular functions and processes has not been clearly elucidated, although several relevant 369 studies have been performed [7,13,16,22,38]. Thus, the pathological role of GDAP1 in this 370 disease development remains to be understood. Cellular and animal models expressing 371 GDAP1 mutations surely represent the most accessible and easier tool to mimic GDAP1-372 induced pathophysiology. Murine and human GDAP1 proteins share 94% of amino acid 373 homology supporting the relevance of this rodent model in expression and localization 374 studies [7,9], structural studies [39], functional studies on knockout animal models [12,13]. 375 Further analyses were conducted using Drosophila, and its GDAP1-ortholog gene 376 (CG4623) [10,11], or yeast models, transfected with the human GDAP1 [14,40]. The high 377

Discussion
More than 80 mutations in GDAP1 gene have already been reported to be responsible for demyelinating, axonal, and intermediate forms of Charcot-Marie-Tooth disease, and associated to heterogeneous phenotypic manifestations [37]. However, GDAP1 role in cellular functions and processes has not been clearly elucidated, although several relevant studies have been performed [7,13,16,22,38]. Thus, the pathological role of GDAP1 in this disease development remains to be understood. Cellular and animal models expressing GDAP1 mutations surely represent the most accessible and easier tool to mimic GDAP1-induced pathophysiology. Murine and human GDAP1 proteins share 94% of amino acid homology supporting the relevance of this rodent model in expression and localization studies [7,9], structural studies [39], functional studies on knockout animal models [12,13]. Further analyses were conducted using Drosophila, and its GDAP1-ortholog gene (CG4623) [10,11], or yeast models, transfected with the human GDAP1 [14,40]. The high intra-species and inter-species variability, together with the complexity of GDAP1 molecular mechanisms involved in Charcot-Marie-Tooth disease, may limit the animal models' reliability. Concerning the cellular models developed for GDAP1, most of them have animal origin (mice, rats) [7,9,15], or are immortalized cell lines, naturally expressing GDAP1 (SH-SY5Y, N1E-115, HT22) [7,8,22,24], or by transfection (HeLa, Cos7) [9,15,20,21]. Indeed, a limited number of cell types express GDAP1, notably neurons and Schwann cells. These cell types cannot be obtained from humans, and used in vitro as cellular models. Given the inability to culture human neural cells, the only model available in these conditions was represented by human fibroblasts [16][17][18]22], which, unfortunately, poorly express GDAP1 [22]. The aim of this study was to go beyond limits imposed by existing animal and cellular models, developing a new solid model of human motor neurons, carrying the homozygous p.Ser194* mutation in GDAP1, to investigate GDAP1 functions and GDAP1-associated mechanisms in CMT disease.
We first questioned about GDAP1 expression. In animal models, such as mice and rats, GDAP1 has been shown to be largely expressed in neurons. In particular, the higher expression was detected in cerebellum, cerebral cortex, hippocampus, olfactory bulb, spinal nerve, but also in sciatic nerve, and motor and sensory neurons [7][8][9]. GDAP1 expression in Schwann cells was controversial, whereas, in non-neural tissues, it was poorly explored [7][8][9]. Here, we compared, for the first time, GDAP1 mRNA expression in four human cell types of the same control subject (Ctrl-1): fibroblasts, hiPSCs, NPs, and MNs. Our study revealed, on Ctrl-1 cells, that GDAP1 is weakly expressed in fibroblasts and hiPSCs, while its expression was significantly higher in NPs, and, above all, in MNs. It is interesting to note that GDAP1 mRNA in fibroblasts represented only 3% of NPs-GDAP1 mRNA, and 1.8% of MNs-GDAP1 mRNA. This is in agreement with Noack et al. work, who showed that control human fibroblasts expressed only 2.6% of GDAP1 mRNA compared to embryonic stem cells-derived motor neurons [22]. In contrast, in patient's cells, presenting the homozygous codon-stop mutation c.581C>G in exon 5, GDAP1 mRNA was only 10-20% of mRNA estimated in each cell type of Ctrl-1, up to 6-and 8-fold smaller than those assessed in Ctrl-1 NPs and Ctrl-1 MNs. Our data seem to suggest that GDAP1 mRNA is degraded in patient's cells. As mutated GDAP1 mRNA contains a premature termination codon (PTC), the nonsense-mediated mRNA decay (NMD) system could be activated and induce its degradation, preventing the synthesis of a truncated, and maybe non-functional, protein [41]. In any case, the NMD system is not always 100% efficient and some PTC-mRNA can escape NMD, and be detected by qPCR, as shown here. Real time qPCR results were also supported by GDAP1 protein expression. Indeed, GDAP1 was not express in fibroblasts, both in controls and patient, while the GDAP1 staining was present in Ctrl's MNs, and, as expected, lacked in patient's MNs. Given the high GDAP1 neural expression, we chose MNs and NPs, as cellular models to investigate its functions, and evaluate its role in Charcot-Marie-Tooth disease development.
However, the weak GDAP1 expression detected in fibroblasts does not exclude a GDAP1 role in this cell type, and the possibility of conducting functional studies on them [16][17][18]. In our case, in the examination of electron transport chain's (ETC) activity, any big difference emerged between controls and patient, neither in fibroblasts, nor in MNs. Only a slight increase in succinate dehydrogenase activity was observed, but did not reach significant levels in MNs, and ATP production was preserved in patient's fibroblasts and MNs. Nevertheless, we cannot exclude that markers used to evaluate oxidative phosphorylation can be limited, and the analysis not-exhaustive. Thus, the measurement of each complex activity, by oxygraphy, could prove more comprehensive in investigation of ETC function.
On the other hand, morphological and oxidative stress analyses, allowed highlighting, exclusively in MNs, two main mechanisms which could play a key role in CMT disease progression: the deregulation of mitochondria morphology and dynamics, and the redox imbalance. These aspects support the relevance of our MNs cellular model in the study of GDAP1-associated pathophysiology. First, we investigated mitochondrial morphology. Electron microscopy revealed, only in patient's MNs, a general disorganization of mitochondrial cristae, which could affect the inner mitochondrial space and subsequent metabolism. The disruption of cristae structure has already been associated to other pathological conditions induced by mutations or lack of proteins involved in mitochondrial dynamics, such as optic atrophy 1 (OPA1) protein [42], or mitofusin 2 (Mfn2) protein [43]. Cristae abnormalities were also reported in nerves' axonal mitochondria of a CMT2 patient, carrying the c.174_176delGCCinsTGTG (p.Pro59Valfs*4) mutation in GDAP1 [44], but also, more recently, in muscular tissue of a patient carrying the c.77T>G (p.Leu26Arg) and the c.505_511del (p.Ser169*) GDAP1 mutations [18].
Interestingly, some additional findings of our cell culture need to be pointed out. We observed in patient's hiPSCs carrying the p.Ser194* mutation a higher spontaneous differentiation and a reduced maintenance of stemness compared to controls' hiPSCs, where GDAP1 seems to be expressed, even if at a low level. Moreover, we have demonstrated that patient's neural cells present a lower proliferation rate, compared to controls. Both these aspects could potentially be related to GDAP1 involvement in mitochondrial dynamics. Indeed, it is now well known that the proper preservation of mitochondrial dynamics is a fundamental condition for cell cycle progression, and fragmentation of mitochondrial network is required during the mitosis phase [45]. Thus, the deficiency of a fission protein, such as GDAP1, could impact the regulation of cell proliferation mechanisms. Indeed, Prieto et al., demonstrated that GDAP1 knockout, altering the fission machinery, impairs OSKM (Oct4/Klf4/Sox2/cMyc) reprogramming, and cell cycle progression, in murine iPSCs [46]. Based on our preliminary results and previous studies, we can assume that GDAP1 protein may be a key component in controlling mitochondrial morphology and dynamics, and its lack may disturb mitochondrial-dependent processes, such as in different cell types.
In the cytoplasm of patient's MNs, round electron-dense structures were observed, suspected to be lipid droplets. LDs could be considered as an accumulation of energetic substrate, such as triglycerides, linked to a defect of mitochondrial beta oxidation, or a hallmark of cellular stress, previously observed in nutrient imbalance, inflammation and oxidative stress [47]. Moreover, several studies have demonstrated that their accumulation is one of earliest events following the induction of cellular apoptosis [48]. Thus, the accumulation of LDs could also be considered as an early signal of apoptotic pathways' activation, explaining the significant reduction of patient's MNs observed in last steps of neural differentiation. The synthesis of lipid droplets in neurons has been observed in pathogenesis of several neurodegenerative diseases, such as amyotrophic lateral sclerosis, Huntington's disease, Alzheimer's disease, Parkinson's disease and Hereditary spastic paraplegia [49]. LDs were also described in the ultrastructural analysis of motor neurons obtained from GDAP1 knockout mice [13], in accordance with our results, corroborating the consistency of our cellular model. In stress conditions, a cytoprotective role against reactive oxygen species (ROS) is also supplied by LDs. As polyunsaturated fatty acids (PUFAs) are more susceptible to ROS-induced peroxidation when they are integrated in cellular membranes, LDs sequester them as triacylglycerols (TAGs) in their core, protecting cellular structures from ROS damage [50]. This phenomenon could probably be responsible for the LDs formation in our cellular model of CMT-motor neurons carrying the GDAP1 p.Ser194* mutation. In fact, GDAP1 has been suggested to have an antioxidant role in cellular homeostasis [10,22]. Consequently, in patient's MNs, the lack of GDAP1 protein increases the amount of generated ROS, proven, in our study, by the MitoSOX™ Red mitochondrial superoxide indicator analysis. However, we have also detected a significant increase of ROS in patient's fibroblasts, where GDAP1 is weakly expressed and LDs lacking. The overproduction of superoxide anion, in GDAP1-mutated fibroblasts, has also been reported in a recent work [17]. Moreover, the same study demonstrated that GDAP1mutated fibroblasts presented also a reduced expression of Sirtuin 1 (SIRT1) enzyme, which activates the PPARgamma coactivator-1alpha (PGC1 α) [17]. PGC1 α is a fundamental factor in mitochondrial biogenesis, and it has been associated to neurological disorders and diabetic peripheral neuropathy [51][52][53][54]. These data could strengthen the idea that GDAP1 protein, even if at lower levels, could also be present in cell types other than neural cells. In this case, nevertheless, other molecular mechanisms and proteins would take part to the cellular antioxidant defense, counterbalancing GDAP1 deficiency.

Conclusions
In conclusion, the role of GDAP1 impairment in Charcot-Marie-Tooth pathophysiology through mitochondrial dysfunction and oxidative stress development was underlined in an original human model of motor neuron from patient's fibroblasts, carrying the homozygous codon-stop c.581C>G (p.Ser194*) mutation. The results underlined that GDAP1 is mostly expressed in neural cell types such as MNs and PNs, but also, at lower levels, in fibroblasts and hiPS cells. In patient's cells, 80-90% of GDAP1 mRNA would be degraded by the NMD system, leading to the considerable reduction of GDAP1 protein. Taken together, these results demonstrated that hiPS cells can be a powerful tool to recreate any suitable cellular model from patients carrying mutations and are essential for understanding the pathophysiological role of the altered protein, but also necessary to develop new therapeutic strategies.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.