YG8R mice were purchased from The Jackson Laboratory Repository (stock no. 008398. Maine, USA). The YG8R mouse is a knock-out with both mouse frataxin alleles deleted (Fxn) and hemizygous for the transgene “YG8” that contains two tandem copies of the human FXN gene with 90 and 190 GAA trinucleotide sequence repeats. The GAA expansions promote a low frataxin expression level but enough to rescue the embryonic lethality in knockout mice (Cossée et al., 2000; Al-Mahdawi et al., 2006). For this study, we used the crossing system and genotyping as previously described (Mollá et al., 2016). We used both hemizygous mice, containing one allele of the transgene, YG8R and homozygous mice, containing two alleles of the transgene, YG8YG8R and C57BL/6J as control mice (control). Animals were group-housed under standard housing conditions with a 12 h light–dark cycle, and food and water ad libitum. All mouse experiments were approved by the local Animal Ethics Review Committee of the Consejo Superior de Investigaciones Científicas (CSIC) and the Centro de Investigación Príncipe Felipe (CIPF).

Adult mice were sacrificed by cervical dislocation between 22 and 24 months old and approximately 50 DRGs per animal were dissected and collected in L15 media. Ganglia were incubated at 37°C with 2 mg/ml collagenase (Worthington Biochemical Corporation, Lakewood, NJ, USA) followed by 0.05% trypsin (Sigma-Aldrich, Madrid, Spain). Approximately 50,000–80,000 neurons per animal were obtained in 1 ml of defined medium, consisting of Ham’s F12 supplemented with 2 mM glutamine, 0.35% albumax (Invitrogen™), 60 ng/ml progesterone, 16 μg/ml putrescine, 400 ng/ml L-thyroxine, 38 ng/ml sodium selenite, 340 ng/ml tri-iodo-thyronine, 60 μg/ml penicillin and 100 μg/ml streptomycin supplemented with NGF, BDNF and NT3 at 10 ng/ml and 5 μM Ara-C. Then, 2000 neurons per well were plated in 24-well plates (Sarstedt AG and Co, Nümbrecht, Germany) with 12 mm-coverslips that had been previously coated with poly-ornithine (0.5 mg/ml, overnight) and laminin (20 μg/ml, 4 h). Isolated neurons were incubated at 37°C in a humidified incubator under 5% CO₂ (Davies et al., 1993).

To monitor mitochondria, DRG neurons were incubated with 200 nM Mitotracker deep red (Molecular Probes™) for 15 min at 37°C and rinsed with defined medium 15 min at 37°C. Later, neurons were fixed for 20 min with 4% PFA and 120 mM sucrose in PB 0.1 M, rinsed three times with PB 0.1 M, and blocked in 10% FBS and 0.2% Triton X-100 in PB 0.1 M for 30 min at room temperature. Neurons were then incubated overnight at 4°C with antibodies against β-tubulin III (1:1000, Sigma-Aldrich), peripherin (1:2000, Millipore), β-APP or beta-amyloid precursor protein (1:500, Abcam), synaptophysin (1:250, Sigma-Aldrich), p62 or sequestosome 1 (SQSTM1; 1:100, Abcam), and LAMP2 or Lysosome-associated membrane protein 2 (1:500, DSHB). After three washes with PB 0.1 M, neurons were incubated with appropriate secondary antibodies Alexa Fluor 488 or 633 (Molecular Probes™) for 1 h at room temperature. Following three washes with PB 0.1 M, the nuclei of neurons were stained with DAPI for 15 min at room temperature and then coated with Fluoromount G (Southern Biotech Assoc. Inc, Birmingham, AL, USA). Images were acquired using a Leica TCS SP8 laser-scanning confocal microscope. Confocal imaging settings were kept constant between genotypes and experimental replicates. For confocal z-axis stacks, series of stack images (1024 × 1024) separated by 0.22 μm along the z-axis were acquired to visualize neuronal structures. 2D reconstructions of the stacks were performed with ImageJ (NIH, Bethesda, MD, USA).

Neurons were seeded at 500 cells/cm² in Permanox chamber (Nalge Nunc International, Naperville, IL, USA) coated with poly-ornithine (0.5 mg/ml, overnight) and laminin (20 μg/ml, 4 h). DRG primary cultures at 5 days in vitro (DIV) were fixed in 2.5% glutaraldehyde for 5 min at 37°C and 1 h at 4°C. Next, steps were performed as previously described (Bolinches-Amorós et al., 2014). Photomicrographs were obtained under a transmission electron microscope FEI Tecnai G2 Spirit (FEI Europe) using a digital camera Morada (Olympus Soft Image Solutions GmbH).

DRG primary cultures at 5 DIV were incubated for 30 min at 37°C with 3 μM MitoSOX™ Red (Molecular Probes™). MitoSOX™ is a cell-permeable probe that accumulates in the mitochondria and fluoresces following oxidation by superoxide. As a positive control for pro-oxidative or mitochondrial depolarization, C57BL/6J DRG neurons were treated with 500 μM H₂O₂ or 4 μM CCCP + 1 μM oligomycin for 30 min prior to and during 30 min incubation with MitoSOX™. The coverslips were fixed for 10 min with 4% PFA and sucrose 120 mM in PB 0.1 M at room temperature and then blocked for 1 h at room temperature. Cultured neurons were identified by immunofluorescence using anti-β-tubulin III and counterstained with DAPI for nuclei visualization. The microscope slides were digitized with a Hamamatsu camera (Tokyo, Japan) connected to a Leica DM RXA2 microscope (Nussloch, Germany). Microscopy imaging settings were kept constant between genotypes and experimental replicates. To quantify ROS production, fluorescence intensity relative to area was measured in neuronal somas with ImageJ. At least 236 neurons were analyzed in three or more independent experiments for each genotype.

Cultured DRG neurons at 5 DIV were incubated at 37°C for 30 min with JC1 dye (3 μM; Molecular Probes™). In healthy neurons, JC1 is taken up by the mitochondria and when a critical concentration is reached, red fluorescent J-aggregates are formed. Loss of Δψ_m causes the dye to diffuse into the cytoplasm where it exists in a green fluorescent monomeric form. As a positive control for mitochondrial depolarization, C57BL/6J DRG neurons were treated before with 4 μM CCCP + 1 μM oligomycin for 10 min and during to 30 min incubation with JC1-dye. The coverslips were fixed briefly for 10 min with 0.5% PFA and sucrose 120 mM in PB 0.1 M at room temperature and blocked for 1 h at room temperature (Liu et al., 2013). Cultures were immunodetected with anti-β-tubulin III to identify neurons and counterstained with DAPI for nuclei visualization. The emission of JC-1 monomers and aggregates was detected through 530–550 nm and 585–650 nm emission wavelength ranges respectively, by confocal microscope using a Leica TCS SP8 laser-scanner. Confocal imaging settings were kept constant between genotypes and experimental replicates. 2D reconstructions of the stacks were performed with ImageJ (NIH, Bethesda, MD, USA). To quantify Δψ_m, fluorescence intensity relative to area was measured in neuronal somas with ImageJ and the fluorescence ratio 590 nm/530 nm was expressed. At least 87 neurons were analyzed in three or more independent experiments for each genotype.

Cytosolic calcium imaging with Fura-2 AM (Molecular Probes™) was performed in live DRG neurons cultured at 5 DIV. Neurons were incubated with 5 μM Fura-2 and pluronic acid 0.06% (Sigma-Aldrich) as previously described (Bolinches-Amorós et al., 2014). Neurons were clearly identified by morphological criteria and changes in fluorescence (emission 510 nm) ratio of Ca²⁺-free (380 nm) to Ca²⁺-bound probe (340 nm) were recorded through a 40× fluo oil-immersion objective. Endoplasmic reticulum (ER) calcium storage was depleted by the addition of 150 μM 2,5-di-(ter-butyl)-1,4-benzohydroquinone (tBuBHQ, Alomone Labs) in Ca²⁺-free HCSS medium and later store-operated calcium entry (SOCE) was induced by the addition of 6 mM CaCl_2. C57BL/6J DRG neurons were pretreated with 4 μM CCCP + 1 μM oligomycin for 10 min as a control of depolarized mitochondria. Fluorescence ratio was analyzed using Leica Metafluor (Universal Imaging). Microscopy imaging settings were kept constant between genotypes and experimental replicates. [Ca²⁺]_cyt was calculated from ratio fluorescence Fura-2 measurements following an intracellular calibration procedure according to the Grynkiewicz equation (Grynkiewicz et al., 1985). At least 31 neurons were analyzed in three or more independent experiments for each genotype.

Calpain protease activity imaging was performed with CMAC, t-BOC-Leu-Met (Molecular Probes™) in live DRG neurons cultured at 3 DIV. Neurons were incubated in a sequential two-step process with 200 nM of Mitotracker deep red for 15 min at 37°C and 10 μM CMAC, t-BOC-Leu-Met for 15 min at 37°C. Fluorescence (emission 430 nm), corresponding to cleaved t-BOC produced by calpain activity, was monitored in live neuronal imaging using a 40× objective on a Leica TCS SP8 laser-scanning confocal microscope. Confocal imaging settings were kept constant between genotypes and experimental replicates. 2D reconstructions of the stacks were performed with ImageJ (NIH, Bethesda, MD, USA). Cultured neurons were identified by morphological criteria and fluorescence intensity relative to area was measured in neuronal somas with ImageJ. At least 46 neurons were analyzed in three or more independent experiments for each treatment and genotype. For experiments using [Ca²⁺]_cyt modulators, DRG cultures at 2 DIV were treated for 24 h with: (i) extracellular calcium chelator, 1 mM EGTA (Sigma-Aldrich); (ii) intracellular calcium chelator, 20 μM BAPTA-AM (Molecular Probes™); (iii) inhibitor of metal-ion-dependent enzymes, 100 μM o-phenanthroline (Sigma-Aldrich); and (iv) two-fold increase physiological extracellular calcium, 3.6 mM CaCl₂.

To performed mitochondrial morphology analysis, DRG neuron mitochondria were labeled with Mitotracker as described above, neurons were immunodetected with anti-β-tubulin III and nuclei were stained with DAPI. C57BL/6J DRG neurons treated previously with 4 μM CCCP + 1 μM oligomycin or 5 μM Br-A23187 for 30 min plus incubation with Mitotracker were used as a control of depolarized mitochondria or intra-axonal overload calcium, respectively. Fluorescence was acquired using a Leica TCS SP8 laser-scanning confocal microscope. Confocal imaging settings were kept constant between genotypes and experimental replicates. 2D reconstructions of the stacks were performed with ImageJ (NIH, Bethesda, MD, USA). Morphometric mitochondrial analysis was performed using mito-morphology macro of ImageJ, created by Dagda et al. (2009) and customized by the Brain Connectivity Laboratory (FISABIO/CIPF), headed by MI-V. The macro will draw individual outlines for each mitochondrion and calculate mitochondria number, circularity, perimeter (μm), area (μm²) and % of neuronal area occupied by mitochondria. Index of elongation (inverse of circularity) is sensitive for fragmented to normal-shaped mitochondria and is validated as mitochondrial fission parameter. The index of mitochondrial interconnectivity (area/perimeter ratio) is sensitive for normal to highly interconnected mitochondria that are visualized as a single interconnected mass of reticular network and is validated as mitochondrial fusion parameter. The index of swollen mitochondria (area/perimeter normalized to the minor axis of mitochondria or circularity) is sensitive for swelling mitochondria that attain a large area and may be confounded with interconnected mitochondria; it is validated as accounting for conditions that can induce mitochondrial swelling. For experiments using [Ca²⁺]_cyt modulators, DRG cultures were treated for 4 DIV with: (i) 1 mM EGTA (Sigma-Aldrich); (ii) 20 μM BAPTA-AM (Molecular Probes™); (iii) 100 μM o-phenanthroline; and (iv) 3.6 mM CaCl₂.

We studied protein expression in DRG tissue by western blot. Protein extraction, electrophoresis, transference and blocking were performed as previously described (Mollá et al., 2016). Membranes were incubated in blocking buffer overnight at 4°C with primary antibodies against human FXN (1:500, Immunological Sciences), actin (1:1000, Sigma-Aldrich), p62 (1:500, Abcam) and mono- and poly-ubiquitinylated conjugates-biotin conjugate (1:1000, ENZO). The blots were incubated for 1 h with the appropriate secondary anti-rabbit or anti-mouse conjugated with horseradish peroxidase (HRP)-linked (1:5000, GE). For biotin signal amplification, Pierce™ high sensitivity NeutrAvidin™-HRP (1:20,000, Thermo Fischer Scientific) was used. Blots were developed using the ECL Prime Western Blotting Detection Reagents as specified by the manufacturer (GE HealthcareBio-Sciences). Chemiluminescent signals were assessed with X-ray film and processor (Kodak, Rochester, NY, USA). Densitometry was measured using ImageJ software (NIH, Bethesda, MD, USA).

GraphPad. PRISM^® 5.00.288 software (GraphPad Software Inc., La Jolla, CA, USA) was used to generate the graphs and statistical analysis. Experiments were performed at least in three animals per genotype. The normal distribution of each sample of data was determined with Shapiro-Wilk normality test. To determine significant values between different experimental groups, the mean data were compared using one-way analysis of variance (ANOVA), followed by a Bonferroni post hoc test. Significant p-values: *p < 0.05; **p < 0.01; ***p < 0.001 were considered.

The studies were performed in 24-month-old YG8R, YG8YG8R and C57BL/6J mice, since we have observed histopathological changes in very old animals (Mollá et al., 2016). In neural tissue, the expression level of human frataxin was proportional to the number of copies of the transgene (Supplementary Figure S1). Primary cultures of DRG were labeled with Mitotracker at 5 DIV and the frataxin-deficient neurons exhibited axonal swelling and spheroid formation that retained a great amount of mitochondria, in a completely different pattern than observed in control neurons (Figure 1A). Electron microscopy showed that the mitochondria retained in the axonal spheroids in YG8R were more rounded and swollen than in the control (Figure 1A, right panel).

The effect of spheroids formation and mitochondrial retention in frataxin-deficient neurons was a disorganization of the axonal cytoskeleton (Figure 1B). By immunofluorescence, we have frequently observed an accumulation of microtubules in axonal spheroids detected with β-tubulin III and even disorganized neurofilaments of peripherin in axonal swellings (Figure 1B).

The cytoskeleton structure is essential for a proper axonal transport. Thus, neurofilament and microtubules alterations suggest an impairment of transport along axons. To investigate this point, we tested the accumulation of β-APP and synaptophysin, which are vesicle transport proteins and markers of transport disruption (Adalbert et al., 2009). The detection of β-APP-positive spheroids and synaptophysin along the axons suggest a disruption of axonal transport (Figure 1C). Moreover, the presence of dense-core vesicles (DCVs) in electron microscopy images supports this idea (Figure 1C).

Autophagy in neurons is highly dependent on axonal transport. To know about autophagic flux in frataxin-deficient neurons, we measured autophagic intermediates. In a primary culture of YG8R neurons, we observed an increase of p62 fluorescence (Figure 2A), indicating high levels of ubiquitinylated proteins or damaged mitochondria marked for recycling. The quantification of p62 and ubiquitination in DRG tissue by western blot (Figures 2B,C) showed a significant increase of ubiquitinylated proteins (p < 0.05). This accumulation of autophagic intermediates suggests the impairment of autophagic flux that was supported by a huge retention of lysosomes labeled with LAMP2 in axonal spheroids (Figure 2D).

The accumulation of mitochondria, LAMP2, β-APP and synaptophysin in axonal swellings suggests the failed axonal transport of different cargoes, including lysosomes, synaptic vesicles (SV) and DCVs.

To study the role of mitochondrial dysfunction in the pathological process of neuronal degeneration in FRDA, sensory neurons were isolated from control, YG8R and YG8YG8R age-matched mice (22/24-month old). Δψ_m and oxidative stress were analyzed in neurons under defined conditions for 5 DIV. We investigated whether frataxin deficiency effects disturb the Δψ_m in DRG neurons using the JC-1 ratiometric probe by confocal microscopy. The JC-1 ratio 590/530 was significantly lower in both frataxin-deficient neurons (YG8R = 1.73 ± 0.06 and YG8YG8R = 1.82 ± 0.04) compared to the control (2.19 ± 0.06), indicating decreased Δψ_m in these neurons (Figure 3A and Supplementary Figure S2). The C57 control neurons treated with CCCP-oligomycin, an uncoupler of mitochondrial membrane potential, exhibited the same reduction in JC-1 ratio (1.76 ± 0.04) as frataxin-deficient neurons (Figure 3A and Supplementary Figure S2) confirming mitochondrial depolarization.

Then, to clarify the role of Δψ_m loss in oxidative stress, we decided to measure the ROS production by loading neurons with the MitoSOX™ probe. MitoSOX™ oxidation was significantly higher in both frataxin-deficient neurons (YG8R = 213.9 ± 2.3 and YG8YG8R = 220.7 ± 2.0) compared to the control (160.4 ± 2.7). Likewise, control neurons treated with H₂O₂ (213.1 ± 3.0) or CCCP-oligomycin (186.9 ± 3.2) showed an increase in ROS production (Figure 3B). Thus, the increase in oxidative stress in control neurons treated with CCCP-oligomycin and frataxin-deficient neurons is associated with the loss of Δψ_m. In addition, the higher ROS levels in frataxin-deficient neurons compared with CCCP-oligomycin control neurons suggest that the lack of frataxin causes oxidative stress, in part by decreasing the Δψ_m (Figure 3B).

Control neurons treated with H₂O₂ or CCCP-oligomycin showed axonal spheroids, which suggests that the pathological changes due to oxidative stress and mitochondrial depolarization in frataxin-deficient neurons are involved in axonal spheroid formation (Figure 3C).

We investigated whether frataxin deficiency induces abnormal changes in neuronal calcium homeostasis. Neurons were loaded with FURA-2 AM and alterations in [Ca²⁺]_cyt were assessed in real-time in the neuronal soma. We observed that in Ca²⁺-free medium, resting cytosolic [Ca²⁺] is mildly increased in YG8R (129.5 ± 17.2 μM), YG8YG8R (144.4 ± 19.7 μM) and control neurons treated with CCCP-oligomycin (121.0 ± 27.2 μM) compared with control neurons (111.2 ± 30.5 μM; Figure 4A). Neurons were induced by the addition of t-buHBQ to the medium, a sarco-endoplasmic Ca²⁺-ATPase (SERCA) inhibitor, to obtain a complete depletion of ER Ca²⁺ stores. After [Ca²⁺]_cyt peaked by ER depletion, [Ca²⁺]_cyt recovered towards baseline in a few seconds in control neurons and CCCP-oligomycin control neurons. However, the recovery of [Ca²⁺]_cyt resting levels in frataxin-deficient neurons was incomplete or slower than in control (Figure 4A), possibly because of the impaired mitochondrial buffer capacity showed in frataxin deficiency (Bolinches-Amorós et al., 2014). The induction of SOCE by ER-calcium released was reduced in the frataxin-deficient neurons compared to control neurons (Figure 4B). These findings confirmed an abnormal buffering of intracellular Ca²⁺ levels and a defective SOCE mechanism in frataxin-deficient neurons, caused in part by mitochondrial depolarization. Insufficient calcium entry to neurons by SOCE after each stimulus results in inefficient ER Ca²⁺ store refilling. It could lead to inadequate neuronal calcium signaling, thereby affecting important processes such as synaptic transmission, neuronal plasticity or sensory stimulation during sense.

Calpain is a Ca²⁺-dependent thiol protease linked mechanistically to the axonal degeneration in neurological diseases. To investigate whether elevated axoplasmic Ca²⁺ levels in frataxin-deficient neurons activate calpain and participate in cytoskeleton alterations and axonal spheroid formation, we examined calpain activity in cultured neurons at 3 DIV. Neurons were labeled with Mitotracker and the calpain substrate CMAC (CMAC, t-BOC-Leu-Met) to visualize mitochondria and calpain activity in vivo, respectively, by confocal microscopy. Mitotracker staining revealed a normal mitochondrial network well distributed along neurites in control neurons and a mitochondrial accumulation in axonal spheroids in frataxin-deficient neurons (Figure 5A). In basal culture conditions, the cleaved fluorescent product t-BOC in neuronal somas was more concentrated in control neurons (32.63 ± 1.43) than in both frataxin-deficient neurons (YG8R = 23.10 ± 1.11 and YG8YG8R = 26.91 ± 1.09; Figure 5B). These results indicated that calpain activity was lower in frataxin-deficient neurons than in control neurons. On the other hand, control neurons pre-treated for 24 h with CaCl₂ showed axonal spheroids (Figure 5A) without increased calpain activity (Figure 5B). This result suggests that the increase in intra-axonal calcium levels, but not calpain activation, is involved in axonal spheroids formation in frataxin-deficient sensory neurons.

We have shown evidence that under frataxin deficiency conditions, dysfunctional mitochondria were accumulated in axonal spheroids as a marker of axonal injury. We performed an analysis of neurons to determine the effect of frataxin deficiency in mitochondrial morphology. In order to evaluate the mitochondrial morphology, we used the mito-morphology macro of ImageJ.

Control neurons and frataxin-deficient neurons were grown for 5 DIV in basal culture conditions. Moreover, control neurons were treated with CCCP-oligomycin or A23 ionophore for 30 min before fixation. At 5 DIV, the mitochondria of control neurons in basal culture showed homogenous distribution in proximal and distal axonal segments (Figure 6A). In contrast, either the frataxin-deficient neurons or the control neurons treated with A23 increased the number of mitochondria and the percentage of area occupied by mitochondria in proximal segments (Figures 6B,C). Both showed lower mitochondrial elongation and higher interconnectivity of the mitochondrial network than in basal control neurons (Figures 6D,E). Also, in both cases, the mitochondrial swelling was reached until double the mitochondrial size of the control neurons in some neurons (Figures 6F,G). Then, frataxin deficiency alters mitochondrial morphology as they become shorter, more rounded and swollen than the control; moreover, mitochondria are aggregated and retained in the proximal axon. This effect was exactly reproduced by Ca²⁺-overload control. However, we found a few changes in the control neurons after treatment with CCCP-oligomycin, indicating that mitochondrial depolarization did not affect mitochondrial morphology like frataxin deficiency. Consequently, we observed similar mitochondrial morphology in intra-axonal Ca²⁺ overload and in frataxin-deficient neurons, suggesting a more relevant implication of cytosolic Ca²⁺ levels in the formation of axonal spheroids.

Our results suggested a relevant role of Ca²⁺ in the pathophysiology of FRDA. Consequently, we looked for pharmacological targets against Ca²⁺-mediated axonal injury. To improve Ca²⁺ handling and attenuate the consequences of Ca²⁺ overload, we modulated calcium levels with Ca²⁺ chelators.

YG8R neurons were grown at 2 DIV and treated for 24 h with EGTA (an extracellular Ca²⁺ chelator), BAPTA (an intracellular Ca²⁺ chelator) or o-phenanthroline (a chelator of metal ions). After treatments, we observed a decrease in the activity of calpain, a metal-ion-dependent enzyme, suggesting that the Ca²⁺ was chelated by BAPTA, EGTA or o-phenanthroline (Figure 7A). In the confocal images, the improvement in mitochondrial distribution along the YG8R neurons after the treatments can be appreciated (Figure 7B), indicating that the modulation of calcium levels could be a critical point to improve the axonopathy observed in these neurons.

To test the effect of a reduction in calcium levels on mitochondrial morphology and axonal spheroids formation, YG8R neurons were grown with EGTA, BAPTA or o-phenanthroline for 4 DIV (Figure 8A). The treatment with EGTA for 4 DIV did not result in a sufficiently robust survival of YG8R neurons to allow a mitochondrial morphology study. Neuronal cultures, after immunolabeling with β-tubulin III and Mitotracker, were analyzed for mitochondrial morphology. The treatments on YG8R neurons with BAPTA and o-phenanthroline improved mitochondrial morphology alterations approaching control values (Figures 8B–G). Both treatments showed higher elongation and lower interconnectivity of the mitochondrial network than in YG8R basal conditions (Figures 8D,E), and decreased swelling mitochondria with values that reach sizes of the control neurons (Figures 8F,G). Computer analysis showed that control neurons treated with CaCl₂ presented similar alterations in mitochondrial morphology than those observed in frataxin-deficient neurons (Figures 8B–G), again suggesting the relevant role of cytosolic Ca²⁺ levels in axonal damage.