High‐dose biotin restores redox balance, energy and lipid homeostasis, and axonal health in a model of adrenoleukodystrophy

Abstract

Biotin is an essential cofactor for carboxylases that regulates the energy metabolism. Recently, high‐dose pharmaceutical‐grade biotin (MD1003) was shown to improve clinical parameters in a subset of patients with chronic progressive multiple sclerosis. To gain insight into the mechanisms of action, we investigated the efficacy of high‐dose biotin in a genetic model of chronic axonopathy caused by oxidative damage and bioenergetic failure, the Abcd1⁻ mouse model of adrenomyeloneuropathy. High‐dose biotin restored redox homeostasis driven by NRF‐2, mitochondria biogenesis and ATP levels, and reversed axonal demise and locomotor impairment. Moreover, we uncovered a concerted dysregulation of the transcriptional program for lipid synthesis and degradation in the spinal cord likely driven by aberrant SREBP‐1c/mTORC1signaling. This resulted in increased triglyceride levels and lipid droplets in motor neurons. High‐dose biotin normalized the hyperactivation of mTORC1, thus restoring lipid homeostasis. These results shed light into the mechanism of action of high‐dose biotin of relevance for neurodegenerative and metabolic disorders.

Introduction

Biotin is an essential B‐complex vitamin that controls energy metabolism through its role as a cofactor for five carboxylases: pyruvate carboxylase (PC), 3‐methylcrotonyl‐CoA carboxylase (MCC), propionyl‐CoA carboxylase (PCC), and the two isoforms of acetyl‐CoA carboxylase (ACC1 and ACC2) (69). ACC1 is the rate‐limiting enzyme that generates malonyl‐CoA, the two‐carbon building block for the synthesis of fatty acids, which are essential components of the myelin sheath (9, 68). PC, PCC, and MCC generate intermediates of the tricarboxylic acid cycle and thus increase the levels of cellular ATP at the rate of one molecule of ATP produced per one molecule of acetyl‐CoA used by the TCA cycle. Indeed, nutritional biotin deficiency and biotin deficiency induced by the loss of the recycling enzyme biotinidase causes severe ATP depletion (24). The importance of biotin in the brain is best manifested by inherited errors of biotin metabolism, which lead to severe neurological impairments involving myelin, as seen in biotinidase deficiency, which is reversible to some extent with biotin treatment (76) in the same manner as biotin‐thiamine responsive basal ganglia disease (53).

Recently, treatment with a formulation of high‐dose pharmaceutical‐grade biotin (MD1003, 10 000 times the recommended daily intake) resulted in improvement of clinical markers in a subset of patients suffering from progressive multiple sclerosis (MS) (70). Thereby, we set out to investigate its possible mechanisms of action as the first step in developing a rationale for broadening its application to other neurodegenerative diseases. We chose an X‐ALD mouse model since its hallmarks, that is, redox imbalance of mitochondrial origin intertwined with energy deficiency that leads to axonal degeneration, are shared by MS and the most common neurodegenerative disorders, such as Alzheimer, Parkinson's disease, and ALS (22).

X‐ALD is the most common peroxisomal disease, with an incidence of 1:14 700 live births (50). It is caused by mutations in the ABCD1 gene (51) located on Xq.28, which encodes a peroxisomal transporter that imports very long‐chain fatty acids (VLCFAs) into the peroxisome for degradation by β‐oxidation (61, 75). As a consequence, VLCFAs, especially C26:0, accumulate in tissues and plasma and constitute pathognomonic biomarkers for diagnosis. There are two main forms of the disease (14, 15): (i) cALD or cAMN, a rapidly progressing cerebral demyelinating leukodystrophy that leads to death (35–40% of cases) and (ii) adrenomyeloneuropathy (AMN), which makes up 60% of cases and affects adult men and heterozygous women over the age of 40 (13) and is characterized by distal axonopathy involving the corticospinal tract in the spinal cord and peripheral neuropathy. For the cerebral form of the disease, current therapeutic options are restricted to bone marrow transplantation (47) and for childhood cALD, hematopoietic stem cell gene therapy (5, 12). No treatment is currently available for the AMN phenotype, although a phase II pilot trial of a combination of high‐dose antioxidants recently showed some promise (6). The mouse model of X‐ALD (Abcd1 ^‐) develops axonopathy and locomotor impairment very late in life, at 20 months of age, and thus resembled AMN, the most common X‐ALD phenotype (58). The closest Abcd1homolog, Abcd2, exhibits overlapping metabolic functions (20), shows a complementary expression pattern (16, 71), and has thus been postulated as a modifier of the biochemical defects (52, 57). Double mutant Abcd1⁻/Abcd2^−/− mice develop more substantial accumulation of VLCFAs and more severe, earlier onset axonopathy starting at 12 months of age, which makes this model better suited for therapeutic assays (57). Indeed, this mouse model was instrumental in pinpointing early oxidative damage, mitochondrial depletion and bioenergetic failure as early, intertwined culprits of axonal degeneration (23, 39, 40) and in identifying therapeutic targets along the way (32, 33, 48, 49, 59). This knowledge aligns X‐ALD physiopathogenesis with that of the most prevalent neurodegenerative disorders, including MS (22).

Here, we used X‐ALD patient's primary fibroblasts and two mouse models of X‐ALD (Abcd1⁻ and Abcd1⁻/Abcd2^−/− mice) to investigate the preclinical efficacy and mode of action of high‐dose biotin on the established biochemical hallmarks of redox dyshomeostasis and energetic failure, as well as axonal damage and locomotor function (41, 49, 59). Moreover, we uncovered novel alterations that results from the loss of the Abcd1 transporter, specifically profound dysregulation of the lipogenesis and fatty acid degradation pathways governed by mTOR/SREBP‐1c in the spinal cord concomitant with global increases in triglycerides and lipid droplets. High‐dose biotin restores metabolic balance and rescues axonal degeneration, indicating that it is a promising therapeutic option for X‐ALD and similar neurometabolic conditions.

Materials and Methods

Cell culture experiments

Human X‐ALD fibroblasts were obtained at the Bellvitge University Hospital after informed consent. Control and X‐ALD cells were grown in DMEM containing 1 g/L glucose, pyruvate, L‐glutamine, 10% foetal calf serum (FCS), 100 U/mL penicillin and 100mg streptomycin and maintained at 37°C in humidified 95% air/5% CO₂ (40). The compounds tested were added to 80%–90% confluent cells in medium containing 10% FCS at the following concentrations: 50 µM C26:0 and 0.25 to 10 μM pharmaceutical‐grade biotin (MD1003).

Evaluation of intracellular ROS

Intracellular H₂O₂ levels were estimated using the ROS‐sensitive probe H₂DCFDA (DCF) as previously described (19). Following incubation with 10 μM DCF for 30 minutes, the cells were washed twice with PBS and scraped into water. Intracellular and mitochondrial superoxide anion levels were estimated using DHE probes and MitoSOX^TM Red (Molecular Probes), respectively (41). Following incubation with 5 µM DHE for 10 minutes or 5 µM MitoSox for 10 minutes, the cells were washed twice with PBS and scraped into water. The homogenate was transferred to a 96‐well plate for fluorescence detection with a spectrofluorimeter. The fluorescence of DCF‐, DHE‐, and MitoSOX‐stained cells was measured with a spectrofluorimeter (excitation wavelength of 493 nm and emission wavelength of 527 nm for DCF; excitation wavelength of 530 nm and emission wavelength of 590 nm for DHE and MitoSOX). The fluorescence values were corrected for protein content. Fatty acids were dissolved in ethanol and added to the medium for 24 h, and biotin was dissolved in DMSO. The final concentration of ethanol and DMSO was, respectively, 0.83% and 0.01% in each well. Antimycin A (Ant A; 200 µM for 1 h) was used as a positive control.

Evaluation of reduced glutathione

The intracellular content of GSH was determined using monochlorobimane, a thiol‐reactive probe. Following incubation for 30 minutes with 100 µM monochlorobimane, the cells were washed twice with PBS and scraped into water. The fluorescence of monochlorobimane‐stained cells was measured with a spectrofluorimeter (excitation wavelength of 380 nm and emission wavelength of 460 nm). L‐Buthionine sulfoximine (BSO) (500 µM) was used as a positive control.

Inner mitochondrial membrane potential quantification by flow cytometry

Treated cells were washed with PBS and incubated with 50 nM TMRE (Molecular Probes) in pre‐warmed PBS for 30 minutes at 37°C. Cells were trypsinized, centrifuged at 1000 × g for 5 minutes, and resuspended in pre‐warmed PBS. All samples were captured by a FACSCanto^TM flow cytometer, which recorded 20 000 cells for each condition and genotype tested. FCCP (200 µM for 10 minutes) was used as a positive control. Histograms showing the inner mitochondrial membrane potential levels (∆ᴪm) were obtained after gating live cells. The data were analyzed with FlowJo Tree Star software.

Quantitative real‐time PCR

Total RNA was extracted using the RNeasy Kit (Qiagen). Total DNA was extracted using the Gentra Puragene Tissue Kit (Qiagen). Gene expression and mtDNA levels were measured by TaqMan quantitative real‐time PCR as previously described (49).

Mouse strains

The methods for generating and genotyping of Abcd1⁻ (Abcd1^Tm1Kds) and Abcd2^−/− (Abcd2^Tm1Apuj) mice were previously described (16, 42, 57, 58). We used male mice of a pure C57BL/6J background. All methods employed in this study were in accordance with the ARRIVE guidelines, the Guide for the Care and Use of Laboratory Animals (Guide, 8th edition, 2011, NIH), European (2010/63/UE) and Spanish (RD 53/2013) legislation. Experimental protocols were approved by IDIBELL, IACUC (Institutional Animal Care and Use Committee) and regional authority (3546 DMAH). IDIBELL animal facility is accredited by The Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC, Unit 1155). Animals were housed at 22ºC on specific‐pathogen‐free conditions, in a 12‐hour light/dark cycle, and ad libitum access to food and water. Cages contained three to four animals.

Mouse experiments

Two X‐ALD mouse models were used in this study. The first model was the Abcd1 ^‐ mouse model, which exhibits molecular signs of pathology, including oxidative stress (19) and altered energy homeostasis (23); however, the first clinical signs of AMN (axonopathy and locomotor impairment) appear at 20 months (57, 58). We characterized the molecular signs of adult AMN in these mice. The second model was a mouse with double gene knockout of both the Abcd1 and Abcd2 transporters (Abcd1⁻/Abcd2^−/−). Compared with the Abcd1 ^‐ mice, the Abcd1⁻/Abcd2^−/− mice display enhanced VLCFA accumulation in the spinal cord (57), higher levels of oxidative damage to proteins (21), and a more severe AMN‐like pathology with an earlier onset at 12 months of age (16, 57, 58); therefore, this is the preferred model for assaying therapeutic strategies. Notably, no disease‐causative role of ABCD2 has been demonstrated; however, its absence induces a partially overlapping fatty acid pattern compared with that of the Abcd1 ^‐‐dependent biochemical phenotype (20, 37). We assessed the clinical signs of AMN in these Abcd1⁻/Abcd2^−/− mice.

Rationale for dose selection

The dose (60 mg/kg/day) was selected for use in mice because it is equivalent to the human therapeutic dose used in clinical trials (300 mg/day) (65).

Treatment duration

Treated animals had free access to AIN‐76A + 0.06% high‐dose pharmaceutical‐grade biotin (MD1003), which was distributed twice a week so that a dose of 60 mg/kg/day was reached. To measure the biochemical alterations, animals were treated for 3.5 months and sacrificed at 13 months of age. The animals were randomly assigned to one of the following dietary groups: group I, WT mice that received normal AIN‐76A chow (n = 18); group II, Abcd1⁻ mice that received normal AIN‐76A chow (n = 18); or group III, Abcd1⁻ mice that received AIN‐76A chow containing high‐dose biotin (n = 18). To measure the clinical signs of the disease (locomotor dysfunction and axonal degeneration), 12‐month‐old animals were treated for 6 months. The animals were randomly assigned to one of the following dietary groups: group I, WT mice that received normal AIN‐76A chow (n = 14); group II, Abcd1⁻/Abcd2^−/− mice that received normal AIN‐76A chow (n = 15); or group III, Abcd1⁻/Abcd2^−/− mice that received AIN‐76A chow containing high‐dose biotin (n = 16).

Morbidity and mortality

Each animal was checked for mortality and morbidity once every 3 days during the acclimation period and treatment period, including on weekends and public holidays.

Body weight

The body weight of the animals was recorded twice a week during the study. No change was observed.

ATP levels

Mice were sacrificed by cervical dislocation, and the spinal cords were immediately frozen in liquid nitrogen and stored at −80°C. ATP was extracted and quantified as reported (23).

Motor function tests

Locomotor function was analyzed in WT and Abcd1⁻/Abcd2^−/− mice at 18 months of age under blinded conditions as previously described (32, 33, 39, 48, 49, 57).

Treadmill test

The treadmill apparatus (Panlab, Barcelona, Spain) consisted of a belt (50 cm long and 20 cm wide) with variable speed capacity (5 to 150 cm/s) and slope (0–25º) enclosed in a plexiglass chamber. An electrified grid was located to the rear of the belt, and footshocks (0.2 mA) were administered whenever the mice fell off the belt. The mice were placed on the top of the already moving belt facing away from the electrified grid and in the direction opposite to the movement of the belt. Thus, to avoid footshocks, the mice had to locomote forward. The latency to fall off the belt (time of shocks in seconds) and the number of shocks received were measured as previously described (32, 33, 39, 48, 49, 57).

Horizontal bar cross test

The bar cross test was carried out using a wooden bar that was 100 cm in length and 2 cm in width (diameter). In each experimental session, the number of hind limb lateral slips and falls from the bar was counted in four consecutive trials (32, 33, 39, 48, 49, 57).

Clasping

For behavioral testing, hindlimb clasping was assessed; WT and Abcd1 ⁻ /Abcd2 ^−/− mice were suspended from their tails until they reached a vertical position but were allowed to grab the grill of the lid of the cage with their forelimbs, as adapted from Dumser et al. (10). Hindlimb reflexes were analyzed for 10 s in three consecutive trials separated by 5 minutes of rest.

Terminal euthanasia

The mice were anesthetized with Dolethal and perfused with 4% paraformaldehyde (PFA) solution.

Immunohistochemistry

Tissue preparation

Spinal cords were harvested from 18‐month‐old WT, Abcd1⁻/Abcd2^−/− and Abcd1⁻/Abcd2^−/− mice treated with high‐dose biotin after perfusion with 4% PFA as previously described (16, 32, 33, 39, 48, 49, 57). The spinal cords were embedded in paraffin, and serial sections (5‐μm thick) were cut in the transverse or longitudinal (1‐cm long) plane. The number of abnormal specific profiles in every 10 sections was counted for each stain. At least three sections of the spinal cord per animal per stain were analyzed.

Staining

The sections were stained with hematoxylin and eosin and Sudan black or processed for immunohistochemistry with a rabbit anti‐Iba1 antibody (diluted 1/1000; Wako; 019‐19741); a rabbit anti‐glial fibrillary acidic protein (GFAP) antibody (diluted 1/300; Dako; Z‐0334); a mouse anti‐synaptophysin antibody (diluted 1/500; Leica; SYNAP‐299‐L‐CE); a rabbit anti‐amyloid precursor protein (APP) antibody (diluted 1/100; Serotec; AHP538); a mouse anti‐cytochrome c antibody (diluted 1/100; BD Pharmigen; 55643); a mouse anti‐non‐phosphorylated neurofilament H (SMI32) antibody (diluted 1/3000; Covance Inc.; SMI‐32P); and a rabbit anti‐malondialdehyde (MDA) antibody (diluted 1/1000; (16, 39, 48).

Microscopic examination

Photos were taken with an Olympus BX51 conventional light microscope coupled to an Olympus DP71 color digital camera. The software used to analyze the photos was Cell^B from Olympus. APP and synaptophysin were quantified as previously described (16, 32, 33, 39, 48, 49, 57). The number of GFAP⁺ cells (astrocytes) and Iba1⁺ positive cells (microglia) per mm² in the spinal cord ventral horn of WT, Abcd1⁻/Abcd2^−/− and high‐dose biotin‐treated Abcd1⁻/Abcd2^−/− mice (n = 5) was determined. The number of brown cells was counted with the Cell Counter ImageJ plugin. The data are presented as the average of the data obtained from two 20x images per animal for each group.

Lipidomic analysis

Lipidomic analyses of the mouse spinal cord were performed according to the modified Folch method (17). For each sample, the solvent volume was adjusted to the weight of the fresh tissue. In brief, 190 μL of CHCl₃/MeOH 2:1 (v/v) and 10 μL of internal standard mixture were added to 10 mg of spinal cord. The samples were vortexed for 60 s and then, sonicated for 30 s using a sonication probe. Extraction was performed after incubation for 2 h at 4°C and centrifugation at 15 000 × g for 10 minutes at 4°C. The upper phase (aqueous phase), which contained ganglioside species and several lysophospholipids, was transferred and dried under a stream of nitrogen. The protein interphase was discarded, and the lower lipid‐rich phase (organic phase) was pooled with the dried upper phase. The samples were then reconstituted in 200 μL of CHCl₃/MeOH 2:1, vortexed for 30 s, and sonicated for 60 s, and these total lipid extracts (TLEs) were then stored at −80°C until analysis. Before analysis, the samples were diluted 100 times in MeOH/IPA/H2O 65:35:5 (v/v/v) before injection.

The samples were separated on an HTC PAL system (CTC Analytics AG) coupled with a Transcend 1250 liquid chromatographic system (Thermo Fisher Scientific, Inc.) using a Kinetex C8 2.6 μm 2.1 × 150 mm column (Phenomenex, Sydney, NSW, Australia). High‐resolution mass spectrometry using a Q‐Exactive mass spectrometer (Thermo Fisher Scientific, Inc.) was performed as previously described (66). The relative amount of each lipid was semi‐quantified as the area of its corresponding chromatographic peak.

LPC 26:0 absolute quantification

A total of 400 µL of CHCl₃/MeOH 2:1 followed by 75 µL of pure water was added to 100 µL of total lipid extract. The samples were then centrifuged at 15 000 × g for 10 minutes. The lower phase, which contained phospholipids, was transferred, dried under a stream of nitrogen, and reconstituted in 200 µL of CHCl₃.

Glycerophospholipids were further purified by solid phase extraction (SPE) using Supelclean™ LC‐NH2 cartridges (Sigma‐Aldrich; ref: 57014) (3). The cartridges were first conditioned in 2 mL of hexane and loaded with 200 µL of samples reconstituted in CHCl₃. Four different washing steps with 2 mL of diethylether, 1.6 mL of CHCl₃/MeOH 23:1 (v/v), 1.8 mL of diisopropyl ether/acetic acid 98:4 (v/v) and 2 mL of acetone/MeOH 9:1.2 (v/v) were then performed. Finally, 2 mL of CHCl₃/MeOH 2:1 (v/v) was added, and the corresponding eluent was dried under a stream of nitrogen and reconstituted in MeOH/IPA/H2O 65:35:5 (v/v/v) before injection. LC‐HRMS experiments were performed in positive ionization mode using the same system as that used for the lipidomic method described above.

Absolute quantification of LPC 26:0 in the mouse spinal cord was performed using the standard addition method. A calibration curve was obtained by adding increasing quantities (0, 10, 40, and 100 ng/mg of tissue in triplicate) of pure LPC 26:0 standard (Avanti Polar Lipids; ref: 855810P) to pooled of mouse spinal cord tissues. The calibration curves were analyzed in duplicate, that is, at the beginning of the analytical batch containing the mouse spinal cord samples and at the end. The endogenous concentration of LPC 26:0 in the mouse spinal cord samples was then evaluated by measuring the area under the curve of the protonated [M + H]⁺ ion and calculated according to calibration curves and the standard addition procedure.

Triglyceride levels

Spinal cords (~10 mg) were homogenized in 100 mL of solution containing 5% NP‐40 in water, and then, the samples were slowly heated to 80–100°C in a water bath for 2–5 minutes or until the NP‐40 became cloudy and cooled to room temperature. The heating was repeated one more time to solubilize all TAG. The samples were centrifuged for 2 minutes (top speed in a microcentrifuge) to remove any insoluble material. The TAG levels were quantified with a triglyceride quantification kit according to the manufacturer's protocol (BioVision, ref: K622‐100). The absorbance was measured at 570 nm. All assays were performed in duplicate.

Oil red O staining and quantification of lipid droplets (LDs)

Spinal cords were harvested from mice after perfusion with 4% PFA and embedded in Tissue‐Tek^® OCT (optimum cutting temperature). Serial sections (12‐µm thick) were then cut in the transverse plane at −20°C using a cryostat.

Oil red O (ORO) staining was used to measure lipid accumulation. A stock ORO solution was generated by diluting 0.35 g ORO (Sigma‐Aldrich; ref: 000625) in 100 mL of isopropanol. To prepare the staining solution, the ORO stock solution was filtered, mixed with dH2O at a 6:4 ratio and then, filtered using a 0.2‐micron syringe filter. Spinal cord sections were incubated in ORO staining solution for 15 minutes at room temperature, washed in dH2O and immediately covered with Fluoromount TM Aqueous mounting medium (Sigma‐Aldrich; ref: F4680). Brightfield microscopy images were acquired using a Nikon Eclipse 80I microscope equipped with a Nikon DS‐Ri1 camera operated by NIS‐Elements BR software. An average of six images per animal was acquired.

To assess LDs, images were edited using Adobe Photoshop CS5 software. The area corresponding to the neuronal soma was selected and saved as the region of interest (an average of 25 selections/animal). Then, the total area of LDs accumulated in each neuronal soma was analyzed using ImageJ software. For each measurement, the data were normalized to the tissue area to exclude unwanted biases and to determine the ratio of LD area to tissue area.

Immunoblotting

Mice spinal cords were homogenized in RIPA buffer (50 mM Tris‐HCl, pH 8, 12 mM deoxycholic acid, 150 mM NaCl and 1% NP40) supplemented with Complete Protease Inhibitor Cocktail (Roche) and PhosSTOP EASYpack Phosphatase Inhibitor Cocktail (Roche), sonicated for 2 minutes in 10‐s intervals and centrifuged for 10 minutes at maximum speed. The protein concentration was determined using a BCA protein assay kit (Thermo Fisher Scientific, Inc.). The samples were heated for 10 minutes at 70°C after adding 4X NuPAGE LDS Sample Buffer (Invitrogen, Thermo Fisher Scientific, Inc.). A total of 30–60 µg of protein was loaded onto a 10% Novex NuPAGE SDS‐PAGE gel system (Invitrogen, Thermo Fisher Scientific, Inc.) and run for 60–90 minutes at 120 V in NuPAGE MOPS SDS Running Buffer (Invitrogen, Thermo Fisher Scientific, Inc.) supplemented with 5 mM sodium bisulfite (Sigma‐Aldrich; ref. 243973). SeeBlue Plus2 Pre‐Stained (Invitrogen, Thermo Fisher Scientific Inc.) was used as a ladder. Immunoblotting was carried out with the avidin–biotin peroxidase method, as reported earlier (59).

Antibodies

The following antibodies were used for Western blotting: phospho‐mTOR (diluted 1/1000; Cell Signaling [Beverly, MA, USA]; 5536S); γ‐tubulin (diluted 1/10000; Sigma‐Aldrich); and horseradish peroxidase‐conjugated goat anti‐rabbit IgG (diluted 1/10000; Invitrogen; 81‐6520) and horseradish peroxidase‐conjugated goat anti‐mouse IgG (diluted 1/10000; Invitrogen; 81‐6120) secondary antibodies.

Quantification and Statistical analysis

Statistical analysis was carried out in GraphPad Prism6. The values are expressed as the mean ± standard deviation (SD). Significant differences between comparing two groups were determined by two‐tailed unpaired Student's t‐test (*P < 0.05, **P < 0.01, ***P < 0.001). When comparing more than two groups, significant differences were determined by one‐way ANOVA followed by Fisher post hoc or Sidak post hoc test (*P < 0.05, **P < 0.01, ***P < 0.001).

Results

Biotin counteracts redox imbalance in human fibroblasts

We first investigated whether biotin is efficient in addressing a core problem of X‐ALD, the increased ROS production generated by an excess of hexacosanoic acid C26:0. We previously showed that excess C26:0 triggers mitochondrial reactive oxygen species (ROS) production by electron transport chain (ETC) complexes while depleting the reduced glutathione (GSH) pool and decreasing membrane potential (ΔΨm) (41).

At doses of 0.5 to 5 µM, biotin successfully prevented ROS production induced by excess 26:0, as visualized using the dichlorodihydrofluorescein diacetate (DCF) and dihydroethidium (DHE) fluorescent probes, to assess total intracellular H₂0₂ and superoxide anion, respectively (Figure 1A,B). The same doses of biotin were efficacious at preventing mitochondrial ROS production assessed by MitoSOX (Figure 1C). Moreover, biotin treatment also reduced the mitochondrial ROS produced by the inhibitor of complex III antimycin, indicating that the precise mechanism of protection against ROS may not be restricted to excess fatty acids (Supporting Figure S1A). The mechanism might be indirect, as incubation with biotin for a short period of time (1 h) was not sufficient to abolish the increase in ROS levels (Supporting Figure S1B).

Figure 1.

Biotin prevents ROS production of mitochondrial origin and normalizes GSH levels and GSH biosynthesis genes in X‐ALD fibroblasts. A. Intracellular H₂O₂ (DCF probe) was quantified in control (CTL) (n = 3) and X‐ALD human fibroblasts (n = 3) after 24 h of treatment with vehicle (as a control), C26:0 or C26:0 with biotin (0.25, 0.5, 2.5 or 5 µM). For this and the experiments in B and C, C26:0 was used at 50 µM. Antimycin A (200 µM for 1 h) was used as a positive control. B. Intracellular superoxide anions (DHE probe) were quantified in CTL (n = 4 to 5 by condition) and X‐ALD human fibroblasts (n = 3 to 5 by condition) after 24 h of treatment with vehicle (as a control), C26:0 or C26:0 with biotin (0.5, 2.5, or 5 µM). C. Mitochondrial superoxide anions (MitoSOX probe) was quantified in CTL (n = 4 to 5 by condition) and X‐ALD human fibroblasts (n = 3 to 5 by condition) after 24 h of treatment with vehicle (as a control), C26:0 or C26:0 with biotin (0.5, 2.5 or 5 µM). D. Relative GSH levels were quantified in CTL (n = 4) and X‐ALD human fibroblasts (n = 4) after 24 h of treatment with vehicle (as a control), C26:0 or C26:0 with biotin (0.5, 2.5, or 5 µM). For this and the subsequent experiments, C26:0 was used at 50 µM. Buthionine sulfoximine (BSO; 500 µM for 24 h) was used as a positive control, since it inhibits the rate‐limiting enzyme on GSH biosynthesis. E. Relative GCLC, GCLM, and GSR1 gene expression was measured in CTL (n = 4 to 5 by condition) and X‐ALD human fibroblasts (n = 3 to 5 by condition) after 24 h of treatment with vehicle (as a control) or C26:0. Biotin (5 or 10 µM) was added 12 h after C26:0 treatment for 12 h. Gene expression normalized to RPLP0. Quantification presented as the fold change relative to vehicle‐treated fibroblasts. The experiments were done in triplicates. The values are expressed as the mean ± SD [one‐way ANOVA followed by Fisher post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001)].

Further, incubation for 24 h with biotin at 0.5 to 5 µM prevented GSH reduction in fibroblasts challenged with C26:0 excess, in both X‐ALD and control fibroblasts (Figure 1D). In contrast, incubation of 1 h was insufficient (Supporting Figure S1C).

Biotin did not exert any effects on the mitochondrial membrane potential decrease provoked by excess C26:0 (Supporting Figure S1D). We next explored whether biotin would influence transcription of genes governing GSH biosynthesis. As earlier reported, incubation with C26:0 was not inducing NRF2‐target genes (GCLC, GCLM, and GSR1) in X‐ALD fibroblasts, caused by a blunted response of the NRF‐2 pathway driven by dysregulated AKT/GSK3β signaling (76). Upon biotin treatment, the expression of the first rate‐limiting enzyme in GSH biosynthesis, glutamate‐cysteine ligase (GCL), was increased. This was true for both subunits, namely, the catalytic subunit (GCLC) and the modifier subunit (GCLM). Additionally, the expression of glutathione reductase (GSR1) was increased in X‐ALD fibroblasts only (Figure 1E). These results suggest that biotin reactivates the NRF2 pathway in X‐ALD fibroblasts.

High‐dose biotin rescues mitochondrial biogenesis and energy failure in Abcd1⁻ mice

Based on these positive results, we felt prompted to treat Abcd1 ^‐ mice (58) with a dose of biotin equivalent to the dose administered to MS patients (70); the mice were fed 60 mg/kg/day biotin for 3.5 months, from the age of 9.5 months to 13 months.

We then evaluated the following parameters of altered pathways in the spinal cord in X‐ALD: (i) target genes of NRF2, a transcription factor that governs the endogenous antioxidant response and shows a blunted response in X‐ALD (59); (ii) mtDNA copy number and the master regulators of mitochondrial biogenesis, which are also reduced in X‐ALD (18, 48); and (iii) ATP levels, which are reduced in X‐ALD (41) (Figure 2). Treatment with high‐dose biotin normalized the expression of NRF2 targets (Hmox1, Nqo1, Gsta3, and Gclc) (Figure 2A). In agreement with previous data on fibroblasts, the enzymes GSTA3 and GCLC may have directly contributed to preserving redox potential (Figure 1E). Next, we quantified mitochondrial biogenesis based on the mitochondrial DNA/nuclear DNA (mtDNA/nDNA) ratio (Figure 2B) and the induction of PGC‐1α and TFAM mRNA in Abcd1⁻ mice (Figure 2C), which were normalized by the treatment. We also observed that high‐dose biotin reversed the drop in ATP levels (Figure 2D).

Figure 2.

High‐dose biotin rescues mitochondrial biogenesis and prevents energy failure in the Abcd1⁻ mouse spinal cord. Experiments were performed on the spinal cord of 13‐month‐old WT, Abcd1⁻ and Abcd1⁻ mice treated with high‐dose biotin (Abcd1⁻ + high‐dose biotin) (A‐D). A. Relative gene expression of NRF2 target genes (Hmox1, Nqo1, Gsta3, and Gclc) was analyzed by quantitative RT‐PCR in WT, Abcd1⁻ and Abcd1⁻ + high‐dose biotin mice (n = 5 to 11 per genotype and condition). B. mtDNA content was analyzed by quantitative RT‐PCR and is expressed as the ratio of mtDNA to nuclear DNA (mtDNA/nDNA) (n = 6 to 7 per genotype and condition). C. Relative gene expression of Pgc‐1a, Tfam, and Nrf1 was analyzed by quantitative RT‐PCR (n = 6 to 11 per genotype and condition). D. ATP levels (n = 7 to 8 per genotype and condition). Gene expression normalized to mouse Rplp0 and presented as the fold change to relative to that in WT mice. The values are expressed as the mean ± SD [one‐way ANOVA followed by Fisher post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001)].

High‐dose biotin prevents locomotor deficits in Abcd1⁻/Abcd2^−/− mice

Our next step was to evaluate the preclinical efficacy of high‐dose biotin in alleviating axonopathy and the associated locomotor dysfunction exhibited by the X‐ALD mouse model (57). We used mice with complete loss of Abcd1 and its closest homolog, Abcd2 (Abcd1⁻/Abcd2^−/−), as they represent an improved model of AMN‐like phenotypes, showing an onset of axonal degeneration at approximately 12 months of age, which is amenable to therapeutic testing. The Abcd1⁻/Abcd2^−/− mice were treated for 6 months with high‐dose biotin and then, challenged with three locomotor tests. In the treadmill test, the total number and duration of shocks received by the Abcd1⁻/Abcd2^−/− mice were higher than those received by the WT mice under conditions of high speed and high slope, indicating that the mutants reached exhaustion earlier than WT. Remarkably, the animals fed high‐dose biotin performed the same as control mice (Figure 3A). In the bar cross experiment, while the double knockout mice slipped off the bar more frequently and needed a longer time to reach the opposite platform, the Abcd1⁻/Abcd2^−/− + high‐dose biotin mice performed the task the same as their WT littermates and were statistically similar (Figure 3B). In the clasping test, the best score for each animal was used for statistical analysis. Abcd1⁻/Abcd2^−/− mice presented a lower score than WT mice, demonstrating a locomotor deficit. In this test, high‐dose biotin treatment also significantly improved this score (Figure 3C). In conclusion, high‐dose biotin rescued locomotor deficits in Abcd1⁻/Abcd2^−/− mice.

Figure 3.

High‐dose biotin halts axonal degeneration and locomotor dysfunction in Abcd1⁻/Abcd2^−/− mice. A. The treadmill test and (B) bar cross test, and (C) clasping test were carried out in 18‐month‐old WT, DKO and DKO + high‐dose biotin mice (n = 13 to 16 by genotype and condition). A. The latency to fall off the belt (time of shocks) and the number of shocks received were computed after 5 minutes. B. The time required to cross the bar and the number of slips were quantified. C. The best score for each animal was used for analysis (10). D‐A′. Immunohistological analysis of axonal pathology was performed in 18‐month‐old WT, Abcd1⁻/Abcd2^−/− (DKO) and Abcd1⁻/Abcd2^−/− plus high‐dose biotin (DKO + high‐dose biotin) mice (n = 6/genotype and condition). Spinal cord immunohistological sections were processed for (D‐F) Iba1, (G‐I) GFAP, (J‐L) synaptophysin, (M‐O) APP, (P‐R) Sudan black, (S‐U) SMI‐32, (V‐X) Cytochrome c, and (Y‐A′) MDA. Representative images (D, G, J, M, P, S, V, and Y) for WT, (E, H, K, N, Q, T, W, and Z) DKO (F, I, L, O, R, U, X, and A′) and DKO + high‐dose biotin mice are shown. Scale bar = 25 µm. The values are expressed as the mean ± SD [one‐way ANOVA followed by Fisher post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001)].

High‐dose biotin prevents axonal damage in Abcd1⁻/Abcd2^−/− mice

We then evaluated the correlation between the observed improvements in locomotor dysfunction and arrested axonal degeneration. Abcd1⁻/Abcd2^−/− mice presented an overt neuropathological phenotype characterized by the following: (i) microgliosis (Figure 3D,E, Supporting Figure S2A) and astrocytosis (Figure 3G‐H, Supporting Figure S2B), as detected by Iba1 and GFAP staining, respectively; (ii) axonal damage suggested by the accumulation of synaptophysin (Figure 3J,K, Supporting Figure S2C) and amyloid precursor protein (APP) in axonal swellings (Figure 3M,N, Supporting Figure S2C); (iii) scattered myelin debris around the axonal ovoids, probably secondary to axonal degeneration, as detected by Sudan black (Figure 3P,Q); (iv) reduced SMI‐32 staining in motor neurons, an indicator of disturbed neurofilament status (Figure 3S,T); (v); a diminished amount of cytochrome c (Figure 3V,W), which is a marker of mitochondrial depletion; and (vi) increased staining for malondialdehyde (MDA), a marker of lipoxidation (Figure 3Y,Z) (16, 32, 33, 39, 48, 49, 57, 59). We found that treatment with high‐dose biotin in Abcd1⁻/Abcd2^−/− mice efficaciously suppressed microgliosis (Figure 3D‐F, Supporting Figure S2A) and astrocytosis (Figure 3G‐I, Supporting Figure S2B). Moreover, synaptophysin (Figure 3J‐L, Supporting Figure S2C) and APP accumulation (Figure 3M‐O, Supporting Figure S2C) and myelin debris were prevented (Figure 3P‐R), demonstrating that high‐dose biotin halted axonal damage in this X‐ALD mouse model. It is worth noting that the current experimental setting does not discriminate between disturbed axonal trafficking or axonal transection. Moreover, the treated cohort presented healthier motor neurons (Figure 3S‐U), normalized mitochondrial content (Figure 3V‐X) and reduced signs of lipoperoxidation (Figure 3Y‐A′), confirming the role of high‐dose biotin in maintaining redox and metabolic homeostasis in the nervous system in this model.

Increased triglyceride levels in the spinal cord of Abcd1 ⁻ mice and restoration by high‐dose biotin

Knowing the effects of biotin on fatty acid synthesis as a cofactor of the enzymes ACC1 and ACC2, we generated an untargeted lipidomic profile of the spinal cord of 13‐month‐old WT, Abcd1⁻ and Abcd1⁻ + high‐dose biotin mice to evaluate the possible global effects of the treatment using liquid chromatography coupled to a high‐resolution mass spectrometer (LC‐HRMS). We identified and annotated 747 unique lipid species belonging to different lipid classes: free fatty acids (FA), fatty acylcarnitines (FA‐Carn), cholesteryl esters (Chol. Ester), diacylglycerols (DG), triacylglycerols (TAG), lyso‐glycerophosphocholines (LPC), lyso‐glycerophosphoethanolamines (LPE), lyso‐glycerophosphoinositols (LPI), lyso‐glycerophosphoserines (LPS), glycerophosphocholines (PC), glycerophosphoethanolamines (PE), glycerophosphoglycerols (PG), glycerophosphoinositols (PI), glycerophosphoserines (PS), ceramides (Cer), hexosylceramides (HexCer), dihexosylceramides (DiHexCer), sphingomyelins (SM), sulfoglycosphingolipids (Su), and gangliosides (GSL). The only class significantly altered in Abcd1⁻ mice was TAG (Figure 4A). This is in line with the increased TAG detected in the peripheral blood mononuclear cells (PBMC) of a cohort of patients with AMN (62) and more recently in the fibroblasts of cALD patients (34). Seventy‐nine different TAG species, ranging from a TAG with 42 carbons and 0 double bonds (TAG 42:0) to a TAG with 75 carbons and four double bonds (TAG 75:4), were annotated. Of these, 61 species of TAG accumulated over 1.3‐fold, with 33 of these 61 species reaching statistical significance. Strikingly, the levels of all of these different species of TAG were normalized by treatment with high‐dose biotin (Figure 4B‐D, Supporting Figure S3A,B).

Figure 4.

High‐dose biotin rescues TAG levels in Abcd1⁻ mice spinal cords. A. Lipidomic experiments were performed in the spinal cord of 13‐month‐old WT (n = 5), Abcd1⁻ (n = 3) and Abcd1⁻ + high‐dose biotin (n = 6) mice, and all lipid species annotated are represented (FA = free fatty acids, FA‐Carn = fatty acyl carnitines, Chol. Ester = cholesterol esters, DG = diacylglycerols, TAG = triacylglycerols, LPC = lyso‐glycerophosphocholines, LPE = lyso‐glycerophosphoethanolamines, LPI = lyso‐glycerophosphoinositols, LPS = lyso‐glycerophosphoserines, PC = glycerophosphocholines, PE = glycerophosphoethanolamines, PG = glycerophosphoglycerols, PI = glycerophosphoinositols, PS = glycerophosphoserines, Cer = ceramides, HexCer = hexosylceramides, DiHexCer = dihexosylceramides, SM = sphingomyelins, Su = sulfoglycosphingolipids and GSL = gangliosides). B and C. Species of TAG annotated by lipidomic analysis which are significantly altered in Abcd1⁻ or Abcd1⁻ + high‐dose biotin mice. They are represented in function of the numbers of both total carbons and double bonds. Relative lipid levels were normalized to WT mice. D. TAG levels in the spinal cord of 13‐month‐old WT (n = 5‐6 per genotype and condition). E. A Venn diagram depicting TAG species that have been annotated by untargeted lipidomic experiments in the spinal cord of Abcd1⁻ mice and Ddhd2^−/− mice (25). The values are expressed as the mean ± [for A, B, C and E: one‐way followed by Sidak post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001); for D: one‐way followed by Fisher post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001)].

We noticed that a similar untargeted lipidomic study was undertaken in brain of adult mice with knockout of the TAG hydrolase DDHD2, the main enzyme that degrades TAGs in nervous tissue (25). The brains of Ddhd2^−/− mice accumulated 13 species of TAGs ranging from TAG 48:0 to TAG 66:18, and 11 of these species were significantly increased (2.07‐ to 5.45‐fold). We observed that a high number of species identified in the previous study were also identified in the Abcd1⁻ mice based on our lipidomic data, with 1.08‐ to 2.33‐fold accumulation in the Abcd1⁻ mice (Figure 4E).

We also detected a tendency for cholesterol esters (CE) to be increased in the Abcd1⁻ mice, albeit nonsignificantly (Figure 4A). Both, CE and TAG alterations were normalized by high‐dose biotin (Figure 4A).

However, when we specifically investigated the effects of high‐dose biotin on lysophosphatidylcholine C26:0, a phospholipid species widely used as a pathognomonic marker for the diagnosis of X‐ALD (72), we found no amelioration of the profile (Supporting Figure S3C).

Increased size and number of LDs accumulated in motor neurons of Abcd1⁻/Abcd2^−/− mice and restoration by high‐dose biotin

TAG and sterol esters are mostly stored in cells as LDs, which act as reservoirs to be used in case of insufficient energy sources (73). We, therefore, decided to visualize possible LD accumulation at the cellular level in the spinal cord by performing classical ORO staining on spinal cord sections. This histological stain is specific for neutral lipids and does not stain the polarized phospholipids of the cell membrane (45), enabling clear visualization of LDs by bright‐field microscopy (46). ORO staining revealed an increase in LDs in neuronal somas in the ventral horn of Abcd1⁻ and Abcd1⁻/Abcd2^−/− mice, with no differential staining in glial cells. Quantitative analysis using ImageJ software to assess the area occupied by LDs relative to the area of the motoneuronal soma indicated an augmented surface in mutant animals, in Abcd1⁻ mice at 12 months of age (Supporting Figure S4), and in Abcd1⁻/Abcd2^−/− mice at 18 months of age (Figure 5A,B,D,E). This accumulation of LDs was normalized in Abcd1⁻/Abcd2^−/− mice after 6 months of treatment with high‐dose biotin (Figure 5A‐G).

Figure 5.

High‐dose biotin normalizes excess lipid droplet accumulation in motorneurons in Abcd1⁻/Abcd2^−/− mice. A‐G. Histological analysis performed in 18‐month‐old (A and D) WT, (B and E) Abcd1⁻/Abcd2^−/− (DKO) and (C and F) Abcd1⁻/Abcd2^−/− plus high‐dose biotin (DKO + high‐dose biotin) mice. Spinal cord histological sections were processed for oil red O (ORO) staining (n = 4 to 5 per genotype and condition). G. Quantification of lipid droplet accumulation in spinal cord histological sections from 18‐month‐old WT, DKO, and DKO + high‐dose biotin mice by ImageJ. Scale bar = 25 μm (A‐C); scale bar = 5 μm (D‐F). The values are expressed as the mean ± SD (one‐way ANOVA followed by Fisher post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001).

Transcriptional dysregulation of lipid homeostasis in the Abcd1⁻ spinal cord and restoration by high‐dose biotin

We set out to gain insight into the origin of the increased TAG levels in X‐ALD and the mechanisms of normalization by high‐dose biotin to assess whether the effects were caused by an increase in biosynthesis or a decrease in the degradation of TAG. We thus measured a set of genes involved in the main lipid homeostatic pathways: (i) the synthesis of fatty acids (Figure 6A) and (ii) the degradation of fatty acids via β‐oxidation (FAO) in mitochondria and peroxisomes (Figure 6B). At baseline, we observed an induction of lipogenesis in the Abcd1⁻ spinal cord, which characterized by an upregulation of sterol regulatory element binding protein 1c (SREBP‐1c) and its targets Fas, Scd1, Scd2, Scd3, Acsl1, Acsl3 Acsl4, Acsl5, and Acsl6 (Figure 6A). SREBP‐1c belongs to the basic‐helix‐loop‐helix‐leucine zipper (bHLH‐LZ) family of transcription factors and controls lipogenesis (67). Moreover, the FAO genes, in particular Cpt2, Acadm, Acadl, Acadvl, Echs1, Hadha, Acaa2, Acox1, and Ehhadh, showed decreased expression (Figure 6B).

Figure 6.

High‐dose biotin restores lipid homeostasis through SREBP‐1c/mTOR in Abcd1⁻ mice spinal cords. Experiments were performed in the spinal cord of 13‐month‐old WT, Abcd1⁻ and Abcd1⁻ + high‐dose biotin mice. Relative gene expression of enzymes involved in (A) lipogenesis (Srebp1c, Fas, Scd1, Scd2, Scd3, Acsl1, Acsl3, Acsl4, Acsl5, and Acsl6) in WT, Abcd1⁻ and Abcd1⁻ + high‐dose biotin mice (n = 7 to 11 per genotype and condition) and (B) mitochondrial and peroxisomal fatty acid β‐oxidation (Cpt1a, Cpt2, Acadm, Acadl, Acadvl, Echs1, Hadha, Acaa2, Acox1, and Ehhadh) in WT, Abcd1⁻ and Abcd1⁻ + high‐dose biotin mice. Gene expression was normalized to mouse Rplp0 and is presented as the fold change to relative to that in WT mice (n = 5 to 7 per genotype and condition). C. P‐mTOR protein level (n = 4 per genotype and condition). Protein content was normalized relative to γ‐tubulin and quantification depicted as fold change to WT mice. The values are expressed as the mean ± SD [one‐way ANOVA followed by Fisher post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001)].

High‐dose biotin exerted a striking effect in modulating the transcription of core lipid homeostatic pathways; the treatment repressed SREBP‐1c and thus reduced the lipogenesis gene expression (Figure 6A) while inducing fatty acid beta‐oxidation by normalizing the levels of enzymes in both peroxisomes and mitochondria (Figure 6B), which may explain the correction of total TAG levels in treated X‐ALD mice (Figure 4A‐D). Intrigued by this concerted dysregulation of both anabolic and catabolic lipid pathways, we investigated mTOR, the mammalian target of rapamycin and nutrient sensor. This kinase is activated by anabolic signals and governs global lipid metabolism, including SREBP‐1c, in response to nutrition (56). When induced, mTOR facilitates the accumulation of TAGs by promoting lipogenesis while inhibiting the beta‐oxidation of fatty acids and lipolysis (4). Here, we detected increased levels of P‐mTOR, the active form of mTOR. Importantly, treatment with high‐dose biotin restored the levels of activated mTOR (Figure 6C).

In summary, in X‐ALD, high‐dose biotin is efficacious in correcting the molecular drivers of disrupted lipid homeostasis in the nervous system, which may eventually lead to halted axonal demise (Figure 7).

Figure 7.

Model recapitulating the mode of action of High‐dose biotin in X‐ALD. High‐dose biotin restores redox through reactivation of the NRF‐2 response, and normalizes the mTOR/SREBP1 transcriptional program, thus re‐establishing lipid homeostasis. The lipid droplet accumulation in X‐ALD may be caused by a combination of factors: (i) An increase in the formation of lipid droplets via increased lipid synthesis driven by mTOR/SREBP‐1; (ii) A decrease in the mitochondrial capacity to degrade lipid droplets caused by fewer mitochondria and lower fatty acid beta‐oxidation: the latter driven by mTOR/SREBP1, and (iii) The decreased capacity of the peroxisome to degrade TAG caused by the loss of ABCD1: shown to be a key element in the tethering of LD to the peroxisomes.

Discussion

This study provides novel insights into the mode of action of high‐dose biotin on redox homeostasis and energy metabolism at multiple levels. Our data shows that treatment with high‐dose biotin halts axonal damage and locomotor dysfunction while preventing a decrease in ATP, mitochondrial DNA depletion, and oxidative damage in the spinal cord of X‐ALD mice. Moreover, it unveils a general dysregulation of lipid metabolism as a novel culprit that drives axonal demise in an X‐ALD model, also restored upon high‐dose biotin treatment. This has direct implications for X‐ALD as well as MS and other metabolic axonopathies since mitochondrial dysfunction has been reported in the chronically demyelinated axons of MS patients (11) and in acute MS lesions (43).

Biotin prevented total ROS and mitochondrial ROS production in vitro. To the best of our knowledge, this is the first time that biotin or high‐dose biotin was shown to induce the endogenous antioxidant response of NRF‐2. A different mechanism of action of the antioxidant role of biotin dependent on the biotinylation of HSP60 at its lysines, a mitochondrial heat‐shock protein that ameliorates oxidative stress in cells, was previously reported (36).

Biotin is a cofactor for five carboxylases. ACC1 is the cytosolic rate‐limiting carboxylase of the synthesis of malonyl‐CoA, the two‐carbon building block for fatty acid synthesis and subsequent lipogenesis (9, 68). Malonyl‐CoA is also a well‐known inhibitor of carnitine palmitoyltransferase I (CPT1) (44), the first enzyme required in the process of mitochondrial FAO. CPT1 inhibition is mainly caused by ACC2‐derived malonyl‐CoA, as ACC2 (1) is bound to the mitochondrial outer membrane where CPT1 acts. Surprisingly, treating Abcd1⁻ mice with high‐dose biotin upregulated the FAO pathway, and repressed the lipogenesis pathway, thereby decreasing triglyceride accumulation. These results suggest that the observed effects go beyond the actions of biotin on ACC1/ACC2 carboxylases in favor of a general effect on lipid homeostasis.

In addition, the recovery of ATP levels may also suggest that high‐dose biotin enhanced FAO in Abcd1 ^‐ mice. To enter the TCA cycle, fatty acid‐derived acetyl‐CoAs are bound to oxaloacetate by citrate synthase. Additional oxaloacetate may be directly provided by other biotin‐dependent carboxylases, such as pyruvate carboxylase (PC) (28) or indirectly provided by refurbishing TCA cycle intermediates through the action of propionyl‐CoA carboxylase (PCC) (77), thereby boosting the efficiency of FAO in producing ATP. Thus, the effects of biotin on PC and PCC rather than the enzymes ACC1/ACC2 could contribute to the restoration of metabolic homeostasis we see in this model.

Beyond the possible impact of biotin on its well‐known carboxylase targets, the effect of high‐dose biotin on the mTOR/SREBP‐1c axis is intriguing, as it may have broad implications beyond the field of neurodegenerative diseases given its role as master regulator of metabolism, cell growth, autophagy, and innate immunity (29, 55). We propose that mTOR overactivity in the nervous system of X‐ALD mice is neutralized by high‐dose biotin, igniting a cascade of compensatory effects that lead to restored lipid homeostasis, improved metabolic function and the axonal maintenance. Restoring the expression levels of SREBP‐1c and its targets, the lipogenic enzymes FAS, Δ9‐desaturases, acyl‐CoA synthetases and, moreover, the genes involved in mitochondrial fatty acid oxidation, may have direct consequences on axonal function. Indeed, knockout of SREBP‐1c in the mouse induced the development of peripheral neuropathy and decreased myelin synthesis, along with blunted fatty acid synthesis and increased fatty acid oxidation, establishing a direct link between SREBP‐1c function and axonal health (7).

Thus, normalization of the mTOR/SREBP‐1c axis by high‐dose biotin led to a reduction in total TAG levels in Abcd1⁻ spinal cords, although the levels of LPC C26:0 were maintained, indicating that there was no direct action on ABCD1 or other peroxisomal transporters potentially bypassing ABCD1 loss. The normalization of TAG levels by biotin was previously observed in a study of WT mice fed pharmacological doses of biotin (31). The treated mice showed a similar reduction in SREBP‐1c‐dependent lipid synthesis in the liver and adipose tissue along with decreased serum levels of TAG. Compared to our study, the experimental setting in that work was shorter (8 weeks), the mice were younger (3 months) and were of a different genetic background (Balbc), and lower doses of biotin were used (13.5 mg/kg/day vs. 60 mg/kg/day in our study), which underscores the robust effect in different organs and treatment conditions. Notably, the hypotriglyceridemic effect of biotin supplementation was also observed in nondiabetic and diabetic human probands in the absence of effects on cholesterol, glucose, or insulin levels (60). Taken together, these findings suggest a broader indication of high‐dose biotin to treat systemic metabolic conditions.

The accumulation of LDs in the spinal motoneurons of X‐ALD mouse models may have detrimental consequences. This appears to be an undescribed pathogenic factor inherent to X‐ALD underappreciated to date, although LDs were recently detected in an in vitro endothelial model mimicking the blood brain barrier (35). Based on this observation, the authors suggested that metabolic lipid dysfunction not only contributes to chronic axonal degeneration in the mouse model (63) but is also implicated in the cerebral ALD phenotype.

There is a long‐held view of LDs as inert intracellular storages for neutral lipids in all living organisms and cells; however, early indications of the accumulation of TAGs and sterol esters as consequence of membrane remodeling in neurodegenerative disorders and neural injury also exist, suggesting that LDs would merely be transient indicators of disease states (78). More recently, however, the direct impact of impaired LD dynamics on energy homeostasis and the overall physiology of organisms have begun to emerge (74). Later, advances have argued for a driving role of LDs in protection against oxidative stress. LDs may allow cells to safely sequester otherwise toxic lipids, in particular in cases of overabundant fatty acids, which may pose a threat to membrane integrity and peroxidability (2). Once these fatty acids become triglycerides, they incorporate into LDs and become relatively harmless (74). Indeed, in Drosophila larvae, LDs are substantially increased in the CNS as a result of oxidative stress originating from either hypoxic conditions or excess free radicals. By abolishing droplets specifically in glia, the capacity of neuronal stem cells to proliferate under oxidative stress milieu is arrested. The role played by these glial LDs is thought to be protective and has been attributed to the sequestration of vulnerable membrane fatty acids away from free radicals, thus avoiding the vicious cycle of lipid peroxidation (2). This view is further supported by work in flies, as elevated mitochondrial ROS in neurons induce lipid synthesis via JNK/SREBP in these cells (38). Neuronal fatty acids are then transferred to astrocytes to form LDs, where they are oxidized in mitochondria (27). The accumulation of peroxidized lipids induces neurodegeneration, supporting LDs as both markers and protective agents rather that causes of neuronal demise. Furthermore, antioxidants reduced LD accumulation, thereby delaying neurodegeneration in mice (38). Moreover, a primary, causative role of dysrupted LD dynamics in neuronal dysfunction is underscored by examining the consequence of loss of function of genes that are responsible for a number of hereditary cortical motoneuron diseases (the hereditary spastic paraplegias), such as atlastin, REEP1, seipin, spartin, and kinesin‐1. These proteins play crucial roles in LD biology, such as mediating the fusion of ER tubules or controlling the size of LDs (30, 54). More precise insights are derived from the study of DDHD2, the principal brain TAG lipase, which is defective in patients with spastic paraplegia 54 (SPG54) (64). DDHD2 patients exhibit a thin corpus callosum and periventricular myelin abnormalities in MRI, features reminiscent of other leukodystrophies, including X‐ALD. In Ddhd2^−/− mice, a greater amount of TAGs in the form of LDs accumulates in the brain and spinal cord (25). The disruption of TAG hydrolase activity impairs the capacity to protect cells from LD accumulation following free fatty acid exposure (26). These results converge into a strong body of evidence directly connecting the dysfunction of LDs to decreased protection from lipid‐caused oxidative stress, leading to motoneuron disease.

In sum, we believe the LD accumulation in X‐ALD may constitute a protective mechanism against excess VLCFAs, which does not rule out a direct contribution of the increased LD to axonal demise when the situation becomes chronical. This increased size of the LD compartment is most likely caused by multifactorial mechanisms: (i) an increase in the biogenesis of LDs through increased lipid synthesis driven by mTOR/SREBP‐1 and oxidative stress; (ii) a decrease in LD degradation capacity through impaired mitochondrial function caused by lower mitochondrial mass and lower fatty acid beta‐oxidation expression, the latter driven by mTOR/SREBP‐1; and (iii) the incapacity of peroxisomes to degrade at least some of the TAG fatty acids caused by the loss of ABCD1, which was shown very recently to be a key element in the tethering of LDs to the peroxisomes (8) (Figure 7). Consequently, LD accumulation in the spinal cord and increased TAG levels, which are also detectable in the periphery (PBMCs) of AMN patients (62), may be considered new biomarkers of X‐ALD.

Taken together, our findings strongly suggest that interventional treatment using high‐dose biotin may be an attractive therapeutic option for patients with X‐AMN. The biological effects could be easily monitored by quantitatively measuring TAG levels in the PBMC of patients with X‐AMN, as previously described (62). Moreover, these data open new perspectives for other diseases that exhibit axonal degeneration as a significant component of clinical progression, together with redox dyshomeostasis, mitochondrial depletion and/or alterations in lipid metabolism.

Conflict of Interest

This study was supported by grants from Medday Pharmaceuticals SA who provided high‐dose pharmaceutical‐grade Biotin. High‐dose Pharmaceutical‐grade Biotin, MD1003, is a not approved investigational product being developed for use in patients with progressive MS.

Funding information

We thank CERCA Program/Generalitat de Catalunya for institutional support. This study was supported by grants from Medday Pharmaceuticals SA to AP, the Autonomous Government of Catalonia [SGR 2014SGR1430; 2017SGR1206], and Instituto de Salud Carlos III [PI17/00916] (Co‐funded by European Regional Development Fund. ERDF, a way to build Europe) to AP; by Instituto de Salud Carlos III through the grants [Miguel Servet program CPII16/00016] to SF (Co‐funded by European Social Fund. ESF investing in your future); and the Center for Biomedical Research on Rare Diseases (CIBERER) to MR and NL. Locomotor experiments were performed by the SEFALer unit F5 (CIBERER) led by AP. Writing and editorial assistance was provided by Jean Fiber funded by MedDay Pharmaceuticals.

Supporting information

Figure S1. (A) CTL (n = 3 to 4 by condition) and X‐ALD human fibroblasts (n = 4 to 5 by condition) were pretreated with biotin (0.5, 2.5, or 5 µM) for 24 h, and then, antimycin A (200 µM) was added to the medium for 1 h. Next, H₂O₂ (DCF probe) was quantified. (B) CTL (n = 4) and X‐ALD human fibroblasts (n = 5) were pretreated with biotin (0.5, 2.5 or 5 µM) for 1 h, and then, antimycin A (200 µM) was added to the medium for 1 h. Next, H₂O₂ (DCF probe) was quantified. (C) Relative GSH levels were quantified in CTL (n = 5) and X‐ALD human fibroblasts (n = 3) after 24 h of treatment with vehicle (as a control) or C26:0. The cells were incubated with biotin (0.5, 2.5, or 5 µM) for the last hour. (D) The relative inner mitochondrial potential (ΔΨm) was measured in CTL (n = 4 to 5 by condition) and X‐ALD human fibroblasts (n = 3 to 4 by condition) after 24 h of treatment with vehicle (as a control), C26:0 or C26:0 with biotin (5 or 10 µM), with C26:0 being added to medium containing 10% FCS. FCCP (200 µM for 10 minutes) was used as a positive control. Quantification presented as the fold change relative to vehicle‐treated fibroblasts. n = 4‐5 per genotype and condition, with experiments done in triplicates. The values are expressed as the mean ± SD (one‐way ANOVA followed by Fisher post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001).

Fig S2

Figure S2. Quantification of (A) Iba1⁺ density (cells/mm²) (n = 4 to 5 per genotype and condition), (B) GFAP⁺ density (cells/mm²) (n = 4 to 5 per genotype and condition), and (C) synaptophysin and APP accumulation (n = 7 to 8 per genotype and condition) in spinal cord immunohistological sections from WT, DKO and DKO + high‐dose biotin mice. The values are expressed as the mean ± SD [one‐way ANOVA followed by Fisher post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001)].

Fig S3

Figure S3. Experiments were performed in the spinal cord of 13‐month‐old WT, Abcd1⁻ and Abcd1⁻ + high‐dose biotin mice. (A‐B) Species of TAG annotated by lipidomic analysis (n = 3 to 6 per genotype and condition) are represented based on the numbers of both total carbons and double bonds. The relative lipid levels were normalized to WT mice. The number is the relative fold change in the corresponding TAG species in (A) Abcd1⁻ and (B) Abcd1⁻ + high‐dose biotin mice compared with WT mice. In bold, species of TAG which are significantly different with (A) WT and (B) Abcd1⁻. Scale from red to blue represents the relative levels of each TAG species compared with WT. (C) LPC‐26:0 levels (ng/mg of tissue) WT (n = 5), Abcd1⁻ (n = 3) and Abcd1⁻ + High‐dose biotin (n = 6). Values are expressed as mean ± SD (One‐way followed by Fisher's post hoc test (*P < 0.05, **P < 0.01 and ***P < 0.001).

Fig S4

Figure S4. (A‐D) Histological analysis performed in 13‐month‐old (A, C) WT (n = 5) and (B, D) Abcd1⁻ mice, (n = 5). Spinal cord histological sections were processed for ORO staining. (E) Quantification of lipid droplet accumulation in spinal cord immunohistological sections from 13‐month‐old WT and Abcd1⁻ mice. Scale bar = 25 μm (A‐B); scale bar = 5 μm (C‐D). The values are expressed as the mean ± SD [Student's t‐test (*P < 0.05, **P < 0.01 and ***P < 0.001)].

Data Availability Statement

All data used and/or analyzed during the current study are available from the corresponding author on reasonable request.