Ataxia with giant axonopathy in Acbd5-deficient mice halted by adeno-associated virus gene therapy

Abstract

Acyl-CoA binding domain containing 5 (ACBD5) is a critical player in handling very long chain fatty acids (VLCFA) en route for peroxisomal β-oxidation. Mutations in ACBD5 lead to the accumulation of VLCFA and patients present retinal dystrophy, ataxia, psychomotor delay and a severe leukodystrophy. Using CRISPR/Cas9, we generated and characterized an Acbd5 Gly357* mutant allele.

Gly357* mutant mice recapitulated key features of the human disorder, including reduced survival, impaired locomotion and reflexes, loss of photoreceptors, and demyelination. The ataxic presentation of Gly357* mice involved the loss of cerebellar Purkinje cells and a giant axonopathy throughout the CNS. Lipidomic studies provided evidence for the extensive lipid dysregulation caused by VLCFA accumulation. Following a proteomic survey, functional studies in neurons treated with VLCFA unravelled a deregulated cytoskeleton with reduced actin dynamics and increased neuronal filopodia. We also show that an adeno-associated virus-mediated gene delivery ameliorated the gait phenotypes and the giant axonopathy, also improving myelination and astrocyte reactivity. Collectively, we established a mouse model with significance for VLCFA-related disorders.

The development of relevant neuropathological outcomes enabled the understanding of mechanisms modulated by VLCFA and the evaluation of the efficacy of preclinical therapeutic interventions.

Introduction

Very long-chain fatty acids (VLCFA) containing 22 or more carbon atoms are key components of several complex lipids, including glycerophospholipids, sphingolipids and ceramides. The synthesis of VLCFA occurs through fatty acid elongation from other saturated and unsaturated fatty acids.¹ The homeostasis of VLCFA is a fine-tuned process that requires the concerted interplay between synthesis and catabolism. Peroxisomes play a crucial role in catabolism, as they are the only organelle capable of handling VLCFA β-oxidation.^2,3 Defects in ABCD1, the peroxisomal transporter of VLCFA, prevent the β-oxidation process and cause the accumulation of these fatty acids.^4-7 In humans, deleterious loss-of-function variants in ABCD1 cause X-linked adrenoleukodystrophy (ALD), a severe disorder with adrenal, cerebral, spinal cord and/or peripheral nerve involvement.^8-11 Three main presentations, namely, Addison-only, cerebral inflammatory disease and adrenomyeloneuropathy (AMN), and variable age of onset from childhood to adulthood, highlight the complexity of understanding how VLCFA accumulation leads to cell and tissue impairments. Additionally, defects in ACOX1, the first enzyme in the three-step β-oxidation process, also specifically impair VLCFA β-oxidation leading to VLCFA accumulation.^12-14 The disease spectrum in ACOX1 deficiency can range from the severe neonatal presentation with hypotonia, seizures and hepatomegaly,^12,14 to an adult presentation with cerebellar and brainstem atrophy,¹⁵ which may be related to differences in residual Acox1 activity. More recently, ACBD5 was identified as another critical factor in VLCFA catabolism.^16-18 ACBD5 is a peroxisomal protein containing an acyl-CoA binding domain (ACBD) and a FFAT-motif that can mediate interaction with endoplasmic reticulum (ER)-resident proteins VAPA and VAPB.^19,20 The ACBD and the observation of VLCFA accumulation in ACBD5-deficient patients consolidated a central role of ACBD5 as a possible handler of VLCFA en route for ABCD1-mediated transport and peroxisomal β-oxidation.^21-24 ACBD5 deficiency was initially discovered in a subgroup of patients presenting rod-cone dystrophy.²⁴ Notwithstanding the retina defects, psychomotor delay, spasticity, ataxia and white matter involvement are also prominent features of the disorder.^21,23

Histological studies of tissues from people affected by ALD significantly contributed to the characterization of the neuropathology caused by the accumulation of VLCFA.^10,25-29 Active myelin destruction with infiltrating monocytes and lipid-laden macrophages are key signatures of the cerebral inflammatory presentation.^30-32 In contrast, spinal cord axonopathy and peripheral neuropathy are the main presentations of adult AMN presentation.^27,31,33 The lack of correlative features between VLCFA accumulation and disease onset or pathology highlights a complex system through which VLCFA mediates cell and tissue pathology.^34,35 The limited availability of mouse models with partial pheno- or pathocopying of the corresponding disease also hindered this understanding.^36-38 The Abcd1 knockout (KO) mouse has a very late onset of axonal degeneration.^36-40 The Acox1 KO mouse presents liver disease without nervous system involvement.^41,42 In both cases, the existence of possible alternative routes for handling the defect is thought to be a principal cause behind the lack of a more severe and targeted effect.^6,43,44 In support of this, it has been observed that the double Abcd1:Abcd2 KO mutant mouse presents a worsened pathology and disease presentation. Despite not reflecting the cerebral inflammatory form of ALD, this mutant mouse allowed significant progress in understanding mechanisms and evaluating candidate therapies.^45-47 In the case of ACOX1, the peroxisomal β-oxidation system is endowed with two additional oxidases, ACOX2 and ACOX3, which, although having primary affinities for bile acid intermediates and branched-chain fatty acids, do have residual activities towards straight-chain fatty acids.⁴⁸ On the other hand, ACBD5 seems to have a more centralized role as it may be critical for handling VLCFA and necessary for the correct presentation of these fatty acids to ABCD1. In support of this, an Acbd5^tm1a mutant mouse with blocked transcription has the characteristic accumulation of VLCFA and displays cerebellar and hepatic changes.⁴⁹

To investigate the consequences of abnormal VLCFA metabolism within the CNS, we developed, using CRISPR/Cas9, an Acbd5 Gly357* mutant mouse model. Here, we show that Gly357* mice faithfully mimic key aspects of the human disorder. Gly357* mutant mice developed progressive ataxia and severe neuropathology affecting the retina, cerebellum and spinal cord, with neurodegeneration, gliosis and myelin loss. Targeted lipidomic analysis revealed that Acbd5 deficiency leads to generalized lipid dysregulation. Proteomic analysis unveiled the cellular defects in degenerating spinal cord and uncovered a disruption in actin dynamics. Despite the multitude of cellular defects, the ataxia and neurodegeneration can be prevented through adeno-associated virus (AAV)-mediated expression of Acbd5. Our results provide mechanistic insights into how impaired VLCFA oxidation induces a severe neuropathology. Additionally, we demonstrate that AAV-mediated gene therapy is capable of preventing the escalating degeneration observed within the CNS.

Materials and methods

Establishment of the Gly357* mouse model using CRISPR/Cas9

Mouse husbandry and all procedures were done according to the European Directive 2010/63/EU, as well as the National legislation (Decreto-Lei 113/2013), and in accordance with Institutional guidelines, i3S Ethical Committee (n° 2017/09) and the Portuguese Veterinarian Board (009966/2018-05-17).

In brief, we used Genious software to analyse potential gRNAs targeting exon 9 of the Acbd5 gene. The target sequence 5′-AAAGGCGAAGTGAAACACGG-3′ displayed high on-target activity and low off-target scores and was used to synthesize gRNAs. Next, the ribonucleoprotein consisting of Cas9 protein (IDT) plus the gRNA were mixed with a ultramer 5′-gatggtagcagtcaaaggaaaaggcgaagtgaaacaCTGAGTGAGTGAGAATTCGGGcggaggagaagatggcagaagtagcagtggagcaccgcaccgtgagacgagaggtggagagagcgaggacttctccagtgtcaggagaggg-3′ modified with phosphorothioate bonds and were co-injected into fertilized eggs of F1 C57BL/6J:129S2 mice. F0 founders were screened for the insertion of 20 bp using PCR followed by sequencing. Founder 9 contained the desired mutation and was used to generate a congenic strain by backcrossing to C57BL/6J. At generation N7, we began breeding heterozygous mice to characterize the novel strain: B6.B6/129-Acbd5^em1#9PB/PB. PCR genotyping with primers Fwd-5′-TCAAACAATGGACACTTTCAGT-3′ and Rev-5′-GCTACTTCTGCCATCTTCTCC-3′ yielded amplicons of 180 bp for the wild-type allele and 201 bp for the mutant Acbd5 Gly357* allele.

Phenotypic evaluation

Limb clasping, horizontal ladder and paw print analysis were assayed in wild-type and Gly357* mice. For limb clasping, mice were scored 0 for the absence of clasping, 0.5 for partial clasping of only the hindlimbs, and 1 for clasping of all limbs. For the horizontal ladder test, mice were video-recorded walking (three runs per animal) along a ladder with evenly spaced metal rungs. During analysis of the videos, the percentage of paw slips was calculated. For paw print analysis, we applied black ink to the hind paws and recorded the locomotion on white paper (four to five sheets per animal). The resulting paw prints were scanned and analysed for stride and step length, toe spreading, base of support and paw area.

Western blotting

Lysates of spinal cord and mouse fibroblasts were prepared by sonication in lysis buffer containing 1% Triton, 0.1% SDS, 140 mM NaCl and 1× Halt Protease and phosphatase inhibitor (Thermo Scientific) in 20 mM Tris 1 mM EDTA pH 8.0. Total protein levels were measured using DC™ Protein Assay (Bio-Rad). Protein samples were separated by SDS-PAGE on Criterion™ TGX Precast (Bio-Rad) or Tris-Glycine NG (NuSep) 4–20% gels and transferred to nitrocellulose (0.45 μm; Amersham). Membranes were blocked with 5% skimmed milk (Millipore) in Tris-buffered saline (TBS), washed and incubated with primary antibody (Supplementary Table 3) diluted in 5% bovine serum albumin (BSA) in TBS with 0.1% Tween 20 (TBS-T). Horseradish peroxidase-conjugated (HRP, Jackson ImmunoResearch Laboratories) or IRDye (LI-COR) secondary antibodies were diluted 1:10 000 in 5% skimmed milk in TBS-T. HRP-labelled membranes were developed as described.⁵⁰ Li-cor IRDye-labelled membranes were scanned on an Odissey CLx Infrared Imaging System and quantified using Image Studio 5.2 software. Total Protein Staining (TPS; LI-COR) was used as loading control.

Liquid chromatography-mass spectrometry proteomics

Lysates of cervical spinal cords were prepared by sonication in lysis buffer (as above). Total protein levels were measured using DC™ Protein Assay (Bio-Rad), and 100 μg of each sample (6 months, wild-type n = 3, Gly357* n = 3; 12 months, wild-type n = 2, Gly357* n = 3) were processed in duplicate for the proteomics analysis that was performed by Proteomics service (i3S, Porto, Portugal), as previously described.⁵¹

Lipid extraction and liquid chromatography-mass spectrometry lipidomics

Spinal cords were collected from wild-type and Gly357* mice and immediately frozen and stored at −80°C. Lipid extraction was performed by Metabolomics Core Facility (EMBL). Liquid chromatography with tandem mass spectrometry (LC-MS/MS) analysis was performed using a Vanquish UHPLC system coupled to a Q-Exactive plus HRMS (Thermo Scientific) in ESI positive mode as follows: the separation of lipids and fatty acids was carried out on Kinetex C18 (100× 2.1; 2.6 uM) column at a flow rate of 0.26 ml/min and maintained at 30°C. The mobile phase consisted of solvent A [0.1% formic acid in isopropyl alcohol:acetonitrile (9:1)] and solvent B [0.1% formic acid water:acetonitrile (6:4)] were buffered with 10 mM ammonium formate for positive mode analysis. Lipid detection and data acquisition were performed as described.⁵² Lipid annotation was performed using BasicLipids and Waters Library hits from Progenesis.

Histological and morphological analysis

Mice were euthanized using a lethal anaesthesia mixture of ketamine/xylazine (125 and 12.5 mg/kg body weight, respectively) and tissues were isolated and fixed by immersion. For ultrastructural analysis, tissues were fixed in 4% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4) for 5 days at 4°C, and processed as described.⁵³ For axonal diameter analysis,⁵⁴ five to eight images per animal at 40× magnification were acquired on an Olympus optical microscope equipped with an Olympus DP 25 camera and Cell B software and imported into Photoshop (Adobe) for quantification. For g-ratio and myelin thickness, five to eight images per animal at ×12 000 magnification were analysed. The Photoshop Quick Selection and the Recording Measurement tools were used to determine the ratio of the diameter of the axon to the diameter of the myelinated fibre. For immunohistochemistry, tissues were fixed either in modified Carnoy’s solution (absolute ethanol:methanol:glacial acetic acid 6:3:1) for 2 h at room temperature, processed for paraffin embedding, sectioned at 4 μm, cleared and rehydrated; or 4% paraformaldehyde (PFA) overnight at 4°C, immersed in 30% sucrose overnight at 4°C, covered and kept in Cryomatrix (Thermo Scientific) at −80°C, sectioned at 25 μm and stored in 1× PBS. Sections (two to three sections per animal) were permeabilized in 100% methanol at room temperature for paraffin or −20°C for cryosections, treated with 0.1% NaBH₄ in 10 mM Tris, 1 mM EDTA at pH 9.0 and blocked with 5% normal donkey serum (NDS) for paraffin or 1% fish gelatin for cryosections in PBS. Primary antibodies were diluted in blocking buffer and incubated for 1 h 30 min at room temperature or overnight at 4°C (Supplementary Table 3). Alexa Fluor-conjugated secondary antibodies (Jackson ImmunoResearch Europe) were diluted in blocking buffer and incubated at room temperature for 1 h. All tissues sections were mounted with Ibidi Mounting Medium with 1× DAPI in microscope slides and visualized by epifluorescence in a Zeiss Axio Imager Z1 microscope equipped with an Axiocam MR3.0 camera and Axiovision 4.7 software. All quantifications, including cell counting and signal intensity measurement, were performed using the imaging software Fiji Image J.

Cortical neuron culture and VLCFA treatment

Cortical neuron culture was performed as described,⁵⁵ with minor alterations. Cortices were isolated from individual wild-type and Gly357* embryos at embryonic Day 17.5 (E17.5) and digested at 37°C in 0.05% Trypsin + 1× EDTA for 8 min. Digestion was stopped with Dulbecco’s modified Eagle medium (DMEM)/F12 medium (Pan-Biotech) containing 10% fetal bovine serum (FBS). After dissociation, cells were filtered and resuspended in culture medium containing Neurobasal medium (Invitrogen) supplemented with 2× B-27 (Invitrogen), 1% penicillin and streptomycin (P/S) and 2 mM L-glutamine (Invitrogen). Cells were plated into 24-well (Costar) or 15 μm-slide eight-well (Ibidi) plates coated with 20 μg/ml poly-L-lysine (Sigma) and 0.4 μg/ml of laminin (Sigma) at a density of 12 500 cells/well or 50 000 cells/well, respectively. Cortical neurons were maintained in culture and used for experiments at days in vitro (DIV)7 and DIV8. For treatment with VLCFA, tetracosanoic acid methyl ester (meC24:0; Cymit Quimica) was used.^56,57 A meC24:0 work solution at 5 mM was prepared in 50% DMSO/50% H2O. Wild-type and Gly357* cortical neurons were treated with 60 μM or the vehicle, which were added to the culture medium at DIV1 and DIV4.

Actin dynamics and super-resolution STED microscopy

For actin retrograde flow analysis, cortical neurons plated in an eight-well plate were transfected using Lipofectamine™ 3000 (Invitrogen) at DIV5 with 0.4 μg of Lifeact-GFP plasmid, and analysis of actin dynamics was performed as previously described.⁵⁸ Immunolabelling and super-resolution imaging was performed as described,⁵⁹ with minor alterations. Cortical neurons were fixed at DIV8 with 4% PFA for 10 min at room temperature, followed by permeabilization with 0.1% Triton X-100 in PBS for 5 min and autofluorescence quenching with 0.2 M ammonium chloride (Merck) for 3 min at room temperature. Non-specific labelling was blocked by incubation with blocking buffer (5% NDS in PBS) for 1 h. Actin staining was performed with 0.3 mM phalloidin 635P (Abberior GmbH) diluted in PBS, which was incubated for 30 min at 37°C. Stimulated emission depletion microscopy (STED) images were acquired with a single point scanning confocal Stellaris 8, equipped with a fully motorized inverted Leica DMI8 microscope, a HC PL APO 93×/1.3STED WHITE glycerol immersion objective equipped with a correction collar (Leica Microsystems), a WLL and Falcon and STED modalities.

Plasmids and AAV therapy

We designed two AAV plasmids, i.e. a control plasmid AAV-GFP expressing green fluorescent protein under the CMV promoter, and the AAV-Acbd5 plasmid expressing under the CMV promoter the turbo green fluorescent protein linked via the 2A self-cleaving small peptide (P2A) to the mouse Acbd5 gene.

Lung-derived fibroblasts were isolated from 6-month-old wild-type and Gly357* mice. Briefly, lungs were dissected, cut into small pieces and briefly digested 0.05% trypsin-1× EDTA (Alfagene) for 30 min at 37°C. The pieces were further disrupted by mechanical dissociation with a P1000 micropipette tip. The resulting suspension was plated on a petri dish and cultured at 37°C, 5% CO₂ in DMEM containing 10% FBS and 1% P/S. For transfection, cells were cultured in a six-well plate at 50%–70% confluency overnight at 37°C, 5% CO₂. Cells were transfected with plasmids AAV-GFP and the AAV-Acbd5 using Lipofectamine™ 3000 transfection reagent. After 24 h, cells were lysed and processed for western blot analysis.

For in vivo gene therapy, plasmids were packaged into AAV-PHP.eB particles (Vectorbuilder) and used to transduce mice at 3 months of age. Gly357* mice were transduced via a single tail intravenous injection of the viral solution containing 1.06 × 10¹² gene copies for AAV-Acbd5 and 1.48 × 10¹² for AAV-GFP. Wild-type untransduced mice served as controls. Mouse behaviour was blindly assessed at 3 months before transduction and at 6 months of age. At this age, mice were euthanized and tissues processed for biochemical and histological analyses. To measure Acbd5 mRNA we used RT-qPCR. Briefly, total RNA was prepared using NZY Total RNA Isolation Kit and following manufacturer’s protocol (Nzytech). cDNA was prepared from 0.9 μg of RNA using NZY First-Strand cDNA Synthesis Kit (Nzytech), as recommended by the manufacturer. SYBR-Green quantitative PCR (qPCR) was performed using a CFX384 Touch™ Real-Time PCR Detection System (Bio-Rad) with primers for Acbd5 (Fwd-5′-GCCGTGAAGGTGATCCAGAG-3′ and Rev-5′-CAGAATCCAGGCCGTGAAAG-3′) and for β-actin (Fwd-5′-GGGCTGTATTCCCCTCCATC-3′ and Rev- 5′-TCTCCATGTCGTCCCAGTTG-3′). The fold change in gene expression was calculated using the ΔΔCt relative expression method (Livak method) and normalized for β-actin.

Statistical analysis

Results are expressed as the mean ± standard error of the mean (SEM) and data were analysed with Prism software (GraphPad 8.1.1 Software). To compare two groups, parametric unpaired Student’s t-test was used. For comparison of more than two groups, one-way ANOVA tests were used, followed by Sidak’s or Tukey’s post hoc tests. P < 0.05 was the criterion used to consider a significant difference.

Results

Generation of Acbd5 mutant mice

To study the impact of Acbd5 deficiency, we delineated a CRISPR-Cas9 gene-targeting approach. A specific gRNA was used in conjunction with homologous directed repair to introduce a 21 bp sequence in exon 9 of Acbd5 (Fig. 1A). The knock-in sequence introduces a premature codon stop causing the mutation p.Gly357* (Fig. 1B). A primer-based approach using primers flanking the knock-in sequence can differentiate between wild-type, heterozygous, and homozygous mutants (Gly357*) and was used for genotyping all animals (Fig. 1C). To eliminate possible off-target mutations, a founder carrying the mutation was selected, and a congenic strain was generated by backcrossing to C57BL/6J. At generation N7, with an estimated 99% of the donor genome eliminated, we began breeding heterozygous mice to generate and characterize Gly357* mice. Genotyping of the offspring from heterozygous matings revealed normal Mendelian ratios indicating that Gly357* mice were viable. Most of the described mutations in ACBD5 patients are nonsense leading to undetectable protein levels.^21,23,24 To characterize the consequences of the mutation, we analysed by western blot the expression of Acbd5 in liver and spinal cord lysates (Fig. 1D). Expression of Acbd5 was undetected in Gly357* mice, indicating that the Gly357* allele functions as a null allele.

Disruption of CNS lipid homeostasis in Gly357* mice

In humans and mice, Acbd5 deficiency impairs the β-oxidation of VLCFA, leading to the accumulation of these fatty acids.^22,23,49 To delineate the lipid changes caused by the mutant allele, we performed an unbiased LC-MS lipidomic analysis in spinal cords from wild-type and Gly357* mice. Gly357* mice had a 2-fold increase in C26:0 levels (Fig. 2A) and we observed a dysregulation of 139 lipids (Supplementary Table 1). The main lipid classes altered were phosphatidylcholines (PC), phosphatidylethanolamines (PE) and sphingolipids (SL). In spinal cords from Gly357* mice, we observed the accumulation of several lipid species containing saturated and unsaturated VLCFA (Fig. 2A). In addition, we also observed the accumulation of lipid species that did not contain VLCFA. Alkyl- and alkenyl-glycerophospholipids were among the lipids with the highest accumulation in Gly357* mice (Fig. 2B). Notably, Gly357* mice were also characterized by multiple lipid deficiencies. Numerous lipids containing VLCFA were reduced in Gly357* mice (Fig. 2C). Additionally, we also observed several defects in lysophospholipids that did not contain VLCFA (Fig. 2D). Overall, the altered lipid composition in spinal cords from Gly357* mice highlight a complex metabolic outcome from impaired Acbd5 function.

Loss of Acbd5 causes retina defects and ataxia

Mutations in ACBD5 were initially identified in patients with retinal dystrophy.²⁴ Histological analysis of retinas from Gly357* mice at 1 year of age (Supplementary Fig. 1A) revealed decreases in photoreceptor outer segment and outer nuclear area, which contains the cells bodies of photoreceptors (Supplementary Fig. 1B and C). In addition, the unchanged inner nuclear layer (Supplementary Fig. 1D) highlights a targeted degeneration of the synaptic terminals of rods and cones with a concomitant loss of neurons.

Notably, the most striking phenotypic presentation in Gly357* mice was limb clasping (Fig. 3A), a reflex defect initially detected at 4 months of age, which displayed complete penetrance by 6 months of age (Fig. 3B). Additional tests revealed that Gly357* mice also performed poorly in the horizontal ladder test. Increased paw slips during locomotion runs were a key feature of Gly357* mice (Fig. 3C). The motor and coordination defects were observed at 6 months of age, and there was a progressive worsening of the performance until 1 year of age (Fig. 3D). Increased severity of the motor presentation combined with the development of tremors contributed to establishing a humane end point at 14 months of age. Comparative analysis of footprint patterns (Fig. 3E) at 6 and 12 months of age also revealed locomotion defects. At 6 months of age, Gly357* mice had decreases in stride and step lengths (Fig. 3F and G). Worsening was observed at 1 year of age and at this stage, Gly357* also had defects in toe spreading (Fig. 3H), base of support (Fig. 3I) and paw area (Fig. 3J). Next, we sought to determine if there was a cerebellar involvement that could contribute to the impaired motor and coordination defects. Analysis of Purkinje cell histology and density revealed focal neuronal loss in 6-month-old Gly357* mice (Supplementary Fig. 2A–C). The pathology was progressive and by 1 year of age, Gly357* mice displayed generalized loss of Purkinje cells in the anterior lobules (Supplementary Fig. 2A–C). Axonal swellings were also a striking feature in surviving Purkinje cells from Gly357* mice (Supplementary Fig. 2B), which is indicative of neuron damage and may be a contributing factor to the progressive degeneration of Purkinje cells. These data support the conclusion that loss of Acbd5 function severely affects Purkinje cells, which may contribute to the ataxic presentation of Gly357* mice.

Extensive gliosis underscores CNS pathology in Gly357* mice

Alterations in astrocyte and microglia reactivity status are often associated with many CNS injuries and pathologies. To delineate the participation of these glial cells upon Acbd5 deficiency, we analysed the gliotic state on spinal cords from 6 and 12-month-old mice, using immunofluorescence with antibodies against the glial fibrillary acidic protein (Gfap) and ionized calcium-binding adapter molecule 1 (Iba1) (Fig. 4A and C). At 6 months, increased Gfap reactivity (Fig. 4B) was detected throughout the grey matter in Gly357* mice. By 1 year of age, increased astrocytosis was observed in both grey and white matter (Fig. 4A and B). Microgliosis was initially observed in the grey matter and progressed to a generalized condition within the spinal cord at 1 year of age (Fig. 4C and D). The initial prevalence of grey matter versus white matter gliosis may demark differential neuronal and myelin defects. The progressive regional expansion of the gliotic status from 6 to 12 months indicates the development of a generalized neuropathological condition.

Loss of Acbd5 causes demyelination in the CNS

Considering the white matter involvement observed in ACBD5 patients^23,60 and the astrocytic condition of Acbd5 mutant mice (Fig. 4A), we next explored myelination in the Gly357* mice. Ultrastructural analysis of spinal cords revealed thinned myelin sheets in Gly357* mice at 6 and 12 months of age (Fig. 5A). Measurements of g-ratio (Fig. 5B) and myelin thickness (Fig. 5C and D) confirmed defective myelination in Gly357* mice. Loss of myelin was not restricted to the spinal cord, as we also observed demyelination in the cerebellar white matter (Fig. 5E). To further characterize myelin defects, we quantified the expression of different myelin markers—myelin basic protein (Mbp), 2′,3′-cyclic nucleotide 3′-phosphodiesterase (Cnp) and tubulin βIV—in wild-type and Gly357* mice. The western blot data revealed decreases of Mbp, Cnp and tubulin βIV levels in Gly357* mice (Fig. 5F to H). Our results demonstrate a demyelinated CNS in Gly357* mice and highlight the relevance of Acbd5 dysfunction towards impaired myelin maintenance in the CNS.

Acbd5 deficiency causes giant axon swellings

The histological analysis of spinal cords revealed the presence of giant axons in the white matter of 6-month-old Gly357* mice (Fig. 6A, top row). By 1 year of age, these giant axons were a predominant feature (Fig. 6A). Analysis of axon diameter highlighted the magnitude of the increased axonal size, which in some cases reached almost 50 µm in diameter (Supplementary Fig. 3A). The analysis also revealed that spinal cords from Gly357* mice had an increased frequency of axons with larger diameters (Fig. 6B). This feature was more pronounced with ageing (Fig. 6B), suggesting a worsening of the pathology. To better understand the structure of these giant axons, we analysed longitudinal sections of the spinal cord. The histological analysis revealed that the giant structures were axonal swellings (Fig. 6C). Notably, individual axons could present several of these giant swellings that were separated by stretches with seemingly normal diameters.

The giant axonal swellings were a generalized pathological feature observed in the brain, spinal cord, optic nerve and cerebellum (Fig. 6D and Supplementary Fig. 3B). The ultrastructural analysis also revealed the accumulation of organelles (Fig. 6D and E). Whereas in some swellings, the organellar structure was intact, we also observed accumulations of organelles with altered structure and appearance (Fig. 6F). Another striking feature observed in axons from Gly357* mice was alterations to the abundance and distribution of the ER in stretches of axons just preceding or just following the giant swellings (Supplementary Fig. 3C). This prompted us to investigate if ER stress was a component of the cellular changes caused by Acbd5 deficiency. Using western blot and immunohistochemistry, we observed increased levels of protein disulphide isomerase (Pdi) and C/EBP homologous protein (Chop), two well established markers of ER stress.⁶¹ Additionally, the giant axonal swellings presented a severe disruption of the cytoskeleton organization with a meshwork of microtubules and neurofilaments (Fig. 6F). We next evaluated the levels of non-phosphorylated neurofilaments as a surrogate marker for axonal damage using SMI32 immunoreactivity.⁶² Analysis at 6 and 12 months of age revealed increased SMI32 in axons of Gly357* mice (Fig. 6G and Supplementary Fig. 3C). We also observed increased SMI32 reactivity in the cerebellar white matter of Gly357* mice, which also allowed the detection of axons with increased diameters (Supplementary Fig. 3D). Combined, our results demonstrate that the loss of Acbd5 causes a severe axonopathy with increased axon diameter and the formation of giant swellings. The disruption of the axonal cytoskeleton architecture is likely a contributing factor leading to defects in organelle transport and their regional accumulation.

Spinal cord proteomic signature mirrors cellular pathology and unravelled actin defects in Gly357* neurons

To investigate the mechanisms underlying the neuropathology, we analysed the proteomic profiles of wild-type and Gly357* spinal cords at 6 and 12 months of age. A total of 2108 and 1904 unique proteins were identified from the two age groups, respectively. At 6 months of age, 52 proteins were downregulated and 31 were upregulated in the spinal cords of Gly357* mice (Fig. 7A and Supplementary Table 2). At 1 year of age, we identified 29 downregulated and 73 upregulated proteins in Gly357* spinal cords (Fig. 7A and Supplementary Table 2). Amongst the differentially expressed proteins, 58 were only detected at 6 months, 77 were only detected at 12 months, and 25 proteins were found in the two age groups (Fig. 7B). Using data from single-cell RNAseq databases,^63,64 analysis of the 160 dysregulated proteins revealed that nine were mainly expressed in neurons, nine in astrocytes, 10 in microglia and 14 in oligodendrocytes (Fig. 7B, bottom). These results highlight the main cell types affected in the CNS pathology of Gly357* mice. Heat map clustering of proteomics expression data revealed that proteins commonly found at both ages displayed a higher dysregulation at 1 year of age (Fig. 7C), mirroring the progressive worsening of the pathology. Independent validation using western blot and immunofluorescence confirmed the dysregulation of 19 proteins from our proteomic data (Fig. 7D and E and Supplementary Fig. 4A).

Next, we aimed to identify potentially altered pathways using gene ontology analysis (Supplementary Fig. 4B). There was an enrichment of dysregulated proteins related to several molecular functions, including extracellular matrix and actin/cytoskeleton organization. Notably, the cluster related to actin binding contained 18 dysregulated proteins, which prompted further analysis. Independent validation of the dysregulation using western blot revealed the upregulation of alpha-actinin (Actn1), transgelin 2 (Tagln2) and transgelin 3 (Tagln3), calponin 3 (Cnn3), heat shock protein beta-1 (Hspβ1), ezrin (Ezr) and moesin (Msn) (Fig. 4F). Considering that some of these proteins are typically expressed in neurons, we investigated if their dysregulation could cause alterations to actin organization and its dynamics.

Using an in vitro approach, we cultured cortical neurons from wild-type and Gly357* mice. To bypass the initially slow progression of the axonal pathology, neurons were exposed to methyl-ester C24:0 (meC24:0). This VLCFA promptly enters cells and can be used by the elongation system to generate other VLCFAs, which contributes to the accumulation of VLCFA in cells with β-oxidation defects.⁵⁷ Next, we used super-resolution STED microscopy and phalloidin staining to visualize the neuronal actin cytoskeleton (Fig. 7F). Under control conditions, the actin cytoskeleton of Gly357* neurons had a similar structure to that of wild-type neurons. However, upon treatment with meC24:0, Gly357* neurons displayed increased actin-enriched filipodia and an abundance of actin patches (Fig. 7F). Next, we investigated actin dynamics by measuring the actin retrograde flow in neurons transfected with LifeAct-RFP.⁶⁵ Under control conditions, wild-type and Gly357* neurons had similar velocities of actin retrograde flow (Fig. 7G). However, upon administration of meC24:0, Gly357* neurons displayed a 20% reduction in actin dynamics (Fig. 7G). Combined, our data identified a signature of protein dysregulations that are delineated by the pathology and highlight neurons, oligodendrocytes, astrocytes and microglia as targets of Acbd5 deficiency. Moreover, we demonstrate that in neurons, VCLFA induced reorganizations of filamentous actin and defective actin dynamics.

AAV-mediated gene therapy ameliorates the presentation and axonopathy of Gly357* mice

We reasoned that Acbd5 gene delivery might be an appropriate approach to rescue the diverse and complex cellular defects observed in Gly357* mice. To test this hypothesis, we designed an AAV vector with a P2A-mediated bicistronic expression of green fluorescent protein (GFP) and murine Acbd5 under the control of the CMV promoter (AAV-GFP-P2A-Acbd5; AAV-Acbd5) (Supplementary Fig. 5A). The control vector also expressed GFP under the CMV promoter. The functionality of the AAV-Acbd5 vector was first assayed in transfection experiments using fibroblasts from wild-type and Gly357* mice (Fig. 8A). Acbd5 expression was not detected in control Gly357* cells or after the transfection with AAV-GFP. Notably, the AAV-Acbd5 vector was able to drive the expression of Acbd5 in Gly357* cells (Fig. 8A, top). Levels of GFP, the GFP-P2A fusion and total ERK1/2 were used to control for transfection and loading (Fig. 8A, middle and bottom).

Using the AAV serotype PHP.eB, which is capable of highly effective CNS transduction via systemic delivery, packed AAV particles of AAV-GFP and AAV-Acbd5 were delivered to Gly357* mice at 3 months of age, and we analysed gene expression, presentation and pathology at 6 months of age (Fig. 8B). The delivery of AAV-Acbd5 was able to normalize Acbd5 mRNA levels in Gly357* mice, whereas the control group maintained near-undetectable levels of Acbd5 (Fig. 8C). Histological analysis of longitudinal and cross-sections of spinal cords from control-treated Gly357*-AAV-GFP mice revealed the disorganization of axons and the presence of giant axonal swellings (Fig. 8D). Remarkably, the AAV-Acbd5 gene therapy was able to prevent the drastic development of the giant axonal swellings and maintain the normal organization of axons within the white matter tracts (Fig. 8D). Next, we quantified the number of dystrophic axons, i.e. axons with giant swellings, or containing the accumulation of organelles (Fig. 6D). After Acbd5 gene therapy, there was a very significant reduction of dystrophic axons in Gly357*-AAV-Acbd5 (Fig. 8E) and we also observed a decreased SMI32 immunoreactivity, indicative of reduced axonal damage (Fig. 8F). The evaluation of the mutant presentation was also significantly improved by the Acbd5 gene therapy, with Gly357*-AAV-Acbd5 mice having normal locomotion on the horizontal ladder, normal step length and significant improvements in limb clasping reflexes and stride length (Fig. 8G–I and Supplementary Fig. 5B). The histological analysis of GFP fluorescence in Gly357*-AAV-Acbd5 mice also revealed that in addition to neurons, oligodendrocytes and astrocytes were also transduced. This prompted us to investigate if the AAV-Acbd5 therapy could also have additional beneficial outcomes. The degree of myelination was improved in Gly357*-AAV-Acbd5 mice with increased myelin thickness (Supplementary Fig. 5D) and decreased g-ratio (Supplementary Fig. 5D). Whereas microgliosis was unaffected by the gene therapy, we observed an improvement in the degree of astrocytosis, with decreased Gfap immunoreactivity (Supplementary Fig. 5E–G). Combined, our data demonstrate that AAV-mediated gene therapy is able to halt the severe progression of the pathology and improve the characteristic disease presentation of Gly357* mice.

Discussion

The impaired metabolism of VLCFA with the ensuing accumulation of these fatty acids has a crucial relevance in several peroxisomal disorders. Understanding the impact of deficient VLCFA β-oxidation on myelin destruction, neuron degeneration and glial activation has been curbed by limited knowledge of the cellular mechanism(s) of disease. A significant contributing factor has been the lack of suitable in vivo models that mimic CNS pathology. In this work, we demonstrate that the Acbd5 Gly357* mouse develops a phenotype and neuropathology similar to that of ACBD5 patients.

We choose an engineered CRISPR/Cas9 strategy to generate an exonic premature translation-termination codon in exon 9 of Acbd5 that genetically mimics mutations found in ACBD5 patients.^21,24 Western blot analysis revealed that expression of Acbd5 is absent in the Gly357* mutant, similar to what is observed with other nonsense mutations.²³ The lack of Acbd5 mRNA expression in Gly357* mice is likely caused by active nonsense-mediated mRNA decay.^60,66 In the Acbd5^tm1a mutant, the knockout-first strategy significantly decreased Acbd5 expression but displayed variable degrees of reduction.⁴⁹ Combined, the Gly357* mutant displayed the characteristics of a null allele. The Gly357* mice develop a severe phenotype with early onset at 4 months of age. Limb clasping characterized the initial signs of impaired reflexes in Gly357* mice, that progressed to defects in motor coordination and locomotion, with penetrance and severity of the defects increasing with ageing. Compared to Abcd1, Abcd1:Abcd2 and Acox1 mutants, Gly357* is the first mutant mouse with an altered lifespan.⁶ From 12 months of age onwards, development of tremors and hindlimb paralysis justified the establishment of a humane end point by 14 months of age. Identifying such a clear and progressive presentation supported further studies to unravel how Acbd5 deficiency could impact the CNS.

Studies in skin fibroblasts and plasma from ACBD5 patients clearly demonstrate the impact on VLCFA accumulation,^23,67 which is also present in the Acbd5^tm1a mutant mouse.⁴⁹ In common with VLCFA β-oxidation defects, targeted lipidomics revealed the overt lipid dysregulation with the accumulation of lipids containing VLCFA. However, other lipid species not containing VLCFA were also accumulated in the spinal cords of Gly357* mice. The reductions in other lipid species with and without VLCFA highlighted the complexity of the dysregulation. The decreased levels of lysophospholipids may be a consequence of the excessive VLCFA and a response to the build-up of free VLCFA in membranes. Lysophospholipids are key intermediates in the remodelling of phospholipids, and the overproduction and accumulation of free VLCFA may hijack their homeostasis. To prevent the effects of the enriched free VLCFA in membranes, their incorporation into phospholipids may provide alternative solutions to handle and store the excess fatty acid content. Nonetheless, a current challenge is to understand whether there exists a connection between specific lipid alterations and cellular dysfunctions that could underlie tissue pathology and, by extension, the disease state. VLCFA-containing lipids might either directly contribute to the disruption of cellular processes or mediate further alterations in lipid composition that result in cellular dysfunctions.

A clear sign of the impact of Acbd5 deficiency in the CNS was the gliotic state observed from 6 months onwards. Reactive astrocytes and microglia were one of the main features of the neuropathology of Gly357* mice. In addition to the increased levels of Gfap and Iba1, the proteomic analysis also uncovered the generalized reactivity of these glial cells, with upregulations of Aldh1l1, Dhrs1, Lgals3, A2m and C1qc, amongst others. Reactive gliosis can be found in many CNS diseases, but we found parallelisms between the disease markers in the spinal cords of Gly357* mice and diseases caused by VLCFA accumulation. Hspb1 and Hspb5 were upregulated in Gly357* mice and astrocyte reactivity with upregulation of four heat shock proteins (HSPs), including HSPB1 and HSPB5, preceded the demyelination in cerebral ALD.⁶⁸ The microglia response in ALD has a particular spatial and temporal progression.^26,69 In Gly357* mice, we observed a progression from grey to white matter regions and the striking observation that white matter microgliosis was concurrent with the onset of myelin loss. Strikingly, in Acbd5 Gly357* mice, the thinning of myelin sheaths observed with ultrastructural analyses in spinal cords and cerebellum, combined with the decreased content of several myelin markers, denotes a relevant feature of the neuropathology. Importantly, myelin loss was never observed in other models of VLCFA accumulation,⁶ which broadens the scope of using Gly357* or Acbd5^tm1a mutant mice to understand how impaired VLCFA β-oxidation affects myelin and oligodendrocytes.

Our findings pinpoint an intricate conjunction of cellular pathologies that may cumulatively participate in the ataxic presentation. Loss of Purkinje cells in the cerebellum of Gly357* mice was progressive from 6 months onwards. Additionally, a striking axonopathy with giant axonal swellings was observed throughout the CNS of Gly357* mice. The sheer size of the swellings is aggravated by their multifocal presence within a given axon, the disruption of the axonal cytoskeleton, and the accumulation of organelles. Additionally, the extent to which the axons enlarge is partially accompanied by the radial expansion of myelin. Despite the ongoing demyelination, oligodendrocytes appear to adjust to the axonal enlargement as the swellings continue to be myelinated. Axonal swellings are a feature of several neurodegenerative disorders, but exemplified by their increased size (∼20 µm in diameter) in giant axonal neuropathy (GAN) and Charcot-Marie-Tooth disease type 2E (CMT2E).^70,71 In GAN, lipid alterations have been identified, which include the upregulation of FABP5 and lipid droplets.⁷² It is tempting to hypothesize that the increased magnitude of the axonal swellings in Gly357* mice is in part attainable by the complex lipidic rearrangement observed in Gly357* mice. Notably, Fabp5 was identified as upregulated in our proteomic analysis and lipid droplets were also increased in Acbd5^tmn1a mice.⁴⁹ Increased fluidity due to VLCFA⁷³ may be necessary to accommodate membrane expansion. Additionally, because of their increased size, VLCFA can span the lipid bilayer leaflets, which may provide increased membrane stability.¹ Nevertheless, the neuronal cytoskeleton should also provide some structural stabilization to withstand axonal integrity. In our proteomic analysis, several actin-binding proteins were found dysregulated in spinal cords from Gly357* mice. Several of these proteins are involved in actin remodelling, filament branching and formation of actin meshes and networks (e.g. Actn1, filamins) and link the actin cytoskeleton to membranes (Ezr, Msn). The complex nature of the dysregulation of actin-binding proteins was also manifested in a reduction of actin polymerization in VLCFA-treated neurons. However, we observed induced filopodia formation in neurons from Gly357* mice, which may be a result of increased Tagln2, Tagln3 and Fmnl2.^74,75 A link between fatty acids and actin organization has also been highlighted by the observation that α-linolenic acid (C18:3) disrupts the actin cytoskeletal network,⁷⁶ and that lignoceric acid (C24:0) can affect the organization of the actin network and its interaction with microtubules.⁷⁷

Despite breakthroughs in halting the progression of the childhood cerebral form of ALD using allogeneic and gene-corrected autologous haematopoietic stem cell transplantation, the development and efficacy of treatment options for peroxisomal disorders associated with VLCFA accumulation has been challenging.⁷⁸ The field has been piloted by efforts to treat ALD using several strategies that range from metabolic modulators and antioxidants^47,56,79 to bone marrow transplantation and gene therapy.^80-82 AAV-mediated gene delivery to the CNS has been one of the most promising tools for treating neurogenerative disorders.⁸³ In the Abcd1 KO mouse, gene therapy using rAAV9 was able to restore Abcd1 expression and lead to a partial improvement in the levels of C24:0, despite the short 15-day post-administration period.⁸⁴ In the PEX1-G844D mouse, a model for Zellweger spectrum disorder, gene delivery to the retina using AAV8 was shown to improve some electroretinogram outputs, and the potential to rescue C26:0 lyso-PC levels if administered before onset of retinal changes.⁸⁵ Our findings provide, to our knowledge, the first proof-of-concept that gene therapy can halt or prevent the neuropathology of Gly357* mice. In vivo efficacy was manifested by typical axonal architecture in the spinal cords, and normalization or improvement in locomotion and reflexes. The characteristic giant axonopathy of Glyc357* mice also showed a drastic reduction. Considering the reduced size of the swellings and their reduced frequency at 6 months of age, we hypothesize that they may represent signs of axonal damage that were already present at the start of the therapy. These initial swellings may not be amenable for a complete rescue, but their progressive degeneration seems to be halted. In addition to the high tropism for neurons, the AAV-PHP.eB capsid can transduce astrocytes and oligodendrocytes, albeit with less efficiency.⁸⁶ Nevertheless, we observed a significant improvement in myelination and decreased astrocytosis. The lack of AAV-PHP.eB tropism for microglia⁸⁶ underscores the persistent microgliosis observed in Gly357* mice, similarly to what was observed using rAAV9 in the Abcd1 KO mouse.⁸⁴ Combined, our results demonstrate that unrestricted delivery of Acbd5 using AAV can target the cellular defects and mitigate the severe and rapid neurodegeneration.

In summary, we report the generation and characterization of the Acbd5 Gly357* mutant mice that mimic several key features of the corresponding human disease, including retinal dystrophy, demyelination, gliosis and cerebellar ataxia with giant axonopathy. Proteomic analysis established key molecular pathways, of which actin dynamics and cytoskeleton showed prominent effects on neurons. The neuropathology of Gly357* mice was amenable to AAV-mediated gene therapy, highlighting a promising avenue for treating the disease.

Data availability

Underlying data can be accessed in the Supplementary material or from the corresponding author upon reasonable request.

Acknowledgements

The authors acknowledge the support of the i3S Scientific Platforms (CCGEN, Animal Facility, Proteomics Facility, Histology, and Electron Microscopy, Advanced Light Microscopy) and the national infrastructure Portuguese Platform of Bioimaging; PPBI-POCI-01-0145-FEDER-022122. We thank Dr Prasad Phapale for all the assistance with the lipidomic assays, Dr Hugo Osório for the proteomic analysis, and Gonçalo Araújo for the help with neuronal assays.

Funding

L.G. was funded by a fellowship from Fundação para a Ciência e a Tecnologia - FCT (SFRH/BD/138968/2018). This work was funded by national funds through FCT, under the project UIDB/04293/2020.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.

References