Introduction
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder caused by the selective loss of motor neurons in the motor cortex, brainstem and spinal cord, leading to the paralysis of voluntary muscles [
51]. Motor symptoms usually appear in midlife and ultimately escalate to death, usually within 2 to 5 years after symptom onset [
51]. Approximately 90% of ALS cases are sporadic, while the remaining 10% are inherited [
51]. The most prevalent genetic causes of ALS are mutations in the genes encoding superoxide dismutase 1 (SOD1), TAR DNA-binding protein 43 (TDP-43) and fused in sarcoma (FUS), as well as hexanucleotide repeats in the ‘
Chromosome 9 open reading frame 72’ (
C9orf72) gene [
2]. Despite the identification of multiple genetic factors, the pathogenic mechanisms underlying the selective degeneration of motor neurons remain incompletely understood. At present, effective therapies are lacking for ALS patients [
1,
9,
59]. As a consequence, there is a strong need for more effective therapies for ALS.
Mouse models manifesting the core symptoms of the disease, including a motor phenotype and motor neuron degeneration, remain crucial for the preclinical investigation of pathological mechanisms and novel therapeutic options. As all therapeutic strategies effective in the commonly used mutant SOD1 model subsequently failed in the clinic, other in vivo models to study ALS are urgently needed. One example is the PrP-hFUS-WT3 mouse model, which shows a strong ALS-like phenotype with severe motor neuron degeneration and a short lifespan, unlike many other FUS models [
40,
62]. FUS is a DNA/RNA binding protein linked to ALS as well as to frontotemporal dementia (FTD), another neurodegenerative disorder belonging to the same disease spectrum as ALS [
21,
69]. Mutations in the
FUS gene account for approximately 4% of familial ALS cases and lead to neuronal and glial cytoplasmic mislocalization of the FUS protein [
32,
70]. The majority of the identified mutations are found in the glycine-rich region and in the nuclear localization signal [
21]. However, mutations in the 3′ untranslated region causing a strong increase in FUS expression have also been identified in ALS patients [
19,
57]. Moreover, FUS aggregates are found in the absence of mutations in the
FUS gene in a subset of FTD cases [
16,
42]. These findings strongly indicate that not only mutant, but also wild-type FUS can be detrimental to neurons. However, it remains unclear which pathogenic mechanisms contribute to the observed neurodegeneration [
40].
Recently, aberrant homeostasis of epigenetic marks was detected in the nervous system of several models of neurodegenerative disorders, including ALS [
31,
34,
53,
58,
68]. One of these deregulated marks is acetylation of lysine residues of histones. Acetylation of histones regulates high-order chromatin folding and thereby the accessibility of transcription factors to their target genes [
22]. Histone acetylation levels are regulated by the counteracting activities of two protein families –the histone acetyltransferases (HATs) and the histone deacetylases (HDACs) [
22]. Accumulating evidence indicates that decreased histone acetylation is a negative determinant for neuronal survival [
34,
53,
58,
68]. A major indication is that global histone hypoacetylation in the nervous system is a feature of multiple neurodegenerative models [
7,
31,
58]. This imbalance tremendously affects the global transcription profile [
53,
54,
58,
68].
HDACs are druggable targets and compounds that inhibit this class of enzymes are used to treat several forms of cancer [
24]. The therapeutic potential of HDAC inhibitors was already studied in a wide range of neurodegenerative disorders, such as Alzheimer’s disease, Huntington’s disease, spinal muscular atrophy and ALS [
5,
6,
37,
52,
54,
56,
64,
67,
76].
Many HDAC inhibitors were developed showing distinctive selectivity toward the 11 zinc-dependent HDACs identified in mammals. Genetic and pharmacological studies suggested that especially the class I HDACs (HDAC1, 2, 3 and 8) as well as the class IIb HDAC6 have the largest potential as treatment strategies for neurodegenerative diseases, including ALS [
18,
34,
75]. Class I HDACs reside in the nucleus and their major targets are the histones. HDAC6 has distinct functions in the cytoplasm.
The therapeutic potential of pan-HDAC inhibitors has been investigated in SOD1 mice, showing modest effects. Principal determinants responsible for these limited outcomes could be the low HDAC inhibitory potency, poor blood-brain-barrier (BBB) permeability, lack of isoform selectivity or toxicity of the compounds used [
54,
56,
64,
76]. These could also explain the negative results of phase II clinical trials with valproate and phenylbutyrate [
13,
47]. Therefore, research on HDAC inhibition as a therapeutic strategy for ALS deserves further attention. Moreover, new HDAC inhibitors with improved characteristics were recently developed.
The aim of this study was to test the efficacy of the HDAC inhibitor ACY-738 in a transgenic
FUS mouse model of ALS and to investigate the contribution of different types of HDACs to the disease phenotype. ACY-738 displays a selectivity profile towards class I HDACs and HDAC6 [
8,
30,
41]. We used ACY-738 as it is unique amongst the HDAC inhibitors, having a high potency, tolerability, and capacity to penetrate the BBB [
30,
37,
41]. ACY-738 treatment restored histone acetylation and slowed down disease progression of the Tg
FUS+/+ mice. This treatment also corrected metabolic pathways in the spinal cord, which were already dysregulated at presymptomatic stage. Overall, our data suggest that global histone hypoacetylation is associated with metabolic dysregulation in FUS-mediated ALS pathology, and that histone deacetylases are potential therapeutic targets to delay disease progression by re-establishing metabolic homeostasis.
Material and methods
Animals
Tg FUS+/− breeding mice (stock no. 017916) were purchased from The Jackson Laboratory (Maine, USA). From the age of 28 to 30 days, crushed Teklad LM-485 (7912) sterilizable rodent chow (Envigo, Cambridgeshire, UK) containing 100 mg/kg ACY-738 (Regenacy Pharmaceuticals Inc., Waltham, USA) and mixed with water was provided ad libidum.
For genotyping, DNA was isolated from ear biopsies, and digested overnight in lysis buffer with 20 mg/ml proteinase K (Roche, Basel, Switzerland) at 55 °C. Genotyping of the mice was done using qPCR with a FAM-labelled probe recognizing the human FUS transgene (forward and reverse primers 5′-CAGCAAAGCTATGGACAGC-3′ and 5′-GTCTTGATTGCCATAACCGC-3′ and Taqman probe 5′-AGCAGAACCAGTACAACAGCAGCA-3′). β-actin was used as a housekeeping gene (forward and reverse primers 5′-CCCTACAGTGCTGTGGGTTT-3′ and 5′-GACATGCAAGGAGTGCAAGA-3′).
All mice were housed according to the guidelines of the KU Leuven. Mice were randomly selected for drug treatments and littermates were used as controls. Sample size and intermediate endpoints were based on the original study and a pilot experiment [
40].
ACY-738 bioavailability
The concentration of ACY-738 in plasma and brain samples was determined by Liquid Chromatography/tandem Mass Spectrometry (LC-MS/MS). Plasma and brain were collected from mice killed after 5 days of ACY-738 therapy. ACY-738 was extracted from the samples by protein precipitation using 50:50 acetonitrile:methanol and analyzed using LC/MS/MS. The concentrations were then calculated using a standard curve.
Electrophysiology
Nerve conduction measurements were performed as described previously [
14]. Briefly, mice were anaesthetized under a 2.2 l/min oxygen flow containing 3% isoflurane and placed on a heating pad (Physitemp Instruments Inc., Clifton, USA) to maintain the body temperature. Motor nerve conduction studies were performed using sub-dermal platinum coated 0.4 mm needle electrodes (Technomed Europe, Maastricht, the Netherlands) for stimulating and recording and a UltraPro S100 monitoring set-up (Natus Medical Incorporated, Pleasanton, USA). Compound muscle action potentials (CMAPs) were determined by measuring the electrical response of the gastrocnemius muscle after stimulation of the sciatic notch. The stimulation intensity was gradually increased until the supramaximal CMAP amplitude was obtained. The amplitude and latency of the signal were quantified. The average of 3 measurements was determined.
Behavioral testing
Muscle force of the mice was measured using a Grip Strength Meter (Columbus Instruments, Columbus, USA) with a triangular bar as a probe to measure the grip strength in the forepaws. The average of five trials per animal was determined. To assess both fore- and hindlimb muscular strength, the hanging wire test was performed. Mice were placed on a wire cage lid. The lid was inverted and the latency to fall was recorded to a maximum of 60 s. The average of three trials per animal was determined. We evaluated motor performances and weight two times a week starting at 28 days of age. Disease onset was considered as the time point at which the mouse showed the highest value for forelimb grip strength. Survival was determined by the loss of righting reflex within 5 s after laying the mouse on its side. This time point was considered as disease end-stage, whereupon the mouse was sacrificed by intraperitonial injection with 200 mg/kg sodium pentobarbital (Dolethal, Vetoquinol) and used for histopathology.
Histopathology
Mice were anaesthetized by intraperitonial injection with 200 mg/kg sodium pentobarbital (Dolethal, Vetoquinol) and transcardially perfused with PBS.
For spinal cord histology, the lumbar part of the spinal cord was harvested and post-fixed with 4% paraformaldehyde (PFA) overnight (4 °C) and dehydrated for 48 h in 30% sucrose (4 °C). To visualize the motor neurons in the ventral horn, cryosections (20 μm) were stained with a 2X thionin solution. Briefly, slides were washed twice with water for 15 s, stained with 2X thionin for 45 s, again washed twice with water and dehydrated in a three-step ethanol series of increasing concentrations (70, 90 and 100%) for 30 s each. Sections were cleared with Histoclear solution for 1 min and mounted with PerTex mounting medium (Histolab, Göteborg, Germany). Images were acquired with a Zeiss Imager M1 microscope (Carl Zeiss, Oberkochen, Germany) using a 10x objective. The area of neuron cell bodies in the ventral horn were analyzed using the AxioVision program and neurons larger than 400 μm2 were considered as motor neurons. To investigate FUS localization in the spinal cord, colocalization of FUS with neurons was assessed. Sections (20 μm) were blocked for 1 h with 10% Normal Donkey Serum (Sigma, St. Louis, USA) in PBS-T (0.1%) and incubated overnight (4 °C) with antibodies for FUS (Proteintech 11,570-A-1P, 1:100) and NeuN (Millipore, MAB377; 1:500). The sections were subsequently washed and incubated with appropriate secondary antibodies (ThermoFisher Scientific, 1:1000). Slides were mounted with DAPI-containing ProLong® Gold antifade reagent (Invitrogen - Life Technologies, Carlsbad, USA). To investigate astrocytosis and microgliosis, a similar protocol using antibodies against GFAP (Sigma, G3893; 1:200), CD11b (Serotex MCA74G; 1:200) was used. Images were acquired using the Leica SP8x confocal microscope.
For muscle histology, the legs were tied with a strip during perfusion. The gastrocnemius muscles were dissected and snap-frozen in isopentane cooled by immersion in liquid nitrogen. To visualize neuromuscular junctions, longitudinal cryosections (20 μm) were fixed with 4% PFA for 10 min and then washed with PBS and PBS-T (0.1% Triton X-100). Subsequently, sections were blocked for 1 h with 10% Normal Donkey Serum (Sigma) in PBS-T. To visualize the nerve axons, the sections were incubated overnight (4 °C) with Neurofilament-L conjugated to Alexa-488 antibody (Cell Signaling Technologies; 1:500), Synaptophysin antibody (Cell Signaling Technologies; 1:500) and Synaptic vesicle 2 (DSHB; 1:100). The sections were then washed and incubated with appropriate secondary antibodies (ThermoFisher Scientific, 1:1000) and α-bungarotoxin conjugated to Alexa-555 (Invitrogen; 1:1000) for 1 h at room temperature (RT). Slides were mounted with ProLong® Gold antifade reagent (Invitrogen - Life Technologies). At least 50 neuromuscular junctions were analyzed for innervation, as determined by the co-localization of the Neurofilament-L/Synaptophysin/Synaptic vesicle 2 and α-bungarotoxin labeling, on a Zeiss Axio Imager M1 microscope (Carl Zeiss) using a 10x objective. To visualize muscle fibers, transversal cryosections (20 μm) were blocked for 1 h with 1% BSA (Serva, Heidelberg, Germany) in PBS, after which they were incubated with WGA conjugated to Alexa-488 antibody (1/100 Invitrogen W11261) in 0.5% BSA in PBS for 1 h at room temperature. After washing three times with PBS, slides were mounted with ProLong® Gold antifade reagent (Invitrogen - Life Technologies).
Western blot analysis
For blotting of acetylated α-tubulin or FUS, tissue samples were lysed in RIPA buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% NP-40, 0.5% Na-deoxycholic acid, 0.5% SDS), supplemented with Complete EDTA-free protease inhibitor cocktail (Roche, Basel, Switzerland). Tissues were homogenized using Lysing Matrix D beads (MP Biomedicals, Solon, USA) and a MagNa Lyser oscillator (Roche) at 6500 rpm for 30 s thrice with 1 min interval on ice. The samples were subsequently centrifuged at 14,000 rpm for 20 min, after which the supernatant was transferred to a pre-chilled tube.
For blotting of acetylated histones and for the subcellular fractionation experiment, tissue samples were extracted using the NE-PER Nuclear and Cytoplasmic Extraction kit (Pierce Biotechnology, Rockford, USA) following the Manufacturer’s instructions. Briefly, tissues were cut in small pieces, washed with PBS and homogenized with a pellet pestle motor in CER I buffer. Samples were vortexed thoroughly for 15 s and left on ice for 10 min. CER II buffer was added, after which the mixture was vortexed and centrifuged at 16,000 g for 5 min. The nuclear pellet was washed with CERI/II buffer, after which it was resuspended in NER buffer. Samples were left on ice for 40 min, vortexing thoroughly for 15 s every 10 min. All samples were sonicated quickly, centrifuged at 16,000 g for 10 min and the supernatant (nuclear extract) was transferred to a pre-chilled tube. For histone blotting, only the nuclear extract was used.
Protein concentrations were measured with the micro BCA kit (Pierce Biotechnology). Reducing sample buffer (ThermoFisher Scientific) was added to samples containing equal amounts of protein (2.5 μg for α-tubulin; 30 μg for FUS; 15 μg for histones) and heated for 10 min at 95 °C before separation on a SDS–polyacrylamide electrophoresis gel (12%, 90 V, 50 min for α-tubulin; 3 h for FUS). Proteins were transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore, Massachusetts, USA) by a semi-dry transfer apparatus (Bio-Rad, Hercules, USA) (180 mA, 1 h 45 min). The membranes were blocked with 5% milk in TBS-T (10 mM Tris-HCl (pH 7.5), 150 mM NaCl and 1% Tween-20) for 1 h at RT. Immunoblots were then incubated with primary antibodies against FUS (rabbit, 1/1000, Bethyl A300–302), acetylated α-tubulin (mouse, 1/5000, Sigma T6793), calnexin (rabbit, 1/2000, Enzo Life Technologies ADI-SPA-860-F), acetyl histone 3 K9/14 (rabbit, 1/500, Cell Signaling 9677), histone 4 (rabbit, 1/500, Abcam ab10158) in TBS-T for 1 to 2 h at RT. The membranes were incubated with the appropriate secondary antibody conjugated with horseradish peroxidase (HRP) (1/5000; Agilent Technologies (Dako)) for 1 h at RT. Protein bands were visualized using enhanced chemiluminescence (ECL substrate, ThermoFisher Scientific) and an ImageQuant LAS 4000 Biomolecular Imager (GE Healthcare, Illinois, USA). For histone blotting, a mild reblotting buffer (Millipore) was applied to strip the blots. Band intensities were quantified using ImageQuant™ TL version 7.0 software (GE Healthcare) and normalized to the appropriate control (calnexin or histone 4).
In situ nuclear HDAC activity
For the determination of in situ nuclear HDAC activity, nuclei of tissue samples were extracted using the Epiquick Nuclear Extraction Kit (Epigentek, Farmingdale, USA), after which HDAC activity was determined using the Epigenase HDAC Activity/Inhibition Direct Assay Kit (Colorimetric) (Epigentek). In brief, tissues were homogenized in NE1 buffer using a motor-driven pestle. After 15 min incubation on ice, the samples were centrifuged for 10 min at 12,000 rpm, after which the supernatant was removed. Two volumes of NE2 containing PIC were added to the pellet, after which the samples were incubated on ice for 15 min. Samples were vortexed, sonicated and centrifuged for 10 min at 14,000 rpm. The supernatant was collected and the protein concentration was measured with the micro BCA kit (Pierce Biotechnology). The HDAC activity in the nuclear extracts was then determined using the HDAC activity assay kit, by adding equal amounts of protein of the nuclear extracts in the wells of a microplate, together with HO1 and HO2 in a final volume of 50 μl. The plate was then incubated at 37 °C for 90 min, after which the reactions were removed from the wells. After thorough washing, primary antibodies were added to the wells for 60 min at RT, after which the secondary antibodies were incubated for 30 min at RT. After thorough washing, DS buffer was added for 5 min, after which SS buffer was added to stop the enzymatic reaction. Signals were detected by absorbance at 450 nm. The amount of deacetylated histone product was extrapolated from a standard curve, after which the HDAC activity in OD/min/mg was calculated.
Transcriptional analysis
Mice were euthanized with CO
2 followed by cervical dislocation. Spinal cord was rapidly dissected and snap-frozen in liquid nitrogen. RNA was extracted using a combined protocol of TRIzol:chloroform and the RNeasy mini kit (Qiagen) and further processed according to the manufacturer’s protocol. For RNA sequencing, samples were sent to the Nucleomics Core (VIB, Leuven, Belgium) and analyzed for RNA integrity (≥ 8) by running on a Bioanalyzer (Agilent) before Trueseq total stranded RNA library preparation and sequencing on a Illumina NextSeq 500 system (Illumina, San Diego, USA). The libraries were sequenced using a high output paired end with 75 bp reads and ~ 50 million reads per sample. Reads were mapped using STAR aligner [
20] on the latest mouse genome build (mm10). Reads were then counted using Salmon to estimate transcript and gene expression of every sample [
80]. Differential expression of coding genes and transcripts was performed with edgeR [
38]. Gene and transcripts with a FDR-adjusted
P-value smaller than 0.05 and with an altered expression of 30% were deemed significantly differentially expressed. Differentially expressed genes were used as input for Panther gene ontology enrichment analysis to identify pathways associated with the differentially expressed genes [
3,
39,
66]. Multiple testing correction was performed using FDR Benjamini-Hochberg correction. The most significant and promising genes were validated with qRT-PCR. For qRT-PCR analysis, first-strand cDNA was synthesized using SuperScript III (Invitrogen). PCR reactions were performed using SybrGreen reagents (ThermoFisher Scientific) with primers specific for the genes-of-interest (Additional file
1: Table S1) [
63]. Expression levels were normalized to two reference genes: AP3d1 (forward, 5′-CAAGGGCAGTATCGACCGC-3′; reverse, 5′-GATCTCGTCAATGCACTGGGA-3′) and Mon2 (forward, 5′-CTACAGTCCGACAGGTCGTGA-3′; reverse, 5′-CGGCACTGGAGGTTCTATATCTC-3′). Analysis was performed using qBase+ (v.3.0, Biogazelle, Zwijnaarde, Belgium).
Proteomics analysis
Mice were anaesthetized by intraperitonial injection with 200 mg/kg sodium pentobarbital (Dolethal, Vetoquinol, UK) and transcardially perfused with PBS (Sigma). Spinal cord was rapidly dissected and snap-frozen in liquid nitrogen. Samples were sent to the Proteomics Core Facility (VIB, Ghent, Belgium) for Mass Spectrometry analysis. Liquid Chromatography - tandem Mass Spectrometry (LC-MS/MS) runs of all samples were searched together using the MaxQuant algorithm (version 1.6.1.0) with mainly default search settings including a false discovery rate set at 1% on both the peptide and protein level. Spectra were searched against the human protein sequences in the Swiss-Prot database (database release version of January 2018 containing 20,234 human protein sequences) (
http://www.uniprot.org). Proteins that were identified in all replicates of at least one condition were kept. For statistical analysis, the missing values were imputed using a mixed approach. Missing not at random (MNAR) values were imputed using quantile regression-based left-censored function (MLE-QRILC), while missing at random values (MAR) were imputed using maximum-likelihood estimation (MLE). MNAR was assumed when the protein was not identified in all replicates of at least one condition. To reveal proteins of which the expression level was significantly regulated between the different conditions, the test_diff function of the DEP R-package was performed to compare intensities of the proteins between the three different conditions (non-Tg, Tg
FUS+/+ vehicle, Tg
FUS+/+ ACY-738) [
78].
Mice were euthanized with CO2 followed by cervical dislocation. Spinal cord was rapidly dissected and snap-frozen in liquid nitrogen. Samples were sent to the Metabolomics core (VIB, Leuven, Belgium) to perform targeted metabolomic profiling by Liquid Chromatography - Mass Spectrometry (LC-MS). Separation of metabolites prior to MS measurement was performed using a Dionex UltiMate 3000 LC System (Thermo Scientific) in-line coupled to a Q Exactive Orbitrap mass spectrometer (Thermo Scientific). Practically, 10 μl of the extract was injected on a C18 column (Acquity UPLC®HSS T3 1.8 μm 2.1x100mm, Waters) using solvent A (H2O, 10 mM Tributyl-Amine, 15 mM acetic acid) and solvent B (100% Methanol). Chromatographic separation was achieved with a flowrate of 0.250 ml/min and the following gradient elution profile: 0 min, 0%B; 2 min, 0%B; 7 min, 37%B; 14 min, 41%B; 26 min, 100%B; 30 min, 100%B; 31 min, 0%B; 40 min, 0%B. The column was thermostatted at 40 °C throughout the analysis. The MS operated in full scan negative ion mode (m/z range: 70–1050 Th) using a spray voltage of 4.2 kV, capillary temperature of 320 °C, sheath gas at 50.0, auxiliary gas at 15.0. The AGC target was set at 3e6 and resolution at 140,000, with a maximum IT fill time of 512 ms. Data processing was performed using the Xcalibur Quan software (Thermo Scientific).
Statistical analyses
Statistical analyses were performed using GraphPad Prism software version 8.0.0 (GraphPad software Inc). For Kaplan-Meier survival curves, the log-rank test was used to determine the statistical significance. Unpaired two-tailed Student’s t-test was used for the comparison of two means. If needed, the results were corrected for multiple testing using the Holm-Sidak method, with Q = 1. One-way and two-way ANOVA were used for multiple group analyses. Data were tested for equal variances using Bartlett’s test, Brown-Forsythe test, F-test. Kruskal-Wallis and Mann-Witney were used when no equal variances were obtained in the data sets. Data are presented as means ± SEM. Statistical significance was set at P < 0.05. *P < 0.05, **P < 0.01, ***P < 0.001 ****P < 0.0001.
Discussion
The pathological mechanisms underlying ALS remain incompletely understood, which could account for the numerous failed clinical trials in this patient population. We provide evidence for the contribution of aberrant histone acetylation in motor neuron pathology and discovered that HDAC inhibitors are an effective therapy in a preclinical model of ALS. Our data show that ACY-738, a potent HDAC inhibitor that crosses the BBB, restored global histone acetylation in the spinal cord, ameliorated the disease phenotype and significantly extended the lifespan of mice overexpressing wild-type FUS. We found that HDAC6 is not the key HDAC mediating this therapeutic effect. Our multi-omics approach revealed that dysregulated metabolism was an early phenomenon correlating with motor neuron pathology, which was largely restored by ACY-738 therapy.
Recent evidence points to a possible role of epigenetic mechanisms, including microRNAs, DNA methylation and histone modifications, in ALS pathology [
7,
44]. Amongst these, we concentrated on the possible contribution of histone acetylation, a post-translational modification that regulates transcription, in a rodent model of ALS. Aberrant activation of HDACs, resulting in histone hypoacetylation, has been associated with several neurodegenerative disorders and with neuronal toxicity [
7,
31,
58]. Pioneering work by Rouaux and collaborators reported evidence for a role of HDACs in ALS, demonstrating histone hypoacetylation in the SOD1 mouse model starting at the age of onset [
53]. In postmortem brain and spinal cord specimens of ALS patients, analysis of HDAC expression levels also revealed altered levels of HDAC2 and HDAC11 [
29]. While the various HDACs were classified and categorized, the role of particular HDACs in ALS pathogenesis has not been fully addressed yet. In more recent studies, it was shown that muscle HDAC4 plays a crucial role in muscle reinnervation in SOD1 mice and in patients with rapidly progressive ALS [
11,
48,
49,
74].
One important question is the molecular mechanism underlying the decrease in histone acetylation observed in Tg
FUS+/+ mice. ALS-causative genes have been associated with various epigenetic modifiers and epigenetic tags [
7,
12]. For example, the DNA/RNA binding protein FUS directly interacts with CBP and p300, two histone acetyltransferases, as well as with HDAC1 [
72,
73]. Therefore, the observed hyperactivity of nuclear HDACs and associated histone hypoacetylation in the Tg
FUS+/+ mice could arise from such an interaction. Further studies are required to confirm this.
The role of HDACs in neurodegenerative diseases has largely been deduced from effects observed after inhibition of their enzymatic activity. In the context of ALS, several studies have shown a modest protective effect of three pan-HDAC inhibitors, trichostatin A (TSA), valproic acid (VPA) and phenyl butyrate (PB) in the SOD1 model [
15,
46,
56,
76]. TSA is a very potent HDAC inhibitor, yet it can be used only in laboratory experiments due to its genotoxic effects [
43]. VPA is used as an antiepileptic drug, but has a low HDAC inhibitory potency, poor BBB permeability and cumbersome side effects. PB is currently used to treat hyperammonemia due to urea cycle disorders and is also known to have a low inhibitory potency. It may therefore not be surprising that phase II clinical trials in ALS patients showed negative results [
13,
47]. In our study, we used ACY-738 to explore the therapeutic potential of HDAC inhibition in a preclinical
FUS model of ALS. ACY-738 is a highly BBB permeable, potent HDAC inhibitor that primarily targets class I HDACs and HDAC6 [
8,
30,
41]. Our data show that the inhibition of the class I HDACs, which corrects global histone acetylation and partially restores transcription, is responsible for the positive effect in the Tg
FUS+/+ model. ACY-738 is an attractive drug candidate, as we could show robust class I HDAC inhibition in vivo with no obvious adverse effects on the treated animals.
The finding that genetic removal of
Hdac6 did not have a beneficial effect in the Tg
FUS+/+ mice is in contrast with our results observed in the SOD1 model, in which HDAC6 removal modestly extended their survival [
65]. This could be due to differences in the pathogenic mechanisms underlying both genetic forms of ALS. Another indication of this diversity is our previous observation that pharmacological inhibition as well as genetic silencing of HDAC6 in a different FUS-related model had positive effects [
27]. Motor neurons differentiated from induced pluripotent stem cells (iPSCs) from
FUS-ALS patients developed axonal transport defects over time and inhibition of HDAC6 using tubastatin A or ACY-738 completely rescued these defects [
27]. In our current study, we excluded that HDAC6 inhibition by ACY-738 was responsible for the positive effect on the survival of the Tg
FUS+/+ mice. One should keep in mind that these different effects of ACY-738 were obtained in two completely different models that both mimic relevant aspects of the ALS disease process. Whatever the exact contribution in ALS patients is of axonal transport defects due to decreased acetylation of α-tubulin or of aberrant transcription caused by hypoacetylation of histones, ACY-738 has the major advantage that it corrects both processes.
Furthermore, we noticed that ACY-738 therapy can significantly decrease cytoplasmic human FUS levels in the remaining motor neurons. It is possible that ACY-738 acts on this process by slowing down several key aspects of disease progression due to its effects on transcription. However, the partial restoration of FUS localization might also be independent from changes in histone acetylation or transcription. The synergistic correction of each of these different processes could contribute to the therapeutic activity of ACY-738.
Using an integrative transcriptomic, proteomic and metabolomic approach, we ultimately discovered that dysregulation of lipid metabolism was an early phenomenon correlating with progressive motor neuron pathology. These results are in line with a recent study that showed drastic morphological alterations in mitochondria at the synaptic terminals of pre-symptomatic Tg
FUS+/+ mice [
62]. Furthermore, expression changes of genes related to (lipid) metabolism were noted in another wild-type FUS overexpression mouse model, as well as in a mutant FUS overexpression and FUS-ΔNLS-knock-in mouse model, suggesting a specific pathogenic mechanism contributing to FUS-mediated ALS [
17,
50,
61]. It is possible that these other FUS models also present with histone acetylation changes that dysregulate the expression of metabolic genes. Strikingly, our multi-omics analysis revealed a substantial rescue of the observed metabolic disturbances by ACY-738 treatment. Although neurons and glial cells express many molecular components of lipid metabolic pathways, the importance of local regulation of lipid metabolism in the CNS has remained neglected for a long time. Only recently, several studies highlight central lipid metabolism homeostasis as a critical factor for neuronal function and survival [
33,
35,
36]. Cholesterol is locally synthesized in the adult CNS, with little or no import from the periphery [
77]. It is synthesized mostly by astrocytes, after which it is transported by apolipoprotein E (APOE) and taken up by neurons via the low-density lipoprotein receptor. Fatty acids (FA) on the other hand are mainly transported into the CNS from the systemic circulation, although some FAs can be synthesized de novo by neurons and astrocytes [
10,
23,
35]
. Disturbances in either the synthesis, transport, or turnover of lipids were shown to trigger synapse loss and neurodegeneration [
10,
23,
25,
35,
36]. The observed disturbances in lipid metabolic pathways in the Tg
FUS+/+ mice could affect membrane biogenesis-dependent processes such as synaptogenesis, synapse maintenance, neurotransmitter release and mitochondrial function [
26,
33,
36]. As these processes are all determinants of proper neuromuscular functionality, an early disturbance in central lipid homeostasis could contribute to ALS pathogenesis [
4,
60]. Accordingly, the rescue of metabolic homeostasis by ACY-738 therapy could account for the preservation of neuromuscular junctions, as appreciated by an increased number of innervated junctions and increased compound muscle action potentials.
It will be important to validate our findings on disturbed histone acetylation and metabolism in other ALS models. In fact, none of the ALS rodent models, including the Tg FUS+/+ mouse model used in our study, fully recapitulates human disease. Therefore, it is reasonable that only the combined knowledge obtained from each of these models will enable us to understand the deleterious processes underlying the selective motor neuron death in ALS.
Overall, our hypothesis illustrated in Fig.
8d is that ALS-associated proteins such as FUS can influence the epigenetic code. These epigenetic alterations, in combination with cytoplasmic accumulation of ALS-associated proteins, will have tremendous implications on the global transcription profile. This will affect different processes, such as metabolic pathways, that are essential for motor neurons, ultimately leading to degeneration. We propose that inhibition of class I HDACs by ACY-738 can partially overcome these transcriptional defects by restoring histone acetylation and FUS mislocalization, thereby ultimately improving the ALS phenotype. It will be important to define which HDAC family member is most crucial for the observed effect, so that safe, selective and potent HDAC inhibitors can be developed.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.