Combination of acamprosate and baclofen (PXT864) as a potential new therapy for amyotrophic lateral sclerosis

Abstract There is currently no therapy impacting the course of amyotrophic lateral sclerosis (ALS). The only approved treatments are riluzole and edaravone, but their efficacy is modest and short‐lasting, highlighting the need for innovative therapies. We previously demonstrated the ability of PXT864, a combination of low doses of acamprosate and baclofen, to synergistically restore cellular and behavioral activity in Alzheimer's and Parkinson's disease models. The overlapping genetic, molecular, and cellular characteristics of these neurodegenerative diseases supported investigating the effectiveness of PXT864 in ALS. As neuromuscular junction (NMJ) alterations is a key feature of ALS, the effects of PXT864 in primary neuron‐muscle cocultures injured by glutamate were studied. PXT864 significantly and synergistically preserved NMJ and motoneuron integrity following glutamate excitotoxicity. PXT864 added to riluzole significantly improved such protection. PXT864 activity was then assessed in primary cultures of motoneurons derived from SOD1G93A rat embryos. These motoneurons presented severe maturation defects that were significantly improved by PXT864. In this model, glutamate application induced an accumulation of TDP‐43 protein in the cytoplasm, a hallmark that was completely prevented by PXT864. The anti‐TDP‐43 aggregation effect was also confirmed in a cell line expressing TDP‐43 fused to GFP. These results demonstrate the value of PXT864 as a promising therapeutic strategy for the treatment of ALS.


| INTRODUC TI ON
Amyotrophic lateral sclerosis (ALS) is a debilitating and fatal disease characterized by the progressive degeneration of upper and lower motoneurons leading to muscle weakness and paralysis. One hallmark of the disease in human patients is cytoplasmic accumulation and aggregation of the RNA-binding protein TDP-43 in the brain and spinal cord (Neumann et al., 2006), indicating an important role of RNA metabolism and transport dysregulation. Several genetic factors have been identified that drive or modify motoneuron degeneration in ALS (Cirulli et al., 2015;Taylor, Brown, & Cleveland, 2016).
The first mutation described for causing familial ALS affects the ubiquitously expressed cytoplasmic enzyme Cu-Zn superoxide dismutase (SOD1), which is involved in reactive oxygen species detoxification (Rosen et al., 1993). Overexpression of mutated SOD1 (SOD1 G93A ) in transgenic mice induced motoneuron degeneration (Gurney et al., 1994). This model has been extensively studied and has revealed pathological mechanisms involved in ALS. Among them, dysfunction and degeneration of neuromuscular junctions (NMJs) was shown to precede motoneuron death in SOD1 G93A transgenic mice and ALS patients (Dadon-Nachum, Melamed, & Offen, 2011;Rocha, Pousinha, & Correia, Sebastião, & Ribeiro, 2013), highlighting the crucial role of NMJ alterations in the disease.
Despite identification of such pathological mechanisms, there are still no neuroprotective therapies that significantly impact the course of the disease, and patients die on average within 3-5 years of disease onset. The only approved treatments for ALS are riluzole and edaravone, but their efficacy is modest and short-lasting, highlighting the urgent need for innovative therapies (Fang et al., 2018;Hogg, Halang, Woods, Coughlan, & Prehn, 2017;Lacomblez, Bensimon, Leigh, Guillet, & Meininger, 1996).
Given the multifactorial and complex molecular nature of ALS, poly-therapeutic interventions could be considered as the most promising approaches (Attarian et al., 2014;Chumakov et al., 2014).
We previously demonstrated the ability of a combination therapy (PXT864), consisting of low doses of acamprosate (ACP) and baclofen (BCL), to synergistically restore cellular and behavioral endpoints affected in Alzheimer's (AD) and Parkinson's disease (PD) models (Chumakov et al., 2015;Hajj et al., 2015). AD, PD, and ALS share common cellular and molecular features. Among them, glutamate-induced excitotoxicity is considered a major pathophysiological factor contributing to neuronal death. Altered excitatory neurotransmission has been reported in all forms of ALS, resulting in aberrant spontaneous discharge of lower motoneurons onto the muscle fibers they innervate (King, Woodhouse, Kirkcaldie, & Vickers, 2016).
We therefore hypothesized that regulation of glutamatergic and GABAergic imbalance may be beneficial in ALS. In light of this hypothesis and previous successful interventions in AD and PD with PXT864 (Chumakov et al., 2015;Hajj et al., 2015), we investigated the effectiveness of this poly-therapeutic intervention for ALS.
To this end, we assessed the activity of the combination (PXT864) of ACP and BCL in in vitro models relevant to ALS pathology. We took advantage of an in vitro NMJ model to investigate neuroprotective properties of PXT864 against glutamate excitotoxicity.
Furthermore, primary motoneuron cultures derived from SOD1 G93A rats were used to assess PXT864 action on neuronal protection and maturation in the presence and absence of glutamate insults. The activity of PXT864 on TDP-43 accumulation and aggregation was also studied in SOD1 G93A rat motoneurons and U2OS human cells line overexpressing TDP-43 respectively. The nonantagonistic activity between PXT864 and riluzole was also assessed in vitro in anticipation for future clinical trials where PXT864 could be added to the standard of care.

| Ethics
Animal care was conducted in accordance with Directive 2010/63/UE of the European Community Council. All experiments and protocols were authorized and approved by the Ministère de l'Enseignement supérieur, de la Recherche et de l'Innovation (MESRI), as well as by local animal welfare committees Neuronexperts and Neuro-Sys.

| Animals
For in vitro experiments, pregnant female Wistar (pregnancy time = E13 for nerve/muscle cells coculture and E15 for cortical neurons primary culture) and SOD1 G93A rats (pregnancy time = E14) were purchased from Janvier Labs (France) and Taconic bioscience (USA), respectively. Animals were euthanized using a deep anesthesia with

Significance
We describe a positive activity of a novel combination of two drugs-acamprosate and baclofen, namely PXT864on a set of endpoints in relevant models of amyotrophic lateral sclerosis (ALS). We show that PXT864 protects neuromuscular junctions and motoneurons in vitro. PXT864 prevents the accumulation of toxic proteins such as TDP-43, a major hallmark of the disease. When PXT864 is added to riluzole treatment, protection against glutamateinduced damages is even improved. We provided the rationale for further testing of PXT864 in animal models of ALS or direct translation in patients, and a proof of principle for the value of repurposing approved and clinically safe drugs, using reproducible preclinical systems of ALS.
CO 2 chamber and a cervical dislocation. To determine SOD1 G93A genotype of embryos, DNA was extracted using the SYBR Green Extract-N-Amp tissue PCR kit (Sigma-Aldrich) and a qPCR was performed using the SYBR Green Master Mix kit (Sigma-Aldrich) and human SOD1-specific primers SOD-i3f (GTGGCATCAGCCCTAATCCA) and SOD-E4r (CACCAGTGTGCGGCCAATGA).

| Cell cultures and treatment
Experiments were performed at Neuronexpert (France) and Neuro-Sys (France) (Details on data origins are available in Table 1).
On day 27 after the beginning of the culture, contracting cocultures were preincubated for 1 hr with either ACP (0.14 nM to 40 nM), BCL (32 nM to 2000 nM), or PXT864 (25 combinations) before glutamate intoxication (60 μM for 20 min), and for a further 48 hr ( Figure 1a). To check the contractile functionality of the NMJs and to avoid any confusion with spontaneous contractions, the cocultures were treated with acetyl choline (100 nM). Only cocultures that immediately after application, showed contractions were retained for the study and distributed into the treatment groups. The effects of riluzole on the same endpoints below were investigated. On day 23, cultures were preincubated for 96 hr with or without riluzole (5 µM) followed by incubation with PXT864 (0.14 nM ACP and 32 nM BCL), riluzole, or both 1 hr prior to glutamate intoxication and for a further 48 hr ( Figure 2a).
Each embryo was dispatched on numerating petri dish (35 mm of diameter). Tail of embryos was cut for genotyping (see 2.2 Animals section). Spinal cords were removed and pooled by genotype in ice-cold medium of Leibovitz (L15) and treated for 20 min at 37°C with a trypsin-EDTA solution (Dutscher, cat#P10-023100) at a final concentration of 0.05% trypsin and 0.02% EDTA. The dissociation was stopped by the addition of Dulbecco's modified Eagle's medium (DMEM) with 4.5 g/L of glucose (Dutscher, cat#P04-03600), containing DNase I grade II (final concentration 0.5 mg/ml) (Dutscher, cat#P60-37780100) and 10% fetal calf serum (FCS, Dutscher, cat#S15898S181B by three forced passages through the tip of a 10-ml pipette. Cells were then centrifuged at 180 x g for 10 min at +4°C on a layer of BSA (3.5%) in L15 medium. The supernatant was discarded, and the pellet was resuspended in a defined culture medium consisting of Neurobasal medium (Fisher scientific, cat#1894793) with a 2% solution of B27 supplement (Fisher scientific, cat#1899722), 2 mM of l-glutamine (Dutscher, cat#3220316), 2% of PS solution (Fisher scientific, cat#4650416), and 10 ng/ml of brain-derived neurotrophic factor (BDNF, Dutscher, cat#H151208). Viable cells were counted in a Neubauer cytometer, using the trypan blue exclusion test and were seeded at a density of 20 000 cells per well in 96-well plates precoated with poly-l-lysine and were cultured at 37°C in a humidified air (95%)/CO 2 (5%) atmosphere. Half of the medium was changed every other day.
Motoneurons were treated from days 1 to 9 with PXT864 (0.05 to 0.32 nM ACP and 12 to 80 nM BCL) alone or with riluzole (1 µM) to assess survival and maturity of SOD1 G93A motoneurons. To assess protective activity against glutamate intoxication, cells were exposed to PXT864 alone or with riluzole (1 µM) on day 13 for 1 hr before glutamate intoxication (5 µM for 20 min) and for a further 24 hr.
Briefly pregnant female rats of 15-day gestation were euthanized by cervical dislocation and the embryos were removed from the uterus.
The cortex was removed and placed in ice-cold medium of Leibovitz Invitrogen, cat#10270). Cells were then mechanically dissociated by three passages through a 10-ml pipette. Cells were centrifuged at 515 g for 10 min at +4°C. The supernatant was discarded and pellet of cells was resuspended in a defined culture medium consisting of Neurobasal (Gibco, cat#21103) supplemented with B27 (2%; Gibco, cat#17504), l-glutamine (2 mM; Pan Biotech, cat#P04-80100), 2% of PS solution, and 10 ng/ml of brain-derived neurotrophic factor (BDNF, Pan Biotech, cat#CB-1115002). Viable cells were counted in a Neubauer cytometer using the trypan blue exclusion test. The cells were seeded at a density of 30,000 cells/well in 96-well plates precoated with poly-l-lysine (Greiner, cat#655930) and were cultured at 37°C in a humidified air (95%)/CO 2 (5%) atmosphere. After 12 days of culture, drugs were solved in culture medium (+0.1% DMSO) and then preincubated with neurons for 1 hr before glutamate injury.
One hour after drug incubation, glutamate was added for 20 min, to a final concentration of 40 μM diluted in the presence of drugs. At the end of the incubation, medium was changed with medium with compounds and without glutamate. The culture was fixed 24 hr after glutamate injury.

| NMJ and neurite network analysis in neuromuscular cocultures
After treatment, NMJ were stained by incubating cells with

| Motoneuron neurite network, maturity, and TDP-43 analysis
Cells were fixed in a cold solution of 95% ethanol and 5% acetic acid. Neuron cell bodies and neurite networks were detected using mouse monoclonal anti-microtubule-associated protein 2 (MAP2) antibody ( Invitrogen, cat#A-11011, RRID:AB_143157) or anti-mouse Alexa F I G U R E 1 Combinations of acamprosate and baclofen act synergistically to protect neuron-muscle cocultures against glutamate toxicity. (a) Schematic of neuron-muscle coculture treatments. WT Wistar E13 embryos and human muscle cell line were used to generate neuron-muscle coculture. ACP and BCL showed dose-dependent protective properties against glutamate toxicity as measured by total NMJ area (b and e), total NMJ number (c and f), and neurite length (d and g). PXT864 combinations were tested for protective activity against glutamate, as measured by NMJ area (h), NMJ number (i), and neurite length (j). Significant increases in protection are represented by green circles, nonsignificant changes by grey circles, and significant decreases by purple circles. Circle diameter represents the effect size of the combination, and synergistic combinations are shown with a white S. Association of subactive doses of ACP (0.14 nM) and BCL (80 nM) had a synergistic activity significantly higher than maximal effects (E max ) of each drug alone on NMJ area and number (k, l), but equal for neurite length (m). Green arrows represent drug doses (ACP 0.14 nM and BCL 80 nM) combined in PXT864 (green bar). In control condition (without any treatment), value of 100 is equivalent to 2,092 µm 2 ± 89 for NMJ area or to 39 ± 1.59 NMJ number (per field). Data are presented as mean ± SD. *p < 0.05; **p < 0.01; and ***p < 0.

| Statistical analysis
All analyses were performed using Prism 6 and 8. synergic combinations, only combinations exhibiting an effect significantly higher than the vehicle effect (Dunnett's test, α = 5%) were considered.
Neuron-muscle coculture data were derived from five independent experiments with six replicates per experiment. Primary motoneuron data were derived from one experiment with six to F I G U R E 3 PXT864 improves SOD1 G93A motoneurons survival and maturity. (a) Schematic of SOD1 G93A motoneuron treatments. WT and SOD1 G93A embryos (E14) were used to generate primary culture of motoneurons (only SOD1 genotype was cultured). PXT864 protected SOD1 G93A motoneurons as assessed by MAP2-positive neuron count (b) and MAP2-positive neurite length (d). PXT864 also improved SOD1 G93A motoneuron maturity, assessed by the length of neurofilament (200 kDa) positive neurites (f). PXT864 (ACP 0.14 nM and BCL 80 nM) and riluzole acted synergistically to improve neurite length as assessed by response additivity analysis (white S symbol) (e) but not motoneurons survival (c). PXT864 (ACP 0.14 nM and BCL 80 nM) and riluzole combination had a higher protective activity on SOD1 G93A motoneuron maturity than each compound alone (g). In control condition (SOD1 without any treatment), value of 100 is equivalent to 48 ± 0.6 (b-d-f) and 47 ± 0.8 (c-e-g) motoneurons (per field  (Figures 1 and 5). Cortical neuron primary cultures were repeated in three independent cultures with six replicates per condition.

| Acamprosate and Baclofen act synergistically to protect neuron-muscle cocultures against glutamate
Defective synaptic transmission at NMJs and glutamate-induced excitotoxicity are common features of ALS (Battaglia & Bruno, 2018;King et al., 2016;Rocha et al., 2013). Therefore, protective potential of

| PXT864 improves SOD1 G93A motoneuron survival and maturity
We then sought to assess PXT864 protective properties in another model of ALS, primary cultures of motoneurons derived from SOD1 G93A transgenic rat embryos. When ALS motoneurons were exposed to PXT864, riluzole, or both (Figure 3a

| PXT864 protects SOD1 G93A motoneurons from glutamate-induced excitotoxicity
As glutamate-induced excitotoxicity contributes to ALS pathogenesis, SOD1 G93A motoneurons were challenged by glutamate ( Figure 4a). PXT864 protected SOD1 G93A motoneurons from glutamate toxicity in a dose-dependent manner and had a significant greater effect on motoneuron survival than riluzole (Figure 4b).
PXT864 was also significantly effective on the protection of neurite length (Figure 4c).
TDP-43 leakage from the nucleus and accumulation in cytoplasmic granules is a hallmark of ALS (Neumann et al., 2006). Glutamate stress-induced TDP-43 cytoplasmic accumulation in motoneurons was counteracted by PXT864 dose dependently with a significant greater activity than riluzole, reaching non-intoxicated vehicle levels ( Figure 4d,e).

| PXT864 prevents TDP-43 stress granule formation
Under various cellular stresses, such as oxidative stress, cytosolic structures composed of assembled riboproteins such as TDP-43 are formed to stop protein translation. To assess PXT864 effect on stress granule formation, U2OS cells overexpressing TDP-43 fused to turbo-GFP were treated with ACP, BCL, or PXT864 before stress granule formation upon induction of oxidative stress by arsenate (Figure 5a). ACP and BCL prevented TDP-43 stress granule formation in a U-shaped dose-dependent manner (Figure 5b,c). Then, ACP (0.14 nM to 200 nM) and BCL (32 nM to 1,000 nM) were mixed in 36 PXT864 combinations. Among these, 63% were significantly active and 30% acted synergistically on TDP-43 stress granule formation, as assessed by the Bliss independence model (Figure 5d). The greatest effect was achieved by PXT864 containing 200 nM ACP and 1,000 nM BCL with an inhibitory activity of 76% that was superior to the maximal effects of ACP or BCL when tested individually (Figure 5e,f).
In this model, ACP and BCL proved again their importance to be combined in PXT864 for the achievement of high effectiveness at low doses.

| PXT864 protects cortical neurons against glutamate induced excitotoxicity
As ALS disease affects both upper and lower motoneurons, we also assessed PXT864 activity on primary cultures of cortical neurons ( Figure 6). PXT864 (ACP 0.32 nM + BCL 80 nM) significantly protected cortical neurons from glutamate injury.

| D ISCUSS I ON
Overall PXT864, a combination of ACP and BCL, protected NMJ, motoneurons, and cortical neurons against glutamate-induced excitotoxicity. Interestingly, PXT864 improved riluzole's protection and reduced TDP-43 stress granule formation.
Although more than 50 drugs targeting various pathophysiological mechanisms of ALS have been studied, only two compounds have come to the market: riluzole and edaravone. Riluzole is a relatively safe drug and is speculated to reduce glutamatergic transmission (Cheah et al., 2010). However, its efficacy has turned to be poor and very short-lasting (Lacomblez et al., 1996). Edaravone is a free radical scavenger mediating elimination of lipid peroxides and hydroxyl radicals by an unknown mechanism. The efficacy of this drug is very limited as well, and was only assessed in a highly selected subset of ALS patients (Group. WGE (MCI-186) A 19 S, 2017). Thus, there is still an unmet medical need for alternative treatments for ALS (Fang et al., 2018).
Excitotoxicity is a complex mechanism involving the dysregulation of extracellular glutamate concentration and overactivation of NMDA receptors resulting in calcium overload and neuronal death (King et al., 2016). Increased glutamate release along with reduced glutamate uptake by astrocytes were widely described as Interneurons regulate motoneuron activity by releasing inhibitory neurotransmitters such as GABA and glycine in the spinal cord (Jonas, Bischofberger, & Sandkühler, 1998). Cortical and spinal cord interneurons are both affected in ALS (Ince et al., 1993;Stephens et al., 2006). GABA inhibition is described to be reduced in ALS patient's cortex, as shown by TMS measurement, and correlates with disease duration and severity (Zanette et al., 2002a(Zanette et al., , 2002b. In ALS patients, spinal cord glycine concentration is also reduced (Malessa, Leigh, Bertel, Sluga, & Hornykiewicz, 1991) and its receptor (GLRA1) expression is decreased in a murine ALS model (Chang & Martin, 2011). Overall, these elements point to the imbalance between excitatory and inhibitory transmission (Kiernan, Ziemann, & Eisen, 2019) contributing in part to motoneuron death. In this context, we adopted a multitargeting strategy combining ACP and BCL (PXT864), two well-known drugs with favorable safety profiles. ACP is currently used for relapse prevention in alcoholism and is thought to attenuate the hyper-glutamatergic state observed in early abstinence (Dahchour et al., 1998). Its precise mechanism of action is still unknown, but it is thought to modulate glutamate receptors and inhibitory glycine-gated ion channels. BCL is a GABA B agonist currently used to treat spasticity in ALS (Yoon et al., 2017 Promoting both motoneuron axonal and NMJ integrity has been shown to delay muscle atrophy and improve muscle performance in SOD1 G93A mice (Seijffers et al., 2014). Communication between motoneurons and muscles is bidirectional and axonal "dying back" degeneration is well described in human ALS and murine models (Dadon-Nachum et al., 2011;Fischer et al., 2004;Rocha et al., 2013).
Maintaining axonal connectivity with the muscle and promoting mo- to improve lower motoneuron activity. We verified in part this hypothesis by testing the effect of PXT864 on cortical neurons and found the protective effect of the combination against glutamate-induced excitotoxicity ( Figure 6). Similar observations were also previously made on the protective ability of PXT864 from oxidative stress (Chumakov et al., 2015).
Overall, in vitro results demonstrated the positive activity of PXT864 in ALS. The combination acted synergistically in vitro whether ACP and BCL were combined each at low inactive doses, or at higher doses close to their maximal effect. These results pointed also to a mechanism of action potentially involving the regulation of protein misfolding and clearance. PXT864 prevented TDP-43′s cytoplasmic accumulation in motoneurons and stress granule formation in vitro.
The propensity of mutant proteins to misfold and form aggregates is a common factor defining neurodegenerative diseases. Strategies aiming in part to rescue abnormal protein clearance could afford protection against motoneuron degeneration and extend survival in animal models of ALS (Medinas, Valenzuela, & Hetz, 2018;Saxena, Cabuy, & Caroni, 2009 (Caldeira et al., 2013;Djakovic et al., 2009) leading to an improvement in disease conditions. Moreover, direct involvement of target proteins modulated by PXT864 in regulation of protein folding/clearance is also not excluded. For example, GABA B receptors have been shown to bind directly to CHOP and ATF4 proteins, two key transcription factors implicated in control of ER stress (Lim & Yue, 2015;Yap et al., 2016); as well, recent publications demonstrated that GABA B receptors are able to suppress cyto-destructive autophagy and downregulate the activity of PERK, ATF4, and CHOP proteins in degenerating neurons under hypoxic conditions (Fu, Wu, Hu, Li, & Gao, 2016;Kim et al., 2018;Liu et al., 2015;Nehring et al., 2000;Sauter et al., 2005). Further investigations of PXT864 effects on neuronal excitation/inhibition balance and on cellular pathways involved in protein clearance, as well as on the regulation of SOD1 and TDP-43 expression, will allow a deeper understanding of its mechanisms of action.

| CONCLUSIONS
In the present study, we brought evidence on the ability of PXT864 to protect NMJs and motoneurons from glutamate excitotoxicity.
Such effects could be partly explained by the capacity of PXT864 to prevent TDP-43 accumulation and stress granule formation. Last, PXT864 interacted positively with riluzole paving the way toward its use in addition on top of the standard of care in patients suffering from this life-threatening disease. The converging evidence from different cellular models relevant for the disease strengthen the therapeutic potential of such a safe combination to be tested in animal models of ALS or directly in patients due to the lack of robust animal models of this devastating disease and the urgent need of new therapeutic options.
These results highlight the importance of combining ACP and BCL to achieve a positive efficacy and demonstrate the potential value of combining polytropic-acting drugs as a promising new strategy to treat ALS.

DECL AR ATION OF TR ANS PAREN C Y
The authors, reviewers and editors affirm that in accordance to the policies set by the Journal of Neuroscience Research, this manuscript presents an accurate and transparent account of the study being reported and that all critical details describing the methods and results are present.

E THI C S APPROVAL
Our manuscript reports animal experimentation that was approved by local ethics committees.

ACK N OWLED G M ENTS
We thank Dr Jonathan Robertson (Scinopsis) for reviewing the English of the manuscript. We thank Dr Nathalie Cholet for her advices on study protocol.

CO N FLI C T O F I NTE R E S T
All the authors except Noëlle Callizot are employee of Pharnext.
RH, SN, and DC are cited in patents held by Pharnext. The authors declare this study was founded by Pharnext. The funder had the following involvement with the study: study design, data collection and analysis, decision to publish and preparation of the manuscript.

AUTH O R CO NTR I B UTI O N S
All mentionned authors contributed to this work. Conceptualization:

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/jnr.24714.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data supporting these findings are available from the corresponding author upon request.