Introduction

SCA3 (spinocerebellar ataxia type 3), also known as Machado-Joseph disease (MJD), is considered to be the most common form of SCA worldwide [1]. SCAs are considered to be rare disorders, with a prevalence ranging from 0.3 to 2.0 per 100,000 [2]. MJD, a polyglutamine expansion disease, is a hereditary autosomal dominant neurodegenerative disease that develops when there is an abnormal expansion of the trinucleotide cytosine-adenine-guanine (CAG) repeats in the causative gene [3,4,5]. MJD is characterized mainly by ataxia, spasticity, peripheral muscular atrophy, and/or other motor-related clinical manifestations along with neurodegeneration in selective regions including the cerebellum, basal ganglia, brainstem, and spinal cord [5,6,7,8]. MJD is more highly observed among people of Portuguese/Azorean with the highest prevalence in the Azorean island of Flores (1/239) [9]. This disease is caused by an unstable expansion of CAG trinucleotide within the coding region of ATXN3 gene, located on chromosome 14 (14q24.3-14q32.45) [3, 10,11,12]. The CAG repeat number in healthy individuals is between 10 and 51, whereas in SCA3 patients, this repeat is expanded to 55–87 [13, 14]. The mutated ATXN3 gene is translated into an abnormal polyglutamine (poly-Q) tract within the ataxin-3 protein (ATXN3) [10, 15]. The abnormal expansion in the ATXN3 gene leads to misfolding of ATXN3 and is associated with numerous harmful impacts such as generation of toxic poly-Q protein species, protein aggregation, disrupted protein homeostasis, loss of deubiquitylase activity, and dysregulation of autophagy, among others [3, 16]. However, the exact mechanisms of disease pathogenesis, as well as molecular pathways, are still not well understood. This restraint knowledge prevents the development of possible clinical therapies to cure or even attenuate the phenotypes associated with this disease.

To investigate aspects of MJD, we turned to the simple animal model, Caenorhabditis elegans, a popular system for neuroscience investigations [17]. C. elegans is especially well-suited for neuroscience and genetic research due to the large number of conserved molecular pathways resembling that of vertebrates and its ease of genetic manipulation [18, 19]. Furthermore, C. elegans has been used to model several neurodegenerative diseases such as MJD, amyotrophic lateral sclerosis (ALS), Parkinson’s disease (PD), and Alzheimer’s disease (AD) [6, 20,21,22,23,24,25,26].

C. elegans is also a useful system for drug discovery and linking drug activity to specific genetic and molecular pathways required for action. It is a powerful tool for the therapeutic discovery of compounds that are able to restore the disease phenotype, such as motility, and also to identify specific molecular pathways [27, 28]. We have previously shown that our transgenic ATXN3 mutant worms display motility defects, along with progressive neurodegeneration and reduced longevity [20]. Based on these findings, the goal of this study was to identify small molecules with neuroprotective effects that could be translated along preclinical and clinical pathways for MJD. To achieve this goal, we conducted a blind drug screen of 3942 compounds and identified lead compounds that suppressed motor deficits of ATXN3 mutant worms, and rescued accompanying neurodegeneration and reduced longevity phenotypes. Among the lead compounds, chenodiol, fenbufen, and sulfaphenazole rescued the ATXN3 toxicity by acting as modulators of the transcription factor EB (TFEB/HLH-30), a transcription factor implicated in autophagy. These molecules are promising tools to improve neuronal health and survival in MJD models, warranting investigation in vertebrate systems.

Materials and Methods

Nematode Strains and Maintenance

Standard methods of culturing and handling worms were used [19]. Worms were maintained on standard nematode growth media (NGM) plates streaked with OP50 E. coli. All strains were scored at 20 °C. Strains used for this study were N2, MAH240 sqls17[hlh-30p::hlh-30::GFP + rol-6(su1006)] and hlh-30(tm1978), all obtained from the Caenorhabditis Genetics Center (University of Minnesota, Minneapolis).

Transgenic ATXN3 Worms and Plasmid Constructs

The development of our transgenic ATXN3 worms is explained in our previous paper [20]. The strain used in this study include: XQ351 unc-119(ed3); ttTi5605mosII; xqIs351(unc-47p::ATXN3-CAG89; unc-119(+)), referred to as ATXN3-CAG89. As previously shown, these transgenic mutants have motility defects, neurodegeneration, and reduced longevity.

Drug Libraries

The Sigma LOPAC library, composed of 1280 bioactive compounds; the MicroSource Discovery library, containing 880 compounds; the Biomol Natural Products library (Enzo Life Sciences, Inc.), composed of 500 compounds; and the Prestwick Chemical library, containing 1282 compounds, were selected for the worm-based motility screenings. All compounds were dissolved in DMSO (dimethyl sulfoxide) and tested at a concentration of 20 μM for 4 h.

Drug Screen

The drug screen was performed by using a WMicroTracker machine (Phylum Tech) [29]. Briefly, ~ 50–60 worms adult day one were exposed to the drugs at a concentration of 20 μM, or to DMSO as the control in a 96-well plate. The swimming movement was tracked for at least 4 h. Measurements were performed in triplicates and the average movement score was compared to the control and the average movement score of the whole plate. If the values were higher than the respective controls for a certain drug, then a secondary screen was performed to validate our observations. Candidates that increased significantly the swimming movement in the secondary screen were considered as positive hits.

Age-Synchronized Populations

To obtain age-synchronized worms, ~ 10 adult hermaphrodites were placed on ten NGM plates for 3–4 days and kept at 20 °C. Once the plates were full of adult worms, they were collected with M9 buffer (1 M KH2PO4, 1 M Na2HPO4, 1 M NaCl, 1 M MgSO4) and centrifuged at 4000 rpm (A-4-81 Rotor) for 2 min at 4 °C. After centrifugation, 3 mL of the supernatant was taken and replaced with 3 mL of a mix solution containing NaOH 5 M and bleach (1:2). A 5-min vortex followed with a high intensity to degrade the worms leaving a pellet containing only eggs. The pellet was washed 3 times with M9 buffer and centrifuged at 4000 rpm (A-4-81 Rotor) for 2 min at 4 °C. The pellet was transferred onto NGM plates without bacteria and kept overnight at 20 °C. The following day, L1 worms were transferred using M9 buffer onto plates streaked with OP50 E. coli and are kept at 20 °C.

Paralysis Assays on Solid Media

Worms were scored for paralysis from adult day one to adult day twelve. Thirty L4 worms were transferred to NGM plates. They were scored as paralyzed if they failed to move after being prodded with a worm pick. Worms were scored as dead if they were unable to respond to tactile head stimulus. All experiments were conducted at 20 °C, and each condition was done in triplicate with 30 worms/plate.

Lifespan Assays

Forty L4 worms were transferred to NGM plates and tested every 2 days from adult day one until death. They were scored as dead if they failed to respond to tactile stimulus and showed no spontaneous movement or response when prodded. Dead worms displaying internally hatched progeny, extruded gonads, or that crawled off the plate were excluded. All experiments were conducted at 20 °C, and each condition was done in triplicates with 40 worms/plate.

Fluorescence Microscopy (Neurodegeneration and Stress Reporter Fluorescence Quantification)

For scoring of neuronal processes for gaps or breakages, worms were collected at adult day five and day nine for visualization of motor neuron processes in vivo. For fluorescence quantification of the stress reporter, hlh-30::GFP, hlh-30::GFP; ATXN3-CAG10 and hlh-30::GFP; ATXN3-CAG89 worms were collected at adult day one, two, three, and five. The nematodes were immobilized with 60% glycerol and mounted on slides with 2% agarose pads. mCherry was visualized at 595 nm and GFP was visualized at 488 nm using a Zeiss Axio Imager M2 microscope. Fluorescent expression was visualized with a DIC microscope Zeiss AxioObserver A1. The software used was AxioVs40 4.8.2.0. One hundred to one hundred fifty worms were scored per condition for the neurodegeneration assays. Approximately 17 to 20 worms were visualized per condition for the fluorescence quantification experiments. Image processing and quantification were done with Adobe Photoshop. For the fluorescence quantification of the stress reporter, we calculated the changes in the ratio (size/intensity of fluorescence). One-way ANOVA test was used for statistical analysis.

Compound Treatment on Solid Media

Worms were exposed from L4 state to compounds at 20 μM or 2 μM (with 1% DMSO or 0.1% DMSO final concentration, respectively) incorporated into an NGM solid medium, or an NGM solid medium only as a control. All the plates were streaked with OP50 E. coli. The five lead compounds were purchased from Sigma-Aldrich (St. Louis, MO), Cayman Chemical Co. (Ann Arbor, MI), or Toronto Research Chemicals Inc. (North York, TO). Briefly, 30–40 worms were picked and plated on the corresponding NGM medium (30–40 worms/plate for each condition and each condition was done in triplicate) in order to complete the paralysis and lifespan assays, the neurodegeneration observations, and fluorescence quantification (fluorescence microscopy).

Cheminformatics

Each pair of compounds to be analyzed for similarity was evaluated for the presence or absence of any of thousands of possible FP2 fingerprints. The Tanimoto coefficient also known as the Jaccard index represents one of the most popular methods for quantifying molecular similarity. Tanimoto coefficient scores were calculated for each hit and compound present in the Prestwick library using OpenBabel (http://openbabel.org) [30, 31].

To enhance the effectiveness of similarity of molecules, Tanimoto scores higher than 0.50 were clustered using an unweighted Euclidean distance similarity metric with complete linkage clustering allowing visualization of the ranked molecules within the context of the dendrogram. Enrichment for the structural similarity of members in each cluster (Tanimoto scores were > 0.50 for majority of members in a cluster) was calculated using Fisher’s exact test (GraphPad Prism Software) [31].

The heat maps were generated using: https://software.broadinstitute.org/morpheus/. This tool uses R software (R Team: R Development Core Team (2005). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, www.R-project.org.).

Statistics

Paralysis and lifespan curves were generated and compared using the log-rank (Mantel-Cox) test. For neurodegeneration and the fluorescence assays, one-way ANOVA tests were used. All experiments were repeated at least three times. For the heat map statistics, Fisher’s exact tests were realized. Prism (GraphPad Software) was used for all statistical analyses.

Results

Chemical-Genetic Screen in ATXN3 Mutant C. elegans

Our previous work demonstrated that the motility defects observed in ATXN3-CAG89 transgenic worms could be rescued using specific small molecules [20]. This promising prescreen encouraged us to go further, so we performed a blind, liquid culture, drug screen of 3942 small molecules (at 20 μM) for the rescue of motility defects in ATXN3-CAG89 transgenic worms, and identified 24 lead molecules (Fig. 1, Supplementary Table S1, Supplementary Table S2).

Fig. 1
figure 1

Chemical-genetic screen in ATXN3-CAG89 mutant transgenics. By using the WMicroTracker machine, 3942 compounds with 20 μM concentration were screened in our MJD strain. These compounds were, from four companies: Microsource Discovery, Sigma Lopac, Prestwick, and Biomol. Measurements were performed in triplicates and the average movement score was compared to the control (ATXN3-CAG89 mutants in DMSO, no compound) and the average movement score of the whole plate. If the values were higher than the respective controls for a certain drug, then a secondary screen was performed to validate our observations. Candidates that increased significantly the swimming movement in the secondary screen were considered as positive hits. From this screen, we identified 24 positive compounds able to significantly correct the impaired movement phenotype of our MJD strain when compared to the controls. After paralysis, neurodegeneration, and lifespan assays, we identified five lead compounds able to rescue all these phenotypes. The lead compounds are: alfacalcidol (ALFA), chenodiol (CHEN), cyclophosphamide (CYCLO), fenbufen (FEN), and sulfaphenazole (SULFA)

Full size image

To validate these findings, the ATXN3-CAG89 mutant worms were then treated with the positive compounds using a solid media approach (Fig. 2A–C), and 21 of the initial 24 compounds maintained rescuing activity (clebopride, lamotrigine, and phorbol 12-myristate 13-acetate were rejected) (Fig. 2A–C). These differences could be due to the acute nature of the liquid drug screen where the animals are exposed to molecules over a period of 4 h, compared to drug exposure over 12 days using solid media assays.

Fig. 2
figure 2

Validation of the positive compounds on solid media assays. (a–c) The motor defect phenotype observed in ATXN3-CAG89 worms was significantly rescued when treated with (a) butaclamol (**P < 0.01), or naphazoline, diacerein, chlorpropamide, sulfaphenazole, etodolac, octoclothepin (****P < 0.0001 for all previous compounds); (b) isotretinoin (***P < 0.001), ibuprofen, alfacalcidol, cylclophosphamide, nomifensine, methiothepin, biotin (****P < 0.0001 for all mentioned compounds); (c) benserazide, digitonin, acetamide, cyclobenzaprine, chenodiol, fenbufen, and cycloheptadine (****P < 0.0001 for all previous compounds) (by log-rank (Mantel-Cox) test, N = 90–100 per trial, and N = 270–300 when all trials combined). No rescue of the locomotion defect is observed when the ATXN3-CAG89 worms were treated with phorbol 12-myristate 12-acetate, lamotrigine, or clebopride. The concentration of the compounds was tested at 20 μM as in the liquid culture. This experiment was done 3 times

Full size image

The compounds were then tested at a lower dose (2 μM) and we observed that 13 of the 21 compounds showed significant suppression of the locomotion deficit in ATXN3-CAG89 transgenics at this concentration (Fig. 3A, B). We further tested these 13 compounds at two additional concentrations (0.2 μM and 40 μM), and we conclude that the optimal concentration for these compounds to suppress the paralysis phenotype in our MJD model, among the ones that have been evaluated for this study, is set at 2 μM.

Fig. 3
figure 3

Rescue of motility deficit, neurodegeneration and reduce longevity when ATXN3-CAG89 worms are treated with positive compounds at 2 μM. (A–B) Mutant transgenic worms showed a rescue of locomotion impairment when treated with 2 μM (a) cyclophosphamide, methiotepin, diacerein, alfacalcidol, butaclamol, sulfaphenazole, chenodiol, fenbufen, or nomifensin (****P < 0.0001 for all the compounds), as well as (b) naphazoline (*P < 0.05), benserazide (**P < 0.01), acetamide or chlorpropamide (*P < 0.05 for both compounds) (by log-rank (Mantel-Cox) test, N = 90–100 per trial, and N = 270–300 when all trials combined). This experiment was replicated 3 times. (c–d) Shown are representative photos of living, adult expressing unc-47p::mCherry; ATXN3-CAG89 transgenic worms at day five and day nine of adulthood with or without drug. The concentration of drugs used for this experience is 2 μM. Images are black and white. Arrows indicate gaps or breaks along neuronal processes. (c) Images of the GABAergic motor neurons from an entire unc-47p::mCherry; ATXN3-CAG89 transgenic worms at adulthood day five with or without drug. Quantification of neurodegeneration in unc-47p::mCherry; ATXN3-CAG89 worms at day five of adulthood. Significant neurodegeneration is observed in unc-47p::mCherry; ATXN3-CAG89 transgenics when compared to the control, unc-47p::mCherry (####P < 0.0001). Significant rescue of the neurodegeneration morphology was observed in unc-47p::mCherry; ATXN3-CAG89 worms when treated with 2 μM diacerein (**P < 0.01), butaclamol, fenbufen, chenodiol, alfacalcidol, sulfaphenazole or cyclophosphamide (****P < 0.0001 for all the previous compounds) (by one-way ANOVA test, N = 150 for each condition). These experiments were done 6 independent times (N = 25 per trial for each condition). (d) Images of degenerating GABAergic motor neurons of unc-47p::mCherry; ATXN3-CAG89 transgenic worms at adulthood day nine with or without drug. Quantification of neurodegeneration in transgenic unc-47p::mCherry; ATXN3-CAG89 worms at day nine of adulthood. Significant neurodegeneration is observed in unc-47p::mCherry; ATXN3-CAG89 transgenics when compared to the control, unc-47p::mCherry (####P < 0.0001). This neurodegeneration is rescued when unc-47p::mCherry; ATXN3-CAG89 worms are treated with fenbufen, chenodiol, alfacalcidol, sulfaphenazole, or cyclophosphamide (****P < 0.0001 for all compounds) (by one-way ANOVA test, N = 150 for each condition). These experiments were repeated 6 times (N = 25 per trial for each condition). (e) Increased lifespan in ATXN3-CAG89 worms when treated with 2 μM CYCLO, ALFA, FEN, or SULFA (****P < 0.0001 for all four compounds), or CHEN (**P < 0.01) (by log-rank (Mantel-Cox) test) (N = 300–360). The experiment was done 3 times

Full size image

Small Molecules Rescue the Neurodegeneration and Extend Lifespan in ATXN3-CAG89 Transgenics

ATXN3-CAG89 transgenics display progressive degeneration of motor neurons as they age [20], and we focused on days five and nine of adulthood to evaluate the neuroprotective capacity of the 13 lead compounds. The health of motor neurons in ATXN3-CAG89 transgenics was assessed with a stably integrated unc-47p::mCherry translated that expresses a red fluorescing protein in the GABAergic motor neurons of living animals. Using this ATXN3-CAG89; unc-47p::mCherry strain, we observed that seven compounds rescued neurodegeneration at day five of adulthood (Fig. 3C). These seven compounds were then tested on ATXN3-CAG89 mutant worms and neurodegeneration was examined at adult day nine, and significant rescue of neurodegeneration was observed when treated with the following five compounds: alfacalcidol (ALFA), chenodiol (CHEN), cyclophosphamide (CYCLO), fenbufen (FEN), and sulfaphenazole (SULFA) (Fig. 3D). Finally, these five compounds rescued the decreased lifespan phenotype observed in ATXN3-CAG89 transgenics (Fig. 3E). Altogether, we identified five molecules that suppressed several negative phenotypes associated with the expression of mutant ATXN3 in vivo.

Heat Map Visualization of the Five Hit Compounds and Their Chemical Classes

Afterward, we wondered to evaluate the structural similarity and similar clusters existing between either the hit or not hit compounds screened from the 1280 molecules present in the Prestwick library (all five identified leads in this study belong to this library). From the hit compounds, several chemical classes were more representative: vitamin D3 derivatives (**P = 0.0078), cyclophosphamides (**P = 0.0078), benzenes (*P = 0.0233), and sulfonamides (*P = 0.0271) in which belong ALFA, CYCLO, FEN, and SULFA, respectively (Suppl. Fig. 1, Supplementary Material). All these chemical classes were statistically overrepresented, except the steroids (ns; P = 0.0610), CHEN, which strongly suppressed the phenotypes but were not statistically overrepresented (Suppl. Fig. 1, Supplementary Material). This could be due to the fact that more compounds were represented in this category (16 compounds) compared to the others, and also, a high structural similarity was observed between different isoforms in this chemical class.

Involvement of TFEB/HLH-30, a Key Regulator of the Autophagy Process in MJD

Autophagy is a major cellular recycling pathway that has an important role in aging. Several studies have highlighted an important link between autophagy and neurodegenerative diseases, and disruption of this pathway may contribute to pathology [3, 32, 33]. It has also been shown that upregulation of this pathway could be beneficial and ameliorate disease pathology [32]. Transcription factor EB (TFEB) is a major regulator of autophagy and lysosomal gene expression, and C. elegans possesses an orthologue named HLH-30 [34]. Based on this, we wondered if upregulation of TFEB/HLH-30 could modify phenotypes in our C. elegans MJD model. We crossed our ATXN3 transgenics with a transgenic, translational hlh-30 reporter strain which uses the hlh-30 promoter to express hlh-30 fused to GFP and measured the level of fluorescence in different adulthood stages: day one, two, three and five. No difference in the level of fluorescence was observed in day one of adulthood between treated or untreated hlh-30::GFP, hlh-30::GFP; ATXN3-CAG10 and hlh-30::GFP; ATXN3-CAG89 transgenic worms. However, we observed increased fluorescence in hlh-30::GFP; ATXN3-CAG89 worms when compared to hlh-30::GFP or hlh-30::GFP; ATXN3-CAG10 controls at adulthood days two, three, and five (Fig. 4A, B, Suppl. Fig. 2). This increase of fluorescence was mostly observed in the head (H), spermatheca (S), and tail (T). These data suggest there is an upregulation of the autophagy process in response to toxicity induced by mutant poly-Q proteins. From here, we investigated the activity of the lead compounds in relation to hlh-30 expression. We found that three out of the five compounds (exceptions were ALFA and CYCLO) decreased the fluorescence level in ATXN3-CAG89 mutant worms at adulthood days two and three (Suppl. Fig. 2), but significantly increased it at adulthood day five (Fig. 4A, B). We observed that in aging worms, hlh-30::GFP expression tends to shift and increase in ATXN3-CAG89 transgenics when treated with CHEN, FEN, and SULFA. We observed that the decreased level of fluorescence related to hlh-30::GFP expression in treated ATXN3-CAG89 worms at adulthood day three is less severe than adulthood day two, and increased at adulthood day five in the treated worms. This suggests that the increased expression of HLH-30 in ATXN3-CAG89 worms may represent an autophagic stress response that is mitigated by the small molecules CHEN, FEN, and SULFA in the early stages of worms (adulthood day two and three) but with aging, in a longer-term matter (adulthood day five), the molecules increase the level of HLH-30 expression in our MJD worms in order to protect the organism and decrease the toxicity caused by mutant ATXN3. We observed no effect of the compounds in hlh-30::GFP or hlh-30::GFP; ATXN3-CAG10 worms at different stages when treated (Suppl. Fig. 2). These data suggest a possible link between mutant ATXN3 toxicity and autophagy, as detected via hlh-30 expression. Although, further experiments would be required in order to be able to answer if the link existing between the action of the compounds and the autophagy in MJD mutants is due to a direct or indirect effect involving additional mechanisms and pathways.

Fig. 4
figure 4

Involvement of TFEB/HLH-30, a key regulator of the autophagy process in MJD. (a) Shown are representative photos of living, adult hlh-30::GFP, hlh-30::GFP; ATXN3-CAG10 and hlh-30::GFP; ATXN3-CAG89 transgenics at day five of adulthood. hlh-30::GFP; ATXN3-CAG89 transgenics showed increase GFP expression compared to hlh-30::GFP, or hlh-30::GFP; ATXN3-CAG10 controls (top panels). Treatment with 20 μM CHEN, 20 μM FEN, or 20 μM SULFA increased fluorescence of hlh-30::GFP; ATXN3-CAG89 mutants. hlh-30::GFP reporter showed increase fluorescence in the head (H), spermatheca (S), and tail (T) of adult animals. (b) Quantification of fluorescence of transgenics with or without treatment with compounds. An increased fluorescent signal was observed in hlh-30::GFP; ATXN3-CAG89 worms compared to hlh-30::GFP or hlh-30::GFP; ATXN3-CAG10 controls (###P < 0.001 and #P < 0.05 respectively). A significantly increased fluorescence was observed in hlh-30::GFP; ATXN3-CAG89 mutants when treated with 20 μM CHEN (*P < 0.05), 20 μM FEN (**P < 0.01) or 20 μM SULFA (*P < 0.05) (by one-way ANOVA test, N = 17–20 per trial for each condition). We observed no change in the fluorescence when hlh-30::GFP; ATXN3-CAG89 mutants were treated with 20 μM CYCLO or 20 μM ALFA. These experiments were replicated 3 times

Full size image

Chemical-Genetic Modulation of hlh-30 Suppresses the Neuronal Toxicity Observed in ATXN3-CAG89 Worms

Next, we examined directly for neuroprotective effects of increased hlh-30 expression in conjunction with the five lead molecules. We observed a decreased rate of paralysis in hlh-30::GFP; ATXN3-CAG89 worms when compared to ATXN3-CAG89 controls (Fig. 5A). Indeed, overexpression of HLH-30 alleviates the motility defect observed in ATXN3-CAG89 mutant worms. We tested the five lead molecules in the hlh-30::GFP; ATXN3-CAG89 worms, and although they all suppressed paralysis (Fig. 5A) and motor neuron degeneration (Fig. 5B) compared to ATXN3-CAG89 controls, only FEN showed an additive effect compared to hlh-30::GFP; ATXN3-CAG89 controls. These data suggest that upregulation of HLH-30 and treatment of the MJD model with FEN could be a novel neuroprotective approach to undertake in order to decrease the neuronal toxicity caused by expanded ATXN3.

Fig. 5
figure 5

Chemical-genetic modulation of hlh-30 suppresses the neuronal toxicity observed in ATXN3-CAG89 worms. (a) Decreased motility defect in hlh-30::GFP; ATXN3-CAG89 worms when compared to the control, ATXN3-CAG89 transgenics (###P < 0.001). When hlh-30::GFP; ATXN3-CAG89 worms are treated with 2 μM CHEN, 2 μM ALFA, 2 μM SULFA, or 2 μM CYCLO a similar profile to non-treated worms is observed. Rather, a significant rescue of motility is observed in hlh-30::GFP; ATXN3-CAG89 when treated with 2 μM FEN (*P < 0.05) (by log-rank (Mantel-Cox) test, N = 90–100 per trial, and N = 270–300 when all trials combined). These experiments were replicated 3 times. (b) Quantification of neurodegeneration in hlh-30::GFP; (unc-47p::mCherry; ATXN3-CAG89) worms at day nine of adulthood. Significant rescue of neurodegeneration in hlh-30::GFP; (unc-47p::mCherry; ATXN3-CAG89) worms was observed when compared to the control at adulthood day nine (####P < 0.0001). Treatment of the worms with 2 μM CHEN, 2 μM ALFA, 2 μM SULFA, or 2 μM CYCLO showed no additional rescue when compared to untreated hlh-30::GFP; (unc-47p::mCherry; ATXN3-CAG89) worms. Treatment of hlh-30::GFP; (unc-47p::mCherry; ATXN3-CAG89) worms with 2 μM FEN (*P < 0.05) (by one-way ANOVA test, N = 100 for each condition) showed a significant rescue of the neurodegeneration profile. These experiments were done 4 times

Full size image

CHEN, FEN, and SULFA Require hlh-30 for Their Neuroprotective Activities

To determine if any of the five lead molecules required hlh-30 for their neuroprotective activity, and in order to confirm if the link existing between hlh-30 and the action of the compounds is due to a direct or indirect effect, ATXN3-CAG89 worms were crossed with loss-of-function (LOF) mutant hlh-30 worms. We observed that CHEN, FEN, and SULFA required hlh-30 for their activity as these compounds were unable to rescue motility deficits (Fig. 6A) and neurodegeneration (Fig. 6B) in hlh-30(tm1978); ATXN3-CAG89 worms. Treatment with CYCLO or ALFA continued to rescue paralysis and neurodegeneration phenotypes in the absence of hlh-30 (Fig. 6), suggesting a possible compensatory effect by other pathways. Also, we observed in the paralysis assays that hlh-30(tm1978); ATXN3-CAG89 worms showed a slightly lower (but not significant) paralysis phenotype when compared to ATXN3-CAG89 mutants, suggesting a possible rescue effect in our MJD model when hlh-30 is absent (Fig. 6A). However, with the neurodegeneration assays, this hypothesis was rejected as we observed a slight increase (without showing significance) of neurodegeneration in hlh-30(tm1978); ATXN3-CAG89 worms compared to ATXN3-CAG89 mutants (Fig. 6B). These data show that the absence of hlh-30 in our MJD model does not necessarily modify ATXN3-CAG89 phenotypes, but confirm that CHEN, FEN, and SULFA require hlh-30 for their neuroprotective activities, as well as a direct link between hlh-30 and the action of these three compounds.

Fig. 6
figure 6

CHEN, FEN, and SULFA require hlh-30 for their neuroprotective activities. (a) hlh-30(tm1978); ATXN3-CAG89 worms showed a similar paralysis phenotype than the control, ATXN3-CAG89 mutants. When hlh-30(tm1978); ATXN3-CAG89 worms were treated with 2 μM CHEN, 2 μM FEN, or 2 μM SULFA, no rescue in the motility deficit was observed. Treatment with 2 μM CYCLO or 2 μM ALFA did rescue the motility defect (*P < 0.05 for each compound) (by log-rank (Mantel-Cox) test, N = 90–100 per trial, and N = 270–300 when all trials combined). These experiments were replicated 3 times. (b) Quantification of neurodegeneration in hlh-30(tm1978); (unc-47p::mCherry; ATXN3-CAG89) worms at adulthood day nine. No rescue of neurodegeneration was observed in hlh-30(tm1978); (unc-47p::mCherry; ATXN3-CAG89) worms at adulthood day nine when treated with 2 μM CHEN, 2 μM FEN, or 2 μM SULFA. Significant rescue of neurodegeneration in these worms when treated with 2 μM CYCLO (**P < 0.01) or 2 μM ALFA (***P < 0.001) (by one-way ANOVA test, N = 100 for each condition). These experiments were repeated 4 times

Full size image

Finally, we tested treatment with the three compounds (CHEN, FEN, and SULFA) and observed that they failed to suppress motility phenotypes in hlh-30(tm1978); ATXN3-CAG89 worms (Fig. 7A), whereas the triple molecule treatment provided robust suppression in hlh-30::GFP; ATXN3-CAG89 worms (Fig. 7B). These results demonstrate a key role for hlh-30 in the suppression of mutant ATXN3 toxicity by small molecules.

Fig. 7
figure 7

Effect of the combination of the three hlh-30 related compounds on MJD mutants. (a) No rescue of the locomotion impairment is observed in hlh-30(tm1978); ATXN3-CAG89 worms when treated with the three compounds simultaneously (by log-rank (Mantel-Cox) test, N = 90–100 per trial, and N = 270–300 when all trials combined). These experiments were replicated 3 times. (b) Significant rescue of the motility defect is observed in hlh-30::GFP; ATXN3-CAG89 worms when treated with the combination of the three compounds (**P < 0.01) (by log-rank (Mantel-Cox) test, N = 270–300). These experiments were repeated 3 times

Full size image

Discussion

By using our ATXN3 transgenic worms, we performed a large comprehensive blind drug screen of 3942 molecules. This screen allowed us to identify five lead compounds that rescued the motor function, neurodegeneration, and reduced longevity observed in our ATXN3 mutant transgenic C. elegans. Besides finding five lead compounds, we also identified a new transcription factor related to ATXN3 toxicity, TFEB/HLH-30. We also showed that chenodiol, fenbufen, and sulfaphenazole, are dependent on this transcription factor for their neuroprotective activities in MJD.

Drug screening in C. elegans, a simple and easily maintained model organism, is a popular tool to identify compounds, as well as pathways contributing to a better understanding of the disease and development of possible therapies [18, 35,36,37]. In this study, we used this approach in order to find lead molecules able to rescue the locomotion deficit observed in our mutant ATXN3-CAG89 transgenics. The aim of this study was to identify molecules and pathways related to MJD, as well as potential compounds for drug development in the treatment of MJD. The drug library we screened contained many FDA-approved compounds allowing for rapid, potential translation to preclinical settings. Also, many of the compounds in this library have been studied and tested in models for other neuropathologies such as Alzheimer’s disease, Parkinson’s disease, and ALS. The screening and other assays (paralysis, neurodegeneration, and lifespan) allowed us to identify five lead compounds: alfacalcidol (ALFA), chenodiol (CHEN), cyclophosphamide (CYCLO), fenbufen (FEN), and sulfaphenazole (SULFA). With the heat map visualization, we were able to identify several chemical classes present in our Prestwick library, in which all five compounds belong: vitamin D derivatives, steroids, cyclophosphamides, benzenes, and sulfonamides, respectively.

ALFA is known to be an active metabolite of vitamin D performing important functions in the regulation of calcium and bone metabolism [38]. Vitamin D hydroxylating enzymes and receptors are located on immune cells and in some key areas of the brain [39]. Epidemiological evidences demonstrated that deficiency of vitamin D is relevant to disease risk in multiple sclerosis (MS), Parkinson’s disease (PD), and Alzheimer’s disease (AD) [38, 40, 41]. Several studies done on neurodegenerative diseases demonstrated that low serum or plasma vitamin D levels bring cognitive impairment, increased risk of dementia, and impaired motor functions [40, 42,43,44,45]. Also, in a study evaluating the mechanisms by which vitamin D influences aging in C. elegans, it has been demonstrated that vitamin D promotes protein homeostasis, supresses protein insolubility and toxicity, and extends lifespan by specific stress response pathway genes [46]. Thus, these findings confirm the importance of keeping appropriate vitamin D serum levels and could explain the reason why many age-related diseases are related to vitamin D deficiency [46]. Moreover, a perturbation of calcium signaling in MJD cases has also been observed and as therapeutic perspectives, an improvement of the calcium homeostasis has been suggested [47]. Thereby, ALFA, being implicated in the calcium hemostasis, might be one of the reasons why this compound figures among the leads. As an off-target effect, the use of vitamin D could also reverse the oxidative stress damage to DNA and protects against neuroinflammation [48, 49]. Treatment with alfacalcidol could decrease the neuronal toxicity caused by expanded ATXN3, which induces a high level of oxidative stress, DNA damage, disturbance in calcium signaling, and neuroinflammation in MJD patients [20, 47, 50, 51]. This compound could act on many different targets beneficial for MJD.

CHEN, also known as chenodeoxycholic acid, is a natural bile acid found in the body that dissolves the cholesterol that makes gallstones and inhibits its production in the liver and absorption in the intestine [52]. It has been noted that this compound displays also some off-target effects decreasing the neuronal toxicity. Indeed, it has been reported in a rat Alzheimer’s disease model study that CHEN decreases the neurotoxicity and cognitive deterioration observed in this disease by activating the cAMP response element-binding protein (CREB), enhancing brain-derived neurotrophic factor (BDNF), and improving insulin sensitivity [52]. As for MJD, CHEN could be an interesting therapeutic approach to undertake in order to study these aspects as well. In this study, we showed a novel role for CHEN which figures among the leads. We have shown that CHEN requires hlh-30, a transcription factor related to the autophagy pathway, to complete its neuroprotective activities and decrease the neurotoxicity in our MJD model.

FEN is a non-steroidal anti-inflammatory drug predominantly used to treat inflammation. This compound prevents prostaglandin’s production, which can cause inflammation by inhibiting cyclooxygenase (COX) [53]. FEN is used in Alzheimer’s disease in order to prevent the inflammation observed in patients [54]. Several studies demonstrated the involvement of neuroinflammation in several neurodegenerative diseases such as Alzheimer’s disease, Parkinson’s disease, amyotrophic lateral sclerosis, Huntington’s disease, and MJD [55,56,57,58,59,60]. Using FEN as a therapeutic compound could have clinical benefits such as decreasing the neuroinflammation in patients suffering from any of these neurodegenerative diseases. As an off-target effect, FEN also acts as an antioxidant. Indeed, this compound is able to scavenge reactive oxygen species (ROS) and reactive nitrate species (RNS) [61]. An interesting target to study in our MJD model in which we have already demonstrated high levels of oxidative stress [20]. In this study, we observed that FEN rescues phenotypes such as motility defects, neurodegeneration, and reduced longevity in ATXN3-CAG89 mutant worms. We also showed that FEN has high potential to rescue the motility deficit and neurodegeneration when TFEB/HLH-30 is overexpressed. This suggests that FEN could have beneficial and important effects on the autophagy pathway.

CYCLO, also known as cytophosphane, is an alkylating agent used in order to suppress the immune system, is an immunomodulator, and is also a potent anti-inflammatory [62]. This compound has been used as a preventative drug in multiple sclerosis disease. It has been shown that CYCLO has the property to permeate the blood-brain barrier, has a good bioavailability in the central nervous system, and, by exhorting its immunomodulation and immunosuppression role, is able to stabilize and prevent the progression of this disease [62, 63]. This drug is also used in individuals with mild to moderate Alzheimer’s disease in order to decrease side effects [64]. This molecule is able to restore the motility deficit and neurodegeneration and enhance longevity in our MJD strain.

SULFA, finally, is a sulfonamide antibiotic that targets bacterial replication by inhibiting folate biosynthesis. It is also a selective inhibitor of the mammalian Cytochrome P450 isozyme CYP2C9 [65, 66]. This compound was also identified in a screen looking for molecules that block light-induced, degenerative loss of photoreceptors that occurs in age-related retinal degenerative diseases [67]. It has also been shown in a Parkinson’s disease study that SULFA acts as a neuroprotective molecule by enhancing the longevity and normal dopaminergic neurons in their C. elegans model organism [66]. As an off-target effect, it has been shown that SULFA induces autophagy and protein kinase C (PKC) activations [68]. They have shown that the protection mediated by SULFA in their model was due to an increase in the autophagy process [68]. It has also been shown the autophagy activation is associated with neuroprotection [69]. As in this study, we showed that SULFA is dependent on hlh-30, an important regulator of the autophagy process, for its neuroprotective activities. We also showed that this compound is able to prolong longevity, as well as suppress neurodegeneration and motility defects in ATXN3 worms.

Autophagy is one of the main and important pathways allowing the degradation of abnormal protein aggregates. Disruption and impairment of this pathway could contribute to many neurodegenerative diseases such as MJD [70,71,72,73,74,75,76,77]. It has also been shown that the upregulation of autophagy is capable of reducing the level of toxic proteins, ameliorate signs of disease, and also delay the disease progression in several neurodegenerative diseases such as Alzheimer’s disease, Parkinson’s diseases, polyglutamine diseases, and amyotrophic lateral sclerosis [32]. Based on these findings, we investigated if an important regulator of the autophagy process, TFEB/HLH-30, was involved in a model of MJD.

When TFEB/HLH-30 is overexpressed, we observed an increased level of hlh-30::GFP expression in our MJD mutants. Also, we observed a shift and increase of hlh-30::GFP expression during aging in our MJD mutants when treated with CHEN, FEN, and SULFA. In the early stages (adulthood days two and three), the compounds tend to decrease the level of hlh-30::GFP expression in ATXN3-CAG89 worms perhaps as an early protective response in order to bring back the homeostasis. But then, at later stages (adulthood day five), with aging, they tend to increase the expression of hlh-30::GFP in our MJD mutants in order to protect and decrease the toxicity induced by mutant poly-Q proteins. Indeed, alterations of autophagy levels during aging have been reported and may be relevant to age-dependent neurodegeneration [78].

Also, it has been observed that the hlh-30::GFP reporter presents an increase in the level of fluorescence in the head (H), spermatheca (S), and tail (T) of adult animals. However, the reason explaining how this elevation could affect the GABAergic neurons (where ATXN3 has been expressed in our transgenic MJD model) is due to the communication existing between cells during stress in order to augment coordination of an organism-wide stress response. Here, we suspect that proteotoxically stressed GABA neurons are communicating to the rest of the organism, something that we have previously reported for our MJD model in the context of ER stress [20].

ATXN3-CAG89 worms overexpressing TFEB/HLH-30 showed a rescue of motility deficits and neurodegeneration. Indeed, overexpression of this protein in our MJD model showed an ability to alleviate these phenotypes caused by the expanded ATXN3. A similar profile was observed when these worms were treated with small molecules. The lead compounds, CHEN, FEN, and SULFA rescued the locomotion defect and neurodegeneration in ATXN3-CAG89 worms overexpressing TFEB/HLH-30 when compared to ATXN3-CAG89 mutants. One compound, FEN, having an additive effect, showed high potential in decreasing remarkably the motor function and neurodegeneration (at adult day nine) in our mutant overexpressing TFEB/HLH-30 worms. Thus, a chemical-genetic approach could be a potential therapeutic strategy for MJD cases. These findings affirmed that TFEB/HLH-30 may be involved in ATXN3 phenotypes.

Finally, we demonstrated that CHEN, FEN, and SULFA require hlh-30 for their neuroprotective activities. Indeed, these compounds are dependent and directly linked to this transcription factor to restore MJD phenotypes such as motility defect and neurodegeneration. In the absence of this gene, no small molecule mediated rescue is observed in our mutants. However, we observed that ALFA and CYCLO are still both able to rescue motility deficits as well as neurodegeneration in our MJD mutants. This rescue might be due to some compensatory effects of other activated pathways. Also, we observed a slight, non-significant decrease of paralysis phenotypes in our MJD model in the absence of hlh-30, suggesting a possible rescue effect. However, in the neurodegeneration assays, this rescue effect was completely absent, and we rather observed a slight increase (not significant) of neurodegeneration in the absence of hlh-30 in ATXN3-CAG89 mutants. Based on these observations, we concluded that the absence of hlh-30 might not be toxic, as expected, in our MJD model, and shown to have a more neutral effect contrary to its overexpression, where we observed a significant decrease of paralysis and neurodegeneration phenotypes. This neutral effect observed in the LOF hlh-30 might be due to a compensatory effect of other autophagy genes present in C. elegans.

In summary, we conducted a high-throughput screen of 3942 compounds using a C. elegans model of MJD. This screen led to the identification of five lead compounds that could be promising candidates for the treatment of MJD. Among the leads, one compound, fenbufen, showed to be the most effective. Indeed, additional studies are required to extend these findings to mammalian models of MJD. However, from a practical perspective, fenbufen, being FDA approved, could be a novel neuroprotective molecule to investigate since it could be translated rapidly into clinical settings for MJD.

Also, in this study, we identified TFEB/HLH-30 as a new potential regulator of MJD pathology. However, further investigation is required to better understand the roles of TFEB/HLH-30 and autophagy in the context of mutant ATXN3 toxicity and MJD. Finally, the drug screening and the mutant ATXN3 transgenic worms are valuable tools to advance MJD research, as the pathogenic mechanisms of disease are still not well understood. Our findings could guide future MJD research, by validating these lead molecules and autophagy regulators in more advanced model organisms and eventually in humans.