Re-establishing ataxin-2 downregulates translation of mutant ataxin-3 and alleviates Machado–Joseph disease

Abstract

Machado-Joseph disease is a progressive neurodegenerative disorder associated with the polyQ-expanded ataxin-3 (encoded by ATXN3), for which no therapy is available. With the aim of clarifying the mechanism of neurodegeneration, we hypothesized that the abnormally long polyQ tract would interact aberrantly with ataxin-2 (encoded by ATXN2), another polyQ protein whose function has recently been linked to translational regulation. Using patient’s samples and cellular and animal’s models we found that in Machado-Joseph disease: (i) ataxin-2 levels are reduced; and (ii) its subcellular localization is changed towards the nucleus. Restoring ataxin-2 levels by lentiviral-mediated overexpression: (i) reduced mutant ataxin-3 levels; and (ii) rescued behaviour defects and neuropathology in a transgenic mouse model of Machado-Joseph disease. Conversely (i) mutating the ataxin-2 motif that enables binding to its natural interactor and translation activator poly(A)-binding protein; or (ii) overexpressing poly(A)-binding protein, had opposite effects, increasing mutant ataxin-3 translation and aggregation. This work suggests that in Machado-Joseph disease, mutant ataxin-3 drives an abnormal reduction of ataxin-2 levels, which overactivates poly(A)-binding protein, increases translation of mutant ataxin-3 and other proteins and aggravates Machado-Joseph disease. Re-establishment of ataxin-2 levels reduces mutant ataxin-3 and alleviates Machado-Joseph disease pathogenesis opening a new avenue for therapeutic intervention in this and potentially other polyQ disorders.

The mechanisms underlying neurodegeneration in Machado-Joseph disease are unclear. Nóbrega et al. show that the polyglutamine expansion in ataxin-3 is accompanied by a reduction in ataxin-2 levels, which results in overactivation of the translational activator PABP and increased expression of ataxin-3. Restoring ataxin-2 levels alleviates pathology in a mouse model.

Machado-Joseph disease, mutant ataxin-3, ataxin-2, translation regulation, PABP

Introduction

The polyglutamine (polyQ) neurodegenerative disorders consist of at least nine diseases caused by CAG-repeat expansions that encode polyQ tracts in the mutant proteins (Zoghbi and Orr, 2009). These diseases share some similarities including progressive neurodegeneration in a disease-specific subset of neurons and the presence of insoluble protein aggregates. Nuclear ubiquitin-positive aggregates of mutant proteins are one hallmark of polyQ disorders and despite the controversial role of inclusions in pathogenesis, nuclear localization of expanded polyQ proteins has been shown to contribute to disease pathogenesis (Paulson et al., 1997; Klement et al., 1998; Perez et al., 1998; Saudou et al., 1998; Peters et al., 1999; Ross et al., 1999; Orr, 2001; Bichelmeier et al., 2007) by altering transcription (Li et al., 2000, 2002; McCampbell et al., 2000) or by disrupting nuclear organization and function (Sun et al., 2007).

The polyQ disorder Machado-Joseph disease or spinocerebellar ataxia type 3 (SCA3) is the second most common polyQ disorder. It is a dominantly-inherited disorder associated with the expansion of a (CAG)n tract in the coding region of the causative gene ATXN3. The CAG repeats range from 10 to 51 in the normal population and from 55 to 87 in those patients with Machado-Joseph disease (Maciel et al., 2001) and translate into a polyQ tract within the protein ataxin-3 (Atx3), a protein whose physiological function has been linked to ubiquitin-mediated proteolysis (Doss-Pepe et al., 2003). Despite important progresses, the mechanisms accounting for neuronal degeneration are still largely unknown and there is currently no treatment available.

Evidence suggests that as it happens with other polyQ disorders, on mutation (trinucleotide expansion), Atx3 gains a new toxic function that is sufficient to cause the disease. Expanded Atx3 accumulates and aggregates in the nucleus, where it may interact with multiple proteins potentially aggravating the pathogenic process.

Ataxin-2 (Atx2), another protein whose mutated form is involved in spinocerebellar ataxia type 2 (SCA2), is one of the candidate proteins for potential interaction and sequestration with Atx3 in intranuclear aggregates. Importantly, the wild-type Atx2 function has been linked to translational regulation of specific transcripts (Liu-Yesucevitz et al., 2011; McCann et al., 2011; Drost et al., 2013; Zhang et al., 2013; Sudhakaran et al., 2014). Accordingly, there is evidence that Atx2 interacts with poly-a-binding protein [PABPC1, which is required for poly(A) shortening and translation initiation] in polyribosomes (Ralser et al., 2005) and in this way could reduce and regulate protein translation (Satterfield and Pallanck, 2006; Eliseeva et al., 2013; Magana et al., 2013). This possible role of Atx2 is also reinforced by the interaction with the putative RNA-binding protein: ataxin 2-binding-protein (A2BP, encoded by RBFOX1) (Shibata et al., 2000), and also with endoplasmic reticulum (van de Loo et al., 2009). Furthermore, it was recently shown that: (i) Atx2 has been associated to SCA1, in a screen for genetic modifiers for this disease (Fernandez-Funez et al., 2000); and (ii) Atx2 polyQ length expansion (below the length causative of SCA2) has also been associated with increased risk for amyotrophic lateral sclerosis (Elden et al., 2010). Overall, there is accumulating evidence of an important role of Atx2 in neurodegenerative disorders pathogenesis.

In Machado-Joseph disease a potential link with Atx2 could also be hypothesized, as this protein is detected in mutant Atx3 (Atx3MUT) nuclear aggregates (Uchihara et al., 2001), despite no direct interaction between the two proteins (Lim et al., 2006). The Atxn2 gene was also identified in a screen for modifiers of degeneration induced by Atx3MUT (Bilen and Bonini, 2007) and the normal activity of Atx2 was shown to be critical to neurodegeneration in a Drosophila model, suggesting that the toxicity of one polyQ disease protein can be modulated by the normal activity of another (Lessing and Bonini, 2008). Nevertheless, how these two proteins interact and how they might contribute to Machado-Joseph disease pathology had not been investigated.

In the present work we aimed at clarifying whether Atx3 and Atx2 interaction contributes to Machado-Joseph disease pathogenesis. We found that Atx3MUT drives an abnormal reduction of Atx2 levels, which overactivates poly(A)-binding protein (PABP), increases translation of Atx3MUT and other proteins and aggravates Machado-Joseph disease. Importantly, re-establishment of Atx2 levels reduces Atx3MUT and alleviates Machado-Joseph disease pathogenesis opening a new therapeutic avenue.

Material and methods

Lentiviral and plasmid vectors

The cDNA encoding for human Atx2 (ATXN2) with 23 CAG (GC-Y4388, Genecopoeia), human Atx2 with 22 CAG (kindly provided by Prof. Stefan Pulst), human PABP (IRAUp969H0777D, ImaGenes), and for human Ataxin-2-like (IRAUp969E05113D, ImaGenes) was cloned in a self-inactivating lentiviral vector as described previously (Deglon et al., 2000). All the viral vectors encoding for the different constructs were produced in HEK 293T cells using a four-plasmid system described previously (de Almeida et al., 2002). Viral titre was assessed by quantification of p24 by enzyme-linked immunosorbent assay (HIV-ELISA, Zeptometrix).

Ataxin-2 mutagenesis of PAM2 motif

The mutagenesis of the PAM2 motif of the plasmid EGFP-Atx2-Q22 was performed using the QuickChange II XL site-directed kit (Agilent) following the manufacturer’s instructions. Briefly, the amino acid change was performed through a site directed mutagenesis by PCR using the primers: forward 5’- GAATCCCAATGCAAAGGAGGCCAACCCACGTTCCTTCTCTC-3’; and reverse 5’- GAGAGAAGGAACGTGGGTTGGCCTCCTTTGCATTGGGATTC-3’. After the DpnI digestion and bacterial transformation several clones were selected for analysis of the mutated site. Positive clones were confirmed by double sequencing using four different internal primers.

Neuroblastoma and human fibroblast culture

Mouse neuroblastoma cell line (N2a cells) obtained from the American Type Culture Collection cell biology bank (CCL-131) were incubated in Dulbecco’s modified Eagle medium supplemented with 10% foetal bovine serum, 100 U/ml penicillin and 100 μg/ml streptomycin. Human fibroblast cells from patients with Machado-Joseph disease and controls were obtained from the Neurology and Pathology Services of University Hospital of Coimbra, Portugal. Cells were kept in culture in Dulbecco’s modified Eagle medium supplemented with 10% bovine serum, 1% non-essential amino acids, 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. Both cell cultures were maintained at 37°C in 5% CO₂/air atmosphere.

Cell viability assay

Cell viability under the different experimental conditions was assessed by a modified alamarBlue® assay. At several time-points after transfection or infection, the cells were incubated with Dulbecco’s modified Eagle medium containing 10% (v/v) alamarBlue® dye. After a 1 h incubation period at 37°C, the absorbance of the medium was measured at 570 nm and 600 nm. Cell viability was calculated as a percentage of the control cells (non-transfected/infected), according to:

Cell viability (% of control) = [(A₅₇₀ − A₆₀₀) of treated cells × 100/(A₅₇₀ − A₆₀₀) of control cells].

Flow cytometry analysis

Ten days post-infection the N2a cells were processed for flow cytometry analysis. Cells were harvested into conic tubes (BD Bio-sciences), washed twice with cold PBS, and resuspended in 0.5–1.0 ml of Hanks buffer saline solution. For apoptosis detection experiments the FITC Annexin V Apoptosis Detection Kit II (BD Pharmingen) was used following the manufacturer’s instructions. All samples were then analysed in a FACS Calibur flow cytometer (BD, Biosciences) and data evaluated by Cell Quest software (BD). Annexin–FITC fluorescence was evaluated in the FL-1 channel and PI in the FL-2 channel. The flow cytometry analysis was performed in triplicate for each experimental set.

Neuroblastoma cells infection and transfection

Cells were plated into 6- or 12-multiwell plates for infection and transfection, respectively. For the transfection experiments using Lipofectamine® LTX the manufacturer’s instructions were followed, with a total concentration of 0.5–1 μg of DNA per well. Cells were lysed for western blot processing at 36–48 h post-transfection. In the infection experiments, cells were infected with lentiviral particles in a total concentration of 30–50 ng (p24, HIV-ELISA Zeptometrix) for each viral vector per 100 000 cells per well and condition. At 8–10 days post-infection cells were lysed for western blot subsequent processing or fixed for immunocytochemistry.

Translation rate assay

We used a previously described method that allows the monitoring and quantification of global protein synthesis based in the incorporation of puromycin during translation (Schmidt et al., 2009). The N2a cells expressing Atx3MUT were plated into multiwell plates and in the next day were transfected with Atx2 or green fluorescent protein (GFP). Twenty-four hours post-transfection the puromycin assay was performed. Briefly, cells were incubated with 10 μg/ml of puromycin (Sigma) for 15 min, and later lysed for western blot posterior processing. As positive control for the translation inhibition, some cells were incubated first with 10 μM of cycloheximide (CHX, Sigma) for 15 min, and then incubated with 10 μg/ml of puromycin (Sigma) for 15 min. Some cells were not treated with puromycin as control for the experiment, and wild-type N2a cells (without Atx3MUT) were also used in these experiments as additional controls.

Human brain tissue

Post-mortem putamen tissue from patients with clinically and genetically confirmed Machado-Joseph disease was obtained from the Tissue Donation Program of the National Ataxia Foundation, Minneapolis, MN, USA. Controls with no evidence of neurological disease were obtained from Neurology and Pathology Services of University Hospital of Coimbra, Portugal. The tissue was fresh when dissected, placed in 10% neutral buffered formalin and kept at 4°C.

Animals

Adult C57BL/6 J mice (Charles River Laboratories), and 3-week-old transgenic Machado-Joseph disease mice and wild-type littermates (Torashima et al., 2008) were used. The animals were housed in a temperature-controlled room maintained on a 12-h light/12-h dark cycle. Food and water were provided ad libitum. The experiments were carried out in accordance with the European Community directive (86/609/EEC) for the care and use of laboratory animals. The researchers received adequate training (Felasa-certified course) and certification to perform the experiments from the Portuguese authorities (Direcção Geral de Veterinária).

Machado-Joseph disease lentiviral and transgenic mice models

The cerebellar cortex of 3-week-old and 2-month-old transgenic mice (Q69 mice) and their wild-type littermates (Torashima et al., 2008) were dissected fresh and kept at −80°C for subsequent processing for western blot or reverse transcription-polymerase chain reaction (RT-PCR) analyses, or alternatively perfused with 4% paraformaldehyde and kept for subsequent histological processing. For the lentiviral mouse model (Alves et al., 2008a), one striatal hemisphere was injected with Atx3MUT with 72 glutamines, whereas the contralateral was injected with wild-type Atx3 with 27 glutamines actin as internal control. Brains were then used for immunohistochemistry, western blot or RT-PCR purposes, at 2, 4 and 8 weeks post-injection.

In vivo injection of lentiviral vectors

For the stereotaxic injection of lentiviral vectors, concentrated viral stocks were thawed on ice and resuspended by vortexing. The animals were anaesthetized with ketamine (75 mg/kg intraperitoneally) and xylazine (10 mg/kg intraperitoneally). Particle content of lentiviral vectors was matched to 400 ng of p24/ml. For the striatal injections animals received a single injection in each hemisphere (0.2 µl/min): in the left hemisphere 1 µl of mutant Atx3 (72 CAG) and in the right hemisphere 1 µl of mutant Ataxin-3 (72 CAG) and 1 µl of Atx2 or PABP. The injections were performed at the coordinates previously described (Alves et al., 2008a; Simões et al., 2012). Animals were kept for 4 weeks post-injection before being sacrificed for posterior analysis. For the cerebellar injections, animals received a single injection of 4 µl of Atx2 or GFP (Nascimento-Ferreira et al., 2013), following the surgical procedure described previously (Nóbrega et al., 2013a). Animals were kept for 8 weeks post-injection before being sacrificed for subsequent analyses, with behaviour tests being performed every 2 weeks.

Behavioural testing

Mice were trained on a battery of motor tests starting at 21–25 days of age (postnatal Days 21–25) and performed every 2 weeks until 8 weeks post-injection by an experienced operator in a blind fashion way following the same procedure described before (Nóbrega et al., 2013b). All tests were performed in the same room after 30 min of acclimatization. Mean values for each measure were calculated and statistical analysis was performed by two-way ANOVA with GraphPad (La Jolla, USA).

Histological processing

Tissue preparation

Animals were killed by sodium pentobarbital overdose, transcardially perfused with 0.1 M phosphate buffer solution and a 4% paraformaldehyde fixative solution (Fluka, Sigma) followed by brain removal. The brains were removed and post-fixed in 4% paraformaldehyde for 24 h and cryoprotected by incubation in 20% sucrose/phosphate buffer for 48 h. The brain was frozen and sectioned using a cryostat (LEICA CM3050 S). For the striatal injections coronal sections of 20 µm were made. For the cerebellar injections, all brain was sectioned in 30 µm sagittal sections. Slices were collected and stored in 48-well trays, free-floating in 0.1 M phosphate buffer solution supplemented with 0.12 mmol/l sodium azide. The plates were stored at 4°C until immunohistochemical processing. For western blot and RT-PCR purposes, animals were killed by sodium pentobarbital overdose and brain directly were removed without perfusion. Brains were dissected fresh and striatal punches were collected with a Harris Uni-Core pen, with 2.0 mm diameter (Ted Pella, Inc.). Samples were kept at −80°C for posterior processing.

Reverse-transcription polymerase chain reaction

Frozen punches were slightly thawed at room temperature and RNA extraction was performed with QIAzol™ lysis reagent and RNeasy® Mini Kit (Qiagen). Briefly, the RNA at the aqueous phase obtained with the QIAzol™ lysis reagent was cleaned up with the RNeasy® Mini Kit using DNase digestion at the RNeasy® spin column, according to the manufacturer’s recommendations. The RNA concentration and purity were determined with NanoDrop 2000 (Thermo Scientific). cDNA synthesis was performed with iScript™ cDNA Synthesis Kit (Bio-Rad) from 1 μg of total RNA according to the manufacturer’s instructions. PCR was performed in real time quantitative PCR with the SsoAdvanced™ SYBR® Green Supermix Kit (Bio-Rad). The primers used for target and housekeeping genes were the following: human ATXN2 (QT01852480), mouse Atxn2 (QT00151249), human ATXN3 (QT00094927), human HPRT (QT00059066), mouse HPRT (QT00166768), all pre-designed (QuantiTect Primer Assays, Qiagen) and codon optimized human ATXN2 (coptAtx2: forward 5’-GGACCAGGGCCTCCAC-3’; reverse 5’-GCAGCGGGTGGTGCT-3’). Briefly, 2.5 μl of the cDNA obtained in the reverse transcription reaction diluted 10-fold with DNase free deionized water were used. The quantitative PCR was performed as follows: one single cycle at 95°C for 30 s, followed by 45 cycles of two steps, first step of 5 s at 95°C, second step of 15 s at 55°C. The melting curve protocol started immediately after the quantitative PCR and consisted of 5 s heating at 65°C with a 0.5°C temperature increase in each step until 95°C is reached. The threshold cycle (Ct) values were generated automatically by the StepOne Software (Applied Biosystems). To each gene, and in each experiment, a standard curve was performed and quantitative PCR efficiency was determined by the software. The mRNA relative quantification with respect to control samples was determined by the Pfaffl method, taking into consideration the different amplification efficiencies of all genes in all experiments.

Western blot

For the protein extraction protocol, brain tissue and cells were lysed in radio-immunoprecipitation assay-buffer solution (50 mM Tris HCl, pH 8, 150 nM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulphate) containing proteases inhibitors (Roche diagnostics GmbH) followed by a 4-s ultra-sound pulse (1 pulse/s). Total protein lysates were stored at −80°C, and protein concentration was determined with the Bradford protein assay (Bio-Rad). Depending on the analysis, 40 or 60 µg of protein extract were resolved in sodium dodecyl sulphate-polyacrylamide gels (4% stacking, 10 or 12% running), transferred onto PVDF (polyvinylidene fluoride) membranes. Immunoblotting was performed using monoclonal antibodies: anti-ataxin 2 clone 22, 1:1000; BD Biosciences), anti-ataxin3 (1H9, 1:1000, Millipore), anti-PABP (clone 10E10, 1:1000; Millipore), anti-Puromycin (clone 12D19; 1:25000; Millipore), anti-actin (clone AC-74, 1:5000; Sigma) or anti-tubulin (clone SAP.4G5, 1:15000; Sigma) and polyclonal rabbit anti-cleaved caspase 3 antibody (Asp175, 1:1000; Cell Signaling). Densitometric analysis was carried out in the same gel using ImageJ software (NIH, USA).

Immunochemical procedure

Immunocytochemistry

The immunocytochemical procedure was initiated by fixating cells into glass coverslips with 4% paraformaldehyde fixative solution for 20 min, and after washing with 0.1 M phosphate buffer solution. Then samples were incubated for 10 min with PBS containing 0.25% Triton™ X-100. Blocking was made for 30 min in PBS with 1% bovine serum albumin (BSA, Sigma). Primary antibody was incubated overnight in blocking solution in the proper dilution, and the secondary antibody (1:200) for 2 h at room temperature. The secondary antibody used was coupled to a fluorophore (Alexa Fluor®, Invitrogen) and followed by a nuclei staining reaction with 4’,6-diamino-2-phenylindole (DAPI, 5 min, room temperature). The coverslips were then mounted in FluorSave™ Reagent (Calbiochem) on microscope slides.

Immunohistochemistry

The immunohistochemical procedure was initiated by incubating brain sections in phenylhydrazine diluted in phosphate buffer solution (1:1000; 15 min, 37°C; light imaging only), followed by a Tris-buffered saline pH 9 antigen retrieval method (30 min, 95°C, human tissue only), blocking in 10% normal goat serum in 0.1% Triton™ X phosphate-buffered solution (1 h, room temperature) and incubation with the respective primary (overnight at 4°C) and secondary antibodies (2 h at room-temperature) diluted in blocking solution. For light imaging, the secondary antibody (1:200) used was biotinylated and followed a reaction with the Vectastain elite avidin-biotin-peroxidase kit and by 3,3’-diaminobenzidine substrate (both from Vector Laboratories). For human tissue, premounted sections were then counterstained with Cresyl violet (2 min in Cresyl violet solution followed by differentiation in ethanol 70%) or directly dehydrated in toluene and coverslipped with Eukitt® mounting medium (O. Kindler). For the fluorescent imaging, the secondary antibody (1:200) used was coupled to a fluorophore (Alexa Fluor®, Invitrogen) and followed a nuclei staining reaction with 4’,6-diamino-2-phenylindole (5 min, room temperature). The sections were washed three times and then mounted in FluorSave™ Reagent (Calbiochem) on microscope slides.

Immunochemical antibodies

Samples were processed with the following primary antibodies: mouse monoclonal anti-ataxin 2 antibody (clone 22, 1:2000; BD Biosciences); anti- ataxin3 (1H9, 1:5000, Millipore); anti-HA (haemagglutinin tag, 1:1000, InvivoGen); anti-neuronal nuclei (clone A60, 1:1000, Millipore); rabbit polyclonal anti-ubiquitin antibody (1:1000, Dako); anti- dopamine-and-cyclic AMP-regulated neuronal phosphoprotein (DARPP-32, 1:1000; Chemicon), anti-Calbindin (clone D-28 K, 1:1000, Chemicon), and anti- cleaved caspase 3 antibody (Asp175, 1:500; Cell Signaling). Fluorescence images were acquired with a Zeiss Axiovert 2000 imaging microscope or LSM Zeiss confocal microscope.

Immunochemistry quantitative analysis

Quantitative analysis of ataxin-2 fibroblasts expression

Quantification of Atx2 immunoreactivity in fibroblasts nucleus and cytoplasm was performed blindly by assessing optical densitometry in each area with ImageJ (NIH, USA). The outline of the nucleus and of the cell was delimitated using freehand selection tool of the program, and then optical densitometry was measured in each zone separately. The immunocytochemistry for fibroblasts of patients with Machado-Joseph disease and controls was made in the same day, and the images acquired in a Zeiss Axiovert 2000 imaging microscope using the same image settings. For each patient with Machado-Joseph disease or control subject at least 30 cells were randomly measured. Values are represented as the mean value of Atx2 optical density per in nucleus or cytoplasm.

Quantitative analysis of ataxin-2 positive cells in post-mortem brain sections

The human brain tissue analysis was made by blindly counting for each acquired field the number of Atx2-positive cells. For each sample, the immunohistochemistry was repeated twice and for each sample several images were randomly acquired. At least nine fields (×100 magnification) were acquired for each sample in a Zeiss Axiovert 2000 imaging microscope. Values are represented as the ratio between the average of Atx2-positive cells in each field and the average of Atx2-positive cells in controls.

Quantitative analysis of ataxin-2 positive cells in the lentiviral model

The quantification of the number of Atx2-positive cells was made in the lentiviral model at 4 weeks post-injection. The number of cells co-localizing Atx2 and wild-type Atx3 (left striatal hemisphere) or Atx3MUT (right striatal hemisphere) was counted in three sections per animal in ×20 objective on a Confocal LSM 710. The mean number of Atx2-positive cells was plotted relative to control condition (wild-type Atx3).

Quantitative analysis of aggregates, DARPP-32 volume and cell counts

The quantification of Atx3 and ubiquitin-positive inclusions was performed blindly by scanning eight coronal sections spread over the anterior–posterior extent of the striatum (inter-section distance: 160 µm), using a ×20 objective on a Zeiss Axiovert 2000 imaging microscope. For each animal, the total number of inclusions in the striatum was calculated as described previously (Alves et al., 2008b). The quantification of DARPP-32 (encoded by Ppp1r1b) depleted volumes was made by scanning eight coronal sections spread over the anterior–posterior extent of the striatum using an ×10 objective (inter-section distance: 160 µm). The volume was estimated as described elsewhere (Alves et al., 2008b), and data were expressed as the evaluated DARPP-32 depleted volume (mm³). The number of cleaved caspase-3-positive cells was assessed by scanning a representative section using a ×20 objective on a Zeiss Axiovert 200 imaging microscope. The number of cleaved caspase-3-positive cells in each section was automatically calculated by ImageJ (NIH, USA), as well as the total number of ubiquitin aggregates in that section. Data were expressed as the ratio of the number of cleaved caspase-3-positive cells by the number of ubiquitin aggregates. Composite images of complete aggregate-containing, cleaved caspase-3 or DARPP32-depleted regions were automatically acquired using the MozaiX function of the AxioVision software and the used for quantifications. For the transgenic Machado-Joseph disease mice, the quantification of the aggregates was performed by scanning 10 sagittal sections spread over the entire cerebellum, using a ×20 objective on a Zeiss Axiovert 200 imaging microscope. The estimated total number of aggregates per lobule (transduced or non-transduced) was assessed for each animal by calculating the number of inclusions in all sections multiplied by the number of sections sampled. Data were expressed as the total number of aggregates per lobule in the entire cerebellum. The number of Purkinje cells was assessed in the transduced and non-transduced lobules of cerebellar cortex in transgenic Machado-Joseph disease mice. Three sagittal sections of the central part of the cerebellar cortex (inter-section distance: 180 µm) were scanned using a ×20 objective a Zeiss Axiovert 2000 imaging microscope. The number of Purkinje cells (calbindin-positive) was counted for each section. Data were expressed as the total number of Purkinje cells in the sampled sections in the transduced or non-transduced lobules.

Results

Mutant ataxin-3 induces decrease in ataxin-2 levels in patients with Machado-Joseph disease and mouse models

Aiming at clarifying the role of Atx2 in Machado-Joseph disease, we first investigated whether Atx2 levels were deregulated in samples from patients with Machado-Joseph disease and models of disease. The analysis of post-mortem brain tissue from patients with Machado-Joseph disease (Fig. 1A) and age-matched controls (Fig. 1B) revealed a significant reduction in the number of Atx2-positive cells (Fig. 1C) and mRNA levels (data not shown). Moreover the analysis of the Atx2 protein levels in Machado-Joseph disease patient’s fibroblasts (Fig. 1D) revealed also a significant reduction compared to fibroblasts from healthy controls (Fig. 1E). We next analysed whether deregulation of Atx2 was also present in Machado-Joseph disease rodent models. In a cerebellar Machado-Joseph disease model, with Purkinje cell expression of a truncated form of human Atx3 with 69 repeats (Torashima et al., 2008), the levels of Atx2 protein (Fig. 1F and G) and mRNA (Fig. 1H) were significantly reduced as compared to the wild-type littermates. This downregulation in Atx2 levels was also observed in another transgenic Machado-Joseph disease mouse model based in the expression of the full-length mutant Atx3 (data not shown). Finally, analysis of a lentiviral Machado-Joseph disease model (Alves et al., 2008a) revealed a cytoplasmic co-localization between Atx2 and Atx3 in the control hemisphere expressing wild-type Atx3 (Fig. 1I). In contrast, in the striatal hemisphere expressing mutant Atx3 (Atx3 MUT) we observed that roughly half of the cells with intranuclear aggregates were almost devoid of Atx2 immunoreactivity (Fig. 1J), while all cells immunoreactive for wild-type Atx3 were also positive for Atx2 in the control hemisphere (Fig. 1K). Consistently, western blot analysis for this model (Fig. 1L) further revealed a significant reduction of Atx2 protein (Fig. 1M) and mRNA (Fig. 1N) levels in the hemisphere expressing Atx3MUT, as compared to the control hemisphere expressing wild-type Atx3. The disease progression in this model resulted in further reduction of the Atx2 levels, as at 8 weeks post-injection almost no Atx2 labelling was detected in the Atx3MUT expressing hemisphere (Supplementary Fig. 1). Altogether, these results show that Atx3MUT reduces Atx2 levels.

Figure 1

Atx2 mRNA and protein levels are reduced in Machado Joseph disease patients and animal models. (A and B) Immunohistochemistry image of Atx2 labelling in brain sections (counterstained with Cresyl violet) from the putamen region of patients with Machado-Joseph disease (B, n = 3), and age-matched controls (A, n = 3), revealing a significantly lower number of Atx2-positive cells per section in patients with Machado-Joseph disease compared to controls (C). (D) Western blot analysis of fibroblast protein lysates, depicting significantly lower levels of Atx2 in patients with Machado-Joseph disease (n = 3), compared to controls (n = 3) (E). (F) Western blot of cerebellar punches protein lysates in Machado-Joseph disease transgenic with 69 glutamines (n = 3), and wild-type controls (n = 5). The detected Atx2 levels were significantly lower in Machado-Joseph disease transgenic mice compared to wild-type controls, both at protein (G) and at mRNA (H) levels. (I and J) Confocal laser microscopy analysis depicting ataxin-3 (green), ataxin-2 (red) and DAPI (blue) in brain sections from a lentiviral striatal Machado-Joseph disease model at 4 weeks post-injection. In the hemisphere injected with Atx3-72Q (J) we observed reduced Atx2 labelling, compared with the hemisphere expressing Atx3-27Q (I), especially in the cells with mutant Atx3 aggregates (n = 3). (K) The number of positive cells co-localizing with Atx3MUT was significantly reduced to the number of Atx2-positive cells co-localizing with wild-type Atx3 in the control hemisphere (n = 3). (L) Western blot of striatal punches from the lentiviral Machado-Joseph disease model at 4 weeks post-injection, depicting lower levels of Atx2 in the hemisphere expressing Atx3-72Q compared to the control hemisphere expressing Atx3-27Q (M). The Atxn2 (Atx2) mRNA levels (n = 6) were also reduced in the hemisphere expressing Atx3-72Q, compared to the contralateral hemisphere (N). *P < 0.05, **P < 0.01, unpaired Student’s t-test.

Mutant ataxin-3 changes the subcellular localization of ataxin-2

We then investigated whether interaction of Atx3MUT with Atx2 would modify its subcellular localization as this could have important consequences on the activity of the protein. Atx2 protein normally localizes in the cytoplasm, where it associates with the Golgi complex, polyribosomes and endoplasmic reticulum (Huynh et al., 2003; Satterfield and Pallanck, 2006; van de Loo et al., 2009), which matched our observations in fibroblasts from control individuals (Fig. 2A and B). On the contrary, in Machado-Joseph disease patient’s fibroblasts (Fig. 2C and D) this subcellular localization was altered, with a significant shift of Atx2 immunoreactivity from the cytoplasm into the nucleus (Fig. 2E). Consistently, in Neuro2a (N2a) cells stably transduced with wild-type Atx3 (Fig. 2F and G) we found a cytoplasmic distribution of Atx2. In contrast, in N2a cells stably transduced with Atx3MUT, we found a subset of cells with nuclear labelling of Atx2 (Fig. 2H and I).

Figure 2

Atx2 subcellular localization in altered in Machado-Joseph disease patient’s fibroblasts and in a cellular model. (A–D) Laser microscopy analysis showing an Atx2 labelling in the nucleus of Machado-Joseph disease patient fibroblasts (n = 4), whereas in fibroblasts from control subjects, Atx2 distribution is strictly cytoplasmic (n = 3). (E) This subcellular localization alteration was found to be significant by quantification of Atx2 immunoreactivity in the cytoplasm and in the nucleus. (F–I) Confocal laser microscopy analysis depicting Atx2 (red) and dapi (blue) in N2a cells stably expressing wild-type ataxin-3 (Atx3WT, F and G) or mutant ataxin-3 (Atx3MUT, H and I). It was only detected the presence of Atx2 inclusions in the nucleus of some cells expressing Atx3MUT (white arrows). (J–M) Laser microscopy analysis of N2a cells stably expressing wild-type Atx3 (J and K) or Atx3MUT (L and M) and transfected with EGFP-Atx2. It was observed a recruitment of Atx2 to the nucleus (white arrows) in the cells expressing Atx3MUT, whereas in N2a cells expressing wild-type Atx3 the Atx2 remained always cytoplasmic. *P < 0.05, **P < 0.01, unpaired Student’s t-test.

To further investigate the role of mutant Atx3 as a driver of nuclear Atx2 localization, we transfected Atx2 with an EGFP tag in N2a cells stably expressing wild-type Atx3 (Fig. 2J and K) or Atx3MUT (Fig. 2L and M). Interestingly, we found that when cells expressed Atx3MUT the EGFP-Atx2 localized in the nucleus, whereas in cells expressing wild-type Atx2, it remained cytoplasmic. Altogether, these data show that in Machado-Joseph disease the subcellular localization of Atx2 is altered into the nucleus, and that Atx3MUT is driving this process.

Reinstating ataxin-2 levels decreases mutant ataxin-3 levels and aggregates

Taking into account the observed reduction of Atx2 levels in Machado-Joseph disease, we next investigated whether reinstating those levels would reduce Atx3MUT levels and aggregates in cellular (Supplementary Fig. 2) and rodent models of Machado-Joseph disease (Supplementary Fig. 3). In N2a cells stably expressing Atx3MUT we found that Atx2 expression (Fig. 3A) led to a significant reduction in Atx3MUT levels, both aggregated (Fig. 3B) and soluble (Fig. 3C), as compared to a control situation. This modulation of Atx3 levels was driven by Atx2, as the expression of ataxin-2-like protein, which is a paralogue of Atx2, with some functional overlap (Kaehler et al., 2012) did not produce any effect in Atx3MUT levels (Supplementary Fig. 4).

Figure 3

Restoring Atx2 levels reduces mutant Atx3 levels and the number of intranuclear inclusions. (A) Western blot for protein lysates from N2a cells expressing Atx3MUT or co-expressing Atx3MUT and Atx2 (n = 3). The restoration of Atx2 depleted levels led to a significant decrease in the levels of Atx3MUT aggregates (B), and soluble Atx3MUT levels (C). (D) Western blot of striatal punches from the lentiviral model (n = 3), in which the co-expression of Atx3MUT and Atx2 in one hemisphere resulted in lower levels of aggregates (E) and soluble (F) Atx3MUT protein, compared to the control hemisphere expressing Atx3MUT. (G and H) Immunohistochemistry image analysis of ubiquitin labelling in sections from mice injected in striatum (n = 5) with lentiviral vectors encoding Atx3MUT (G), and in the contralateral hemisphere co-injected Atx3MUT and Atx2 (H). (I) On Atx2 expression, the number of ubiquitin-positive aggregates was significantly decreased. *P < 0.05, **P < 0.01, ***P < 0.001, unpaired Student’s t-test.

Based on these results we then analysed whether restoration of Atx2 levels would modulate Atx3MUT levels in a lentiviral Machado-Joseph disease mouse model. Briefly, we co-injected Atx3MUT and Atx2 in one striatal hemisphere, and as control the contralateral hemisphere was injected with Atx3MUT. At 4 weeks post-injection, the analysis of protein levels (Fig. 3D) revealed that Atx2 expression led to a significant decrease in the Atx3MUT aggregates (Fig. 3E), and to a tendency of reduction in soluble Atx3MUT levels (Fig. 3F).

An important hallmark of Machado-Joseph disease is the presence of ubiquitinated intranuclear aggregates of Atx3MUT in the affected areas (Paulson et al., 1997; Schmidt et al., 1998; Rub et al., 2008). Thus, we further analysed the effects of Atx2 expression in the lentiviral Machado-Joseph disease mouse model by counting the number of ubiquitin-positive inclusions upon immunohistochemical staining. At 4 weeks post-injection, we found that compared to the control hemisphere expressing only Atx3MUT, the co-expression of Atx2 led a significant decrease in the number of ubiquitin-positive inclusions (Fig. 3G–I). Altogether, these results suggest that Atx2 reduces Atx3MUT protein levels and aggregation.

Ataxin-2 attenuates mutant ataxin-3 induced neuronal dysfunction and degeneration

We next analysed whether restoration of Atx2 levels could also mitigate neuronal dysfunction and degeneration in models of Machado-Joseph disease. In N2a cells expressing Atx3MUT (Fig.4A) the expression of Atx2 led to a significant reduction in cleaved caspase-3 levels compared to the control cells expressing Atx3MUT (Fig. 4B). Moreover, the expression of Atx2 significantly rescued cell viability, impaired by Atx3MUT expression (Fig. 4C). Using apoptosis detection probes for flow cytometry we found that Atx2 expression mediated a significant increase in viable cells (Fig. 4D), and reduction in the percentage of dead or apoptotic cells (Fig. 4E), compared to a control situation.

Figure 4

Restoring Atx2 levels alleviates mutant Atx3 induced toxicity and neurodegeneration. (A) Western blot for protein lysates from N2a cells expressing Atx3MUT or co-expressing Atx3MUT and Atx2 (n = 3). (B) The restoration of Atx2 depleted levels led to a significant decrease in the cleaved caspase-3 levels. (C) The N2a cell viability measured by alamarBlue® assay revealed a significant decrease in viability upon Atx3MUT expression, which was significantly mitigated by Atx2 co-expression (n = 4). (D and E) The N2a cell viability assessment by flow cytometry analysis using an apoptosis detection kit revealed an increased number of viable cells (D) and decreased number of apoptotic or dead cells (E) on Atx3MUT and Atx2 co-expression, compared to control situation (n = 3). (F and G) Immunohistochemistry image analysis of DARPP-32 labelling in sections from mice injected in striatum (n = 5) with lentiviral vectors encoding Atx3MUT (F), and in the contralateral hemisphere co-injected Atx3MUT and Atx2 (G). (H) On Atx2 expression the total volume of neuronal loss was significantly reduced. (I–L) Laser microscopy analysis of immunohistochemistry labelling of ubiquitin (I and J) and cleaved caspase-3 (K and L) in sections from mice injected in striatum (n = 5) with lentiviral vectors encoding Atx3MUT (I and K), and in the contralateral hemisphere co-injected Atx3MUT and Atx2 (J and L). (M) The co-expression of Atx2 with mutant Atx3 led to a significant decrease in the number of cleaved caspase-3 positive cells per section, compared to the control hemisphere-expressing mutant Atx3. *P < 0.05, ***P < 0.001, unpaired Student’s t-test; *P < 0.05, **P < 0.01, one-way ANOVA.

We then analysed if Atx2 expression in vivo could mitigate the local depletion of the neuronal DARPP-32 marker induced by the expression of Atx3MUT (Fig. 4F). The injection of lentiviral Atx2 in the striatum did not cause any toxicity, as no loss of DARPP-32 staining was observed on Atx2 expression (Supplementary Fig. 5). Importantly, the co-expression of Atx2 significantly reduced the loss of DARPP-32 immunoreactivity induced by Atx3MUT, when compared to the control hemisphere (Fig. 4G and H). The expression of Atx3MUT in the striatum leads to a positive labelling for cleaved caspase-3 in the cells with ubiquitin aggregates (Fig. 4I and K). In line with previous findings, expression of Atx2 led to a significant reduction in cleaved caspase-3 positive cells (Fig. 4J, L and M). Overall these results indicate that restoration of Atx2 levels in vivo reduces toxicity and neurodegeneration associated with the expression of Atx3MUT.

Ataxin-2 alleviates motor and cerebellar deficits of a Machado-Joseph disease transgenic mouse model

To investigate whether Atx2 expression would alleviate motor and cerebellar deficits after the onset of the disease, we used the transgenic Machado-Joseph disease model (Torashima et al., 2008) in which we found reduced Atx2 levels (Fig. 1). The animals were injected at 3 weeks of age with lentivirus encoding Atx2 (LV-Atx2), or GFP as control (Supplementary Fig. 6), as previously described (Nóbrega et al., 2013b). At the time of the injection these transgenic mice already display a marked phenotype characterized by an ataxic movement, with difficulty to walk and to equilibrate when compared to wild-type mice. Moreover, at 3 weeks of age these transgenic mice, compared with wild-type littermates, already display a disorganization of Purkinje cell layer, resulting from the expression of Atx3MUT (Supplementary Fig. 7). On injection, lentiviral-mediated expression was found at cerebellar lobules I–II and IX–X, efficiently transducing molecular layer neurons, Purkinje cells and also granular neurons (Supplementary Fig. 8). We monitored motor behaviour every 2 weeks post-injection by analysing rotarod performance and footprint patterns in both groups of injected animals. The latency to fall in the rotarod was similar between both groups until 4 weeks post-injection. At 6 weeks post-injection the animals injected with LV-Atx2 were able to equilibrate for a drastically longer period in the rotarod than control animals GFP-injected, either at constant velocity (Fig. 5A) or with increased velocity (Fig. 5B). The improved performance of Atx2-injected animals was maintained until the time of sacrifice by 8 weeks post-injection. This improvement in motor function was further investigated by the quantitative analysis of the footprint patterns. Accordingly, from 4 weeks post-injection we detected a longer stride length in Atx2-injected animals compared to controls, which further increased and became statistically significant at 6 and 8 weeks post-injection (Fig. 5C). Moreover, these animals also presented a smaller footprint overlap than GFP-injected mice, which was significant at 6 weeks post-injection (Fig. 5D).

Figure 5

Restoring Atx2 levels mitigates motor and cerebellar deficits in a Machado-Joseph disease transgenic model. (A and B) Rotarod performance of the Machado-Joseph disease transgenic mice injected with lentiviral vectors encoding Atx2 (n = 5) or GFP as control (n = 7). The restoration of Atx2 levels lentiviral-mediated led to a significant improved performance in the rotarod from 6 weeks post-injection, either a constant (A) or at an increasing velocity (B), compared to controls. (C and D) Footprint patterns of transgenic injected mice. The animals Atx2-injected displayed better measures of footprints patterns, namely stride length (C) and (D) footprint overlap from 4 weeks post-injection, compared to GFP-injected controls. (E and F) Confocal microscopy image analysis of haemagglutinin (HA) labelling for Atx2-injected mice (n = 5) (E) and GFP-injected controls (n = 3) (F), in which is depicted a significant reduction in the number of intranuclear inclusions on Atx2 expression in the cerebellum (G). (H and I) Confocal microscopy image analysis of calbindin labelling for Atx2-injected mice (H) and GFP-injected controls (I), in which is depicted a significant increase in the number of Purkinje cells upon Atx2 expression (J). *P < 0.05, **P < 0.01, two-way ANOVA; *P < 0.05, unpaired Student’s t-test.

Next we asked if this improvement in motor behaviour mediated by Atx2 correlated with Atx3 aggregation and cerebellar neuropathology, which is marked and severe in this transgenic Machado-Joseph disease model. In accordance with the behavioural data, the total number of nuclear inclusions (HA-tagged) in the transduced cerebellar lobules revealed a significant and robust decrease in the animals injected with Atx2 compared to the controls (Fig. 5E–G).

This Machado-Joseph disease transgenic model is characterized by degeneration in Purkinje cells, which we could observe in control animals (GFP-injected) (Fig. 5H). Interestingly, in the Atx2-injected mice there was a significant preservation of those cells (Fig. 5I and J). To verify if these differences resulted from Atx2 expression we also analysed the number of intranuclear aggregates and the number of Purkinje cells in non-transduced lobules. No significant differences were found in the lobule VI, between the two different injected groups (Supplementary Fig. 9). Altogether these data show that restoration of Atx2 levels mediates a robust and significant rescue of motor phenotype deficits and neuropathological abnormalities in a severe-progressive Machado-Joseph disease transgenic model.

Ataxin-2 acts as a translation regulator of mutant ataxin-3 through interaction with PABP

It has been reported that Atx2 could regulate the translation of transcripts, with or without interaction with PABP (McCann et al., 2011; Yokoshi et al., 2014). PABP is required for poly(A) shortening and translation initiation in polyribosomes (Ralser et al., 2005), and its depletion reduces dramatically the rates of translation (Kahvejian et al., 2005). To investigate whether Atx2 is modulating Machado-Joseph disease neuropathology through interaction with PABP we: (i) overexpressed PABP; (ii) overexpressed a mutated Atx2 without the capacity to interact with PABP; and (iii) used a translation inhibitor to further analyse whether ataxin-2 acts as a translational inhibitor.

We found that in N2a cells expressing Atx3MUT, the overexpression of PABP leads to a significant increase in Atx3MUT levels (Fig. 6A and B). Importantly, in a Machado-Joseph disease lentiviral rodent model the co-expression of Atx3MUT and PABP (Fig. 6D) led to a significant increase in the number of intranuclear aggregates, compared to the hemisphere-expressing mutant Atx3 (Fig. 6C and E).

Figure 6

The Atx2-PABP interaction regulate translation of mutant Atx3 and its levels. (A) Western blot for protein lysates from N2a cells expressing Atx3MUT or co-expressing Atx3MUT and PABP (n = 5). (B) The expression of PABP led to an increase in the Atx3MUT levels. (C and D) Immunohistochemistry image analysis of 1H9 (ataxin-3) labelling, with striatal injection of Atx3MUT in one hemisphere (C), and in the contralateral hemisphere co-injection of Atx3MUT and PABP (D) (n = 3). (E) The total number of Atx3MUT intranuclear inclusions was significantly higher upon PABP expression. (F) Western blot for protein lysates from N2a cells expressing Atx3MUT, co-expressing Atx3MUT and Atx2 or co-expressing Atx3MUT and Atx2ΔPam2 (with mutation in PAM2 motif). (G) The levels of Atx3MUT are not altered upon expression of Atx2ΔPam2, which abolish the Atx2-PABP interaction, contrary to the significant reduction in those levels upon normal Atx2 expression (n = 6). (H) Western blot for protein lysates from N2a cells or N2a cells transduced with Atx3MUT for puromycin as measure for the translation rate. (I) Upon Atx2 overexpression it was observed an overall reduction in protein synthesis, similar to the reduction observed using a translation inhibitor (CHX = cycloheximide) (n = 4), whereas no difference was found upon GFP overexpression. *P < 0.05 unpaired Student’s t-test.

The Atx2–PABP interaction has previously been reported to be mediated through the PAM2 motif of Atx2 (Lessing and Bonini, 2008; Kozlov et al., 2010), and could prevent translation by inhibiting the PABP-eIF4F complex formation and mRNA circularization (Satterfield and Pallanck, 2006; Wang and Proud, 2006; Eliseeva et al., 2013; Magana et al., 2013). Thus, in the light of the previous results, the inhibition of this interaction might abolish the ability of Atx2 to reduce Atx3MUT levels. To test this hypothesis, we blocked the interaction of ataxin-2 with PABP by mutating the PAM2 motif of Atx2 through change of the 13th amino acid from a phenylalanine to alanine (Supplementary Fig. 10), as previously reported (Huntzinger et al., 2010; Mishima et al., 2012). In N2a cells expressing Atx3MUT (Fig. 6F) the transfection of Atx2 carrying the mutated PAM2 motif (Atx2ΔPam2), was unable to induce alterations in Atx3MUT levels, as compared to the control situation and contrasting to the significant reduction upon normal Atx2 expression (Fig. 6G). These results suggest that Atx2 may reduce the levels of Atx3MUT through a regulation of translation, which could also lead to a reduction in overall protein synthesis.

To further validate this hypothesis, we used a puromycin-based assay to investigate whether Atx2 is inhibiting global protein translation. Strikingly, in N2a expressing Atx3MUT we observed that on Atx2 transfection overall protein translation was reduced (Fig. 6H), as compared to a strong labelling of puromycin in non-transfected N2a cells expressing Atx3MUT (or normal N2a cells). The reduction in overall protein translation induced by Atx2 is comparable to the one observed using cyclohemixide (CHX), which is a known inhibitor of translation (Fig 6I). This reduction in protein synthesis was specific for Atx2, as GFP transfection did not alter the translation rate. Altogether, these data indicate that Atx2 interacts through its PAM2 motif with PABP to reduce Atx3 translation.

Discussion

In this study we report that Atx2 levels are reduced in patients with Machado-Joseph disease and animal models, and it subcellular localization is altered towards the nucleus. Atx3MUT aggregation in the nucleus is associated with significant decrease in Atx2 mRNA and proteins levels and recruitment of Atx2 to aggregates, further decreasing the cytoplasmic levels of the protein. This decrease in Atx2 levels may then lead to a deregulation in translation, by freeing PABP to an overactive protein translation. Accordingly, we show that overexpression of PABP leads to an increase in number of Atx3 nuclear aggregates, which may contribute decisively to the pathology. Importantly, we found that restoration of Atx2 levels reduced Atx3 levels and rescued neuropathological Machado-Joseph disease-related abnormalities, an effect that is lost upon mutation of the PABP interacting motif (PAM) within Atx2. This indicates a clear function of Atx2 in the regulation of the translation of specific mRNAs, particularly Atx3. Altogether these data point to a key physiological role of Atx2 in Machado-Joseph disease pathogenesis.

In polyQ disorders several mechanisms can simultaneously contribute to neurodegeneration and to the disease progression. The gain of toxic function of the mutated protein is one of such mechanisms. To this contributes the expanded polyQ tract, which leads to the formation of aggregates and interacts and sequesters several proteins with important cellular functions, especially of the RNA metabolism. Thus, the combination of several impairments due to the mutant protein may lead to a cascade of events culminating in the neurodegeneration and cell death observed in polyQ disorders. Therefore, it is important to identify new molecular targets that could help to develop new therapeutic approaches for polyQ disorders.

In the present work we found that in Machado-Joseph disease, there is an important reduction in Atx2 mRNA and protein levels, while simultaneously the accumulation of Atx3MUT drives Atx2 from the cytoplasm into a nuclear localization, in accordance with previous reports (Uchihara et al., 2001). These two findings suggest a possible deregulation of Atxn2 transcription due to the presence of the expanded form of Atx3 in the nucleus. In fact, the accumulation of expanded polyQ proteins in the nucleus has been proposed to affect gene expression by deregulating transcription (Bauer and Nukina, 2009). In Machado-Joseph disease this could be even more relevant as Atx3 function was linked to regulation of transcription influencing the expression of many genes (Li et al., 2002; Evert et al., 2003; Rodrigues et al., 2007; Matos et al., 2011). As Atx2 has previously been shown to have a relevant function in mRNA translation (Eliseeva et al., 2013; Magana et al., 2013; Dansithong et al., 2015; Fittschen et al., 2015), this depletion was suggestive of a new pathway to Machado-Joseph disease pathogenesis, which we studied in this work.

We therefore investigated if restoration of depleted Atx2 levels in Machado-Joseph disease was able to reduce or abolish the neuropathology associated with the expression of Atx3MUT. It has previously been shown that Atx2 modulates the levels of several components in the endocytosis complex, and other transcripts (Drost et al., 2013; Dansithong et al., 2015; Fittschen et al., 2015). The Atx2 modulation of specific proteins could be regulated through: (i) interaction with PABP (McCann et al., 2011); or (ii) by direct interaction with mRNA, without PABP intervention (Yokoshi et al., 2014). Strikingly, on Atx2 lentiviral-mediated expression we found that the levels of aggregates and soluble species of Atx3MUT were reduced, but not for other proteins such as actin, tubulin or GFP. Moreover, this was somehow a specific effect of Atx2, as we found that ataxin-2-like protein did not alter the levels of Atx3MUT.

The local expression of Atx3MUT mediated by lentiviral vectors is characterized by nuclear aggregates formation and neurodegeneration (Alves et al., 2008a; Nóbrega et al., 2013a). In a presymptomatic striatal mouse model of Machado-Joseph disease, already extensively used (Alves et al., 2008b, 2010; Nascimento-Ferreira et al., 2011; Simões et al., 2012; Gonçalves et al., 2013) we show that Atx2 expression leads to a decrease in the number of aggregates, which are a hallmark of Machado-Joseph disease. We also found that restoration of Atx2 levels through lentiviral vectors transduction leads to the rescue of Atx3MUT induced toxicity both in N2a cells and in a Machado-Joseph disease mouse model. In this mouse model, Atx2 led to a reduction in the volume of neuronal loss and in the number of cleaved caspase-3-positive cells, strengthening the neuroprotective role of Atx2. Moreover, we did not find any toxicity in the Atx2 expression in vivo, as it was reported in Drosophila studies (Al-Ramahi et al., 2007; Lessing and Bonini, 2008), but in line with the absence of toxicity of wild-type Atx3 expression in the striatum and cerebellum (Alves et al., 2008a; Nóbrega et al., 2013a).

We next evaluated the restoration of Atx2 levels after the onset of the disease, by lentiviral-mediated expression in a transgenic mouse model of Machado-Joseph disease. This model is characterized by an expression of a truncated Atx3 form with 69 CAG repeats in cerebellar Purkinje cells (Torashima et al., 2008), in which we already tested other therapeutic strategies (Nascimento-Ferreira et al., 2013; Nóbrega et al., 2013b). Atx2 is widely expressed in the brain, particularly in Purkinje cells (Nechiporuk et al., 1998), and therefore its downregulation may have specific toxic effects in these cells. In fact, the restoration of Atx2 depleted levels in the cerebellum of these mice led to an improvement in motor performance compared to GFP-injected control animals. Moreover, analysis of the transduced cerebellar lobules revealed a significant decrease in the number of aggregates and a preservation of Purkinje cells on Atx2 expression. These data suggest that targeting Atx2 could help to mitigate the neuropathological and behavioural deficits associated to Machado-Joseph disease, after the onset of the disease.

Based on these results and building on a previously proposed model (Satterfield and Pallanck, 2006) we suggest a mechanistic explanation, in which Atx3MUT reduces Atx2 levels, freeing PABP to an overactive translation state, which contributes decisively to the pathology. Thus, restoring depleted Atx2 levels would be expected to allow normal interaction with PABP, therefore inhibiting translation, and reducing the Atx3MUT levels. Accordingly, it was shown that Atx2 and PABP interact together (Mangus et al., 1998; Ciosk et al., 2004; Satterfield and Pallanck, 2006; van de Loo et al., 2009; Zhang et al., 2013), and that PABP is required for poly(A) shortening and translation initiation in polyribosomes (Kahvejian et al., 2005; Ralser et al., 2005). Moreover, there is evidence that Atx2 regulates the intracellular concentration of PABP, and endogenous PABP levels are increased in the presence of low intracellular Atx2 levels (Nonhoff et al., 2007). In fact, we observed that in opposition to the effects of Atx2, the overexpression of PABP increased the levels of Atx3MUT and also the number of intranuclear aggregates, suggesting a PABP-mediated Atx3 translation. Importantly, the prevention of Atx2-PABP interaction through PAM2 mutation blocked the Atx2 regulation of the levels of Atx3MUT. Altogether and in line with previous studies, our results are in accordance with several studies providing evidence that Atx2 may function in the translational regulation of specific transcripts, by interaction with PABP (Liu-Yesucevitz et al., 2011; McCann et al., 2011; Drost et al., 2013; Zhang et al., 2013; Sudhakaran et al., 2014; Dansithong et al., 2015). Nevertheless, the role of Atx2 in microRNA stability (McCann et al., 2011), in the direct interaction with transcripts (Yokoshi et al., 2014) or the interaction with other proteins of RNA metabolism (Lim et al., 2006) could also be important to the observations here reported, as well as the overall decrease in protein synthesis detected upon Atx2 expression. Thus, further studies should be performed in order to fully understand the molecular mechanisms of Atx2–Atx3 interaction.

Overall the data of this work provide evidence that Atx2 acts as a negative translational regulator reducing the neuropathology and motor deficits in Machado-Joseph disease. It also suggests that regulation of Atx3MUT translation might constitute a possible therapeutic approach to Machado-Joseph disease and might be also explored in other polyQ disorders or in combination with other therapies. However, important questions should be addressed in future works to validate Atx2 as a therapeutic target for polyQ disorders, such as long-term expression safety, the effect of Atx2 polyQ expansions and analysis of Atx2 levels in other polyQ disorders.

Acknowledgements

Authors wish to thank Dr Cláudia Cavadas for the critical reading of this manuscript, Isabel Onofre for the fibroblast samples, and Dr Rui Nobre for control brain sections.

Funding

This work was supported by the Portuguese Foundation for Science and Technology (FCT), E-Rare4/0003/2012 Joint call for European Research Project on Rare Diseases, the French Muscular Dystrophy Association (AFM), the National Ataxia Foundation (NAF), and the Richard Chin and Lily Lock Machado-Joseph disease Research Fund. Project was co-funded by FEDER (QREN), through Programa Mais Centro under projects CENTRO-07-ST24-FEDER-002002, CENTRO-07-ST24-FEDER-002006 and CENTRO-07-ST24-FEDER-002008, and through Programa Operacional Factores de Competitividade - COMPETE and National funds via FCT – Fundação para a Ciência e a Tecnologia under project(s) Pest-C/SAU/LA0001/2013-2014. C.N., S.C.M., D.A., A.V.F., and L.M. have grants from FCT.

Supplementary material

Supplementary material is available at Brain online.

Abbreviations

Abbreviations
 
• Atx2/3

  ataxin-2/3

 
• atx3MUT

  polyQ-expanded ataxin-3

 
• PABP

  poly(A) binding protein

 
• polyQ

  polyglutamine