Background
Spinocerebellar ataxia type 7 (SCA7) is an inherited autosomal dominant neurodegenerative disorder. Patients present with cerebellar ataxia due to moderate to severe neuronal loss and gliosis in the cerebellum, especially Purkinje cells, inferior olivary, dentate nucleus and pontine nuclei, and to a lesser extent in the
globus pallidus,
substantia nigra and
red nucleus. They also present with visual impairment due to degeneration of cone and rod photoreceptors [
1‐
3]. SCA7 is caused by an unstable CAG repeat expansion in the coding region of the
SCA7 gene conferring a toxic gain of function to the ataxin-7 (ATXN7) protein which accumulates aberrantly in neurons, a mechanism also involved in a family of eight other inherited neurodegenerative polyglutamine (PolyQ) diseases, including Huntington’s disease (HD), spinobulbar muscular atrophy (SBMA), dentatorubral pallidoluysian atrophy (DRPLA), spinocerebellar ataxia (SCA) types 1, 2, 3, 6 and 17 [
4]. ATXN7 is ubiquitously expressed in the brain and is a component of the highly conserved transcriptional coactivator Spt/Ada/Gcn5 acetylase (SAGA) chromatin remodelling complex with histone acetyltransferase activity and deubiquitinase activity [
5]. It has been shown that the ubiquitin protease activity of SAGA is important for the expression of tissue-specific and developmental genes [
5]. Recently, it was shown that SAGA acetylates the promoters and deubiquitinates the transcribed regions of all expressed genes [
6]. ATXN7 has been described to be cleaved by caspase-7 at two sites [
7], generating N-terminal fragments containing the polyQ tract, resulting in MUT ATXN7 fragments that accumulate in the nucleus. Indeed, a ∼ 55 kDa ATXN7 amino-terminal fragment was previously identified in SCA7 transgenic mice and in SCA7 patients [
8]. Interestingly, it has been reported that post-translational modifications at lysine 257, adjacent to the caspase-7 mediated cleavage site of ATXN7 at position 266, mitigate fragment accumulation in vitro and in vivo, thus regulating SCA7 toxicity [
9,
10]. ATXN7 expanded polyQ stretches result in conformational modifications, finally leading to the formation of insoluble aggregates, hallmarks of SCA7 [
11]. The exact mechanism by which polyQ aggregates mediate toxicity is still debated, but one strong hypothesis is the fact that they may be prone to trap multiple binding partners such as transcription factors, important to the maintenance of cell homeostasis, that will in turn be progressively depleted [
3], or RNA-binding proteins (RBPs), leading to dysregulation of alternative splicing of target mRNAs [
12,
13]. The generation of murine genetic models that closely recapitulate the human neuropathology are extremely valuable for the dissection of disease mechanisms and evaluation of therapeutic strategies. In the case of SCA7, the cloning of the
SCA7 gene allowed the creation of transgenic and knock-in mouse models, in which cerebellar neuronal dysfunction and progressive retinal degeneration were directly associated to the accumulation of mutant ATXN7 [
8,
14‐
16], despite poor neuronal degeneration [
8,
15]. Alternatively, local overexpression of mutant proteins using viral vectors has been a successful strategy to model polyQ pathologies of the central nervous system (CNS), such as HD [
17] and SCA3 [
18], generating robust in vivo genetic models leading to neuronal degeneration in well-defined brain regions.
Here, we generated an in vivo model of SCA7 by overexpressing truncated MUT ATXN7 in the mouse cerebellum using a locally injected lentiviral vector (LV). The truncated construct we used corresponds approximately to the caspase-7 cleavage fragment [
7]. The SCA7-LV mice developed an ataxic phenotype and this model further allowed investigating whether specific RBPs could be involved in the pathogenesis of SCA7. We initially focused on the RBP
Fused in sarcoma (FUS/TLS), found to be mutated in familial amyotrophic lateral sclerosis (ALS) [
19], since it was shown to be a major component of nuclear aggregates in several polyQ disorders, such as HD, SCA1 and SCA3 [
20,
21]. We next looked for the transactive response DNA binding protein 43-kDa (TDP-43), pathologically associated to ALS and frontotemporal lobar degeneration with ubiquitinated inclusions (FTLD-U) [
22] that was also shown to be sequestered in polyQ aggregates [
23]. Finally, we investigated in this SCA7-LV model, the cellular localization and expression of two evolutionarily conserved RBPs that constitute part of the muscleblind-like protein family (MBNL1 and MBNL2) and are expressed in a wide variety of adult tissues including brain, heart and skeletal muscle [
24]. MBNL1 and MBNL2 associate and bind to expanded CUG and CAG repeats, which accumulate as discrete nuclear foci in both DM1 and DM2 (myotonic dystrophy type 1 and type 2) [
24‐
26], therefore suggesting their implication in these disorders.
We therefore investigated in the LV-based model of SCA7 the impact of MUT ATXN7 accumulation over the expression of the RBPs referred above. This was complemented with immunohistochemical and biochemical analyses performed in the cerebellum of a SCA7 knock-in mouse model [
16] and in cerebella from SCA7 affected patients. In summary, we set up a new LV-based model of SCA7, alternative to transgenesis, in which specific RBPs were found to be accumulated, suggesting their role in SCA7 pathogenesis.
Discussion
When expressed in the cerebellum of transgenic/knock-in mice, mutant ATXN7 was reported to cause severe ataxic symptoms, but no significant cell loss [
8,
15]. The local overexpression of mutant proteins using viral vectors has been successfully used to model neurodegenerative diseases [
17,
18,
31‐
33]. In the present study, we demonstrated that LV-mediated delivery of mutant ATXN7 provided an alternative animal model of SCA7, with the development of an earlier, more robust and aggressive behavioral /histopathological phenotype than that observed in other SCA7 models [
8,
14,
34]. It is, therefore, a time- and cost-effective animal model of SCA7 that partially avoids the limitations previously encountered, such as the slow development of the disease and the absence of neurodegeneration.
Remarkably, truncated fragments were shown to accumulate in SCA7 transgenic models, even when they express the full-length expanded form of ATXN7. An ∼ 55 kDa ATXN7 amino-terminal fragment was identified in both SCA7 transgenic mice and SCA7 patients [
8]. Importantly, it has also been observed, in in vitro studies on SBMA, that full-length polyQ proteins aggregate, but at a much slower rate than their proteolytic fragments [
35]. Protein proteolytic cleavage mediated by caspases produces small, polyQ-containing fragments with increased cellular toxicity [
36]. The products of proteolytic cleavage are often found in aggregates, hallmarks of polyQ diseases, observed in both in vitro and in vivo models [
37] and also in patients post-mortem tissue [
38]. Therefore, we overexpressed, locally in the mouse cerebellum, an ATXN7 fragment similar in size to the cleaved fragments previously reported [
8,
14,
34]. Indeed, the local LV-mediated overexpression of an ATXN7 fragment with an expanded polyQ repeat induced a cascade of events with progressive cell loss and a severe neuropathological phenotype. Furthermore, the local overexpression of mutant ATXN7 allowed us to dissect the specific contribution of SCA7-induced cerebellar pathophysiology, since other brain regions, such as the brainstem and
substantia nigra, were also reported to be affected in this pathology [
28]. This model may therefore be of use not only to dissect molecular mechanisms of SCA7 pathogenesis, but also to investigate in vivo new therapeutic strategies acting on cell degeneration and behavioral abnormalities. Importantly, SCA7 overexpression could be investigated at different time-points and SCA7 severity could be modulated by varying the dose of LV to be injected, as previously described in a LV-based rat model of SCA3 [
17,
18].
Our first results showed that intracerebellar injection of LV encoding either wild-type or mutant truncated human ATXN7 resulted in strong and widespread transgene expression in the cerebellum. However, wild-type truncated ATXN7 was diffusely distributed throughout the Purkinje and granule cell nuclei, whereas mutant ATXN7 progressively accumulated in dense neuronal intranuclear aggregates and was depleted from the cytoplasm, as in previous SCA7 transgenic/knock-in models [
8,
15]. Importantly, nuclear localization of mutated polyQ protein was proposed to be critical for the initiation of neuronal death in rodent SCA1 and SCA3 models [
39,
40], suggesting that the nuclear environment of ATXN7 might be important for disease progression. Our study is in agreement with previous studies, as only transgenic nuclear mutant ATXN7 leads to PC loss, reduced thickness of the granule cell and molecular layers, 12 weeks post-injection, and co-aggregation with RBPs, whereas transgenic wild-type ATXN7, although nuclear, did not affect PC or other cerebellar layers.
In the present LV-based model of SCA7, the initial step in ATXN7 aggregation was visible 2 weeks post-injection in still well preserved PCs and GCL whereas, at the late stage of the disease, the mutant protein was completely aggregated in GCL but not in PCs, suggesting that: (i) the progressive accumulation of insoluble ATXN7 leads to time-dependent neuronal demise; (ii) the dynamics and toxicity of aggregate formation may vary considerably in different cell types from the same brain region; (iii) PCs appear to be more vulnerable than cells from the GCL, thus reproducing general features of human SCA7 pathophysiology [
28]. Strong ATXN7 immunoreactivity is found widely throughout the GCL but not in PCs in SCA7 patients, whereas nuclear inclusions are infrequent, probably due to privileged PC loss [
28]. Indeed, PCs can be injured by slight insults in comparison with other cells [
41], and functional deficits of these cerebellar nerve cells, in particular, calbindin loss, are directly associated with compromised control of motor coordination [
42,
43], such as in mice infected with mutant ATXN7. In this LV-SCA7 model we could also highlight the loss of synaptic markers, which has not been previously demonstrated in animal models of SCA7, as well as the increased immunoreactivity found in astrocytic and glial cells that may be interpreted as associated to pathogenesis and/or an adaptative immune response against excitotoxicity induced by mutant ATXN7.
This LV-based model of SCA7 also enabled investigation of neurodegenerative mechanisms. Indeed, the most studied mechanisms of pathogenesis were centered on the abnormal aptitude of mutant proteins to attract cellular proteins, such as ubiquitin, proteasome components, transcription factors and chaperones, in aggregates causing loss of the homeostasis by means of primary proteinopathy. As in our LV-based model of SCA7, several SCA7 rodent models [
14,
15], which replicate many features of the human condition, and brains from SCA7 patients [
11,
44] display abundant inclusions that consistently stain positively for proteasome subunits, ubiquitin, and molecular chaperones. Indeed, overexpressed mutant ATXN7 sequesters autophagy-related proteins, as previously described in a SCA7 knock-in model [
29], and molecular chaperones (data not shown), important for the maintenance of cell homeostasis, depleting them from neurons. This clearly reflects impaired protein clearance pathways.
In the last few years, it has been shown that RNA-related mechanisms may play an important role in polyQ disorders [
13]. However, information is scarce concerning events implicating specific RBPs in animal models of polyQ disorders, including SCA7. Our findings suggest a direct link between impaired protein degradation and accumulation of misfolded ATXN7 that sequesters other molecules, such as particular RBPs, depleting them [
21] and/or promoting the expression of aberrant, misregulated isoforms [
45]. Studies in in vitro models and transgenic mouse models demonstrate that expanded polyQ proteins are more toxic when translocated into the nucleus [
46], suggesting that the nucleus is a crucial site of pathogenesis in polyQ disorders.
Recently, the RBPs FUS/TLS and TDP-43 were shown to co-localize in nuclear Gems implicated in spliceosome maintenance [
47]. FUS/TLS shares several structural and functional properties with TDP-43; both are genetically related to ALS and FTLD, and are nuclear proteins with RNA and DNA binding abilities that play a role in RNA splicing- reviewed in [
48]. FUS/TLS binds strongly to SCA1, SCA2, HD and DRPLA aggregates [
20,
21,
49]. In addition, several variants of the gene have been identified as risk factors for ALS and rare forms of FTLD [
50], suggesting that FUS/TLS plays a role in neurodegenerative diseases. In our LV-based model and in SCA7
100Q/5Q knock-in mice, we observed a strong co-localization between ATXN7 and FUS/TLS in aggregates. To our knowledge, this is the first study reporting an association between ATXN7 and FUS/TLS and a decrease in FUS/TLS expression that may result from sequestration in inclusions. This suggests that an inadequate supply of this protein could result in abnormal transcription, RNA processing and transport, and potentially cause instability of dendritic spines, as observed in R6/2 HD mice [
51]. Notwithstanding, further studies are needed to elucidate the exact mechanism. Importantly, primary neurons from FUS/TLS-deficient mice have a decreased number of spines, and those remaining have a non-standard morphology [
52], indicating that FUS/TLS is important for neuronal function. Remarkably, we also show that FUS/TLS is preferentially trapped in ATXN7 inclusions compared to TDP-43. In accordance with our results, FUS/TLS co-localized with polyQ proteins in neuronal intranuclear inclusions in SCA2 whereas TDP-43 did not [
53]. TDP-43 also co-localized with huntingtin in dystrophic neurites and intracellular inclusions, but not in intranuclear inclusions [
54]. Secondary TDP-43 proteinopathies have been described in other CAG repeat disorders, such as SCA2 [
55], SCA3 [
56] and HD [
54], suggesting that these disorders might share with ALS some pathological mechanisms involving TDP-43.
In our LV model and in SCA7 patients, a few neurons stained positively for p-TDP43 pS409/410, which was confirmed to be a valuable tool for detecting abnormal TDP-43 in patients and to evaluate TDP-43 proteinopathies in animal models of neurodegenerative disorders [
57]. Furthermore, phosphorylation of aggregated TDP-43 at S409/410 is a defining hallmark of TDP-43 proteinopathies, including ALS and FTLD-TDP [
58,
59]. Phosphorylation of TDP-43 at serines 409 and 410 was recently reported to promote TDP-43 toxicity in vivo [
60]. Excitingly, in SCA7
100Q/5Q knock-in mice, we observed a hyperphosphorylated ~25-kDa species that could potentially be generated from alternative translational or splicing mechanisms, as previously suggested [
57,
61]. Further studies will be needed to better understand the contribution of TDP-43 to SCA7 pathogenesis.
Nevertheless, whether the accumulation of the RBP MBNL1 contributes to pathology or is simply an epiphenomenon, in addition to other features of polyQ pathophysiology, will need further investigations. In this LV model of SCA7, as well as in
Atxn7
100Q/5Q
KI mice and SCA7 patients, MBNL1 co-localized with ATXN7 inclusions. In addition, the specific accumulation of MBNL1 in ATXN7 inclusions was associated with an increased level of MBNL1 in both SCA7 mouse models compared to wild-type mice. In contrast, MBNL2 did not co-localize with ATXN7 in inclusions and its level remains unchanged. The consequences of MBNL1 accumulation in abnormal proteinaceous inclusions remain to be elucidated. It may be associated with neurotoxicity, although recent data also suggest that MBNL proteins might be potential modifiers of polyQ disorders, given that MBNL1 suppresses the expression of polyQ-containing proteins [
62]. The mechanism leading to the specific accumulation of MBNL1 in ATXN7 inclusions remains unknown. It would be of interest to modulate the LV-SCA7 model by overexpressing or knocking-out these RBPs to assess their impact on SCA7 pathology, in particular on aggregate formation, but also to determine how they affect neuronal markers and inflammation. To conclude, the complete elucidation of these mechanisms will be important for understanding SCA7 and related polyQ disorders and the development of potential therapeutics.
Acknowledgements
We thank Prof. H. Zoghbi (Baylor College of Medicine, Houston, Texas, USA) for providing SCA7100Q/5Q KI mice. We thank the Vector Core facility of Genethon (Evry, France), that produced the lentiviral constructs. We are very grateful to Pr. Charles Duyckaerts and the Laboratoire de Neuropathologie Escourolle-Hôpital de la Pitié Salpêtrière (Paris, France) for kindly providing post-mortem human SCA7 tissue. We would like to thank the Cellular Imaging Platform of the Pitié Salpêtrière, especially Dr. Aurélien Dauphin, for confocal imaging advice.