Altered Levels of Proteins and Phosphoproteins, in the Absence of Early Causative Transcriptional Changes, Shape the Molecular Pathogenesis in the Brain of Young Presymptomatic Ki91 SCA3/MJD Mouse

Spinocerebellar ataxia type 3 (SCA3/MJD) is a polyQ neurodegenerative disease where the presymptomatic phase of pathogenesis is unknown. Therefore, we investigated the molecular network of transcriptomic and proteomic triggers in young presymptomatic SCA3/MJD brain from Ki91 knock-in mouse. We found that transcriptional dysregulations resulting from mutant ataxin-3 are not occurring in young Ki91 mice, while old Ki91 mice and also postmitotic patient SCA3 neurons demonstrate the late transcriptomic changes. Unlike the lack of early mRNA changes, we have identified numerous early changes of total proteins and phosphoproteins in 2-month-old Ki91 mouse cortex and cerebellum. We discovered the network of processes in presymptomatic SCA3 with three main groups of disturbed processes comprising altered proteins: (I) modulation of protein levels and DNA damage (Pabpc1, Ddb1, Nedd8), (II) formation of neuronal cellular structures (Tubb3, Nefh, p-Tau), and (III) neuronal function affected by processes following perturbed cytoskeletal formation (Mt-Co3, Stx1b, p-Syn1). Phosphoproteins downregulate in the young Ki91 mouse brain and their phosphosites are associated with kinases that interact with ATXN3 such as casein kinase, Camk2, and kinases controlled by another Atxn3 interactor p21 such as Gsk3, Pka, and Cdk kinases. We conclude that the onset of SCA3 pathology occurs without altered transcript level and is characterized by changed levels of proteins responsible for termination of translation, DNA damage, spliceosome, and protein phosphorylation. This disturbs global cellular processes such as cytoskeleton and transport of vesicles and mitochondria along axons causing energy deficit and neurodegeneration also manifesting in an altered level of transcripts at later ages. Electronic supplementary material The online version of this article (10.1007/s12035-019-01643-4) contains supplementary material, which is available to authorized users.


Introduction
Spinocerebellar ataxia type 3 (SCA3), also called Machado-Joseph disease (MJD), is a dominantly inherited genetic disease resulting from the special type of mutationexpansion of CAG repeats in the ATXN3 gene [1] (MJD and ATXN3: OMIM 109150 and 607047). The presence of mutant allele evokes motor abnormalities, such as gait ataxia, ocular symptoms, and later cognitive disturbances, all characteristics for a symptomatic phase of the disease and usually occurring in the third or fourth decade of life [2]. The causative protein to these symptoms is ataxin-3, a protease/deubiquitinase [3,4], and its mutant version which contains a prolonged stretch of glutamines inside the protein structure [1]. Its natural protease function does not directly imply transcriptional regulation as a mechanism in SCA3 but rather suggests the influence of the mutant ataxin-3 on the level of various kinds of proteins [5]. A number of already identified SCA3 mechanisms also propose affected regulation of processes related to changes in protein level or pathway activation largely independent of transcription. SCA3 pathogenic processes are based on the toxicity of the polyQ tract, proteolytic cleavage of mutant ataxin-3 protein, and accumulation of intranuclear inclusions [6]. Such accumulation is related to dysfunction of proteasome and transport of the protein into the nucleus [7,8]. In addition, the SCA3 mechanism involves autophagy defects, metabolism and mitochondrial impairment, defective transport along the axons, and dysregulation of intracellular calcium turnover [9][10][11]. The mechanisms of pathogenesis in SCA3 have been previously thoroughly discussed [12,13]. Moreover, expanded polyQ stretches alone reside in the nucleus and may bind transcription factors and influence transcriptional activity. Several works also have reported the interaction of ataxin-3 with TBP, CREB-binding protein, p300, MMP-2, and HDAC3 (for review, see [12]).
The already described disease mechanisms and the late occurrence of symptoms in SCA3 patients [6] and models [14,15] and other neurodegenerative disorders underline the existence of presymptomatic disease phase which consists of molecular and cellular events important to the onset and mechanism of disease. In general, the types of molecular events which probably contribute to the presymptomatic phase and lead to disease pathogenesis are transcriptional changes and alterations in the levels of proteins and their phosphorylation state which all can influence cellular processes [6]. Unfortunately, the disease early presymptomatic phase is presently unknown, and it is also unclear what is the contribution of transcriptional and proteome changes in young SCA3 carriers. To detect potential transcriptional and protein initiators of SCA3 pathology resulting from a direct influence of ataxin-3, we used adult but presymptomatic, young, 2-month-old Ki91 knock-in mouse model homozygous for mutant Atxn3 gene. The heterozygous version of the Ki91 SCA3/MJD mouse model was already published by us [14]. The present Ki91 mouse is homozygous and contains a higher number of CAG triplet repeats in the mutant Atxn3 gene. Now, we have tested the cohort of the 2month-old homozygous Ki91 animals using several behavioral tests and found no significant motor symptoms at this early stage. Using the animals, we profiled the transcriptome by RNAseq and proteome by mass spectrometry (MS) and phosphoproteome enriched in the cerebellum and cerebral cortex where we search for protein and mRNA changes defining the onset of SCA3. In addition, we performed qPCR profiling of proteomic-based markers and cellular markers in search for late, symptomatic or cell type-dependent transcriptomic dysregulations characteristic for neurodegenerative disease. For targeted late transcriptomic profile, we used 10-14-month-old homozygous Ki91 animals. Subsequently, in the bioinformatics part, we used our dysregulated proteomic hits and performed a systematic identification of processes, pathways, subcellular localization, and discovery cell origin of the dysregulated molecules by Cytoscape/ClueGO/ Consensus Path DB and other tools. Our work prioritizes the proteomic changes in response to mutant ataxin-3 as the first molecular events occurring in the brain which lacks changes in the levels of mRNA resulting from the presence of mutant ataxin-3. We demonstrated that the transcriptome changes are secondary in SCA3 and appear later in the disease progression in 10-14-month-old Ki91 animals. Moreover, based on early proteomic changes, we discovered several groups of processes that define the onset of SCA3/MJD in Ki91 animals. Among the processes, we identified disturbed termination of translation, spliceosome phosphorylation, chromatin remodeling, protein phosphorylation, mitochondria organization, DNA damage, axon development, and transport of organelles along the axon.

Results
Two-Month-Old Ki91 SCA3/MJD Mice Show Normal Motor Performance, While 14-Month-Old Animals Show Incoordination, Decreased Body Weight, and Symptoms in Scoring Test In order to determine if there are any signs of motor incoordination in young 2-month-old Ki91 mice, we performed several tests measuring motor skills. We observed no motor deficits in 2-month-old Ki91 mice on the rotarod and parallel rod floor test (Fig. 1e, h). The scoring test of 2-month-old Ki91 mice demonstrated no signs of incoordination, gait disturbances, kyphosis, muscle weakness, or uncontrolled muscle contraction (Fig. 1g). In elevated beam walk, we evaluated two parameters: time to turn of 180°at the end of the rod and time to traverse the rod, each performed on six rods with decreasing diameter: 35, 28, 21, 17, 10 and 9 mm. We did not detect any statistically significant differences between 2month-old Ki91 and control mice, except the time to turn on the last rod diameter 9 mm (p < 0.01; Bonferroni post hoc test) (Fig. 1a, c). We attributed this result to a novelty effect, not related to the first sign of pathology since the observation was not reproduced on this rod for 14-month-old animals. Furthermore, there were no differences in body weight in young 2-month-old animals (Fig. 1f). In contrast to young animals, 14-month-old Ki91 mice demonstrated significant motor deficits in elevated beam walk test on rod diameter 28 mm in time to turn (p < 0.05; Bonferroni post hoc test; Fig. 1b) and on rod diameters 35 mm and 28 mm (p < 0.05; Bonferroni post hoc test) and rod diameters 21 and 17 mm in time to traverse (p < 0.01; Bonferroni post hoc test) (Fig. 1d). In the scoring test, old animals also demonstrated disturbed gait and coordination, kyphosis, and mild muscle contraction (p < 0.001; two-sample t test; Fig. 1k). From motor tests, only the rotarod demonstrated no alterations in 14-month-old animals (Fig. 1i). Body weight was significantly lower in 14month-old animals compared to age-matched controls (p < 0.001; two-sample t test) (Fig. 1j).
The Cerebellum and Cerebral Cortex of Young Ki91 SCA3/MJD Mouse Demonstrate the Occurrence of Cells with Accumulation of Atxn3-Positive Signal in the Nuclei One of the cellular hallmarks of SCA3 pathogenesis is the nuclear localization of ataxin-3. Therefore, we investigated ataxin-3 staining of Ki91 2-month-old mouse brain and found cells with nuclei positive for ataxin-3 stained with monoclonal 1H9 anti-ataxin-3 antibody in the white matter of the cerebellum ( Fig. 2d-f) and cerebral cortex (Fig. 2j-l). Of note, ataxin-3 accumulation co-localizes with the staining of the nucleus and was restricted to a subset of cells in both regions. Occasionally, the nuclei of young Ki91 mouse cells reveal ataxin-3-positive intranuclear aggregates (Fig. 2e, k). In turn, the localization of ataxin-3 in the cerebellum and cortex of WT animals is uniformly distributed throughout the cytoplasm and cell nucleus (Fig. 2a-c, g-i).
Transcriptional Changes Related to Mutant Ataxin-3 Do Not Occur in the Cerebellum and Cortex of Young Homozygous Ki91 SCA3/MJD Mouse RNA sequencing was performed using the cerebellum and cortex of four Ki91 mice and four control mice at the age of 2 months. Tables 1 and 2 summarize differentially expressed genes in the cerebellum and cortex, fold changes, p values, and adjusted p values. The genes significantly dysregulated versus control C57/BL/6 mice were all located within two relatively small "hot spots" on two mouse chromosomes. In the first hot spot located on chromosome 12, we identified dysregulated expression of Serpina3n, Slc38a6, Ccdc88c, Fbln5, and Ttc8, where all the dysregulated genes are located very close to the targeted Atxn3 locus. In the second hot spot on chromosome 19, we found dysregulated expression of Btaf1, Ide, and Ablim1 which are all located in close proximity in the genome. Such clustering of dysregulated genes in the hot spot loci also closely surrounding the targeted locus may indicate that the loci have different genetic origin than the rest of the Ki91 mouse genome (congenic; C57BL/6) and may stem from a 129sv background which was the source of stem cells where the Ki91 construct was introduced. We have performed a comparison of SNP signatures of Serpina3n gene using the RNAseq data on Ki91 mouse and using the MGI database on 129sv, FVB, and C57BL/6 strains. It turned out that the SNP signatures of the Serpina3n gene were the same on 129sv, FVB, and Ki91 mice and different from C57BL/6 mouse. This suggested that the region on chromosome 12 near the transgenic Atxn3 locus was segregating together with the transgene and was selected by genotyping of transgenic Atxn3 Ki91 animals. As an additional conclusion to our analysis, we would like to state that we no longer consider Serpina3n as a SCA3 transcriptional marker as stated previously in [14]. The next step was the investigation of identified transcriptomic changes using real-time qPCR. Therefore, to assess the validity of transcriptional dysregulation in Ki91 mouse and the influence of foreign 129sv mouse background on the expression of dysregulated genes, we examined the cerebellum and cortex from homozygous Ki91 mouse versus a double set of controls including the cerebellum and cortex from FVB mouse strain (same SNP signatures as 129sv) and from C57BL/6 mouse strain (Figs. 3 and 4). The transcriptional dysregulations in Ki91 mouse tissues investigated versus mRNA isolated from C57BL/6 mouse tissues again demonstrated the dysregulation of these genes. A similar analysis did not reveal any change in the level of transcripts when tissues of Ki91 mouse were investigated versus mRNA isolated from tissues of FVB mouse strain. Particularly, a pronounced difference between C57BL/6 and FVB mouse strains was detected for Serpina3n where the cerebellum and cortex from C57BL/6 mice demonstrate almost no expression of Serpina3n and the cerebellum and cortex from FVB mice demonstrated high levels of Serpina3n similar to the one found in Ki91 mice. Other dysregulations of genes in Ki91 mouse revealed by RNAseq also demonstrated dependency on FVB levels in the qPCR cross-control experiment. Detailed fold changes and p values are summarized in Table 3. The analysis confirmed that the transcriptomic dysregulation of genes in presymptomatic cerebellum and cortex of Ki91 mouse is the result of genetic background occurring in proximity of transgenic locus and in the locus spot on chromosome 19. In addition, the slightly dysregulated levels of Fgfbp3, Btaf1, and Ide may be the result of known CNV characteristic for C57BL/6 (but not FVB) and covering this region of chromosome 19 [16]. Moreover, the Btaf1 and Ablim1 expression was not altered in symptomatic 14month-old Ki91 animals ( Supplementary Fig. 1 Since the early transcriptional changes dependent on mutant ataxin-3 do not occur in young Ki91 mice, we sought to determine the transcriptional phenotype in later stages of SCA3. We also asked if the transcriptional changes in the symptomatic phase of SCA3/MJD are markers of any specific cell types in the affected cerebellum and cortex. Identification of such markers could indicate changes in the number of cells in population relevant for brain pathogenesis. Therefore, we made use of old homozygous Ki91 animals at the age of 10 and 14 months. According to our behavioral experiments, 14month-old Ki91 animals are symptomatic and demonstrate altered motor performance, altered scoring test, and decreased body weight (Fig. 1). Decreased body weight observed in Ki91 animals is one of the hallmarks of SCA3 phenotype in patients [17][18][19][20]. Based on preliminary pilot proteomic experiments in old Ki91 mice, we have selected 16 most promising dysregulated protein hits (Atp2b1, Psat1, Ppp2r1a, Idh1, Akr1b1, Srsf2, Plp1, Glul, Ca2, Ndufa9, Pea15a, Tuba1a, Psmd4, Omg, Cox7a2, and Qdpr) to examine potential transcriptional events in old Ki91 animals at 10 and 14 months old. In addition to these 16 genes, we also included gene candidates selected based on the scRNAseq database (http:// celltypes.brain-map.org/rnaseq/human) as markers of differentiating oligodendrocytes (Olig1, Olig2), mature oligodendrocytes (Mag, Omg, Cldn11, Plp1), microglia (Cd68), neurons (Reln, Npy, Sst), and markers of energy metabolism [21,22]. Table 4 summarizes gene names (total number of 25 genes), fold changes, and p values of the qPCR investigation in 10-and 14-month-old Ki91 animals. We found that the mRNA of markers of differentiating oligodendrocytes such as Olig1 and Olig2 is elevated in both the cerebellum and cortex of Ki91 mouse indicating an increase of demand for new oligodendrocytes in the SCA3/MJD brain (Fig. 5). Moreover, we have detected decreased markers of myelination such as Plp1 and Cldn11 in both the cerebellum and cortex also related to tight junctions and all being markers of mature oligodendrocytes. In addition, we detected dysregulated markers of metabolism such as Psat1, Ndufa9, Qdpr, and Pea15a (Fig. 5). There was no change in the expression of neuronal markers such as Npy, Sst, and Reln (Fig. 5). In general, we observed a higher number of dysregulaed genes in animals at the age of 14 months as compared to animals at 10 months old. Together, our observation of transcriptional changes in young and old animals indicates that observed transcriptional changes are related to cell damage and not directly to ataxin-3 effects. The fold changes of all transcriptional changes in old animals did not exceed 1.5 value with the exception of Olig1 in the cerebellum. Since the transcriptional changes first appear in older mice and after changes in the proteome (see next sections), we can conclude that changes in mRNA appear in the brain probably when the compensation of cell degeneration and loss is no longer possible and as a result of previous pathogenic processes.
Transcriptional Changes Occurring in Old Ki91 SCA3/MJD Mouse Also Occur in Neurons from SCA3/MJD Patients Since the transcriptome changes in our Ki91 mouse model occurred in the older brain, we asked if such changes would also occur in terminally differentiated SCA3 neural culture from SCA3 iPSC. Therefore, we used SCA3 iPSCderived neural cultures from two SCA3 patients and two genetically related healthy individuals [23]. Neural differentiation comprising 6 weeks therefore represented adult neural cells [24]. The mRNA isolation and qPCR revealed the changes in the genes related to neuronal precursors and oligodendrocytes (Olig1, Olig2, and Plp1) and energy metabolism (Psat1) (Fig. 6). SCA3/MJD mice. Each behavioral test consisted of one training day and three consecutive days of measurement, except of scoring test. In the elevated beam walk (a-d), two parameters "time to turn" and "traverse time" were tested on each rod (diameter of rods is indicated by Ø in mm). Two-month-old Ki91 mice demonstrated significant differences compared to C57BL/6 (C57) in "time to turn" only on the 9-mm rod on the 3rd day of testing (a) and no differences in "traverse time" (c). Fourteen-month-old Ki91 mice needed significantly more "time to turn" on the 28-mm rod on the first day of testing (b) and significantly more "time to traverse" on four rods: 35, 28, 21 and 17 mm (d). There were no significant differences for both 2-and 14month-old mice in rotarod setup which accelerated from 4 to 40 rpm in 9.5 min (e, i In the next set of experiments, we asked if any molecular presymptomatic changes exist in the proteome of young Ki91 animals. The processing of mass spectra detected 2753 unique proteins in eight samples of both the cortex and cerebellum (FDR < 0.01, at least two different peptides per protein). After filtering and selection for proteins presenting valid values (see "Methods" section), the number of proteins was reduced to 1098 and 917 proteins in the cerebral cortex and cerebellum, respectively, and these proteins were considered for further analysis in order to detect dysregulated proteins (Supplementary Table 6). Comparing Ki91 mouse samples to control mouse samples, significant differences (p < 0.05; two-sample t test) in protein levels were observed for 133 proteins in the cerebral cortex and 93 in the cerebellum (Supplementary Table 6  Moreover, the micrograph demonstrates localization of microaggregates in the nucleus of cells in Ki91 mice. The figure demonstrates micrographs of three cells per genotype and brain region; cerebellum of C57BL/6 (C57) (a-c), cerebellum of Ki91 mice (d-f) cortex of C57BL/6 (g-i), and cortex of Ki91 mice (j-l). Each cell is presented as green and blue fluorescent channel in addition to micrograph with merged channels cerebellum, the vast majority of proteins were upregulated (77) and 16 proteins were downregulated in Ki91 mice compared to controls, whereas in the cortex, there were slightly more proteins which were upregulated (76) as compared to downregulated (57) (Supplementary Table 6). The heatmap constructed after filtering of dysregulated proteins demonstrates distinct clustering of the datasets ( Supplementary  Fig. 2). Interestingly, 14 dysregulated proteins were identified both in the cerebellum and cerebral cortex (Hist1h1d, Trim28, Tubb3, Tubb4a, Tubb5, Rac1, Ran, Arpc3, Uba1, Syngr3, Gnai2, Rap1gds1, Kpnb1, Gcsh). The lists of all upregulated and downregulated proteins were subjected to separate analysis of GO terms (p value cutoff < 0.001) and pathways (at least 10% of dysregulated hits) in CPDB (Tables 5 and 6). In the cerebellum, downregulated proteins are associated mainly with apoptosis, whereas upregulated proteins are involved in metabolism, in particular the citric acid cycle and electron transport chain and carboxylic acids (Table 5). In addition, upregulated proteins in the cerebellum are also implicated in Parkinson disease and microtubules (Table 5). On the contrary, in the cerebral cortex, energy metabolism (oxidative phosphorylation, NADH dehydrogenase activity), Alzheimer disease, and retrograde endocannabinoid signaling involve downregulated proteins (Table 6). Meanwhile, upregulated proteins of the cerebral cortex are associated with microtubules, neuronal projections, membrane trafficking, and transport of synaptic vesicles ( Table 6).

Early Dysregulations of Phosphoproteome in Young Ki91 SCA3/MJD Mouse Cerebellum and Cortex
Following an analysis of total proteome, we asked whether there are any alterations regarding phosphorylation of proteins. Since there are no transcriptional changes in young 2-month-old presymptomatic Ki91 mice, we speculated that dysregulation of protein levels could be related to other cellular mechanisms of protein amount control, such as phosphorylation, which is the most abundant posttranslational modification of proteins. Analysis of mass spectra of enriched samples enabled total identification of 4034 unique proteins. Similar to processing data of total proteins, filtration of valid values and statistical tests were performed on the results obtained in the analysis of the phosphoproteome. The analysis revealed significant differences in the phosphorylation level of 82 individual proteins in the   Table 6). Part of those proteins had altered modification level in more than one site which increased the number of hits (total number of altered phosphorylation sites: cerebellum = 95, cortex = 481). In both the cerebral cortex and cerebellum, we observed that the phosphorylation levels of dysregulated Genes identified by RNAseq were analyzed by qPCR using control brain tissue collected from C57BL/6 (C57) and FVB mouse strains to exclude the influence of the genetic background (BG) on the level of gene expression. The differences in expression levels of genes would be considered statistically significant if the tested gene demonstrated p ≤ 0.01 for each of the brain tissue controls (unpaired Student's t test; error bars: SEM; total number of samples n = 12, n = 4 per experimental group). However, none of the tested genes from the a cerebellum and b cerebral cortex consistently reached such significance value across controls and brain tissues. The qPCR results indicate that the differences in expression level measured by RNAseq are the result of genetic background and are not the result of the influence of the mutant ataxin-3. Hence, the presymptomatic cerebellum and cortex from Ki91 mouse do not demonstrate SCA3 causative changes in mRNA levels proteins were decreased and phosphorylation of only three residues in three proteins was increased (cerebellum: Hist1h1d; cerebral cortex: Prkcg and Pclo). Notably, the majority of dysregulated phosphoproteins in the cerebellum overlap with the list of dysregulated phosphoproteins in the cerebral cortex (68% of individual cerebellar phosphoproteins overlap with the cortex; N = 56 and 73% cerebellar phosphosites overlap with the cortex; N = 70) (Supplementary Table 6). We have constructed a network of dysregulated kinases with the highest number of substrates among dysregulated The presymptomatic changes in mRNA levels of genes identified by RNAseq were examined by qPCR, however using brain tissue from 4month-old Ki91 mice. The differences in expression levels of genes for the a cerebellum and b cerebral cortex did not reach consistent statistical significance across controls and tissues (p ≤ 0.01 for each of the brain tissue controls in unpaired Student's t test; total number of samples n = 12, n = 4 per tissue; C57BL6 (C57) or FVB mouse tissue was the control for Ki91 mouse tissues; error bars: SEM) proteins (altered total levels or phosphorylation pattern) using the CPDB tool, protein-protein interaction networks (induced network) visualized in Fig. 7. The network of proteins of the cerebral cortex separates the most important kinases, which are arranged in the outer space, circularly around their substrates ( Fig. 7a). This arrangement demonstrates that most of the identified dysregulated kinases are interactors (CPDB) and have an influence on common protein substrates which were identified by us as dysregulated in total proteome or phosphoproteome analysis. The smaller network organizing proteins of the cerebellum places the kinase Pak1 in the center of protein-protein interactions (Fig. 7b). Pak1 has a role in cellular pathways related mainly to cytoskeleton organization and transport, proliferation, and apoptosis [25]. In the cerebral cortex, Pak1 is also one of the most compelling kinases, together with Brsk1 and Brsk2 also playing role in axon guidance, Src, Braf, and multifunctional Mapk and Gsk3b (Fig.  7a). Other important kinases of the network are involved in trans-synaptic signaling: calcium/calmodulin-dependent kinases Camk2a and Camk2b and isoforms of protein kinase c: Prkca, Prkcd, and Prkcg (Fig. 7a). On the other hand, substrates which may be modified by the highest number of kinases in the network are MAPT and MAP2-both regulating functions of the microtubule of axons and dendrites. Furthermore, we have performed another analysis with the use of PHOSIDA in order to find those kinases, which are not necessarily dysregulated, but which phosphorylate dysregulated proteins at identified residues within motifs. The analysis identified five kinases, namely Ck1 and 2, Pka, Cdk2, and Camk2 for both the cerebellum and cerebral cortex (Fig.  7c, d), which are mainly responsible for the altered level of phosphorylation on most residues and motifs (77% in the cerebellum, 78% in the cerebral cortex).

Western Blot Analysis of Dysregulated Proteins and Phosphoproteins in Presymptomatic Ki91 SCA3/MJD Mice
The selected proteins and phosphoproteins identified in label-free mass spectrometry analysis were subjected to western blot analysis using commercial antibodies. The criteria for validation of proteins were the fold change of dysregulation, the affected processes, and in the case of phosphoproteins, also the commercial availability of the phospho-specific antibodies. p-Darpp32, p-Tau, Pabpc1, Mbp, Tubb3, Ddb1, and Nefh were assayed versus αactin as loading control (Fig. 8). We confirmed the downregulation of p-Darpp32 (phosphosite Ser97; fold change (FC) = 0.62; p = 0.045; two-sample t test) in the cerebral cortex and p-Tau in the cerebral cortex (FC = 0.5; p = 0.0012; two-sample t test) (Fig. 8a, c) and cerebellum (FC = 0.58; p = 0.0035; two-sample t test) (Fig. 8b, d). Furthermore, we selected several relevant proteins related to such biological processes as translation [26] (Pabpc1), DNA damage and repair [27] (Ddb1), myelin formation [28] (Mbp), and neuronal microtubule function [29,30] (Tubb3, Nefh). We confirmed increased levels of Pabpc1 (FC = 1.5; p = 0.013; two-sample t test) and decreased levels of Mbp (FC = 0.78; p = 0.0012; two-sample t test) and Tubb3 (FC = 0.78; p = 0.011; two-sample t test) in the cerebral cortex (Fig. 8a, c) as well as increased levels  Transcriptomic changes occur in 10-and 14-month-old symptomatic homozygous Ki91 SCA3/MJD mice and are also related to changes associated with particular cell types. In 10-month-old Ki91 mice, the analysis revealed the elevated level of Psat1 and Olig1 in the cerebellum and cortex (a). Mag gene related to oligodendrocytes revealed upregulated level in the cortex. Plp1 demonstrated decreased level in the cortex of 10-month-old Ki91 mice suggesting loss of adult oligodendrocytes during disease progression. Fourteen-month-old Ki91 mice demonstrate more pronounced alterations in tested mRNA levels (b). In the cerebellum, the Cd68, a microglial marker, demonstrated upregulated expression level. The metabolism-associated genes, Apt2b1 and Ca2, are downregulated. The gene highly expressed in adult oligodendrocytes Plp1 is also downregulated in the cerebellum. In the cortex, the level of transcripts, characteristic for oligodendrocyte precursors, is upregulated (Olig1, Olig2) and, on the other hand, decreased the level of transcripts characteristic for adult oligodendrocytes (Plp1 and Cldn11) and also increased the level of Mag. We did not detect transcriptional changes characteristic for neuronal markers both in 10-and 14-month-old Ki91 mouse brains. We also observe the changed level of genes characteristic for metabolism (Psat1, Qdpr, and Psmd4). p ≤ 0.05, using unpaired Student's t test; total number of samples n = 8 per age per cerebellum or cortex; n = 4 for the control group per individual tissue of 10 or 14 months, n = 4 for the SCA3 group per tissue of 10-month-old. In the case of 14-month-old Ki91 mouse, n = 3 or n = 4 depending on the gene tested: n = 3 in the Ki91 mouse group for the following genes in the cerebellum: Srsf2, Ppp2r1a, Idh1, Glul, Atp2b1, Ca2, Plp1; n = 3 in Ki91 mouse for the following genes in the cortex: Olig1, Olig2, Cd68, Cox7a2, Reln, Cldn11, Mash, Plp1; n = 4 in the Ki91 mouse group in the cerebellum: Olig1, Cd68, Mash1, Olig2, Sst, Mag, Reln, Plekhb1, Pdgfra, Npy, Cldn11; n = 4 in the Ki91 mouse group for the following genes in the cortex: Omg, Ndufa9, Srsf2, Psat1, Pea15a, Sst, Mag, Npy, Psmd4, Tuba1a, Qdpr (error bars: SEM) of Ddb1 (FC = 1.34; P = 0.011; two-sample t test) and Nefh (FC = 1.7; p = 0.007; two-sample t test) in the cerebellum (Fig. 8b, d) (p < 0.05; two-sample t test).

Dysregulations of Total Proteome and Phosphoproteome Reveal Three Major Arbitrary Groups of Disturbed Biological Processes in the Brain of Young KI91 SCA3/MJD Mouse
Based on the CPDB and ClueGO analysis (GO term analysis (B, MF, level 5) and pathway enrichment), we have selected three groups of biological processes, which are enriched among dysregulated proteins in both the cerebral cortex and cerebellum. The complete lists of GO terms and pathways related to dysregulated proteins and phosphoproteins are included in Supplementary Tables 3-5. Of note, results obtained with both bioinformatic tools (CPDB and ClueGO) showed overlap in the majority of biological processes and pathways (see Supplementary Tables 3-5). The analysis is visualized in Fig. 9 where the processes were groped arbitrarily and in Supplementary Figs. 3 and 4 with no arbitrary grouping. Selection of the groups was performed according to mutual relations and similarity of biological function of pathways and GO terms. We termed the first group "Disturbed mechanisms of modulation of protein levels and DNA damage" associated with either protein ubiquitination, translation initiation, splicing, or chromatin organization (Tables 7 and 8; Fig. 9, group I). The second group consists of biological processes that likely result from the initial dysregulations in the first group. The second group was termed "Disturbed formation of neuronal cellular structures: organelles and macromolecules" and is related to aberrant protein folding and affected organelle biogenesis and maintenance, which includes microtubule and actin cytoskeleton organization, the formation of axons and dendrites, axon guidance, the formation of gap junctions, and cellular vesicles (Tables 9 and 10; Fig. 9, group II). The third group was termed "Neuronal cell functionality affected by processes following perturbed cytoskeletal complex formation and apoptosis" (Tables 11 and 12; Fig. 9, group III). The processes affected in this group consist of axonal transport along microtubules, including synaptic vesicles and mitochondria, and consequently, mitochondrial respiratory chain complex formation and integration of energy metabolism (related to trans-synaptic signaling). Of note, several proteins participating in the processes belonging to this group were previously associated with other neurodegenerative disorders like HD, PD, and AD (Supplementary Tables 3-4). In this group, we also included programmed cell death as the final effect of all affected processes.

Subcellular Localization of Dysregulated Proteins
We performed analysis of "cellular component" GO terms (p value cutoff < 0.01) in CPDB to examine putative cellular localization of dysregulated proteins and phosphoproteins (Supplementary Fig. 4; Supplementary Table 7). The majority of the differentially expressed total proteins ( Supplementary  Fig. 4A, C) were assigned to localize in "extracellular exosome" (42% cerebellum and 46% cerebral cortex) or assigned as cytoplasmic (39% cerebellum and 44% cortex). A large number of dysregulated proteins were also identified as mitochondrial (28% cerebellum and 20% cortex) and cytoskeletal (26% cerebellum and 24% cortex). There were also proteins which were assigned to cellular structures which are typical for neurons, such as dendrites (14% cerebellum and 6% cortex) and axons (13% cerebellum and 8% cortex). In addition, in the cerebellar cortex, 9% of proteins are associated with lytic vacuoles.
In apparent contrast to the total proteome, a prominent number of 26% of all phosphorylated proteins in the cerebellum ( Supplementary Fig. 4B) were associated with the nucleus. In the cortex (Supplementary Fig. 4D), there were 22% of phosphoproteins localized in the nucleus and 4% associated with the cell membrane. Decreased phosphorylated proteins predicted to be localized in the nucleus included Sirt2, Srrm2, Thrap3, Acin1, Bclaf1, Ctr9, Rbm39, Numa1, Npm1, Trim28, Srsf9, Matr3, Sept4, Ndrg2, and Ppp6r3 (Supplementary Table 7). These nuclear proteins are implicated in cell cycle, DNA damage, and splicing; however, no GO terms containing dysregulated nuclear phosphoproteins were identified which relate to transcription factors or direct gene expression control. In coherence with the analysis of biological processes, we found no proteins with altered level of phosphorylation, which localized in the mitochondria, neither in the cerebellum nor in the cerebral cortex.

Cellular Identity of Dysregulated Proteins
The analysis of cellular markers among dysregulated proteins and phosphoproteins was performed using the BrainMap tool of the Allen Brain Atlas for cerebral cortex ( Supplementary  Fig. 5A) and DropViz for cerebellum ( Supplementary  Fig. 5B). The analysis revealed a characteristic pattern of cellular markers of inhibitory neurons, which consisted of Gad1  Table 6). In the cerebellum (Supplementary Fig. 5B), the analysis demonstrated several cellular markers with the ratio of relative (log2 FC = 0.74, total), and one protein specific for oligodendrocytes: Enpp6 (log2 FC = 0.65, total). Among proteins with altered level and assigned log fold change, we did not identify conclusive markers for microglia and endothelial cells. However, Hist1h1d (cerebellum log2 FC = − 0.96, total; 1.77, phospho; cortex = − 1.04, total), a microglia marker (with a low percentage of presence in microglia and no expression (0%) in any other cell types), is dysregulated in three analysis (total proteome and phosphoproteome of the cerebellum and total proteome of the cerebral cortex) (Supplementary Table 6). Altogether, proteomic changes occur mainly in neurons and oligodendrocytes in the cerebral cortex ( Supplementary  Fig. 5A), and inhibitory neurons, Bergmann glia, and astrocytes in the cerebellum (Supplementary Fig. 5B).

Discussion
In the present work, we have defined the early molecular signs of SCA3/MJD polyQ neurodegenerative disease. We used a homozygous Ki91 mouse knock-in model, which mimics the genetics and pathological SCA3 situation in patients including presymptomatic phase and later disease onset [14]. Taking advantage of the model, we challenged the question of which molecular events take place before the disease outbreak in the brain by selecting a set of high-throughput methods to define the brain transcriptome first in 2-month-old and later at 10-14month-old animals by qPCR. Subsequently, we identified the brain proteome and phosphoproteome in 2-month-old Ki91 animals. Our goal was to identify the molecular triggers of the disease which may be evident early in life and early in disease pathology and therefore are not "contaminated" by secondary molecular signs originating from neuronal dysfunction and death. In our work, we included the cerebellar cortex as an important source of early signs in SCA3 brain [35]. Recent findings also emphasize the role of communication between the cortex and cerebellum as one of the pathology hallmarks of SCA3 [36][37][38][39].
The first essential finding is the sequence of general molecular events, which governs the disease onset process in SCA3/MJD. We found that Ki91 homozygous animals do not contain mutant ataxin-3-related transcriptional changes in cerebellum and cortex tissue at 2 months of age and do not show any behavioral changes at that stage; however, in the cerebellum and cortex, there are sparsely occurring cells with nuclear localization of mutant ataxin-3. At the same time, 2-month-old Ki91 animals already demonstrate prominent changes both at the level of total proteome and at the level of phosphoproteins. Importantly, the general level of phosphorylation of proteins is greatly decreased in our Ki91 SCA3/MJD mouse model. Of note, many of the changes exceed the decrease of log2 FC − 3 and lower. Furthermore, we demonstrate that later ages of 10-14-month-old Ki91 animals do contain transcriptional changes in both the cerebellum and cerebral cortex, and the mature postmitotic neural cultures from patient neurons from iPSC also contain some transcriptional changes identified in our 10-14-month-old Ki91 animals. The type of neural cultures that we used was reported to be positive for GABA and GAD67 and originate from cells that display differentiation profile toward hindbrain identity [24], whereas the mouse data are generated from the cortex and cerebellum. For example, Olig1 is elevated both in mouse tissue and MJD iPSC-derived neural cells, but there are a number of different cells (including different types of neurons) that evolve from OLIG-positive progenitors depending on the brain region. Thus, concordant expression changes might not necessarily argue for the same changes at the cellular level. Comparison of data from WT and Ki91 mouse hindbrain (pons, medulla) versus MJD iPSC-derived neural cells and isogenic controls requires further studies. Together, the findings demonstrate that the possible sequence of events that leads to full-featured brain disease is composed of the changes at the level of many important proteins and phosphoproteins in the initial absence of mRNA changes dependent on mutant ataxin-3 (depicted in Fig. 10). These early events may further lead to transcriptional changes at later ages. Therefore, transcriptional changes in SCA3 can be classified as secondary and more severe disease signs. Previous reports demonstrated Cytoplasm Cell nucleus ATXN3 DNA damage 1.

Misfolded proteins, microaggregates
Lack of bulk pre-symptomatic transcriptional changes

Disturbed splicing
Ataxin-3 Ataxin-3  Fig. 10 The diagram of processes affected by molecules identified in young Ki91 mouse brain during the early SCA3/MJD pathogenesis that the SCA3 disease pathogenic process includes relatively direct transcriptomic changes by binding of mutant ataxin-3 to chromatin and transcription factors [40][41][42]. Considering widespread expression of ataxin-3 across cells and tissues of Ki91 mouse [14], such direct interaction could occur early and readily and evoke transcriptional changes. However, the scenario that one cannot exclude is the direct influence of ataxin-3 on transcriptional changes which may start only at certain cellular lineages in the brain and that the present tissue-based NGS resolution does not allow for identification of transcriptional changes restricted to the initially small number of cells. We performed the analysis of the dysregulated protein markers to identify types of cells relevant to pathogenesis in the brain. Based on the analysis, we conclude that one of the affected cell types is the parvalbumin-positive neurons. Such neurons are usually GABAergic interneurons in the cortex, Purkinje cells, and other GABAergic neurons of the cerebellum [43,44]. Another prominently affected cell type was astrocytes in the cerebellum and, to a lesser extent, in the cerebral cortex. In addition, according to our analysis of cell types using BrainMap, oligodendrocytes and excitatory neurons are also affected in the cerebral cortex. Another important possibility for the lack of transcriptional changes at the presymptomatic stage is the direct or indirect ataxin-3 effect on transcriptional changes, which start during later adult life as a response to another factor such as aging, stress, DNA damage, or protein interactions. The transcriptomic changes in SCA3 were so far investigated in the symptomatic phase of the disease such as in patient blood or in older brain tissue collected from mouse models [45][46][47][48]. One of the interesting outcomes from our analysis of transcriptome in older Ki91 animals suggests that precursors of oligodendrocytes are upregulated. This may suggest that demyelination and myelin repair responses occur later in the disease course. Since neuronal activity promotes myelination [49], the start of demyelination in SCA3 may be the result of the collapse of neuronal and axonal function which occurs later in the disease course.

Premature termination of translation
On the other hand, elevated Olig1 level suggests some developmental abnormalities in SCA3. One possibility could be an increase in the prolonged existence of the Olig1-positive progenitor pool, which could disrupt the lineage segregation process. Another part of the results which indicate a developmental component in SCA3 pathogenesis is a group of dysregulated proteins involved in such processes as axon guidance and dendrite development.
In view of the fact that we did not find transcriptional changes in the brain of young Ki91 SCA3/MJD mouse model, we reasoned that proteomic changes were the other type of general molecular events that may be disturbed during the disease onset and presymptomatic SCA3 phase. We found a number of proteins with an altered level indicating that the dysregulations in protein level occur earlier and with greater intensity as compared to transcriptional changes during the disease progression.
The changes in protein level in the cell can be evoked by posttranslational modifications such as protein phosphorylation. We tested the global phosphorylation of proteins and we found numerous decreased phosphoproteins in the cerebellum and cortex of young Ki91 animals (most of them exceeded log FC − 2) even reaching − 7 in Ki91 animals versus WT animals in the cerebral cortex. In addition, several common proteins were identified for both total proteome and phosphoproteome.
Commonly dysregulated kinase for both analyzed cerebral regions is p21-activated kinase 1 (Pak1) with highly downregulated levels of phosphorylation at Ser-223 (cerebellum) and Ser-174 (cortex). Phosphorylation of Ser-223 is required for Pak1 activation and is performed by casein kinase (Csnk2a1), a known interactor of ataxin-3 [52,55]. In general, Pak1 is implicated in neurodegenerative disorders, such as Alzheimer disease (AD), Huntington disease (HD), and Parkinson disease [56,57]. Previously, Pak1 was shown to promote polyQ aggregation by enhancing its toxicity like in cellular models of HD, and this process was rescued by Pak1 knockdown [58]. Conversely, cell death was observed when the activity of Pak1 was downregulated in the 6-OHDA rat model of Parkinson disease [56]. In addition, it was demonstrated that oxidative stress reduced the active, phosphorylated form of Pak1 in dopaminergic neurons, which directly led to a reduction of anti-apoptotic Bcl2 protein levels via the ubiquitin/ proteasome pathway. Moreover, an upstream phosphatase of Pak1 modulating its activity is Pp2b-A (Ppp3ca), which is upregulated in the cortex of Ki91 mouse. In addition, a core regulator of phosphatases and also PP2A (Ppp2r1a) [59], phospho Arpp19/Ensa (S62/67) is downregulated in the cerebellum.
We have clustered the dysregulated proteins and phosphoproteins into three groups reflecting the cellular processes and mechanisms in which the proteins are involved. Group 1 comprises basic cellular processes such as translation and its possible premature termination, disturbed assembly of the ribosome, ubiquitination but also dysregulation of proteins that are characteristic for nucleus and take part in nuclear regulation such as nucleosomes forming, chromatin remodeling, splicing, DNA damage, and repair. Most likely, many of those initial processes result from cellular stress related to the excessive abundance of mutant ataxin-3 in the nucleus, which we demonstrate in young Ki91 animals. The transfer of mutated ataxin-3 into the nucleus could also be related to altered levels of Trim28 (log2 FC 0.94 in the cerebellum, log2 FC 1.59 in the cerebral cortex), which was shown to drive the nuclear accumulation of two distinct proteins: α-Syn and Tau implicated in Parkinson disease and Alzheimer disease, respectively [68]. Notably, 26% of dysregulated phosphoproteins in the cerebellum and 22% in the cerebral cortex were predicted to localize in the nucleus, having a role in cell cycle, DNA damage, and splicing.
DNA damage was already demonstrated in SCA3, which was linked to inactivation of polynucleotide kinase 3′-phosphatase (Pnkp) by ataxin-3, an important enzyme for repair of DNA damage [69,70]. In our experiments, we detected upregulated damage-specific DNA binding protein 1 (Ddb1) in the cerebellar tissue of Ki91 mouse model (log2 FC 1.65). The Ddb1 together with Chk1 is an ataxin-3 interactor; however, mutant ataxin-3 does not impair the interaction with Chk1 [53].
A molecule which may also be involved in the regulation of the level of other proteins is Pabpc1 which is highly upregulated in the cerebral cortex (log2 FC 2.02). Its major function is its role in the regulation of translation initiation and mRNA stability and is part of "stress granules" [71,72].
The second group of dysregulated proteins is related to altered turnover of cytoskeleton which influences neuronal cellular structures and organelles. It was shown that expression of expanded polyQ proteins (ataxin-1, ataxin-3, and huntingtin) substantially affects the dynamics of microtubule cytoskeleton, by nucleation of new microtubules and rebuilding microtubule rods in neurons of a Drosophila model [73]. The cytoskeleton dysregulation potentially disturbs cell division, regulation of growth, structure, and guidance of neuronal projections and gap junctional communication. However, a particularly important resulting process detected among the dysregulated proteins seems to be the transport along microtubules and axons. Moreover, it was demonstrated that aggregates of the mutated form of ataxin-3 form inclusions inside the axon, which might perturb axonal transport [7]. The aberrant transport process may lead to disturbed synaptic transmission and energy metabolism and ultimately cause cell dysfunction and cellular death, and proteins involved in those processes were clustered in group 3.
An important part of SCA3 pathogenesis comprises impairments of the mitochondria, which include altered localization of ataxin-3 inside the mitochondria, enhanced interaction of ataxin-3 82Q with mitochondrial proteins (Sdhb and Ndufa4 which are also dysregulated in our analysis), and mitochondrial DNA damage documented in SCA3 models [51,61,[74][75][76]. In line with those aberrations, in young Ki91 animals, we observed a number of dysregulated mitochondrial proteins belonging to group III such as Mt-co1 and mt-Co3 (cortex) and Uqcrh (cerebellum). Strikingly, in the cerebellum, proteins related to the mitochondria are all upregulated, whereas in the cerebral cortex, GO terms and pathways related only to upregulated proteins are mainly associated with mitochondrial electron transport chain and metabolism. Interestingly, it was shown that mitochondrial precursor accumulation may cause cellular stress influencing translation machinery [77]. Remarkably, there are no GO terms or pathways for dysregulated phosphoproteins referring to the mitochondria, which suggest that phosphorylation is not the mechanism responsible for altered levels of mitochondrial proteins.

Conclusions
In the current study, we demonstrate that early transcriptional changes influenced by mutant ataxin-3 do not occur in presymptomatic Ki91 SCA3/MJD mouse model; however, we identify prominent dysregulation of protein levels and phosphorylation. Based on the identified proteins and phosphoproteins, we dissected a set of most early events including impaired phosphorylation of proteins indicating the influence of mutant ataxin-3 on many kinases, proteins involved in DNA damage, and mechanisms playing a role in translation initiation. Moreover, GO term and pathway analysis indicates that dysregulated proteins may further impair cellular pathways and processes such as autophagy, energy metabolism, and transport of molecules and vesicles along axons. Subsequently, this may lead to severe effects such as defective projections and synaptic transmission eventually leading to neuronal dysfunction, subsequent demyelination, and neurodegeneration. For each of the identified pathways, we propose key molecules which are affected by the change of levels in the cerebellum and cortex. Altogether, we conclude that early triggers of the disease act on the level of pathways and processes engaging proteins rather than the transcription stage, whereas late transcriptomic changes most likely result from degeneration of particular populations of cells in the brain such as GABAergic neurons and oligodendrocyte precursors. The relatively high number of dysregulated proteins and phosphoproteins in Ki91 SCA3/MJD mice indicates that there are important modifiers of disease and biomarkers or even target molecules for therapies among those proteins. The most prominent new candidate molecules are kinases, such as Pak1, which have a vast influence on distinct cellular processes. Therefore, a particularly promising area for further studies on disease mechanism and potential therapies in SCA3 is the proteome and phosphoproteome.

Animals
Maintaining and breeding were performed at standard conditions with an 18-/6-h light/dark cycle and water and food provided ad libitum. The animals were marked using numerical ear tags (National Band & Tag Company, Newport, USA). The animals were sacrificed according to AVMA Guidelines for the Euthanasia of Animals by placing them in the programmable CO 2 chamber (Rothacher Medical, Heitenried, Switzerland). The stress level of the animals was minimized throughout all the procedures and animal handling. The animal experimentation and handling were approved and monitored by the Local Ethical Commission for Animal Experiments in Poznan. The Ki91 SCA3/MJD mouse model was bred on C57BL/6 mouse genetic background for 10 generations and the animals were further maintained on the C57BL/6 genetic background. The homozygous (mut/mut) transgenic animals both from Ki91 knock-in mouse were generated by breeding heterozygous (mut/wt) animals. Homozygous Ki91 animals contained between 98 and 132 CAG repeats on a single mutant ataxin-3 allele. For RNAseq, Ki91 animals contained 100-110 CAG repeats. For total proteomics, Ki91 mice contained 98-106 CAG and Ki91 mice for phosphoproteomics contained 98-123 CAGs in mutant Atxn3 gene. For behavioral studies, Ki91 mice contained 103-132 CAGs. The total number of 80 animals of various ages was used for collecting brain tissues and 36 animals at the age of 2 months were used for the behavioral experiments. The cortex and the cerebellum for proteomic and transcriptomic analysis were always collected from brain tissue of a typical experimental group consisting of four mutant mice versus four nontransgenic or C57BL/6 mice. For qPCR validation of RNAseq analysis, 2-and 4-month-old samples from the FVB mouse strain have been used as a second control since the FVB strain demonstrates the same SNP profile (MGI in chromosome 12 )(e.g., Atxn3 and Serpina3n locus) as the 129sv genetic background originating from stem cells to which the Ki91 transgene was introduced. Supplementary Table 1 summarizes the number of animals by strain, genotype, age, experiment type, and tissue.

Behavioral Studies
Assessment of motor function was performed as previously described [14]. Tests included accelerating rotarod (4-40 rpm in 9.5 min), elevated beam walk (diameter of rods-35, 28, 21, 17, and 9 mm), and parallel rod floor test in which the number of foot slips and locomotor activity in the experimental cage were analyzed and measured during 10 min. Each test consisted of one training day (T) and three consecutive days of measurement. In addition, we also performed scoring tests designed for evaluation of ataxia phenotype in mouse models [78]. All mice were weighed during each testing session. Graphing and statistics were performed on PrismVR software (San Diego, CA, USA), using ANOVA with Bonferroni post hoc test.

RNAseq Analysis
NGS QC Toolkit (v 2.2.3) [79] was used to generate quality metrics for assessment of fastq input files. We used the statistics generated by the NGS QC Toolkit and prepared charts ( Supplementary Fig. 6) which show the average quality score per base positions for each library. Fastq files were then aligned to the GRCm38.91 reference genome using STAR software (v 2.5.3a) [80] with parameters suggested by the QoRTs software (v 1.3.0) [81]. The parameters are used by default for estimation of genes and exon hit counts. The alignment statistics are demonstrated on charts in Supplementary  Fig. 6A where the percentage of unique and multiply aligned reads are shown. The data were analyzed using three different experimental approaches which included the alignment/ quantification of transcripts by Star/Deseq and HISAT/ StringTie/Ballgown [82,83]. In brief, the differential gene expression was calculated using DESeq2 software (v 1.14.1) [84]. To countercheck the results, the second software pipeline, namely HISAT2 (v 2.0.5) → StringTie (v 1.3.1c) → Ballgown (v 2.6.0) [82,83,85] was used. The results from the HISAT2 pipeline were consistent with the primary analysis and are not shown in the results section. Additionally, JunctionSeq analysis has been applied in the identification of altered splicing variants [86].

Quantitative Real-Time PCR
Reverse transcription was performed with 500 ng of RNA using Maxima H Minus Reverse Transcriptase according to the manufacturer's protocol using random hexamers. qPCR was performed according to MIQE Guidelines where relative gene expression and splicing events were estimated using the ΔΔCt method with ActinB, Pgk1, and Tfrc as control [87]. qPCR reaction was carried out on the BioRad CFX96 thermocycler using 5x HOT FIREPol EvaGreen qPCR Mix Plus (Solis Biodyne, Tartu, Estonia) with the following parameters: 95 at 15 min and 45 cycles of (95°C 15 min, 60°C

Protein Extraction Digestion and Enrichment for Proteomics
Mouse brain tissues were lysed in buffer containing 1 M triethylammonium bicarbonate (TEAB), 0.1% SDS, and 1 mM sodium orthovanadate (NaVO 3 ) in 2-ml tubes with stainless steel beads (Retch, Germany) followed by automatic homogenization using a Mixer Mill MM400 (Retch, Germany). Subsequently, the material was subjected to a threefold cycle of freezing and thawing followed by bath sonication for 3-min repeated three times while cooling the tube on ice in between the sonication. Protein concentration in the clear lysate was estimated using Pierce BCA protein assay kit (Thermo Scientific, Rockford, lL, USA) according to the manufacturer's instructions. Ten-microgram aliquots of proteins were diluted with 15 μl of 50 mM NH 4 HCO 3 and reduced with 5.6 mM DTT for 5 min at 95°C. Samples were then alkylated with 5 mM iodoacetamide for 20 min in the dark at RT. Subsequently, the proteins were digested with 0.2 μg of sequencing-grade trypsin (Promega) overnight at 37°C. For labeled free quantitative proteomics, 10 μg of digested protein per sample was used for analyses on LC/ MS. A similarly prepared set of samples was used for phosphoproteomics. For each sample, 300 μg of digested protein was used for phosphopeptide enrichment. Phosphorylated peptides were enriched using two different kits, one exploiting titanium dioxide (TiO 2 ) spin tips and the other high-capacity Fe-NTA spin column (Thermo Scientific, Rockford, lL, USA). Elution fractions containing phosphopeptides were further desalted on a C18 column (J.T. Baker, Center Valley, PA) prior to mass spectrometry quantitative measurements. To perform a global analysis of phosphoproteomic changes, we have collected another set of cerebellum and cortex samples. The final analysis was performed on eight cerebella of four young 2-month-old homozygous Ki91 and four age-matched C57BL mice and six cerebral cortices of three young 2-month-old homozygous Ki91 and three age-matched C57BL mice. Collected tissues were further subjected to protein isolation, trypsin fragmentation, and phospho-enrichment procedure (see "Methods" section).

Mass Spectrometry Analysis of the Proteome
The analysis was performed with the use of Dionex UltiMate 3000 RSLC nanoLC system connected to Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific). Peptides derived from in-solution digestion were separated on a reverse phase Acclaim PepMap RSLC nanoViper C18 column (75 μm × 25 cm, 2 μm granulation) using acetonitrile gradient (from 4 to 60%, in 0.1% formic acid) at 30°C and a flow rate of 300 nl/min (for 230 min). The spectrometer was operated in data-dependent MS/MS mode (tandem mass spectrometry) with survey scans acquired at a resolution of 70,000 at m/z 200 in MS mode and 17,500 at m/z 200 in MS2 mode. Spectra were recorded in the scanning range of 300-2000 m/z in the positive ion mode. Higher energy collisional dissociation (HCD) ion fragmentation was performed with normalized collision energies set to 25. Protein identification was performed using the Swiss-Prot mouse database with a precision tolerance 10 ppm for peptide masses and 0.08 Da for fragment ion masses. All raw data obtained for each dataset were imported into MaxQuant 1.5.3.30 version for protein identification and quantification. Protein was considered as positively identified if at least two peptides per protein were found by Andromeda search engine, and a peptide score reached the significance threshold FDR = 0.01.
Obtained data were exported to Perseus software ver. 1.5.3.2 (part of MaxQuant package). Numeric data were transformed to a logarithmic scale, and each sample was annotated with its group affiliation. Proteins only identified by site, reverse database hits, and contaminants were removed from the results. Next, data were filtered based on valid values for proteins. Proteins which contained valid values in 75% of samples in at least one group (the inclusion threshold: 3 or 4 values for control and 3 or 4 values for Ki91 mice) were included as valid hits. In addition, for supplementary analysis, we have selected proteins which produced valid values only in control or only in Ki91 SCA3/MJD mouse group (4 values inclusion). Prior to statistical analysis, normalization of data was performed by subtracting median from each value in a row. A two-sample t test was performed on analyzed sample data with p value < 0.05 being considered significant, and differentiating proteins were normalized using the Z-score algorithm for hierarchical clustering of data.

Bioinformatic Analysis of Proteomic Data
The power of the analyses was increased by using two separate set of tools, namely Consensus Path Database (version 32) (CPDB) [90] and Cytoscape version (version 3.6.0) containing the ClueGO plugin (academic version 2.3.5) [91,92]. CPDB consolidates information from a considerably higher number of databases compared to ClueGO; however, the advantage of ClueGO is its ability to merge the same or similar information about pathways or GO terms which forms general categories. The analysis paradigm included the discovery of affected pathways, molecular function of dysregulated proteins, biological processes affected by the dysregulated proteins, and prediction of subcellular localization of dysregulated proteins. For each analysis, names of genes corresponding to the names of dysregulated proteins or phosphoproteins were used.
Proteins were grouped using the Consensus Path Database according to the pathways (pathway enrichment p value cutoff < 0.01, minimum overlap with input list = 10% of total number of dysregulated proteins), GO term analyses by molecular function, biological process, and cellular component (GO term B, MF level 5, CC levels 4 and 5, p value cutoff < 0.0001).
The second analysis in CPDB was performed with lists of top dysregulated proteins (log2 FC ≤ − 1.0 for downregulated protein levels or log2 FC ≥ 1.0 for upregulated protein levels) for pathway enrichment and GO terms with the same p value and minimum overlap with input list restrictions.
Subsequently, using ClueGO and Cytoscape, we have performed GO term annotation analysis and pathway enrichment (p value cutoff < 0.05, GO tree interval = 3-5, kappa score = 0.5). In ClueGO, the number of GO terms was restricted by the number of dysregulated proteins per GO term in proportional relation to the total number of dysregulated proteins in a dataset (51-100 dysregulated proteins-inclusion of GO terms containing a minimum of 3 proteins from the input list, 101-150-minimum of 4 proteins, 301-350-minimum of 8 proteins). The analysis in ClueGO was based on "biological process," "molecular function," and KEGG, Reactome, and WikiPathways [93][94]. The compatibility of results obtained with the Consensus Path Database and ClueGO was confirmed by identification of common GO terms with similar p values using both tools.
In a separate analysis, we performed identification of cell types in the brain which are affected by pathogenesis using the BrainMap tool of Allen Brain Atlas [22] for the cerebral cortex and the DropViz tool [21] for the cerebellum. This was accomplished by grouping of dysregulated proteins according to cell type in the brain in which they are expressed. Significantly dysregulated (p < 0.05) proteins identified from label-free proteomic and phosphoproteomic LCMS experiments were included together into two tissue groups, based on origin from the cerebellum and cortex. Relative expression is presented in BrainMap as counts per million in log10 scale and in DropViz as the amount of transcripts per 100,000 in the cluster. Moreover, for the phosphoproteomic data, we have predicted kinases which phosphorylate dysregulated phosphoproteins using CPDB and PHOSIDA (using known motifs) [95]. For the analysis with PHOSIDA sequences of peptides with altered phosphorylation level, the FASTA format was used.

Immunohistochemistry
The animals were deeply anesthetized and transcardially perfused using saline followed by 4% PFA. The brains were removed, postfixed in 4% PFA for 48 h, and cryopreserved with graded sucrose (10-20-30%) over 72 h. The 20-μm parasagittal mouse brain sections were cut using a cryostat at − 20°C and collected on SuperFrost Plus slides (Thermo Scientific). The sections were processed immediately. The HIER procedure was applied by incubation of the sections in citrate buffer (pH 9.0) for 30 min at 60°C. The sections were blocked via incubation in 4% normal goat serum in TBS for 1 h. For immunofluorescence staining, the sections were incubated overnight at 4°C with the primary mouse anti-ataxin-3 antibody 1H9 (1:200) [] and, subsequently, with the anti-mouse antibody labeled by AlexaFluor488 (1:400; Jackson ImmunoResearch, Suffolk, UK). The sections were end-stained with Hoechst 33342 (Sigma) nuclear stain at 1:1000 and embedded in Fluoroshield (Sigma) mounting medium. Fluorescent confocal images were acquired using fixed excitation and detection parameters using the TCS SP5 II (Leica Microsystems, Poland).

Statistics
The data regarding behavioral experiments performed within 4 days (rotarod, elevated beam walk, parallel rod floor test) were subjected to a two-way ANOVA, followed by Bonferroni posttests. Scoring test and body weight were evaluated with unpaired Student's t test. The two-group comparisons of the gene expression data by qPCR were conducted using the unpaired Student's t test. p values less than 0.05 were considered significant with the exception of qPCR using C57BL/6 and FVB controls, where p values less than 0.01 were considered statistically significant. Identification of proteins on raw proteomic data was performed by Andromeda search engine in Mascot using the following inclusion criteria: 1. At least two different peptides per protein were identified per sample, and a total peptide score reached the significance threshold FDR = 0.01. Identified proteins matching the inclusion criteria were subjected to further statistical analysis with two-sample t test, and dysregulation of protein level reaching p value < 0.05 was considered as significant. One asterisk indicates p value ≤ 0.05, two asterisks indicate p value ≤ 0.01, and three asterisks indicate p value ≤ 0.001.