Atxn2 Knockout and CAG42-Knock-in Cerebellum Shows Similarly Dysregulated Expression in Calcium Homeostasis Pathway

Spinocerebellar ataxia type 2 (SCA2) is an autosomal dominantly inherited neurodegenerative disorder with preferential affection of Purkinje neurons, which are known as integrators of calcium currents. The expansion of a polyglutamine (polyQ) domain in the RNA-binding protein ataxin-2 (ATXN2) is responsible for this disease, but the causal roles of deficient ATXN2 functions versus aggregation toxicity are still under debate. Here, we studied mouse mutants with Atxn2 knockout (KO) regarding their cerebellar global transcriptome by microarray and RT-qPCR, in comparison with data from Atxn2-CAG42-knock-in (KIN) mouse cerebellum. Global expression downregulations involved lipid and growth signaling pathways in good agreement with previous data. As a novel effect, downregulations of key factors in calcium homeostasis pathways (the transcription factor Rora, transporters Itpr1 and Atp2a2, as well as regulator Inpp5a) were observed in the KO cerebellum, and some of them also occurred subtly early in KIN cerebellum. The ITPR1 protein levels were depleted from soluble fractions of cerebellum in both mutants, but accumulated in its membrane-associated form only in the SCA2 model. Coimmunoprecipitation demonstrated no association of ITPR1 with Q42-expanded or with wild-type ATXN2. These findings provide evidence that the physiological functions and protein interactions of ATXN2 are relevant for calcium-mediated excitation of Purkinje cells as well as for ATXN2-triggered neurotoxicity. These insights may help to understand pathogenesis and tissue specificity in SCA2 and other polyQ ataxias like SCA1, where inositol regulation of calcium flux and RORalpha play a role. Electronic supplementary material The online version of this article (doi:10.1007/s12311-016-0762-4) contains supplementary material, which is available to authorized users.


Introduction
Spinocerebellar ataxias (SCAs) comprise a group of autosomal dominantly inherited disorders that are characterized by massive neurodegeneration preferentially in the cerebellum but also in other areas like spinal cord and brainstem [1]. The main shared symptom of these neurodegenerative processes is motor incoordination (gait, stance, and limb ataxia) [2,3]. Additionally, there are non-motor symptoms that are characteristic for SCA subgroups like dementia or macular degeneration. Thus, the genetic basis and the initial phenotypes of diverse SCAs are heterogeneous, but their clinical and pathological features converge at later stages [4]. An important shared feature and a determinant of patients' motor incoordination is the early vulnerability of cerebellar Purkinje neurons, which are known to function via integration of glutamatergic inputs like climbing fibers and parallel fibers Melanie Vanessa Halbach and Suzana Gispert are joint first authors Electronic supplementary material The online version of this article (doi:10.1007/s12311-016-0762-4) contains supplementary material, which is available to authorized users. [5]. Indeed, some emerging pathogenic pathways are shared among the SCAs and seem to involve the glutamate calciummediated excitation at the dendritic spines of cerebellar Purkinje neurons [6,7]. However, for none of the SCAs, the preliminary insights into molecular pathogenesis have led to neuroprotective therapy.
Physiologically, the ATXN2 protein shows abundant expression in many cell populations throughout the organism, such as cerebellar Purkinje neurons, hippocampal pyramidal neurons, or liver hepatocytes [32]. In cervical cancer HeLa cells, ATXN2 was identified as a protein with direct RNA binding [33]. It has been described as a cytoplasmic protein associating with the rough endoplasmic reticulum (ER) and colocalizing with ribosomes [34]. ATXN2 is involved in the regulation of global RNA processing and ribosomal translation by binding via its PAM2 motif to PABPC1 and via its LSM/LSMAD domains directly to RNA [35,36]. Furthermore, a role in stress response exists, with ATXN2 relocalizing to the RNA quality control machinery in stress granules together with the seeding factor TIA-1 [37]. Finally, a small number of studies implicated ATXN2 also in trophic signaling, cytoskeletal reorganization, or nuclear transcription [38][39][40][41].
It is clear that polyglutamine (polyQ) diseases involve neurotoxicity due to a gain-of-function of the mutant proteins, but controversy exists about the pathogenic roles of fibril formation and aggregation as well as the contributions of a partial loss-of-function [42][43][44][45]. In SCA2 mouse models, the polyQ expansion of ATXN2 leads to sequestration of its physiological interaction partner poly(A)-binding protein C1 (PABPC1) into insolubility [46]. This should affect global mRNA turnover and protein synthesis in many neuron populations but cannot stand alone to explain the selective neurodegeneration pattern of SCA2. The genetic ablation of Atxn2 in mice does not lead to a neurodegenerative process with weight loss such as SCA2, but instead to obesity, lipid anomalies, and insulin resistance [47,48]. Given that ATXN2 binds to the poly(A)-tail of mRNAs, we decided to perform an unbiased global survey of the cerebellar transcriptome in Atxn2 KO mice and then to test if the particularly strong expression dysregulations are mirrored in our previously characterized SCA2 mouse model with knock-in (KIN mice) of a CAG42 expansion into the Atxn2 gene [46]. This approach allows us to discern loss-of-function and gain-of-function effects on mRNAs by ATXN2 mutations. We observed a prominent interesting dysregulation in calcium homeostasis pathways, validated them independently at the RNA and protein level, and assessed these effects in different brain regions and at various ages. Mechanistically, a progressive accumulation of ITPR1 in the relatively insoluble tissue fraction occurred selectively in the SCA2 mouse model, but the expected sequestration of this calcium release protein by endogenous amounts of Q42-ATXN2 was not detectable in coimmunoprecipitation and colocalization studies.

Animals
Generation and characterization of Atxn2 KO and Atxn2-CAG42-KIN mice has been described formerly [46,47]. For KO mice, deletion of Atxn2 exon 1 was achieved through Cre-Lox recombination and confirmed by quantitative real-time reverse transcriptase PCR (RT-qPCR) or by Western blot in the mice and all tissues under study. For KIN mice, the single CAG typical for the murine sequence of Atxn2 exon 1 was expanded to 42 CAGs by homologous recombination. Mice were backcrossed from mixed 129/Ola (KO) and 129Sv/ Pas × C57BL/6 (KIN) background into C57BL/6 for more than eight generations. Animals were housed in individually ventilated cages with fixed light cycle under routine health monitoring at the FELASA-certified Central Animal Facility (ZFE) of the Goethe University Medical School, Frankfurt am Main. They were fed ad libitum and bred in heterozygous matings. For dissection, mice were sacrificed by cervical dislocation. Subsequently, cerebella were removed and frozen immediately in liquid nitrogen. Tissue was stored at −80°C until further use. All procedures were in accordance with the German Animal Welfare Act, the Council Directive of 24 November 1986 (86/609/EWG) with Annex II, and the ETS123 (European Convention for the Protection of Vertebrate Animals).

Global Transcriptome Profiling
After dissection, cerebellar tissue from 6-week-and 6-monthold Atxn2 KO and WT animals (4 Atxn2 +/+ vs. 4 Atxn2 −/− for each age) was sent to MFT Services (Tübingen, Germany). RNA was extracted and its quality was verified. Then, 100 ng of total RNA was amplified, labeled, and biotinylated with the GeneChip HT 3′IVT Express Kit (Affymetrix, Santa Clara, CA, USA). From the labeled and fragmented cRNA, 15 μg was hybridized with the GeneChip HT Mouse Genome 430 2.0 Array Plates (Affymetrix) and then washed, stained, and scanned automatically in a GeneTitan instrument (Affymetrix). The microarray chips can detect more than 39, 000 transcripts covering 34,000 genes. Hybridization artifacts and proper grid alignment were controlled by visual inspection of the scanned images. Raw data were obtained with AGCC 3.0 software (Affymetrix) and stored in CEL files. The software platform R 2.14.0 and Bioconductor (www. bioconductor.org) were used for further analysis starting with background correction of the complete expression information and Robust Multichip Average (RMA) normalization. F-statistics was applied (empirical Bayes model) and the resulting P values were further corrected for multiple testing using the BBenjamini-Hochberg^method. After this correction, transcripts with P values <0.05 were considered as significantly dysregulated. The complete microarray transcriptome results were deposited in the public database GEO (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE55177).

Unbiased Transcriptome Bioinformatics
STRING database interaction analysis The factors from Table 1 were loaded via the Multiple Names entry (http:// string.embl.de/). Automatized visualization of connections between calcium regulators was kept. Remaining random factors were manually grouped by function.
Gene Set Enrichment Analysis Gene symbols and M-values of microarray data from 12-week-old cerebellum were subjected to nonspecific filtering, and the remaining genes were analyzed using Gene Set Enrichment Analysis (GSEA) and the Java-based version GSEA-P [49,50]. For each comparison, the probe IDs were ranked according to the t test statistic. Probe IDs were collapsed to gene symbols. For duplicate entries, the maximum value was used. Permutations were performed on gene sets due to the low number of biological replicates. We used the c2 (online pathway databases, PubMed publications, expert of domain knowledge) genesets from the MSigDB database (v4.0, May 2013, http://www. broadinstitute.org/gsea/msigdb/index.jsp) to analyze the data sets.

RNA Isolation and Expression Analysis
For expression analysis, RNA was extracted from cerebellar tissue (25 mg) of the relevant mice with TRIzol ® reagent (Invitrogen) according to the manufacturers' protocol. Remaining DNA was digested with DNase I Amplification Grade (Invitrogen) before cDNA synthesis. Reverse transcription was performed with SuperScript III Reverse Transcriptase (Invitrogen) and expression levels were measured by RT-qPCR with the StepOnePlus Real-Time PCR System (Applied Biosystems). Therefore, cDNA from 25-ng RNA, 10 μl of FastStart Universal Probe Master (Rox) Mix (Roche), and 1 μl of one of the following TaqMan Assays (Applied Biosystems) were used for each reaction: Atp2a2 (Mm01201431_m1), Inpp5a (Mm00805812_m1), Itpr1 (Mm00439907_m1), Rora (Mm01173766_m1), as well as Tbp (Mm00446973_m1) and Hprt1 (Mm00446968_m1) as endogenous controls. PCR conditions were 50°C for 2 min, followed by 10 min at 95°C and 40 cycles of 95°C for 15 s

Microarray Transcriptome Profile of Atxn2 KO Cerebellum Shows Dysregulation of Several Factors Involved in Calcium Homeostasis Pathways
Attempting to gain deeper insight into the function of the RNA-binding protein ATXN2 and the cerebellar consequences of its deficiency, we performed global transcriptome analysis of Atxn2 KO mice. Cerebella from 6week-, 3-month-, and 6-month-old animals, in each case four wild-type (WT, Atxn2 +/+ ) versus four homozygous knockout (KO, Atxn2 −/− ) mice, were examined for mRNA level changes using a total of 24 Affymetrix oligonucleotide microarray chips. The data confirmed the absence of Atxn2 mRNA at six spots for all three ages with changes ranging from −51.99-fold to −1.36-fold (Table 1), with variance probably due to technical imprecisions in independent assessments. Apart from Atxn2, the analysis of significant downregulations with consistency across ages revealed a list of 126 transcript coding for 109 genes. Table 1 shows the 32 non-anonymous genes with changes stronger than −1.31-fold (corresponding to a decrease to 0.66, or log2-fold-changes/M-values ≤ −0.39) among them, ordered by significance of the adjusted P value. Interaction analysis at the String Heidelberg database highlighted the calcium homeostasis pathway to be affected (Supp. Fig. 1), but multiple factors of RNA processing (Riok3, Rbm28, Adarb1), bioenergetics (Aldh7a1, Gpd1), cell adhesion (Tanc1, Tmem56), growth (Icmt, Gdf10, Foxo3, Rassf3, Sbk1), and lipid signaling (Pip4k2b, Plcxd2, Rgs7bp) were also present in this short list. Further bioinformatics assessment of the complete transcriptome of 12-week-old cerebella by Gene Set Enrichment Analysis (GSEA) at the Broad Institute server also revealed significant downregulations in the KEGG adherens junction pathway of cell adhesion, the BIOCARTA PPARA pathway of lipid signaling, the BIOCARTA EGF pathway of growth signaling, and the BIOCARTA Creb pathway of nuclear response to extracellular triggers with Camk2b (calcium/calmodulindependent protein kinase II beta) as the strongest downregulation (Suppl. Fig. 2-5). These results are in good agreement with previous observations of phenotypic effects in the lipid metabolism, insulin, and EGF signaling [38,39,47]. The most significant change after Atxn2 was observed for Rora as a transcription factor that regulates several genes involved in calcium homeostasis besides having other effects (Table 1), Inpp5a and Itpr1 as two regulators of calcium flux and signaling besides other functions, as well as Prkcd and Anxa11 as calciumregulated targets were present among the top 32 genes, and age-consistent significant downregulations (−1.2fold) were observed at two individual microarray spots for the calcium flux regulator Atp2a2 (Table 1, bottom), as a possible correlate of altered calcium signal transmission at Purkinje neuron dendrites. Thus, this unbiased analysis of cerebellar transcriptome dysregulations with consistency through three ages advanced previous knowledge on altered RNA processing/bioenergetics/cell adhesion/growth/lipid signaling and documented novel prominent downregulations of mRNAs for proteins involved in calcium homeostasis and signaling that are caused by Atxn2 loss.

Downregulation of Factors Involved in Calcium Homeostasis Pathways is Confirmed Independently by qPCR
To validate the results of the transcriptome analysis and to investigate even younger ages as well as other brain regions in Atxn2 KO homozygous and heterozygous animals, an approach by RT-qPCR was used. The downregulation of Atp2a2, Inpp5a, Itpr1, and Rora mRNA levels in cerebellum of 6-month-old Atxn2 −/− animals was thus confirmed by an independent technique (Table 2; n ≥ 4 Atxn2 +/+ vs. ≥3 Atxn2 −/− ). Fold changes were similar to those found by the microarray profiling, apart from Itpr1 that showed a much smaller downregulation.
To test the temporal dynamics of these effects, 6-weekold Atxn2 −/− mice were studied (n ≥ 4 Atxn2 +/+ vs. ≥4 Atxn2 −/− ). To further analyze the effects independently in animals with reduced mutation load, heterozygous (Atxn2 +/− ) mice at the age of 6 weeks (n ≥ 4 Atxn2 +/+ vs. 4 Atxn2 +/− ) were analyzed. Finally, in order to test the earliest appearance of these effects in animals with reduced mutation load, heterozygous pups at~3 days of age (n = 4 Atxn2 +/+ vs. 4 Atxn2 +/− ) were assessed. As shown in Table 2, consistent results for all these conditions were observed, with significant downregulations of similar strength for Atp2a2 and Inpp5a, as well as downregulations of weaker strength for Rora. These observations indicate that changes in calcium homeostasis pathways are present very early even in heterozygous Atxn2 KO cerebellum.

Atxn2-CAG42-KIN Mice Show Subtle Transcript Level Changes in Calcium Homeostasis Pathways
Aware that RORA has been shown to be an important modifier of the cerebellar ataxia in SCA1 mouse models [52], we extended the analysis to the cerebellum of SCA2 mouse models. The Atxn2-CAG42-KIN (Atxn2 CAG42/CAG42 ) mouse cerebellum at an age of 18 months was assessed by RT-qPCR. At this age, the Atxn2-CAG42-KIN mice show the first behavioral deficits and visible aggregates of the ATXN2 protein in cerebellar Purkinje neurons [46]. In contrast to KO mice at this age, only Atp2a2 and Itpr1 were slightly but significantly downregulated in homozygous Atxn2-CAG42 KIN mice while Inpp5a and Rora did not show any change at all (n ≥ 4 Atxn2 +/+ vs. ≥4 Atxn2 CAG42/CAG42 ) ( Table 3).

Dysregulated Protein Levels in Calcium Homeostasis Pathways in Atxn2 KO and Atxn2-CAG42-KIN Mice
Despite the modest fold changes of the expression dysregulations observed, we attempted to validate the corresponding protein levels. In view of the limited availability of specific antibodies, only INPP5A and ITPR1 levels were assessed via quantitative immunoblots in cerebellar tissue from 6-month-old mice (8 Atxn2 +/+ vs. ≥7 Atxn2 −/− ). In the RIPA-soluble fraction (Fig. 1a) where mostly cytosolic factors will be detected, both proteins were significantly downregulated (fold changes −1.21-fold for INPP5A and −1.61-fold for ITPR1). In the SDS-soluble fraction (Fig. 1b), where membrane-associated proteins such as cortical actin or also aggregating proteins may be solubilized, only INPP5A showed significant downregulation (−1.31-fold). These data demonstrate that the relatively subtle transcript dysregulations are indeed resulting in substantially altered abundance of the corresponding proteins, making an impairment of calcium flux regulation in Atxn2 KO mouse cerebellum more credible.
In the KIN mouse line, we focused on ITPR1 only, since the Inpp5a mRNA levels were not dysregulated. Protein lysates from cerebellar tissue (18-month-old animals) were studied with quantitative immunoblots (8 Atxn2 +/+ vs. 7 Atxn2 CAG42/CAG42 ), demonstrating significantly diminished (−1.98-fold) ITPR1 protein levels in the RIPA-soluble fraction of Atxn2-CAG42-KIN mice (Fig. 2a). Conversely, significantly increased (1.28-fold) ITPR1 protein levels were observed in the SDS-soluble fraction of Atxn2-CAG42-KIN mice (Fig. 2b). In view of the previously demonstrated aggregation of ATXN2 protein in the SDS/urea fraction [46] and reports on a protein interaction between ATXN2 and ITPR1 [53,54], this accumulation of ITPR1 in relative insolubility with concomitant depletion from the soluble tissue fraction might reflect a sequestration of ITPR1 by polyQ-expanded ATXN2.

Normal and 42Q-ATXN2 do not Interact with ITPR1 at Endogenous Protein Levels
Given that a protein interaction between ITPR1 and overexpressed 58Q-expanded ATXN2 was previously observed [54], we went on to study both endogenous proteins in their association via Co-IP in the cerebellum of Atxn2-CAG42-KIN and WT mice at age 18 months, using Atxn2 KO tissue as negative control. Pulling was done with anti-ATXN2 on the one hand or with anti-ITPR1 on the other hand, performing pull-downs with unspecific IgG or using beads only as negative controls. ITPR1 (Fig. 3, panel above) and ATXN2 (panel below) at endogenous levels were readily detectable as input in tissues of all genotypes and were concentrated by its immunoprecipitation, providing the necessary quality control. However, ITPR1 bands were not brought down in detectable amounts by WT or Q42-ATXN2. These results suggest that ATXN2 polyQ expansion sizes such as Q42, which are found in most SCA2 patients, are unable at endogenous levels within the first years of pathology progression to mediate a strong sequestration of ITPR1.

ITPR1 Localization in Purkinje Cells is not Visibly Altered due to ATXN2 Expansion or Loss
In order to study the basis of the ITPR1 accumulation in relative insolubility among membrane-soluble factors in Atxn2-CAG42-KIN mice and of the ITPR1 abundance reduction in Atxn2 KO mice, immunohistochemical double stainings of ATXN2 and ITPR1 were performed in both mouse lines. Several sections from sagittal brain slices of 24-month-old WT and Atxn2-CAG42-KIN as well as of 6month-old WT and Atxn2 KO mice (two animals per line) were applied and stained for ATXN2 (red) and ITPR1 T statistical trend with P value between 0.05 and 0.10, n.s. non-significant, Cbll cerebellum *P < 0.05; **P < 0.01; ***P < 0.001 (green) (Fig. 4a, b). Qualitative assessment without sophisticated quantitative techniques was carried out. ATXN2 showed cytoplasmic distribution in Purkinje cells of both WT mice. In Atxn2-CAG42-KIN Purkinje neurons, ATXN2 was aggregated (arrow), as previously reported [46]. As expected, no specific fluorescent signal was detected for ATXN2 in Atxn2 KO cerebellum. ITPR1 exhibited a granular distribution in the cytoplasm as well as in the dendrites of Purkinje cells. Colocalization of ATXN2 and ITPR1 was therefore detected throughout the cytoplasm in WT cells, but the ITPR1 localization appeared largely unchanged when ATXN2 was expanded or lost. No obvious accumulation, relocalization, or markedly decreased intensity in the staining was apparent in the KIN or KO. Furthermore, no change in Purkinje cell soma size was detectable neither in Atxn2-CAG42-KIN nor in Atxn2 KO cerebellum. Thus, although less salt-soluble ITPR1 and more detergent-soluble ITPR1 was observed in Atxn2- ) were analyzed in the RIPA-soluble (a) and SDSsoluble (b) fraction. ITPR1 was decreased in the RIPA fraction, while it accumulated in SDS fraction CAG42-KIN mice, a sequestration of ITPR1 to aggregates or inclusion bodies was not detected, suggesting that the ITPR1 accumulation occurs in its normal membraneassociated form. Fig. 3 No interaction of ITPR1 with normal and expanded ATXN2 in Atxn2-CAG42-KIN cerebellum. Co-IPs pulling either with beads only, ATXN2, ITPR1, or unspecific IgG control managed to precipitate both ITPR1 (panel above) and ATXN2 (panel below) at endogenous levels, but showed no interaction of ITPR1 with wildtype or Q42expanded ATXN2 in immunoblots detecting ITPR1 Fig. 4 Colocalization of ITPR1 and ATXN2. Fluorescent double immunohistochemical stainings of ATXN2 (red) and ITPR1 (green) in a 24-month-old Atxn2-CAG42-KIN and b 6-month-old Atxn2 KO cerebellum. ATXN2 and ITPR1 colocalized in the cytoplasm of Purkinje cells. While ITPR1 staining intensity varied considerably among individual Purkinje neurons of the same cerebellum, the ITPR1 localization was not detectably influenced by ATXN2 polyQ expansion or loss

Discussion
Our global transcriptome survey of Atxn2 KO cerebellum defined significant downregulations of key molecules in several pathways where previous studies had demonstrated phenotypic changes, namely extracellular growth/adhesion signals via lipid and transcription changes. As a prominent novel effect with selectivity for cerebellum, the mRNA levels of calcium homeostasis pathway components Atp2a2, Inpp5a, Itpr1, and particularly Rora showed varying decreases. This insight may provide parameters to measure how the loss of physiological function of ATXN2 affects the integration of calcium signals in the dendritic trees of Purkinje dendrites. It is also useful to understand how the physiological function of ATXN2 is altered by polyQ expansions in SCA2. Thus, our data for the first time show clearly significant dysregulations in the abundance of several factors involved in calcium homeostasis pathways, in the cerebellum of a Q42-knock-in mouse model of SCA2, but these changes are mild even at old age and are probably not the sole basis for the marked pathology already manifest. These factors define a pathway of Purkinje neuron signaling. Climbing fiber and parallel fiber afferents excite Purkinje cell dendrites via glutamate receptors, resulting in the hydrolysis of phosphatidylinositol bisphosphate (PIP 2 ) into diacylglycerol (DG) and inositol 1,4,5-trisphosphate (InsP 3 ). InsP 3 then binds ITPR1 which thereupon releases Ca 2+ from the endoplasmic reticulum into the cytosol. The stimulator InsP 3 itself can be hydrolyzed by inositol polyphosphate-5-phosphatase (INPP5A) producing InsP 2 that is unable to activate the InsP 3 receptor. This mechanism inhibits further stimulation of calcium release [53]. The SERCA2 protein, encoded by Atp2a2, counteracts ITPR1 by translocating Ca 2+ back into the endoplasmic reticulum lumen [55]. Rora is of specific interest as a transcription factor that regulates several genes involved in calcium homeostasis and signaling like Calb1 (encoding Calbindin), Grm1 (encoding the metabotropic glutamate receptor, which influences calcium signaling through excitatory synaptic neurotransmission), and Itpr1 [56], among other effects. Overall, calciummediated excitation influences the duration, direction, extent, and type of synaptic plasticity, thus underlying the pathomechanism of diverse neurological diseases [57].
It is interesting to meditate about the varying mechanisms how this calcium pathway can be affected by different SCA disease proteins. An interaction with the transcription factor RORA was demonstrated for the nuclear factor Ataxin-1 [52]. In contrast, Ataxin-2 is a cytoplasmic protein with concentration at the rough endoplasmic reticulum. There, its overexpressed 58Q-expanded variant was observed to bind to ITPR1 [7,54,67]. Although these studies found the overexpression of INPP5A by adeno-associated virus over 17 weeks and until 10 months to alleviate the Purkinje neuron phenotypes and motor incoordination in mice with stable transgenic overexpression of a human CAG58-ATXN2 cDNA, our data in the GAG42-knock-in mouse model of SCA2 show no significant upregulation of the Inpp5a transcript, but instead consistent downregulations of the Atp2a2 mRNA to occur at ages before and after the manifestation of motor incoordination. It is impossible to predict the net effect of these changes in abundance of antagonistic calcium modulators without future functional studies of electrophysiology and calcium imaging in knock-in mice. The mild changes at the RNA level translate into substantial changes at the protein level: INPP5A protein is downregulated in the KO cerebellum. Furthermore, ITPR1 is markedly altered in its levels and solubility both by ATXN2-Q42 expansion and by ATXN2 deficiency. Given that a sequestration of ITPR1 into insolubility through protein interaction with expanded ATXN2 could not be substantiated in our coimmunoprecipitation and colocalization studies, the likeliest explanation would predict that the reduced amounts of Itpr1 mRNA result in reduced amounts of newly synthesized ITRP1 at ribosomes, soluble in the RIPA fraction, but the cells accumulate ITPR1 in its ER membrane-associated form, which is represented by the SDS fraction. This might be a compensatory stabilization effort of the cells to maintain calcium homeostasis. The previous observation of protein interactions between ITPR1 and ATXN2 by Liu and coworkers [54] may be explained by the longer polyQ expansion in ATXN2, by its overexpression, and by different experimental conditions with a more stringent lysis buffer.
An alternative pathomechanism instead of protein interaction is also conceivable. Expanded ATXN2 could associate with the Itpr1 mRNA and influence its translation efficiency. Recently, it was demonstrated that ATXN2 is involved in the stability of mRNA and protein expression by binding to AUrich sequences in the 3′UTR [68]. These authors furthermore found that a depletion of ATXN2 resulted in decreased mRNA stability and protein levels of specific targets. A polyQ expansion of ATXN2 did not abolish its function in this process but decreased its efficiency. In our analysis, the downregulations of Atp2a2, Inpp5a, Itpr1, and Rora mRNA levels in microarray transcriptomics and in RT-qPCR was specific for cerebellum in KO mice and was robustly detectable from 3 days to 6 months of age, even upon partial loss-of-function. Atp2a2 and Inpp5a transcript level changes were the first to be detected in KO cerebellum, even in heterozygous tissue. Conversely, in KIN cerebellum, the downregulation of Itpr1 levels was the most pronounced and earliest detectable change, while Atp2a2 was significantly downregulated from age 6 months onward. Thus, in Atxn2 KO mice, the loss of ATXN2 may result in decreased transcript and protein levels due to a decreased mRNA stability, while in Atxn2-CAG42-KIN mice, the mRNA stability may be partially decreased, compensatory efforts may stabilize the protein through modulation of its degradation speed, and additional pathology may occur in later disease stages or in stronger expansions due to ATXN2 accumulation/aggregation/sequestration.
The accumulation of ITPR1 in the membrane fraction in association with polyQ-expanded ATXN2 would affect Ca 2+ release, a feature that could be tested in young mice but is cumbersome to evaluate in aged brain tissues. This effort seems hard to justify, given that SCA2 involves not only the cerebellum but progresses into a multisystem-atrophy of the nervous system, and that our data suggest the calcium homeostasis pathway alteration not to be prominent in tissues like midbrain and motoneurons (data not shown). It will be particularly important to determine if the effects of ATXN2 polyQ expansions on the motor neuron degeneration in amyotrophic lateral sclerosis has shared pathogenesis pathways.
Our present findings clarify the position of ATXN2 effects in the emerging shared molecular pathways of spinocerebellar ataxias. Previous data had suggested a role of ATXN2 in SCA1, given that double mutant Drosophila melanogaster flies showed increased dAtx2 levels to enhance Ataxin-1[82Q]-induced neurodegeneration, while decreased dAtx2 levels suppress the neurotoxicity [69]. Furthermore, ATXN2 appears to have also a role in SCA3, given that dAtx2 activity hastens the onset of nuclear inclusions associated with SCA3 [70]. The dependence of ITPR1, INPP5A, ATP2A2, and RORA levels on ATXN2 mutations provides proof-ofprinciple of the existence of shared molecular changes early in the course of SCA2, SCA1, and SCA3.
Further assessment of additional Atxn2-KO mouse tissues by global deep proteomics substantiated the downregulation of iso-valeryl dehydrogenase (IVD), which had shown the second strongest transcript downregulation in this microarray transcriptomics analysis. The IVD protein reduction had nominal significance in cerebellum and showed also actual significance after multiple testing correction in liver tissue, where a decrease to 7 % of WT levels was observed in quantitative immunoblots. This effect was part of a systematic downregulation of amino acids and fatty acids pathways, which is thought to affect excitation and growth signals [71]. Given that deep proteomics is not sensitive toward quantitative changes of factors with very low abundance like transcription factors, and that tissue extractions even with 8 % SDS may solubilize nuclear proteins only partially, it is not surprising that RORalpha changes were not detected by this approach.