Insights From Cerebellar Transcriptomic Analysis Into the Pathogenesis of Ataxia

Abstract

IMPORTANCE

The core clinical and neuropathological feature of the autosomal dominant spinocerebellar ataxias (SCAs) is cerebellar degeneration. Mutations in the known genes explain only 50% to 60% of SCA cases. To date, no effective treatments exist, and the knowledge of drug-treatable molecular pathways is limited. The examination of overlapping mechanisms and the interpretation of how ataxia genes interact will be important in the discovery of potential disease-modifying agents.

OBJECTIVES

To address the possible relationships among known SCA genes, predict their functions, identify overlapping pathways, and provide a framework for candidate gene discovery using whole-transcriptome expression data.

DESIGN, SETTING, AND PARTICIPANTS

We have used a systems biology approach based on whole-transcriptome gene expression analysis. As part of the United Kingdom Brain Expression Consortium, we analyzed the expression profile of 788 brain samples obtained from 101 neuropathologically healthy individuals (10 distinct brain regions each). Weighted gene coexpression network analysis was used to cluster 24 SCA genes into gene coexpression modules in an unsupervised manner. The overrepresentation of SCA transcripts in modules identified in the cerebellum was assessed. Enrichment analysis was performed to infer the functions and molecular pathways of genes in biologically relevant modules.

MAIN OUTCOMES AND MEASURES

Molecular functions and mechanisms implicating SCA genes, as well as lists of relevant coexpressed genes as potential candidates for novel SCA causative or modifier genes.

RESULTS

Two cerebellar gene coexpression modules were statistically enriched in SCA transcripts (P = .021 for the tan module and P = 2.87 × 10⁻⁵ for the light yellow module) and contained established granule and Purkinje cell markers, respectively. One module includes genes involved in the ubiquitin-proteasome system and contains SCA genes usually associated with a complex phenotype, while the other module encloses many genes important for calcium homeostasis and signaling and contains SCA genes associated mostly with pure ataxia.

CONCLUSIONS AND RELEVANCE

Using normal gene expression in the human brain, we identified significant cell types and pathways in SCA pathogenesis. The overrepresentation of genes involved in calcium homeostasis and signaling may indicate an important target for therapy in the future. Furthermore, the gene networks provide new candidate genes for ataxias or novel genes that may be critical for cerebellar function.

Autosomal dominant cerebellar ataxias, also referred to as spinocerebellar ataxias (SCAs), are clinically heterogeneous, with an onset usually in adulthood. The SCAs are characterized by progressive cerebellar dysfunction, manifesting as progressive gait and limb incoordination, and can be associated with a wide range of neurological and nonneurological manifestations, including peripheral neuropathy, ophthalmoplegia, retinopathy, pyramidal and extrapyramidal signs, dementia, and epilepsy.^1,2

Approximately 35 SCA loci have been mapped (Table), but only 24 causative genes have been identified to date. The SCAs can be classified into the following 3 major categories according to the mutation type: (1) polyglutamine ataxias caused by exonic (CAG)n expansions encoding for polyglutamine tracts, (2) noncoding repeat ataxias caused by expansions of repeat motifs located in noncoding regions, and (3) ataxias caused by conventional mutations and copy number variants, including missense, nonsense, and splice-site mutations, deletions, and duplications.³ However, 40% to 50% of patients with SCAs do not have mutations in the known genes.⁴

Table. Spinocerebellar Ataxia Loci, Genes, and Mutations.

Ataxia	Locus	Gene	Type of Mutation	Protein or Complex	OMIM Accession No.
Coding Repeat Expansions a
DRPLA	12p13.31	ATN1	(CAG)n	Atrophin 1	125370
SCA1	6p23	ATXN1	(CAG)n	Ataxin 1	164400
SCA2	12q24	ATXN2	(CAG)n	Ataxin 2	183090
SCA3/MJD	14q32.1	ATXN3	(CAG)n	Ataxin 3	109150
SCA6	19p13	CACNA1A	(CAG)n	Calcium channel, voltage-dependent, P/Q-type, α-1A subunit	183086
SCA7	3p14	ATXN7	(CAG)n	Ataxin 7	164500
SCA17	6q27	TBP	(CAG)n	TATA box binding protein	607136
Noncoding Repeat Expansions b
SCA8	13q21.33	ATXN8, ATXN8OS	(CTG*CAG)n	Ataxin 8 and ATXN8 opposite strand (nonprotein coding)	608768
SCA10	22q13.31	ATXN10	(ATTCT)n	Ataxin 10	603516
SCA12	5q32	PPP2R2B	(CAG)n	Protein phosphatase 2, regulatory subunit B, β	604326
SCA31	16q21	BEAN1	(TGGAA)n	Brain expressed, associated with NEDD4, 1	117210
SCA36	20p13	NOP56	(GGCCTG)n	NOP56 ribonucleoprotein homologue (yeast)	614153
Other Types of Mutations c
SCA5	11q13.2	SPTBN2	Point mutations	Spectrin, β, nonerythrocytic 2	600224
SCA11	15q15.2	TTBK2	Point mutations	Tau tubulin kinase 2	604432
SCA13	19q13.33	KCNC3	Point mutations	Potassium voltage-gated channel, Shaw-related subfamily, member 3	605259
SCA14	19q13.42	PRKCG	Point mutations	Protein kinase C, λ	605361
SCA15/16/29	3p26.1	ITPR1	Point mutations, large deletions	Inositol 1,4,5-trisphosphate receptor, type 1	606658, 117360
SCA18	7q31.1	IFRD1	Point mutations	Interferon-related developmental regulator 1	607458
SCA19/22	1p13.2	KCND3	Point mutations, small deletions	Potassium voltage-gated channel, Shal-related subfamily, member 3	607346
SCA20	11p11.2-q13.3	Not applicable	Genomic duplication	Region with ≥12 genes	608687
SCA23	20p13	PDYN	Point mutations	Prodynorphin	610245
SCA26	19p13.3	EEF2	Point mutations	Eukaryotic translation elongation factor 2	609306
SCA27	13q34	FGF14	Point mutations	Fibroblast growth factor 14	609307
SCA28	18p11.21	AFG3L2	Point mutations	AFG3 adenosine triphosphatase family gene 3–like 2 (Saccharomyces cerevisiae)	610246
SCA35	20p13	TGM6	Point mutations	Transglutaminase 6	613908
Unknown Mutations
SCA4	16q22.1	?	?	?	600223
SCA21	7p21.3-p15.1	?	?	?	607454
SCA25	2p21-p13	?	?	?	608703
SCA30	4q34.3-q35.1	?	?	?	613371
SCA32	7q32-q33	?	?	?	613909
SCA34	6p12.3-q16.1	?	?	?	133190
SCA37	1p32	?	?	?	HGNC 43726

Abbreviations: DRPLA, dentatorubral-pallidoluysian atrophy; HGNC, HUGO Gene Nomenclature Committee; MJD, Machado-Joseph disease; OMIM, Online Mendelian Inheritance in Man; SCA, spinocerebellar ataxia.

^aPolyglutamine ataxias.

^bNoncoding repeat ataxias.

^cAtaxias caused by conventional mutations.

The paradigm to investigate rare mendelian disorders is evolving as a result of the availability of next-generation sequencing, which has facilitated the identification of causal genes. Several SCA genes were recently identified using next-generation sequencing (eg, SCA19/22 and SCA35).^5-7 The interpretation of next-generation sequencing data represents a major ongoing challenge, particularly in distinguishing causal mutations from the thousands of benign variants present in every exome (often >20 000 variants). Therefore, it is important to develop frameworks to prioritize candidate mutations of monogenic diseases in these large data sets.

One way forward is to use the expression and function of known SCA genes to predict novel genes. Although neuronal loss can be widespread in some SCAs (eg, SCA3), certain brain regions (eg, cortical regions) are usually less affected (reviewed by Seidel et al⁸). The pathological features of SCAs are most prominent in the cerebellum, which is characteristically atrophic despite the fact that SCA genes are ubiquitously expressed in the brain. This would suggest that, in common with many other neurodegenerative disorders, regional differences in gene expression within the human central nervous system are insufficient to explain the observed selective vulnerability of neurons. Furthermore, SCA genes have a wide range of functions, including ion transport, deubiquitination, dephosphorylation, phosphorylation, transcriptional regulation, translational elongation, and others, rendering it surprising that mutations in this diverse set of genes can give rise to a consistent phenotype. A study by Lim et al⁹ revealed that some ataxia-associated proteins share interacting partners, suggesting that phenotypes shared among ataxias may arise from their involvement in common molecular pathways. We hypothesized that expression patterns of 24 known SCA-associated genes in the human brain and relevant molecular pathways may be used to prioritize novel candidate genes and identify disease modifiers. In addition, this strategy may be used to identify potential therapeutic targets, as was found for the α-subunit of eukaryotic translation initiation factor (eIF2) in prion neurodegeneration.¹⁰

To date, limited information exists regarding the expression and function of SCA genes within the human brain. We performed expression profiling of 788 brain samples obtained from 101 neuropathologically healthy individuals as part of the United Kingdom Brain Expression Consortium and used weighted gene coexpression network analysis (WGCNA) to group genes into modules in an unsupervised manner.^11-14 This approach has proven useful in identifying modules of biologically related genes that are not only coexpressed but also coregulated.^13,15-17 We used a systems biology approach^18,19 based on whole-transcriptome gene expression analysis to address possible relationships among known SCA genes, predict their functions, and propose novel candidate genes to aid the discovery of new SCA genes. Our data revealed 2 significant SCA transcript–enriched coexpression modules that provide testable hypotheses about the function of proteins that may interact with known SCA genes.

Methods

Human Brain Samples and Analysis of Gene Expression Arrays

All 101 samples had received fully informed consent and were authorized for ethically approved scientific investigation by the National Hospital for Neurology and Neurosurgery and Institute of Neurology Research Ethics Committee. Brain samples were collected by the Medical Research Council Sudden Death Brain and Tissue Bank.²⁰ The 101 individuals had been neuropathologically healthy. Total RNA was isolated and processed for array analysis (Exon 1.0 ST; Affymetrix UK Ltd) as previously described (National Center for Biotechnology Gene Expression Omnibus GSE46706).^21,22 Samples were randomized for all experimental procedures, from RNA extraction to array hybridization, and array data were thoroughly investigated for technical quality as detailed in the eMethods on the authors’ website (http://www.ucl.ac.uk/ukpdc/publications/data/Bettencourt_et_al_2014_JAMA_Neurol_Supplement.docx).

Gene expression patterns throughout development and aging were assessed using data from the Human Brain Transcriptome project (http://hbatlas.org).^23,24 The Biological General Repository for Interaction Datasets, version 3.2.106 (http://thebiogrid.org),²⁵ was used for the analysis of SCA interactors.

Weighted Gene Coexpression Network Analysis

The SCA genes and transcripts were assigned to modules (arbitrary colors) identified through WGCNA on whole-transcriptome gene expression. In total, 15 409 transcripts (13 706 genes) passing quality control were used to identify modules as previously described,²² and 3743 additional transcripts (3541 genes) were assigned to modules based on their highest module membership (MM).

Briefly, the WGCNA network^11,14 was constructed for each tissue using a signed network with β power of 12 to achieve a scale-free topology. A dissimilarity matrix based on topological overlap measure, a pairwise measure of node similarity, was used to identify gene modules (ie, densely interconnected and coexpressed genes) through a dynamic tree-cutting algorithm.

Module preservation statistics²⁶ were calculated (z score) to assess how well modules from one tissue are reproducible (or preserved) in another tissue. Previously proposed thresholds²⁶ were considered (z score of <2 indicates no evidence of module preservation, z score between 2 and <10 indicates weak to moderate evidence, and z score of ≥10 indicates strong evidence).

The hypergeometric distribution was used to evaluate SCA transcript–enrichedmodules(P < .05 was considered significant). Only modules with at least 2 SCA transcripts were considered.

Gene Enrichment Analysis

DAVID Bioinformatics Resources, version 6.7 (http://david.abcc.ncifcrf.gov/home.jsp),^27,28 was used to evaluate the biological and functional relevance among SCA genes and genes within SCA transcript–enriched modules. The overrepresentation of gene ontology (GO) categories and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways was examined (Bonferroni-corrected P < .05 was considered significant). oPOSSUM, version 3.0 (http://opossum.cisreg.ca/oPOSSUM3),^29-31 was used for enrichment analysis on conserved transcription factor binding sites (TFBSs) in SCA genes (cutoffs were a z score of ≥10 and a Fisher exact test score of ≥7).

An overview of the coexpression analysis work flow and associated data sets is shown in eFigure 1 on the authors’ website. Further details are given in the eMethods.

Results

Forty Percent of SCA Genes Are Most Highly Expressed in the Cerebellum

Globally, SCA genes are highly expressed in the human brain throughout development and aging. We found that approximately 40% of SCA genes (SPTBN2, CACNA1A, ATXN7, KCNC3, ITPR1, TBP, KCND3, FGF14, NOP56, and ATN1) have the highest expression in the cerebellum, but 2 genes (PPP2R2B and PDYN) exhibit among the lowest levels in this region (eTable 1 on the authors’ website). Several genes have highly variable expression levels across brain regions, including all SCA genes with ion channel activity (CACNA1A, KCNC3, ITPR1, and KCND3) and modulators of ion channel activity (SPTBN2 and FGF14),^32,33 which exhibit the highest expression levels in the cerebellum, followed by cortical regions, and the lowest expression in the white matter, suggesting that expression patterns may relate to gene function.

NR4A2 TFBSs Are Enriched Around SCA Genes

We had hypothesized that a feature common to SCA genes might be the coregulation of gene expression. Analysis of conserved TFBSs showed shared TFBSs across many SCA genes (up to 21 of 23 genes), suggesting that common transcription factors may be involved in the coregulation of these genes. We found that only TFBSs for NR4A2 are significantly overrepresented nearby SCA genes (z score, 10.4; P = .0003), with a mean of 7 TFBSs per gene (21 of 23 genes, absent only around AFG3L2 and BEAN1).

WGCNA Identifies 2 SCA Transcript–Enriched Modules in the Cerebellum

To gain insights into the functional organization of the brain transcriptome, we used WGCNA²² focusing on SCA genes (Table). Except for TGM6 (SCA35), which was not robustly detected, all other SCA genes were assigned to gene coexpression modules in 10 distinct brain regions (eTable 2 on the authors’ website). We were particularly interested in cerebellar gene MM because this region is most affected in these conditions.

Twenty-three gene coexpression modules were identified in the cerebellum, 14 of which contain at least 1 SCA transcript. Two of these modules include 4 SCA genes (eTable 2 on the authors’ website), namely, the tan module (ATXN3, ATXN10, TTBK2, and AFG3L2) and the light yellow module (ATXN8OS, PRKCG, ITPR1, and KCND3). This clustering was statistically significant (P = .021 for the tan module and P = 2.87 × 10⁻⁵ for the light yellow module).

Whereas the tan module seems to be preserved across all brain regions (mean z score, 23.19), the light yellow module is poorly preserved and seems to be unique to the cerebellum (mean z score, 2.26) (Figure 1). Consistent with these findings, we note that SCA genes contained within the light yellow module give rise to more pure ataxia phenotypes despite being ubiquitously expressed in the human brain, whereas those in the tan module give rise to more complex phenotypes.

Figure 1. z Scores Indicating Whether Genes in the Coexpression Modules Identified in the Cerebellum Cluster in Other Brain Regions.

z Scores of less than 2 (bottom red line) indicate no evidence of preservation (the case for the cerebellar light yellow module in most brain regions), while scores exceeding 5 (green line) and exceeding 10 (upper red line) indicate moderate and strong module preservation across brain regions, respectively.

SCA Genes Have High MM Within SCA Transcript–Enriched Modules

To determine the relevance of each gene in the SCA transcript–enriched modules, we estimated the MM for all genes within these modules (eTable 3 and eTable 4 on the authors’ website). The MM measures how well the expression pattern of each gene within the module correlates with the eigengene (first principal component of gene expression for the module). Except for KCND3 (17th quantile), all SCA genes have an MM above the median in the tan module (TTBK2 [78th quantile], AFG3L2 [75th quantile], ATXN10 [61st quantile], and ATXN3 [61st quantile]) and in the light yellow module (ITPR1 [96th quantile], PRKCG [83rd quantile], and ATXN8OS [79th quantile]), with the ITPR1 gene being a hub gene. These data suggest that genes with high MM may constitute good candidate genes for ataxias of yet unknown cause. For example, these data indicate that 2 members of the light yellow module, C7orf16 (7p15 [94th quantile]) and GPR63 (6q16.1-q16.3 [78th quantile]), which we note are located within the SCA21 and SCA34 loci, respectively, are particularly promising candidate genes.

SCA Transcript–Enriched Modules Contain Additional Ataxia-Related Genes

The tan module contains genes involved in other autosomal dominant and recessive ataxia syndromes, including PRNP (prion disease gene, OMIM 176640), SACS (autosomal recessive spastic ataxia of Charlevoix-Saguenay, OMIM 604490), AFG3L2 (both SCA28 and autosomal recessive spastic ataxianeuropathy syndrome,³⁴ OMIM 604581), and MTPAP (autosomal recessive spastic ataxia 4, OMIM 613669). Also present in the tan module are genes associated with other neurodegenerative disorders (eg, APP [OMIM 104760], VAPB [OMIM 605704], and VCP [OMIM 601023]), some of which exhibit cerebellar degeneration as a pathological feature (eg, POLR3A [OMIM 614258], POLR3B [OMIM 614366], EIF2B1 [OMIM 606686], and EIF2B2 [OMIM 606454]). The light yellow module also includes genes associated with additional ataxia syndromes (eg, CA8 [OMIM 114815] and TRPC3 [OMIM 602345]) and with other neurological diseases (eg, LARGE [OMIM 603590] and KCNMA1 [OMIM 600150]).

We investigated gene interconnections within SCA transcript–enriched modules using VisANT (http://visant.bu.edu).³⁵ Genes with topological overlap measure values exceeding 0.01 were considered connected. For the tan module, no SCA genes passed this cutoff. For the light yellow module, among the genes in the network (Figure 2) are CA8 (gene with the top MM) and TRPC3 (57th quantile), both previously known to interact with ITPR1 and associated with ataxia syndromes.^36,37

Figure 2. Network Representation of the Light Yellow Module in the Cerebellum.

Shown are all genes connected with a topological overlap measure exceeding 0.01. The spinocerebellar ataxia genes in this module are highlighted in pink, together with all genes that are directly connected to it based on the topological overlap measure cutoff used. Genes previously reported as interactors of spinocerebellar ataxia genes are highlighted in blue.

SCA Transcript–Enriched Modules Contain Established Granule and Purkinje Cell Markers

We used previously published cell-specific expression signatures³⁸ to obtain insights into the cell types relating to SCA transcript–enriched modules in the human cerebellum. These signatures indicated that only genes specifically associated with granule cells (C12orf24 [91st quantile], THYN1 [86th quantile], ZDHHC13 [85th quantile], BLOC1S2 [66th quantile], PARP2 [57th quantile], and C14orf101 [53rd quantile]) show an MM above the median in the tan module. On the other hand, only genes specifically associated with Purkinje cells show high MM in the light yellow module (SMPDL3B [89th quantile], LARGE [87th quantile], and PDE5A [73rd quantile]), and the enrichment in these cell markers was statistically significant (P = 3.752 × 10⁻⁴). No astrocyte-associated or oligodendrocyte-associated genes were present in the 2 SCA transcript–enriched modules. Therefore, while the tan module is more closely related to gene expression signatures of granule cells, the light yellow module is associated with Purkinje cells (eFigure 2 on the authors’ website) and may contribute to the specificity of the neuropathology.

SCA Interactome Network vs SCA Transcript–Enriched Expression Modules Were Searched

We next searched whether genes in these SCA transcript–enriched modules have already been deposited in the Biological General Repository for Interaction Datasets²⁵ as interactors of SCA genes. Several genes in the modules are previously known SCA interactors (eTable 5 on the authors’ website), some of which are also associated with neurodegenerative diseases, including ataxias (eg, APP, VCP, CA8, and TRPC3).

Some of these interactors (eg, BECN1 and PICK1)^39,40 are established modulators of SCA-related phenotypes, which suggests that the SCA interactome network may aid the identification of genetic modifiers and candidate genes for SCAs.

SCA Transcript–Enriched Modules Implicate the Ubiquitin-Proteasome System and Calcium Signaling in Disease Pathogenesis

Enrichment analysis for the tan module shows an overrepresentation of multiple GO biological processes mostly related to protein catabolic processes, the regulation of ubiquitin protein ligase activity, and intracellular transport (eTable 6 on the authors’ website). Two KEGG pathways, proteasome and ubiquitin-mediated proteolysis, are overrepresented. The latter pathway is also enriched in previously reported SCA interactors (eTable 5 on the authors’ website). To date, no SCA genes have been assigned to these pathways. However, ATXN3 is known to act as a deubiquitinating enzyme in the ubiquitinproteasome pathway,⁴¹ and other ataxia genes are involved in this pathway (eFigure 3 on the authors’ website).

In the light yellow module, neuron projection (GO cellular component) and gated channel activity (GO molecular function) are overrepresented. The KEGG calcium signaling pathway is also significantly enriched. The KEGG long-term depression includes the same SCA genes but did not reach statistical significance (Bonferroni-corrected P = .051). However, many SCA genes have not been assigned to these canonical pathways (only 6 of 24 genes according to results from DAVID Bioinformatics Resources). Based on cerebellar light yellow module genes and their functions, our data suggest that a single signaling pathway interconnects many SCA genes and other ataxia-related genes (Figure 3).

Figure 3. Proposed Molecular Mechanisms Contributing to Intracellular Calcium Homeostasis and Signaling in Neurons Depicting the Involvement of Genes Known to Cause Ataxia and Genes in the Cerebellar Light Yellow Coexpression Module.

Genes are shown in italics: red indicates human ataxia genes (underlined if in the light yellow module) and purple indicates additional relevant genes in the light yellow module (underlined if involved in ataxia phenotypes in mice). The red star represents calcium ions (Ca²⁺). Red boxes highlight proteins encoded by SCA genes; yellow boxes represent proteins encoded by other genes in the light yellow module; pale yellow boxes show relevant proteins although not found in our data. KCN indicates Potassium channels; CACN, Voltage-gated calcium channels; EAATs, Excitatory amino-acid transporters; GluR, Glutamate receptors; GPCR, G protein-coupled receptors; SOC, Store-operated calcium channels; PLC, Phospholipase C; PKC, Protein kinase C; IP3, Inositol 1,4,5-triphosphate; IP4, Inositol 1,3,4,5-tetrakisphosphate; IP3R, Inositol 1,4,5-triphosphate receptors; DAG, diacylglycerol; STIM, Stromal interaction molecule; SERCA, Sarcoendoplasmic reticulum (SR) calcium transport ATPase; ER/SR, Endo/sarcoplasmic reticulum; mut, mutated; PolyQ^mut, proteins with expanded polyglutamine tracts, namely those encoded by mutated ATXN1, ATXN2, ATXN3, CACNA1A, ATXN7, TBP, and ATN1.

Discussion

Different SCA subtypes share a core set of clinical and pathological features. To determine whether molecular pathways can explain such similarities, we have performed the first comprehensive transcriptomic analysis to date of 24 SCA genes in the human brain. We have shown that SCA genes have specific patterns of expression among brain regions, with those involved in ion channel activity having the highest expression in the cerebellum.

Using genome-wide expression data and WGCNA, we assigned all known SCA genes to expression modules without making a priori assumptions regarding their functions. Our results show that 2 gene coexpression modules are enriched for SCA genes and relate to expression profiles within granule and Purkinje cells of the cerebellum. These modules contain known SCA interactors, validating the importance of these expression modules in SCAs.

Because little is known about the function of most SCA genes, many have not been yet assigned to canonical pathways. The present gene coexpression network analysis using a hypothesis-free approach implicates roles for the ubiquitinproteasome system (UPS) in granule cells and calcium homeostasis and signaling in Purkinje cells. The significant enrichment of TFBSs for NR4A2 (or NURR1) near SCA genes also links to calcium regulation. NR4A2 encodes for an orphan nuclear receptor that is regulated by neural activity through voltage-dependent calcium channels and calcineurin and has been implicated in the pathogenesis of Parkinson disease.^42,43 In addition to data revealed on the known SCA genes, this work is important to future studies identifying new variants and proving their pathogenicity.

The UPS is involved in the turnover of multiple cellular proteins, enabling cells to dispose of biologically nonuseful proteins (eg, mutant, misfolded, and overaccumulated proteins). In addition, the UPS is implicated in controlling gene transcription and protein expression in signal transduction systems as well as in neural synapse (reviewed by Lehman ⁴⁴). Our network analysis revealed that biological processes related to the regulation of transcription and protein catabolic process are relevant for SCAs. Protein misfolding is recognized as a key feature of neurodegenerative diseases. Alzheimer disease, Parkinson disease, amyotrophic lateral sclerosis, prion disease, polyglutamine diseases (including polyglutamine SCAs), and possibly others (eg, SCA11) are proteinopathies, in which a particular protein or set of proteins misfolds and aggregates. In such aggregates, besides the mutated protein, additional proteins are sequestered, including ubiquitin, proteasome components, and transcription factors.⁴⁵ This suggests that, in these diseases, the UPS function may be compromised in the presence or absence of mutations in its components. The phenotype of UPS-associated ataxia genes is of a complex cerebellar ataxia with additional features, such as prominent cognitive dysfunction or pyramidal signs.

Disruption of calcium signaling has been proposed as a common mechanism in the pathogenesis of SCAs^46-48 and other neurodegenerative diseases.^49-51 Our data support this hypothesis, showing that this pathway is important in the SCA transcript–enriched coexpression module relevant to Purkinje cells (light yellow module) and may explain selective neuropathology. Purkinje cells are important target cells in the pathogenesis of many SCAs, and calcium signaling is crucial for their normal cellular function. Abnormal calcium levels in Purkinje cells are thought to uncouple plasticity and activate toxic cascades, resulting in cell death.⁴⁷ In this module, ITPR1 is a hub gene, and ITPR1-dependent signaling has been previously proposed as an important link among several ataxias.⁴⁸ The CA8 (quadruple-gait ataxia gene) product, also a highly interconnected gene in the same module, is thought to be an ITPR1 antagonist. Evidence also exists that the PRKCG product may be activated by robust activation of ITPR1.⁴⁸ Although located in a different coexpression module, the CACNA1A gene, which encodes P/Q-type calcium channel, also modulates ITPR1-dependent plasticity.⁴⁸ It has further been shown that mutant ataxin 1 (ATXN1), ataxin 2 (ATXN2), ataxin 3 (ATXN3), and possibly other pathogenically expanded polyglutamine proteins interact with ITPR1,⁴⁷ suggesting that proteins not normally involved in this pathway can also disrupt ITPR1-dependent signaling. The phenotype of the ataxias associated with calcium signaling abnormalities is usually a pure ataxia, with a few exceptions (eg, CA8 mutations, which are associated with mild mental retardation and an unusual gait). Drugs targeting the stabilization of calcium levels, the ITPR1 directly, or even the trial of calcium agonists in cell models may be useful therapies and might represent potential modifiers of this disease group regardless of the genetic cause.

Conclusions

Normal expression of SCA genes in the human brain reveals pathways linking SCA genes and other ataxia genes. Some of these links have been previously proposed, while others suggest potential new targets for genetic and functional testing. An important next step will be to analyze the expression in different brain regions from individuals affected with the common SCA subtypes. These networks provide valuable lists of candidate genes for diseases with overlapping phenotypes, which are good candidates for novel genetic modifiers.