Clinical, genetic and neuropathological characterization of spinocerebellar ataxia type 37

Abstract

The autosomal dominant spinocerebellar ataxias (SCAs) consist of a highly heterogeneous group of rare movement disorders characterized by progressive cerebellar ataxia variably associated with ophthalmoplegia, pyramidal and extrapyramidal signs, dementia, pigmentary retinopathy, seizures, lower motor neuron signs, or peripheral neuropathy. Over 41 different SCA subtypes have been described evidencing the high clinical and genetic heterogeneity. We previously reported a novel spinocerebellar ataxia type subtype, SCA37, linked to an 11-Mb genomic region on 1p32, in a large Spanish ataxia pedigree characterized by ataxia and a pure cerebellar syndrome distinctively presenting with early-altered vertical eye movements. Here we demonstrate the segregation of an unstable intronic ATTTC pentanucleotide repeat mutation within the 1p32 5′ non-coding regulatory region of the gene encoding the reelin adaptor protein DAB1, implicated in neuronal migration, as the causative genetic defect of the disease in four Spanish SCA37 families. We describe the clinical-genetic correlation and the first SCA37 neuropathological findings caused by dysregulation of cerebellar DAB1 expression. Post-mortem neuropathology of two patients with SCA37 revealed severe loss of Purkinje cells with abundant astrogliosis, empty baskets, occasional axonal spheroids, and hypertrophic fibres by phosphorylated neurofilament immunostaining in the cerebellar cortex. The remaining cerebellar Purkinje neurons showed loss of calbindin immunoreactivity, aberrant dendrite arborization, nuclear pathology including lobulation, irregularity, and hyperchromatism, and multiple ubiquitinated perisomatic granules immunostained for DAB1. A subpopulation of Purkinje cells was found ectopically mispositioned within the cerebellar cortex. No significant neuropathological alterations were identified in other brain regions in agreement with a pure cerebellar syndrome. Importantly, we found that the ATTTC repeat mutation dysregulated DAB1 expression and induced an RNA switch resulting in the upregulation of reelin-DAB1 and PI3K/AKT signalling in the SCA37 cerebellum. This study reveals the unstable ATTTC repeat mutation within the DAB1 gene as the underlying genetic cause and provides evidence of reelin-DAB1 signalling dysregulation in the spinocerebellar ataxia type 37.

Introduction

The autosomal dominant spinocerebellar ataxias (SCAs) are rare inherited movement disorders mainly characterized by progressive cerebellar ataxia variably associated with ophthalmoplegia, pyramidal and extrapyramidal signs, dementia, pigmentary retinopathy, seizures, lower motor neuron signs, and peripheral neuropathy (Matilla-Dueñas et al., 2006; Durr, 2010; Klockgether, 2011; Sun et al., 2016). Disease onset is typically in adulthood, albeit some clinical signs can appear earlier. Over 41 well-defined SCA subtypes have been described evidencing the high clinical and genetic heterogeneity (Nibbeling et al., 2017). Primary disease signs include cerebellar ataxia, dysarthria and oculomotor abnormalities, and secondary signs include parkinsonism, dysphagia, retinal degeneration, psychiatric disturbances or manifestations, sensory deficits and deafness, with each varying depending on the genetic SCA subtype. A minority of SCAs exhibit signs and degenerative patterns allowing for a clear and unambiguous diagnosis of the disease; for instance, retinal degeneration in SCA7, tau aggregation in SCA11, dentate calcification in SCA20, protein depositions in the Purkinje cell layer in SCA31, azoospermia in SCA32, and neurocutaneous phenotype in SCA34 (reviewed in Seidel et al., 2012). Genetics has a prominent role in the physiopathology of the SCAs having identified polyglutamine expansions, non-coding DNA expansions, and conventional mutations as the main triggering defects contributing to the gain- and loss-of-function pathogenic mechanisms underlying neurodegeneration (Durr, 2010; Sánchez et al., 2013, 2016; Matilla-Dueñas et al., 2014).

We recently described a new spinocerebellar ataxia subtype, SCA37, in a large Spanish kindred and localized the causative genetic deficit to an 11-Mb genomic region on 1p32, which includes the gene encoding for the reelin adapter protein disabled-1 (DAB1) (Serrano-Munuera et al., 2013). Affected patients presented with a pure cerebellar phenotype and severe alterations of vertical eye movements, including saccade accuracy and pursuit, and optokinetic nystagmus velocity. These distinctive features appeared early at the pre-symptomatic stage preceding horizontal eye abnormalities or even the ataxic signs.

In this study, we demonstrate the unstable intronic ATTTC pentanucleotide repeat expansion mutation within the DAB1 gene as the genetic cause in our cohort of SCA37 patients in four independent families originating from the south of Spain. We describe the first neuropathological findings and demonstrate the effects of the ATTTC repeat mutation on the dysregulation of DAB1 expression in the SCA37 cerebellum. Moreover, we discuss the genetic and clinical correlations, compare our results with the previous neuropathological studies reported in other SCA subtypes, and provide novel insights into the role of the reelin-DAB1 signalling pathway in the molecular pathogenesis underlying cerebellar pathology in SCA37.

Patients and methods

This study was conducted according to the ethical principles for medical research involving human subjects according to the Declaration of Helsinki.

Patients

Detailed clinical data from the AT-901 SCA37 family was previously reported (Serrano-Munuera et al., 2013). We later identified a second pedigree with the SCA37 phenotype linked to the same locus (AT-9012 family). Family AT-9012 was assessed following the same clinical protocol applied to Family AT-901. Clinical information was obtained from 27 members, whereas 18 were clinically examined, five affected and 13 healthy. More recently, two additional pedigrees (AT-59 and AT-90) from the same area in the south of Spain with 10 affected patients with a similar clinical phenotype and three asymptomatic relatives were identified and included in this study. The four SCA37 pedigrees studied are shown in Supplementary Fig. 1.

Cerebellar volume and mid-sagittal vermis relative diameter quantification

To quantify cerebellar volume, we used the CERES algorithm (Romero et al., 2017) to calculate the percentage of the cerebellar volume relative to the total intracranial volume on T₁-weighted 1.5 T MRI images from five patients (Patients III:8, IV:1 and III:1 from Family AT-90; and Patients IV:4 and IV:5 from Family AT-901). Normal boundaries were used from 30 age- and gender-matched controls (age range: 24–75 years) randomly selected from the open access IXI dataset previously reported (Manjón and Coupé, 2016). To quantify the mid-sagittal vermis relative diameter, the total posterior cranial fossa diameter was measured in a linear segment from the posterior commissure to the opisthion and the largest sagittal diameter of the cerebellum parallel to the previous linear segment in three patients (Patients IV:4 and IV:5 from Family AT-901; and Patient III:6 from Family AT-90) and three asymptomatic subjects carrying the SCA37 mutation (Subjects V:3 and VI:2 from Family AT-901, and Subject IV:8 from Family AT-9012). The ratio of the cerebellar vermis diameter over the total posterior cranial fossa diameter was calculated (Serrano et al. 2015) and compared to 16 age- and gender-matched controls.

Genetic and genomic studies

Informed consent was obtained for all individuals included in the study, which was approved by the ethical board of the University Hospital Germans Trias i Pujol (HUGTiP) in Badalona, the University Hospital Virgen del Rocío in Seville, or the Fundación CIEN in Madrid. DNA samples were obtained and studied from 60 individuals, 25 affected and 35 healthy or asymptomatic, from the four SCA37 families, and 96 DNA samples from our cohort of ataxia patients.

DNA was extracted with the Chemagen Magnetic Separation Module I automated system (Perkin Elmer) from peripheral blood. Whole-genome sequencing studies were performed in two SCA37 patients (Patients IV:9 and V:9) from Family AT-901 (detailed methodology is described in the Supplementary material). Two SCA37-linked SNPs identified by whole-genome sequencing, rs79992829 and rs146472695, both with a low minor allele frequency (MAF) of 0.0056 were used for the identification of three additional SCA37 pedigrees (Families AT-9012, AT-59, and AT-90) from our cohort of ataxia kindreds. The genetic haplotypes were determined using genetic markers on the 1p32 region (Supplementary Table 1). All members from the four SCA37 pedigrees were genotyped for the ATTTC expansion by PCR Sanger sequencing with the BigDye® Terminator v3.1 cycle Sequencing Kit (Thermofisher Scientific) as described (Seixas et al., 2017).

Neuropathology

Post-mortem brains from two clinically and genetically confirmed SCA37 relatives (Supplementary Fig. 1, Subjects IV:9 and IV:10, Family AT-901) from the large previously reported SCA37 Spanish pedigree (Serrano-Munuera et al., 2013) and one gender- and age-matched control brain with no medical history of neurological disease were studied. Brain samples were processed according to a common post-mortem protocol followed by the Spanish Banco de Tejidos de la Fundación CIEN (BTCIEN, Madrid, Spain) as described in the Supplementary material. Primary antibodies used were anti-Calbindin-D28K (EG-20; Sigma-Aldrich), anti-DAB1 (PA5-62538; Thermo Fisher Scientific), anti-human neurofilament protein (clone 2F11; Dako), anti-GFAP (Z0334; Dako), and anti-Ubiquitin (Z0458, Dako). After adding primary antibody diluted in Lab Vision™ Antibody Diluent OP Quanto (Thermo Fisher Scientific), sections were incubated overnight at 4°C. Sections were then rinsed in 0.1% Tween in phosphate-buffered saline (PBS) before incubation with secondary biotinylated rabbit anti-mouse IgG (H+L) Superclonal™, biotinylated goat anti-rabbit IgG (H+L) Superclonal™ antibody (Thermo Fisher Scientific) or secondary fluorescent, Alexa Fluor® 594 donkey anti-rabbit or Alexa Fluor® 488 goat anti-mouse (Jackson ImmunoResearch), for 1 h at room temperature and rinsed with 0.1% Tween in PBS. For immunohistochemistry studies, slides were incubated 30 min with Ultra-Sensitive ABC Peroxidase Standard Staining Kit (Thermo Fisher Scientific) and rinsed with 0.1% Tween in PBS followed by a 30-min incubation with 3,3′-diaminobenzidine (DAB) and stable peroxide substrate buffer. Sections were washed in water and incubated 5 min with DAB enhancer (Dako), counterstained with haematoxylin and eosin, dehydrated, cleared, and mounted. For immunofluorescence, a 10-min incubation with Hoechst was performed for nuclear staining. Visualization was performed using a Carl Zeiss Axioscope 2 or an Axio Observer Z1 microscope coupled with a LSM710 ZEN confocal module and processed with ZEN imaging software (Zeiss). The molecular layer thickness was measured and 200 Purkinje cells from the cerebellar vermis and hemisphere regions were counted on calbindin-stained sections from each SCA37 patient and one control. For each cerebellar region three independent measurements were taken from a total of 17 images with the ImageJ software (Schindelin et al., 2015).

SDS-PAGE and immunoblotting

Proteins were extracted from human cerebella by homogenization following standard protocols summarized in the Supplementary material. Primary antibodies used were beta-actin (AC15; Sigma-Aldrich), pan anti-AKT (#4691; Cell Signalling), anti-phospho-AKT (Ser473) (#9271; Cell Signalling), anti-Calbindin-D28K (EG-20; Sigma-Aldrich), anti-DAB1 (PA5-62538; Thermo Fisher Scientific), anti-GFAP (Z0334; Dako), anti-PI3 Kinase p85 (05-212; Merck), and anti-reelin (MAB5366 clone 142; MilliporeSigma). Infrared-dye conjugated secondary antibodies anti-mouse IRDye® 800CW and anti-Rabbit IRDye® 700CW (Li-Cor Biosciences) were used. Signals were detected and analysed with Odyssey analyser software (Li-Cor Biosciences).

Analysis of human DAB1 gene expression

Total RNAs were obtained from cerebellar vermis and hemispheres from SCA37 Patients IV:9 and IV:10 of Family AT-901, and an age-matched control using RNeasy® Mini Kit (Qiagen). RNA integrity was measured using the RNA ScreenTape assay (Agilent Technologies). cDNAs were synthesized from 1 µg total RNA with PrimeScript™ RT reagent Kit (Takara). Partial transcript fragments were analysed by standard PCR conditions with exon-specific primers (Supplementary Table 1). PCR amplicons were purified and sequenced by Sanger. To quantify the expression levels, cDNAs were mixed with SYBR® Green PCR master mix (Applied Biosystems) and specific primers. GAPDH was amplified as an internal control. Cycles and analysis were performed on the LightCycler 480 (Roche) and relative cDNAs fold-changes were normalized to GAPDH cDNA and calculated using the 2^−ΔΔCt method (Livak and Schmittgen, 2001).

ENCODE DAB1 RNA sequence reads were compared on BAM files from Purkinje cells from a 6-year-old male child and a 20-year-old adult male (Bernstein et al., 2012). Aligned data were sorted and indexed using SAMtools (Li et al., 2009). Read counts for DAB1 RNA were analysed with HTseq (Anders et al., 2015) using GAPDH reads counts to normalize and compare intersample differential expression. RBPmap (Paz et al., 2014) and PROMO (Messeguer et al., 2002) algorithms were used to identify putative splicing and transcription factor binding sites in the human DAB1 gene RNA and DNA sequences, respectively. The gene structure image was generated with GSDS 2.0 (Hu et al., 2015).

Statistical analysis

Statistical analyses were performed using R v3.1.2 with significance set at P < 0.05. Data between samples were analysed using the Shapiro-Wilk test to assess for normal distribution, followed by the parametric t-test or alternatively with the non-parametric Mann-Whitney U-test. Spearman’s rank correlation was used to estimate the contribution of the repeat expansion size and gender on disease age of onset.

Results

Molecular characterization of the SCA37 critical region

We recently described a novel spinocerebellar ataxia subtype, SCA37, in a large ataxia pedigree (AT-901) originating from the province of Huelva in the Andalusia region in the south of Spain (Serrano-Munuera et al., 2013). Affected patients presented with falls, dysarthria or clumsiness evolving to a pure cerebellar syndrome distinctively characterized by early-altered saccadic, pursuit and optokinetic vertical eye movements. By using whole-genome linkage studies, we localized the causative genetic deficit in this family to an 11-Mb genomic region on 1p32, which includes the disabled-1 (DAB1) gene encoding for the reelin adapter protein. Whole exome analysis did not lead to the identification of the causative mutation, which suggested its localization in a non-coding genomic region. By using a combination of whole-genome sequencing and genotyping studies with two SNPs identified within the SCA37 critical region, rs79992829 and rs146472695, found in linkage disequilibrium, the candidate mutation was narrowed by two recombination events to a 1.742 Mb interval spanning the 5′ non-coding regulatory region of the human DAB1 gene (Supplementary Fig. 2). These SNPs allowed for the identification of three additional families (Families AT-9012, AT-59 and AT-90) with similar phenotype from our cohort of ataxia patients of unknown genetic cause. The four kindreds (Supplementary Fig. 1) were clinically and genetically investigated. Likewise, as recently shown in four Portuguese ataxia families (Seixas et al., 2017), we found that an unstable ATTTC pentanucleotide repeat mutation located in the SCA37 1p32 region within intron 11 of the gene encoding the disabled adapter protein DAB1 segregated with the disease in these four families presenting with the SCA37 phenotype (Table 1 and Supplementary Fig. 1).

Clinical and genetic correlation in SCA37

All four SCA37 families included in this study shared a common geographical origin in the south of Spain with the same distinctive genetic haplotype within the critical region on 1p32, suggesting a mutation founder effect (Fig. 1A and Supplementary Fig. 1). Family AT-901 had been described previously (Serrano-Munuera et al., 2013). A total of 25 patients with SCA37 from the four families were clinically diagnosed by a neurologist and a detailed clinical exam was performed in 23 of them. Age at onset ranged from 25 to 64 years old (mean age 43.3 ± 9.9). Initial clinical presentation included dysarthria, falls, clumsiness or a combination of them, evolving to a generally slow progressive and pure cerebellar syndrome with scanning speech, mild trunk ataxia and severe dysmetria mostly in legs (Table 1). Eye movements at onset of the disease could be studied in three patients (Patients IV:4 and IV:5 from Family AT-901; Patient III:6 from Family AT-90) and the examination disclosed dysmetric vertical ocular saccades and irregular vertical ocular pursuit, whereas horizontal eye movements appeared normal. Those findings were confirmed by electrooculographic studies in Patients IV:4 and IV:5 from Family AT-901, who showed abnormal vertical saccades accuracy, diminished velocity and gain in vertical smooth pursuit, and slow velocity in vertical optokinetic nystagmus. Similar but milder abnormalities in the horizontal axis were identified during follow-up. Remarkably, saccade velocity remained normal along the course of the disease. All patients with longer standing disease, 4 to 42 years from onset, showed dysmetric saccades and irregular pursuit in both vertical and horizontal axes. Dysphagia, hand postural and action tremor, oscillopsia, nystagmus, and saccadic eye intrusions variably appeared with disease progression (Table 1). Patients became wheelchair-bound in 10 to 33 years from onset with the exception of one patient who showed an aggressive course becoming wheelchair-bound in 5 years from onset. None of them showed motor or sensory deficits, extensor plantar reflexes, fasciculations, epileptic seizures, or cognitive impairment.

Scale for the Assessment and Rating of Ataxia (SARA) score increase assessed severity of the disease and varied among patients and also in the same patient along disease progression. When SARA scores from 17 patients were normalized to the time from disease onset, a variable rate of progression from 0.38 to 2.05 points/year in SARA score was obtained (mean normalized SARA score 0.97 ± 0.34). For those 12 patients with two or more SARA score assessments, the gradient of the SARA score line was calculated and yielded a wider range (0–4 points/year). This approximation assumed a linear progression of SARA score as previously shown in SCA37 (Serrano-Munuera et al., 2013) and other SCAs (Jacobi et al., 2011). Maximum SARA score obtained was 32 points (Patient IV:21 from Family AT-901) at 42 years from onset and no ceiling effect of the scale was observed in our patients. Nerve conduction tests, transcranial magnetic stimulation, evoked potentials, electrocardiograms, echocardiograms and audiometric tests yielded normal results in all affected patients in agreement to the described clinical findings in SCA37 (Serrano-Munuera et al., 2013).

Brain MRIs at onset of the disease in affected patients showed either vermis atrophy or generalized cerebellar atrophy. Vermis atrophy progressed to generalized cerebellar atrophy in 2 years. In those patients with long standing disease, established cerebellar atrophy with preserved pons was detected. Cerebellar volumetric assessment in five patients (Patients III:8, IV:1 and III:1 from Family AT-90; Patients IV:4 and IV:5 from Family AT-901) revealed a 24 ± 6% decrease in cerebellar volume relative to the total intracranial volume (6.96 ± 0.71%) compared to 30 age- and gender-matched controls (9.17 ± 0.52%) randomly selected (age range: 24–75 years) from the open access IXI dataset (Manjón and Coupé, 2016). Mid-sagittal vermis relative diameter quantification in three patients (Patients IV:4, IV:5 and III:6 from Family AT-901) showed a 17.8% decrease of the midsagittal vermis relative diameter ratio (60.75 ± 1.21) compared to the mean ratio (73.93 ± 3.73) of eight age- and gender-matched controls.

Thirty-three asymptomatic relatives from all four families were clinically investigated. Among them, four subjects (Subjects V:3 and VI:2 from Family AT-901; Subjects IV:4 and IV:8 from Family AT-9012) showed hypometric downwards vertical eye saccades in the first neurological exam at ages 33, 21, 34, and 39 years of age, respectively. Subject V:3 also showed irregular vertical ocular pursuit. Neurological exam was otherwise normal. Oculographic registries from Subjects V:3 and VI:2 showed hypometric vertical saccades, diminished velocity and gain in vertical smooth pursuit and decreased vertical optokinetic nystagmus velocity as described (Serrano-Munuera et al., 2013). No abnormalities were found in the horizontal axis. All four subjects were later found to carry the causative mutation (Table 1). Mid-sagittal vermis relative diameter quantification of MRI in Subject V:3 from Family AT-901 and Subject IV:8 from Family AT-9012 showed a 9.9% decrease in the mean ratio compared to eight age- and gender-matched controls (67.2 ± 3.9 versus 74.6 ± 5.6, respectively). Subject VI:2 from Family AT-901 showed normal results as expected because of his younger age. Vermis atrophy was qualitatively recognizable by MRI in Subject V:3 from Family AT-901 and Subject IV:8 from Family AT-9012, while the vermis was considered normal in Subject VI:2 from Family AT-901 in agreement with the quantitative findings.

Three additional asymptomatic subjects were found to carry the causative mutation (Subject VI:1 from Family AT-59; Subjects IV:2 and IV:3 from Family AT-90). Neurological exams were normal in all subjects at ages 20, 35, and 31 years old, respectively. No oculographic registries or MRI could be obtained. In summary, 25 affected and seven at-risk asymptomatic individuals presented with the ATTTC repeat mutation (average 52.1), ranging from 46 to 71 and from 49 to 62 repeats, respectively (Fig. 1B and Table 1). The average age of disease onset was 40.4 years in females and 49 in males (P < 0.021). Remarkably, a gender-specific contribution of the ATTTC expansion size to the age of onset was identified in males (r = −0.96; P < 0.0001; n = 9), but not in females (r = −0.09; P > 0.75; n = 14) (Fig. 1C). Increments ranging from two to five ATTTC pentanucleotide repeats were identified in four out of the six transmissions studied (Supplementary Fig. 3). No allele contractions were observed. None of the 28 healthy relatives studied presented with the ATTTC mutation.

Neuropathological findings

In the present study, we were able to investigate a complete brain from two patients with SCA37 (Patients IV:9 and IV:10) from the large SCA37 AT-901 Spanish family previously reported (Serrano-Munuera et al., 2013). Both patients had inherited the risk genetic haplotype for SCA37 and later revealed the ATTTC repeat mutation containing each 48 and 46 repeats within the DAB1 gene.

Under macroscopic inspection, the cerebral hemispheres and brainstems of the brains from both patients were unremarkable except for moderate enlargement of the ventricular system in coronal cerebral sections in Patient IV:9. The cerebellum appeared atrophic with prominent fissures and folia (Fig. 2A and B). The cerebellar white matter appeared reduced and dense in both patients. Light microscopy examination of the cerebral cortex showed, in general, a preserved cytoarchitectural pattern, discrete and diffuse gliosis in the primary motor cortex and no evidence of neuronal loss (data not shown). Juxtaglomerular vessels showed moderate hyaline sclerosis. Immunostaining for amyloid-β and tau proteins in CERAD areas revealed a moderate frequency of neuritic plaques with only few neurofibrillary tangles in both patients according to their age. Despite these findings, cytoarchitectural and neuronal populations were well preserved in these areas. Cortical microinfarctions were detected in dorsolateral frontal and temporal inferior cortices in Patient IV:10 related to known vascular risk factors. No α-synuclein immunostaining was found in Patient IV:10 whereas a moderate amount of Lewy bodies and neurites were found in amygdala and entorhinal and cingulate cortices in Patient IV:9 (data not shown). No TDP-43 inclusions were detected (data not shown).

The subcortical regions such as the caudate and lentiform nuclei, nucleus basalis of Meynert, posterior hypothalamus, mammillary bodies, thalamus and subthalamic nuclei were preserved, and the unique abnormality found was mild small vessel pathology in Patient IV:10. The hemispheric white matter of the frontal and occipital periventricular areas was unremarkable unaltered. The brainstem appeared normal except for the moderate (Patient IV:10) to intense (Patient IV:9) neuronal loss and gliosis identified confined to the inferior olive (Fig. 2C). No pathological α-synuclein immunostaining was found in this region in Patient IV:10. Conversely, a few Lewy bodies and neurites were detected in the motor nucleus of the X cranial nerve in Patient IV:9 without cell loss.

The cerebellar cortex showed an extensive and generalized Purkinje cell loss (Fig. 2D) with the remaining exhibiting severe nuclear changes such as lobulation, irregular shape, and hyperchromatism together with Bergman and astrocyte gliosis (Fig. 2E). Phosphoneurofilament immunoreactivity revealed many empty baskets with stained perikarya (Fig. 2F and G). All these changes were diffuse and intense along the whole cerebellar cortex with a relative sparing of the cerebellar amygdala in Patient IV:10. Patient IV:9 showed more extensive neuronal loss in lobule VII of the vermis (Fig. 2H). The cerebellar white matter appeared normal and the deep nuclei offered a homogeneous aspect with preserved neuronal populations and marked gliosis (Fig. 2I). Ubiquitin staining of Purkinje cells revealed the presence of multiple punctate inclusions located in the peripheral cytoplasm suggestive of perisomatic granules in both patients, which later revealed immunostaining for DAB1 protein (Fig. 2J–L, arrows) and leptomeningeal vessels (Fig. 2M). No intranuclear inclusions were observed. The granular layer was relatively spared and basket and Golgi cells were well preserved. Occasional axonal spheroids were seen in Patient IV:10 (Fig. 4E and F). The molecular layer showed mild diffuse gliosis and few hypertrophic fibres in Patient IV:9 while gliosis was specifically severe in the flocculus in Patient IV:10 (Fig. 2N). The cervical spinal cord was unremarkable.

Immunostaining with anti-Calbindin-D28K revealed a significant reduction of the molecular layer thickness and Purkinje cells density more predominantly in the vermis (Fig. 3B, C, E and F) than in cerebellar hemispheres (Fig. 3G–K) in both patients (Fig. 3 and Supplementary Fig. 4). Remarkably, a subpopulation of Purkinje cells showed tangential orientation (24% in Patient IV:9 and 20.5% in Patient IV:10; Figs 3J–L and 4C) with aberrant arborization (Fig. 4D), and ectopic mispositioning within the granular layer (6% in Patient IV:9 and 6.5% in Patient IV:10) or the molecular layer (6.5% in Patient IV:9 and 7% in Patient IV:10) (Fig. 4E and F). GFAP immunostaining showed diffuse cerebellar cortical disorganization in both SCA37 patients (Fig. 4H–K). Calbindin-D28K immunostaining of the cerebellar cortex was decreased in line with Purkinje cell loss in both SCA37 patients (Fig. 5G and H). GFAP protein was shown upregulated by immunoblotting (data not shown) in agreement with reactive astrogliosis.

Dysregulated DAB1 expression and altered reelin-DAB1 signalling in the SCA37 cerebellum

Immunostaining with anti-DAB1 revealed specific overexpression in SCA37 cerebellum compared to age-matched control (Fig. 5). In control cerebellum, DAB1 was present in Purkinje cell soma and dendrites (Fig. 5A and B), whereas in the SCA37 cerebellum it appeared overexpressed and exhibited perisomatic and perinuclear punctate staining in the remaining Purkinje cells (Fig. 5C–E). A few ectopic Purkinje cells showed intense DAB1 staining in the cerebellar molecular layer (Fig. 5F). The expression of DAB1 protein was found increased in both SCA37 patients. Overexpression of DAB1 80- and 63-kDa isoforms was identified in the SCA37 cerebellar vermis and hemisphere (Fig. 5G and Supplementary Fig. 5). Furthermore, dysregulated expression of reelin proteins along with increased levels of PI3K-p85 and phosphorylation of AKT were identified in the SCA37 cerebella indicating upregulation of reelin-DAB1 signalling (Fig. 5G and H and Supplementary Fig. 5). Remarkably, both SCA37 cerebella revealed significant upregulation of the uncleaved 410-kDa form of reelin (Fig. 5G), which is an isoform known to induce DAB1 phosphorylation, with the subsequent interaction with PI3K and AKT phosphorylation and activation (Kohno et al., 2009).

To investigate whether transcriptional dysregulation in the DAB1 gene contributed to the detected increase in DAB1 protein expression observed in SCA37 we characterized the cerebellar DAB1 transcripts from SCA37 patients. Nine alternative non-coding DAB1 transcripts, five of them not previously described (Bar et al., 2003), originating from alternative combinations of 10 5′ UTR non-coding exons within the human DAB1 gene were identified dysregulated in human SCA37 post-mortem cerebellum (Supplementary Fig. 6). ENCODE RNAseq analysis confirmed expression of these 10 5′ UTR non-coding exons in normal human adult Purkinje cells (Supplementary Fig. 7). Remarkably, four different partial alternative coding transcripts originating from the human DAB1 gene were also identified in the post-mortem human cerebellum, two of them (DAB1-1 and DAB1-3) found overexpressed in SCA37 but not in control (Fig. 6A–C). DAB1-1 generates the longest DAB1 protein isoform (80 kDa; Fig. 5G), which contains most of the DAB1 phosphorylation sites involved in PI3K interaction and reelin signalling. The expressed DAB1-2 transcript generates the shortest 63 kDa DAB1 isoform (Fig. 5G and Supplementary Fig. 8). The remaining transcript DAB1-4 containing the non-coding exon 11, adjacent to the ATTTC intronic mutation, and coding exons 20 and 21 were uniquely present in SCA37 but absent in control cerebellum (Fig. 6B). Overexpression of these two evolutionary conserved alternative coding exons from the mouse Dab1 gene have been related to migration deficits of Purkinje cells (Yano et al., 2010). No RNA-sequence reads were identified for exons 20 and 21 in ENCODE normal Purkinje cell RNA data (Supplementary Fig. 7), revealing the absence of DAB1 transcripts containing these two coding exons in normal human cerebellum. These data indicate that RNA switch and selective expression of DAB1 exons 20 and 21 may underlie the cerebellar pathology observed in the SCA37 cerebellar cortex. Finally, to address how the SCA37 mutation dysregulated DAB1 expression, we analysed the SCA37 mutation with in silico algorithms and found that the ATTTC repeat mutation within intron 11 of the DAB1 gene creates putative new XBP1 transcription factor binding motifs (Supplementary Fig. 9).

Discussion

In the present study, we demonstrate that the ATTTC pentanucleotide repeat mutation within the non-coding regulatory region of the gene encoding for the reelin adapter protein DAB1 as the genetic cause underlying spinocerebellar ataxia type 37. We describe the clinical-genetic correlations in four Spanish SCA37 families and, for the first time, the neuropathological findings in two SCA37 brains, which confirm the neurological findings and confine them to a pure cerebellar syndrome.

Since the discovery of pathogenic repeat expansions as a mechanism of disease in the 1990s, the list of neurodegenerative and neuromuscular disorders characterized by unstable repeat expansions has grown to over 40 (reviewed in Paulson, 2018). The disease-causing mechanisms in-clude protein gain-of-function, protein loss-of-function, toxic RNA gain-of-function, non-ATG-initiated translation (RAN) peptides, and transcriptional dysregulation (Groh et al., 2014; Matilla-Dueñas et al., 2014). Non-coding expanded repeat sequences have been implicated in transcriptional dysregulation by altering transcription factor binding (Lin et al., 2010), and epigenetic alterations (Naumann et al., 2009). RNA mediated-toxicity has emerged as a recurrent toxic gain-of-function mechanism in the pathogenesis of several other SCAs and neurodegenerative diseases and has been associated with SCA8 (Zu et al., 2011), SCA10, SCA31 (Donnelly et al., 2013), amyotrophic lateral sclerosis and frontotemporal dementia (Ishiguro et al., 2017), Huntington’s disease (Bañez-Coronel et al., 2015), myotonic dystrophy types 1 and 2 (Gourdon and Meola, 2017), and fragile X-associated tremor ataxia syndrome (Kong et al., 2017). Short-tandem repeats of various nucleotide motifs are frequently found in transcripts. These sequences might configure into alternative or pathological RNA structures depending on the motif type and number of reiterations playing an important role in RNA processing regulation (Galka-Marciniak et al., 2012). A myriad of pathogenic mechanisms has been assigned to toxic RNA repeats including aberrant alternative splicing, the inhibition of nuclear transport and export, induction of the innate immune response, alteration of microRNA biogenesis, and abnormal activation of RNA interference. For instance, the non-coding intronic expanded hexanucleotide GGGGCC repeat associated with ALS/FTD is able to alter C9orf72 RNA processing, in terms of transcription, splicing, and localization (Barker et al., 2017). Herein, we provide evidence that the unstable intronic ATTTC repeat mutation dysregulates DAB1 expression and induces RNA switch, with a consequent upregulation of reelin-DAB1 signalling in the SCA37 cerebellum. Remarkably, a gender-specific contribution of the ATTTC repeat mutation to the age of onset was observed in males, but not in female patients. In this regard, sexually dimorphic expression of reelin in the cerebellum was observed in reeler mouse models and a few human neurodegenerative disorders, highlighting the role of sex hormones in the modulation of the reelin-signalling response (HadjSahraoui et al., 1996; Gross et al., 2012).

A remarkable feature in the ataxic patients of our four SCA37 families is the presence of early vertical eye movement alterations, which often precede the ataxia symptoms. Impaired horizontal eye movements are regarded as a well-known clinical sign in several SCAs subtypes including SCA1 (Genis et al., 1995; Klostermann et al., 1997), SCA2 (Wadia et al., 1998; Velázquez-Pérez et al., 2004; Seifried et al., 2005), SCA3 (Bürk et al., 1999), SCA4 (Nagaoka et al., 2000), SCA5 (Bürk et al., 2004), SCA6 (Gomez et al., 1997; Christova et al., 2008), SCA7 (Oh et al., 2001), SCA8 (Day et al., 2000), and SCA17 (Hubner et al., 2007). These abnormalities may appear even in presymptomatic stages, such as in SCA1 (Genis et al., 1995; Klostermann et al., 1997), SCA2 (Velázquez-Pérez et al., 2009), SCA3 (Raposo et al., 2014), SCA6 (Christova et al., 2008), SCA7 (Oh et al., 2001), SCA17 (Hubner et al., 2007) and have been considered an endophenotype in few subtypes including SCA1 (Wessel et al., 1998), SCA2 (Wadia et al., 1998), and SCA7 (Oh et al., 2001). However, vertical eye movement abnormalities have rarely been reported, to the best of our knowledge only in SCA6 (Christova et al., 2008), SCA26 (Yu et al., 2005), and SCA30 (Storey et al., 2009), and by the time they appear, clear horizontal axis abnormalities are also found.

The main eye movement findings in our SCA37 patients consisted of abnormal saccade accuracy, especially downwards, with slow optokinetic nystagmus and pursuit, all of them initially found in the vertical axis (Serrano-Munuera et al., 2013). The observed normal saccade velocity suggests relative sparing of the pons (Henn et al., 1982), already suspected by the neuroimaging findings in the AT-901 SCA37 family, and later confirmed by our neuropathological investigations. Saccade accuracy is controlled by lobules VI and VII of the posterior vermis and their projections through the caudal pole of the fastigial nucleus to the brainstem saccade-generating system in primates (Voogd et al., 2012). In humans as well, transcranial magnetic stimulation, functional MRI, mapping lesions based on MRI in patients with cerebellar infarctions (Kheradmand and Zee, 2011) as well as clinical data from degenerative ataxic disorders (Wessel et al., 1998), confirm the participation of the posterior vermis in the definition of saccade properties, such as accuracy, latency, trajectory, and dynamics. Additionally, the superior colliculus is involved in the generation of saccades by several routes (Voogd et al., 2012). Optokinetic nystagmus and smooth pursuit are mediated, but not exclusively, by the flocculus in primates (Kheradmand and Zee, 2011). Among several projections, the climbing fibres input to the flocculus has been well systematized in rabbits and again is mediated by the inferior olive (Voogd et al., 2012). As for pursuit, two parallel cortical-cerebellar-olive loops have been proposed, both include inferior olive structures (Voogd et al., 2012). Consistently, Patient IV:10 showed marked neuronal loss in lobule VII and severe gliosis in the flocculus. Remarkably, both siblings revealed prominent neuronal loss and gliosis in the inferior olive. We propose that these anatomical structures and physiological Purkinje cell behaviour may be involved in the specific clinical disease pattern seen in our patients.

Considerable overlap in the neurodegenerative pattern has been shown in post-mortem neuropathology for the polyglutamine (polyQ) ataxias, in which intranuclear [SCA1, SCA3, SCA7, SCA17, dentatorubral-pallidoluysian atrophy (DRPLA)] or intracytoplasmic (SCA2, SCA6) ubiquitin-immunostained polyQ inclusions are a common feature (Matilla et al., 1997; Skinner et al., 1997; Seidel et al., 2012; Rüb et al., 2013). In SCA12, ubiquitinated nuclear inclusions resembling Marinesco bodies are identified in the cerebellum, striatum, and motor cortex (O’Hearn et al., 2015). For the remaining non-polyglutamine SCAs, post-mortem neuropathology data is limited to SCA28 (Smets et al., 2014), SCA31 (Niimi et al., 2013; Yoshida et al., 2014), and SCA36 (Obayashi et al., 2015). Remarkably, a typical feature of SCA31 is the presence of a halo-like amorphous material surrounding Purkinje cells, which contain small ubiquitin-positive granules consisting of synaptophysin-positive vesicles and a fragmented Golgi apparatus (Yoshida et al., 2014). None of these subcellular hallmarks were observed in the SCA37 cerebellum of our patients.

Our data reveal remarkable dysregulation of DAB1 expression in the SCA37 cerebellum, and because normal DAB1 expression in the cerebellum is mostly restricted to Purkinje cells, our neuropathological findings are consistent with main Purkinje cell pathology. Mouse DAB1 is an adaptor molecule mediating signalling from reelin receptors, which upon binding underpin migration and lamination of neurons in the developing cortex, hippocampus and cerebellum (Miyata et al., 2010; Hirota and Nakajima, 2017). In the post-natal cerebellum, reelin signalling triggers the dispersal and positioning of Purkinje cells into their final location within the adult Purkinje cell monolayer and disposes their dendritic deployment by binding with the Purkinje cell receptors, very low-density lipoprotein receptor (VLDLR) and apolipoprotein E receptor 2 (APOER2). Upon reelin binding, DAB1 is tyrosine-phosphorylated by SRC tyrosine kinases FYN and SRC thus activating PI3K/AKT, ERK and CRK/CRKL signalling pathways before it is degraded by the ubiquitin-proteasome pathway. Remarkably, the cerebella in mutant reeler, scrambler and yotari mice, all having loss-of-function mutations in reelin (Reln) or Dab1 genes, are anatomically very similar, with lack of foliation and Purkinje cells failing to migrate and largely remaining in several ectopic clusters deep in the cerebellar cortex, while the granule cells appear to migrate normally (Goffinet et al., 1984; Goldowitz et al., 1997). In this regard, our data support that a gain-of-function mutation by upregulation of DAB1 alternate isoforms underlies SCA37 with mislocalization of Purkinje neurons in the cerebellar cortex reflecting improper positioning and selective migration deficits. DAB1 has been shown to regulate the intracellular trafficking of reelin receptors and intracellular reelin levels after the reelin/receptor complex internalization (Morimura et al., 2005). The identification of significantly increased uncleaved 410 kDa reelin protein, which was previously shown to induce DAB1 phosphorylation (Kohno et al., 2009), may be responsible for the cerebellar overactivation of reelin-DAB1 signalling in SCA37. This in turn would dysregulate reelin-DAB1 function in the regulation of neuronal migration through the upregulation of PI3K and AKT (Jossin and Goffinet, 2007). To address how the SCA37 mutation dysregulated DAB1 expression, we used in silico algorithms, which showed that the ATTTC mutation within intron 11 of the DAB1 gene creates new putative XBP1 transcription factor binding motifs. XBP1 is a transcription factor known to be upregulated in embryos, downregulated during adulthood, and activated during ageing by the unfolded protein response (UPR) to endoplasmic reticulum stress (Lee et al., 2003), and has been implicated in Purkinje cell degeneration in SCA17 (Yang et al., 2014). Remarkably, DAB1-related pathology studies in Nova2 null mice identified overexpression of a Dab1-isoform containing two evolutionary conserved exons homologues to human exons 20 an 21 in the DAB1 gene (Yano et al., 2010). Both exons were responsible for the Purkinje cells migration defects identified in Nova2 null mice. Therefore, the study demonstrated the role of the RNA-binding protein NOVA2 in regulating Dab1 RNA switch by alternative splicing to mediate neuronal responsiveness to reelin signalling. Our data showing overexpression of DAB1-4 transcript containing exons 20 and 21 in the SCA37 cerebellum compared to un-detectable levels in gender- and age-matched control, supports a role for DAB1 RNA switch slipping away from these two coding exons as seen by NOVA2 inhibition. This is supported by in silico predictive algorithms that identified enrichment of the YCAY NOVA motif within the human DAB1 exons 20 and 21 (Supplementary Fig. 10), and the fact that no RNA-sequence reads were identified for exons 20 and 21 in the ENCODE database for normal Purkinje cell RNA data (Supplementary Fig. 7). Therefore, binding of active XBP1, upregulated in the embryo or by the UPR during ageing, to the ATTTC repeats mutation in SCA37 should upregulate DAB1 expression of specific transcripts containing exons 20 and 21, slipped away from NOVA2 inhibition. We propose that this molecular mechanism possibly underlies the neuronal migration deficits and progressive neurodegeneration of Purkinje cells in the SCA37 cerebellum (Fig. 7).

In conclusion, in this study we demonstrate the intronic ATTTC unstable repeat mutation within the non-coding regulatory region of the DAB1 gene as the genetic defect causing SCA37. We report the first neuropathological post-mortem findings of SCA37, which are confined to the cerebellum in agreement with a pure cerebellar syndrome. Importantly, our data show DAB1 overexpression and RNA switch, accompanied by dysregulation of the reelin-DAB1 and PI3K/AKT signalling pathway as the main mechanism underlying cerebellar neurodegeneration in SCA37.

Acknowledgements

We thank all SCA37 patients and family members for participating in this study. We acknowledge the IGTP-HUGTP Biobank integrated in the Spanish National Biobanks Network of the Instituto de Salud Carlos III (PT13/0010/0009), and the IGTP Microscopy and Genomics Core Facilities and staff for their contribution to this publication. We are indebted to CNAG Staff for support in WGS, Dr. Ellen Gelpí at the Banc of Teixits Neurológics of the Hospital Clínic-IDIBAPS for helpful comments and suggestions, and Lluís Pujadas at the Vall d'Hebron Institut de Recerca and CIBERNED for the anti-reelin antibody 142 clone (Merck Millipore). We also thank Marta del Pozo and Kerrie Adrian of the Neurogenetics Unit at the IGTP, and Miguel A. Labrador for assistance in the assessment of cerebellar volumetry.

Funding

The research of this work was funded by the Spanish Health Institute Carlos III grants CP08/00027 and CPII14/00029 (to A.M-D.), FIS PI14/01159 (to V.V.), FIS PI14/00136 (to A.M-D.), and FIS PI17/00534 (to A.M-D. and I.S.). A.M-D. was a Miguel Servet Investigator in Neuroscience supported by the Spanish Health Institute Carlos III (ISCIII; CPII14/00029).

Supplementary material

Supplementary material is available at Brain online.

Abbreviations

Abbreviations
 
• SARA

  Scale for the Assessment and Rating of Ataxia

 
• SCA

  spinocerebellar ataxia

References