The Cyprus Institute of Neurology and Genetics, Nicosia, Cyprus
Corresponding author: Kyproula Christodoulou, The Cyprus Institute of Neurology and Genetics, 6 International Airport Avenue, Nicosia, Cyprus. Tel: +357-22392649; Fax: +357-22392615; E-mail: firstname.lastname@example.org
Autosomal recessive cerebellar ataxias (ARCA) encompass a heterogeneous group of rare diseases that affect the cerebellum, the spinocerebellar tract and/or the sensory tracts of the spinal cord. We investigated a consanguineous Cypriot family with spastic ataxia, aiming towards identification of the causative mutation. Family members were clinically evaluated and studied at the genetic level. Linkage analysis at marker loci spanning known ARCA genes/loci revealed linkage to the APTX locus. Thorough investigation of the APTX gene excluded any possible mutation. Whole genome linkage screening using microsatellite markers and whole genome SNP homozygosity mapping using the Affymetrix Genome-Wide Human SNP Array 6.0 enabled mapping of the disease gene/mutation in this family to Chromosome 9p21.1-p13.2. Due to the large number of candidate genes within this region, whole-exome sequencing of the proband was performed and further analysis of the obtained data focused on the mapped interval. Further investigation of the candidate variants resulted in the identification of a novel missense mutation in the GBA2 gene. GBA2 mutations have recently been associated with hereditary spastic paraplegia and ARCA with spasticity. We hereby report a novel GBA2 mutation associated with spastic ataxia and suggest that GBA2 mutations may be a relatively frequent cause of ARCA.
Autosomal recessive cerebellar ataxias (ARCA) comprise a heterogeneous group of disorders that affect the cerebellum, the spinocerebellar tract and/or the sensory tracts of the spinal cord. They are characterized by impaired walking with lack of gait and limb coordination, and usually by an early age of onset (Harding, 1983; Palau & Espinos, 2006). The prominent features may often be accompanied with spasticity and the disease is called spastic-ataxia (de Bot et al., 2012; Hammer et al., 2013). More than 20 ARCA genes and loci have been identified in the last 15 years thus demonstrating genetic heterogeneity (Anheim et al., 2010; Doi et al., 2011). Friedreich ataxia and ataxia telangiectasia are the two most frequent forms of ARCA (Anheim et al., 2010).
We have been studying families and sporadic cases of ataxia in the Cypriot population for the past 20 years. The majority of our patients have been excluded for the most common known mutations (Votsi et al., 2012) and they remain genetically undiagnosed. Thus far, the most frequent mutation in our population is the GAA repeat expansion in intron 1 of the FXN gene (Dean et al., 1988; Zamba-Papanicolaou et al., 2009). In addition, a novel c.5308_5311delGAGA deletion in the SETX gene has been identified in a large family with five patients (Nicolaou et al., 2008). We hereby describe our clinical and genetic findings in a consanguineous Cypriot family with spastic ataxia. Through an initial investigation of the family at nine known ARCA loci, linkage to the APTX locus was confirmed. Thorough investigation of the APTX gene excluded any possible mutation. Genome-wide linkage analysis and homozygosity mapping combined with exome sequencing led to the identification of a novel missense homozygous GBA2 gene mutation. The GBA2 gene encodes β-glucosidase 2, an enzyme with glucosylceramidase activity. In the course of our study, GBA2 mutations have been reported in patients with hereditary spastic paraplegia (Martin et al., 2013) and ARCA with spasticity (Hammer et al., 2013).
Material and Methods
Subject and Samples
Family 903 is a two-generation, six-member Cypriot ARCA kindred with three affected individuals in one generation (Fig. 1). The parents originate from the same village and are distant relatives. Detailed clinical histories of the patients were obtained and neurological and neurophysiological evaluation was carried out. All three patients had brain MRI (Fig. 2). The elder affected brother died at the age of 53 years old.
Blood samples were obtained from individuals V-1, V-2, VI-1, VI-2, VI-4 and VI-5. DNA was extracted using standard salting out procedures.
Our research programme on ataxias in the Cypriot population has been approved by the National Bioethics Committee of Cyprus. Written informed consent was obtained from all study participants.
ARCA Loci Genotyping
Family members were genotyped at short tandem repeat (STR) markers spanning known ARCA loci (APTX, SCAR3, SYNE1, TDP1, SACS, SCAR7, SETX, POLG and C10orf2) in order to investigate linkage of the family to any of these loci.
Investigation for an APTX Gene Mutation
Genomic DNA sequencing of the proband at the APTX gene was performed. Primers amplifying the sequence of the nine exons and flanking intronic sequences of the APTX gene were designed by us and are available upon request. Amplification products were sequenced in both directions with CEQ Dye Terminator Cycle Sequencing. Sequence traces were automatically compared with the normal APTX gene sequence as listed in the GenBank database (Table S1) using the CEQ8000 investigator software (Beckman Coulter, Nyon, Switzerland).
MLPA analysis of the proband was performed using the P316 Kit on Recessive ataxias (MRC Holland, Amsterdam, The Netherlands) covering all exons of the APTX, SETX and FXN genes. Two normal control samples and one heterozygous control with a complete deletion of the APTX gene were used as references for this analysis.
Southern blot analysis
Southern blot analysis was performed for detection of any possible intronic large-scale mutation in the APTX gene. Proband and normal control samples were compared. Different probes to cover all the areas of the gene were generated using genomic DNA and polymerase chain reaction (PCR) primers designed by us. The APTX gene region from the promoter to the end of exon 6 was investigated with five probes and a PstI digestion of the genomic DNA. The remaining APTX gene region was investigated with two probes and an EcoRI digestion of the genomic DNA. Labelling of the probes, hybridization, stringency washes and detection were carried out using the Dig High Prime DNA Labelling Kit and Detection Starter Kit II (Roche, Mannheim, Germany), and following the manufacturer's instructions.
cDNA synthesis from total RNA
Total RNA was extracted from lymphoblastoid cell lines of the proband and a normal control, using the RNeasy Mini Kit (Qiagen, Hilden, Germany). cDNA was synthesized using the Protoscript First Strand cDNA Synthesis Kit (New England Biolabs, Ipswich, MA, USA). The synthesized strand was used as substrate for the amplification of the APTX mRNA. Four pairs of primers that were designed by us and are available upon request were employed to amplify the full length of the APTX isoform 1 mRNA.
Genome-Wide Screening, Homozygosity Mapping and Linkage Analyses
All family members were genotyped at 138 STR markers from the Research Genetics Screening set 4A, using previously described methodology (Christodoulou et al., 1995). Refined mapping was performed with more dense STR markers selected from candidate regions. Fragment analysis was carried out using the Beckman Coulter CEQ 8000 Genetic Analyzer or the Applied Biosystems 3130xl Genetic Analyzer. Haplotypes of individuals were constructed with the Cyrillic program. More extensive genome-wide SNP genotyping was carried out by Atlas Biolabs using the Genome-Wide Human SNP Array Version 6.0 (Affymetrix, Santa Clara, CA, USA). Samples were analysed with the Birdseed v2 algorithm of the Genotyping console 4.0 and a mean final call rate of 99.0% was reached. Homozygosity mapping was performed with the Homozygosity Mapper software (Seelow et al., 2009). Linkage analysis was performed using the LINKAGE package of programs (Lathrop et al., 1985). Two point and multipoint analyses were carried out using MLINK and LINKMAP. Lod scores were calculated under an autosomal recessive inheritance model and a disease allele frequency of 1 in 10,000.
Screening of candidate genes
Genes located within mapped regions were identified using GeneDistiller and the NCBI database. The AQPEP, LOC644100, COMMD10, DTWD2, DMXL1, TNFAIP8, HSD17B4 and PRR16 genes located within the Chromosome 5 region and the ACO1, DNAJA1, UBAP1 and SIGMAR1 genes located in the linked Chromosome 9 region were sequenced using the proband sample. Primers amplifying the sequence of all exons and flanking intronic sequences of the above genes were designed by us and are available upon request. Furthermore, identified intronic regions with repeat motifs were sequenced using the proband and normal control samples. Amplification products were sequenced in both directions using the CEQ Dye Terminator or the Big Dye Terminator v1.1 Cycle Sequencing kit. Sequence traces were automatically compared with the normal gene sequences as listed in the GenBank database (Table S1) using the CEQ8000 investigator software (Beckman Coulter) or the Seqscape software (Applied Biosystems, Foster City, CA, USA).
Whole-exome sequencing was carried out by Oxford Gene Technology, Begbroke, Oxfordshire, UK on the DNA of the proband. The sample was prepared according to Agilent SureSelect protocols Version 1.2, enrichment was carried out according to Agilent SureSelect protocols and Sequencing was performed on the Illumina HiSeq2000 platform using TruSeq v3 chemistry. Read files (Fastq) were generated from the sequencing platform via the manufacturer's proprietary software. Reads were mapped to their location against the reference human genome hg19/b37 using the Burrows-Wheeler Aligner (BWA) package, version 0.6.1. Local realignment of the mapped reads around potential insertion/deletion sites was carried out with the Genome Analysis Tool Kit (GATK) version 1.6. Duplicate reads were marked using Picard version 1.62. Additional BAM file manipulations were performed with Samtools 0.1.18. Base quality scores were recalibrated using GATK's covariance recalibration. SNP and indel variants were called using the GATK Unified Genotyper for each sample and SNP novelty was determined against dbSNP release 135. Known variants reported as certified polymorphisms in public databases (1000 Genomes, dbSNP 135) were excluded. The remaining novel variants were visually inspected with the Integrative Genomics Viewer (IGV), and then further investigated in the family and in the normal Cypriot population, by Sanger sequencing as described above or by a PCR-restriction fragment length polymorphism (RFLP) assay (primers designed by us and enzymes are available upon request).
Clinical Features of Patients
On examination, both parents and one brother were neurologically and cognitively normal. All three patients had unremarkable prenatal, delivery and early neonatal history. Their psychomotor development was normal and they walked at the appropriate age. The proband had an early disease onset (14 years old) and died at the age of 53 due to disease complications (aspiration pneumonia). The younger brother also had an early onset (13 years old) and the affected younger sister had later age of onset, at the age of 20. The first symptom in all patients was marked spasticity in the lower extremities resulting in gait disturbances. Gait was spastic with dragging and ataxic features, with truncal ataxia and wide gait base. Foot deformities were present with pes cavus and tendon contractures along with spontaneous extensor plantar response. None of the patients had mental impairment. At the early stages of the disease, jerky pursuit and nystagmus were present without evidence of oculomotor paresis in all patients. Progression of the disease was slow, with increasing asymmetric spasticity affecting both upper and lower limbs and truncal muscles, weakness in the lower limbs, severe mainly spastic dysarthria and dysphagia. Patients became wheelchair bound within 20–25 years from disease onset. Overall, there was no significant variability of the clinical phenotype in this family, with the sister having a slightly milder form (Table 1). Cerebellar symptoms were present with finger to nose examination and alternate movements disturbance. Vibration sense was reduced distally in the lower limbs. At advanced stage from disease onset ( >25 years), the neurological examination was similar in the three patients: they all had mild cognitive impairment, hearing loss, urinary incontinence, severe spastic dysarthria and dysphagia. The ocular movements were restricted in the upper gaze with signs of ocular apraxia only at the horizontal lateral gaze. Early cataract at the age of 48 was present only in patient VI-4. They were wheelchair bound, with severe spasticity of the limbs, the neck muscles and the truncal muscles. Electrophysiological evaluation of the patients with nerve conduction studies revealed no findings of polyneuropathy at the early stages of the disease and minimal findings of sensory axonal neuropathy at the advanced stages of the disease (absence of sural sensory action potential (SAP) in patient VI-4 and still preserved but low sural SAP in patient VI-5). The upper and lower limbs somatosensory evoked potentials revealed unobtainable cortical responses bilaterally (normal cervical and brachial plexus potentials). The central motor conduction time after magnetic stimulation, performed at the early and the advanced stages of the disease course, showed definitely prolonged central motor conduction time from the lower limbs in patient VI-4 and prolonged cortical responses. Brain MRI, performed 15 years from disease onset, showed no focal lesions but marked cerebellar atrophy. Lysosomal enzymes (hexosaminidase a&b), Vitamin E, AFP levels, serum albumin, total cholesterol and triglycerides were all within the normal ranges.
ARCA loci analysis and investigation for an APTX gene mutation
Through the initial investigation for linkage at known ARCA loci, family 903 presented with an indication for linkage at the APTX locus. A lod score value of 1.81 at recombination fraction 0 was obtained by two-point linkage analysis. Homozygosity in all affected members was observed for the four analysed marker loci (D9S1118, D9S304, D9S1788 and D9S1845). None of the healthy members was homozygous for these loci (Fig. 1). Sequence analysis of the proband excluded any point mutation in the APTX gene. A deletion or duplication of the APTX gene exons was excluded by MLPA analysis. Southern blot analysis covering the APTX gene promoter, exonic and intronic sequences further excluded the possibility of any intronic large-scale mutation. RNA studies also excluded the possibility of a mutation that controls the expression of the APTX gene.
Genome-wide screening, homozygosity mapping and linkage analyses
Following exclusion of the APTX gene, two genome-wide approaches were employed in order to map the gene; whole genome screening using STR markers followed by two-point and multipoint linkage analyses, and whole genome homozygosity mapping using SNP markers. Two significant regions were identified by the first approach (4p13-q32.1 and 9p21.2-q21.12), which were further investigated with additional STR markers. Linkage to the 4p13-q32.1 region was excluded and linkage to the 9p21.2-q21.12 region was further confirmed within a reduced critical interval. Homozygosity was observed in the interval between loci D9S43 and D9S1862. The disease haplotype covering marker loci between D9S43 and D9S1862 (D9S1118-D9S304-D9S1788-D9S1845-D9S165-D9S1878-D9S1817-D9S1805-D9S1804-D9S1791-D9S1859-D9S50-D9S1874-D9S2148) is 1-2-2-1-1-2-3-2-2-3-1-2-1-1 (Fig. 1). Positive lod scores with maxima at recombination fraction 0 were obtained for all loci mapped within the region of homozygosity. Multipoint linkage analysis resulted in the maximum lod score value of 2.9 at locus D9S1878 (Fig. 3).
Homozygosity mapping by genome-wide SNP analysis revealed the existence of significant chromosomal regions bearing the highest homozygosity score, on Chromosomes 5 and 9, respectively (Fig. 4). These results were consistent with the data of the first approach. The homozygous region identified on Chromosome 9 was within the already mapped critical interval and the homozygous region on Chromosome 5 was within one of the nonexcluded regions of the first approach.
Further analysis of the Chromosome 5 region was performed with four additional STR markers. Multipoint linkage data did not exclude this interval although no significant lod score was obtained (lod ≤ 0.5). The Chromosome 5 candidate region was 3.1 Mb with eight protein coding genes located within the critical intervals. Sequence analysis of these genes revealed no pathogenic mutation, neither in the coding or noncoding mRNA, nor in the flanking intronic sequences. Expansion of intronic simple repeat motifs was investigated and also excluded.
The critical interval of the Chromosome 9 candidate region was further refined by SNPs rs16917452 and rs1555481. The corresponding chromosomal region of homozygosity extended from 9p21.1 to 9p13.2 and defined an interval of 6.49 Mb encompassing 164 genes of which 96 are protein coding. The remaining genes are pseudogenes or code for microRNAs.
Identification of the causative mutation
Investigation for the identification of the causative mutation within the Chromosome 9p21.1-p13.2 region was initially performed by Sanger sequencing of four selected candidate genes (ACO1, DNAJA1, UBAP1 and SIGMAR1). Selection was based on the function and expression of these genes. No small-scale mutation was detected in these genes. Due to the large number of the remaining genes, whole-exome sequencing was employed as it is a more robust method. Overall, more than 115 million sequencing reads were produced (99.17% of these were aligned to the hg19, 91.92% of target bases had at least 20×depth and 83.37% had at least 30×depth).
Analysis of obtained data focused on the 9p21.1-p13.2 region. After excluding the known variants reported as certified polymorphisms in public databases, only three homozygous variants in the GBA2 gene remained as candidates and were confirmed by Sanger sequencing of all family members (Fig. 5A). These variants are the novel nonsynonymous mutation c.1780G > C [p.Asp594His] in exon 11 that was found in complete cosegregation with the other two variants c.2054 + 62G > A (novel in intron 13) and c.2201G > A [p.Arg734His] in exon 15 (rs142621039:C > T) and also with the disease in the family. Nucleotide and protein positions of the reported variants are based on RefSeq accession numbers NM_020944.2 and NP_065995.1. Out of the three mutations, only c.1780G > C is strongly predicted to alter the protein structure (Polyphen and Raptor-X programs), and to cause deleterious consequences (Sift and Condel Algorithms). Screening of the three mutations in control Cypriot chromosomes revealed that: (1) The c.1780G > C mutation (tested with a PCR-RFLP assay) was absent from 264 control Cypriot chromosomes. (2) The c.2054 + 62G > A mutation (tested with a PCR-RFLP assay) was identified in 4 out of 208 control Cypriot chromosomes (subgroup of the 264 cohort). (3) The c.2201G > A variant (tested with Sanger sequencing) was identified in the same four individuals as the c.2054+62G > A variant out of 52 control Cypriot chromosomes (subgroup of the 264 cohort). Both the c.2054 + 62G > A and the c.2201G > A variants cosegregated in a heterozygous state in the four control individuals. Moreover, according to the dbSNP and Ensemble databases, the c.2201G > A variant was detected in the European-American population in a heterozygous state only (5/4300 individuals), with an average genotype frequency of 0.001.
We hereby describe the investigation of a Cypriot consanguineous family with spastic ataxia. We report mapping of the disease locus in this family to a 6.49 Mb region on Chromosome 9p21.1-p13.2 and the identification of a novel GBA2 gene missense mutation (c.1780G > C [p.Asp594His]). All the patients in this family presented with mixed features of cerebellar ataxia and spasticity. Spasticity was increased during the disease progression affecting initially the lower limbs and truncal muscles and later the upper limbs. Some additional features (cognitive impairment, hearing loss, urinary incontinence and dysphagia) often observed in other genetic diseases with spastic-ataxia as the predominant clinical feature were observed in this family as well.
Mapping of the disease locus was confirmed by two independent approaches; genome-wide STR genotyping followed by linkage mapping and genome-wide homozygosity mapping using SNP microarray data. The APTX gene, mutations in which cause ataxia with oculomotor apraxia type 1 (AOA1) (Moreira et al., 2001), is located within the mapped region and has been excluded after thorough investigation. This finding was also in agreement with the absence of hypoalbuminaemia and hypercholesteronemia in the patients of this family as compared to other AOA1 patients. Oculomotor apraxia which is another frequent feature of the AOA1 patients was observed in a mild form in our patients, at the advanced stage of the disease (Palau & Espinos, 2006; D'Arrigo et al., 2008; Anheim et al., 2010). Exome sequencing, which has been used widely in recent years and proves to be the most robust method for the identification of mutations, enabled the identification of the causative GBA2 gene c.1780G > C mutation. This mutation is located within the six-hairpin glycosidase-like domain and affects a highly conserved amino acid (Fig. 5B). This mutation is predicted to alter the protein structure and also to cause deleterious consequences. We also report the identification of two additional GBA2 variations (c.2054 + 62G > A and c.2201G > A) that cosegregate with the c.1780G > C mutation and with the disease in the family. Most probably, the c.1780G > C mutation has occurred on the background of the cis c.2054 + 62G > A and c.2201G > A variants. Even though there are no strong predictions for deleterious consequences, the fact that the two variations were found in unaffected individuals only in a heterozygous state leads to the hypothesis of a possible pathogenic effect in a homozygous state. Therefore, a possible involvement of a combination of mutations in the disease pathogenesis cannot be excluded and will be clarified after further investigation.
In the course of our study, mutations in the GBA2 gene were published by other groups (Hammer et al., 2013; Martin et al., 2013), showing that the glucosylceramide metabolism is associated with the disease. The GBA2 gene encodes the nonlysosomal β-glucosidase 2 protein which performs similar glucosylceramidase activity to β-glucosidase 1, a lysosomal membrane protein involved in the primary catabolic pathway for glucosylceramide. However, the two enzymes have no homology between them, they have different expression and subcellular localization and they also differ in other features such as their substrate and inhibitor specificities. β-Glucosidase 2 is a ubiquitously expressed enzyme with high expression in the brain, heart, skeletal muscle, kidney and placenta (Matern et al., 2001). It is localized at the endoplastic reticulum and Golgi complex cytosolic surface (Korschen et al., 2013) and is strongly associated with membranes. It was first recognized as a microsomal bile-acid β-glucosidase (Matern et al., 1997), while more recently, it was reassigned to sphingolipid metabolism (Yildiz et al., 2006), a critical process for the nervous system (Mencarelli & Martinez-Martinez, 2013). More specifically, it catalyses the hydrolysis of the sphingolipid glucosylceramide to glucose and ceramide which is further rapidly converted into sphingomyelin. In the opposite direction, it can act as a glucosyltransferase by adding glucose to lipid substrates (Yildiz et al., 2006). Contrary to β-glucosidase 2, β-glucosidase 1 is placed in the lysosomal department and ceramides produced by this enzyme are further degraded into sphingosine and fatty acids. A defect in β-glucosidase 1 is known to cause the most common lysosomal storage disease, Gaucher disease (Tsuji et al., 1987) which is characterized by accumulation of glucosylceramide in the lysosomes of macrophages in different organs (Grabowski et al., 1990). Moreover, mutations in the GBA gene have been associated with Parkinson's disease and related disorders (Sidransky & Lopez, 2012; Tsuang et al., 2012).
In the course of the final stages of our study, defects in GBA2 were reported to be associated with two monogenic disorders; hereditary spastic paraplegia (Martin et al., 2013) and ARCA with spasticity (Hammer et al., 2013). Five of the described mutations lead to the formation of stop codons with predicted truncation of the protein, while two single nucleotide substitutions alter the protein structure and affect conserved residues. The c.1780G > C mutation that we identified is located in a highly conserved amino acid across many mammalian species indicating its importance in the function of the protein.
The role of this enzyme in the central nervous system development was confirmed through the validation of the functional phenotype of some of the identified mutations in vivo (Martin et al., 2013), thus demonstrating the connection of sphingolipid metabolism and neurodegeneration. Sphingolipids constitute one out of the three major classes of plasma membrane lipids and their important involvement in cellular biological processes has been well recognized (Mencarelli & Martinez-Martinez, 2013). Their synthesis and degradation are significantly related to ceramide metabolism. Glycosphingolipids are mainly abundant in neural cellular membranes thus playing a significant role in neuronal normal functions. They serve as structural components that affect their dynamic properties and as signalling receptors (Korschen et al., 2013; Mencarelli & Martinez-Martinez, 2013). Their concentration and distribution are decisive for their actions. Deficiency or loss of β-glucosidase 2 function leads to glycolipid accumulation and alteration of the sphingolipid-ceramide pathway in the brain and other tissues, thus probably leading to neurological and/or other additional symptoms (Martin et al., 2013; Mencarelli & Martinez-Martinez, 2013).
In conclusion, we report a novel missense mutation in the GBA2 gene that is associated with spastic ataxia in a Cypriot consanguineous family, thus further confirming the association between GBA2 mutations and ARCA. We also hypothesise the possible involvement of a combination of GBA2 mutations in the disease pathogenesis in this family, which needs to be investigated further.
The authors thank the patients and family members for participation in this study. This work was partly supported by grants from the Cyprus Research Promotion Foundation (ΥΓΕΙΑ/ΔΥΓΕΙΑ/0308(BΕ)/05) and the A.G. Leventis Foundation.