Application of a custom NGS gene panel revealed a high diagnostic utility for molecular testing of hereditary ataxias

Hereditary ataxias (HA) are a rare group of heterogeneous disorders. Here, we present the results of molecular testing of a group of ataxia patients using a custom-designed next-generation sequencing (NGS) panel. Due to the genetic and clinical overlapping of hereditary ataxias and spastic paraplegias (HSP), the panel encompasses together HA and HSP genes. The NGS libraries, comprising coding sequences for 152 genes, were performed using KAPA HyperPlus and HyperCap Target Enrichment Kit, sequenced on the MiSeq instrument. The results were analyzed using the BaseSpace Variant Interpreter and Integrative Genomics Viewer. All pathogenic and likely pathogenic variants were confirmed using Sanger sequencing. A total of 29 patients with hereditary ataxias were enrolled in the NGS testing, and 16 patients had a confirmed molecular diagnosis with diagnostic accuracy rate of 55.2%. Pathogenic or likely pathogenic mutations were identified in 10 different genes: POLG (PEOA1, n = 3; SCAE, n = 2), CACNA1A (EA2, n = 2), SACS (ARSACS, n = 2), SLC33A1 (SPG42, n = 2), STUB1 (SCA48, n = 1), SPTBN2 (SCA5, n = 1), TGM6 (SCA35, n = 1), SETX (AOA2, n = 1), ANO10 (SCAR10, n = 1), and SPAST (SPG4, n = 1). We demonstrated that an approach based on the targeted use of the NGS panel can be highly effective and a useful tool in the molecular diagnosis of ataxia patients. Furthermore, we highlight the fact that a sequencing panel targeting both ataxias and HSP genes increases the diagnostic success level.


Introduction
Hereditary ataxias (HAs) are a group of complex and progressive neurodegenerative disorders that present various modes of inheritance. Accordingly, they can be classified as spinocerebellar ataxias (SCA -AD), episodic ataxias (EA -AD), autosomal recessive spinocerebellar ataxias (SCAR), and spastic ataxias (SPAX -AD/AR). To date, 48 genes have been associated with SCA, 61 with SCAR, 9 with EA, and 6 with SPAX. In addition, seven genes are known to be involved in X-linked cerebellar ataxias, and 4 are involved in ataxias with mitochondrial disorders (Bird 2019). Despite the fact that many ataxia genes have already been discovered, novel candidate genes are still emerging (Valence et al. 2019).
Clinically, the phenotype may include variable features, but generally, the disease is characterized by progressive cerebellar syndrome with balance and coordination problems, gait and limb ataxia, hypotonia, clumsiness, or speech and eye movement abnormalities, whereas the latter can display early or young adult onset of the disease. Moreover, some patients can present with overlapping features, and there may be intrafamilial/interfamilial phenotypic variability (including different ages at onset and severity of symptoms) or incomplete penetrance (Angelini et al. 2019). Additionally, mutations in the same SCA genes may lead to distinct phenotypes or exhibit different modes of inheritance. Cerebellar atrophy is a neuropathological and neuroimaging hallmark of ataxia.
Genetically, the molecular background of ataxias varies widely, ranging from point and small mutations to dynamic ones. The latter account for 45% of all autosomal dominant cerebellar ataxia cases (Durr 2010). Therefore, due to this clinical and genetic heterogeneity, the establishment of an ataxia diagnosis may be very complicated and challenging. Recent studies have shown that many patients remain undiagnosed after screening for the most common repeat expansions (Brusco et al. 2004). The implementation of next-generation sequencing (NGS) approaches, including whole-genome sequencing (WGS), whole-exome sequencing (WES), and targeted gene panel sequencing (TGP), enables comprehensive screening of many genes simultaneously. This can greatly increase the possibility of identifying the genetic cause of these conditions. Currently, custom gene testing is one of the most widely used diagnostic tools for heterogeneous neurodegenerative disorders in clinical practice.
In this study, we present the results of a genetic analysis using a custom-designed panel encompassing all known genes for hereditary ataxia and hereditary spastic paraplegia (HSP) in a cohort of Polish patients. Furthermore, we assessed the efficiency of our NGS panel as a diagnostic tool for the possible expansion of the phenotypic and genotypic spectrum of ataxias in clinical practice.

Materials and methods
This study was approved by the Ethics Commission of IPiN. All individuals participating in the research gave their written informed consent.

Methods
Genomic DNA was extracted from the peripheral blood using the MagNA Pure Compact Nucleic Acid Isolation Kit I -Large Volume (Roche), following the manufacturer's instructions. The quantity and quality of the isolated DNA were assessed with a UV/VIS Spectrophotometer Nan-oDrop2000 (Thermo Fisher Scientific) and Qubit fluorometer (Invitrogen, Thermo Fisher Scientific).
The library of patient DNA was prepared from 250 ng genomic DNA with a KAPA HyperPlus Kit (Roche), according to the manufacturer's instructions. The protocol included the following steps: library preparation, hybridization, bead capture, washing, amplification enrichment QC, sequencing, and pre-and post-capture multiplexing. Quantification analysis and assessment of the average size and length of the NGS libraries were performed using a Bioanalyzer assay (Agilent).
Sequencing on the NGS libraries was performed by a MiSeq (Illumina) paired-end 2 × 75-bp DNA sequencing platform with a MiSeq Reagent Kit v3 (150-cycle, Illumina), according to the manufacturer's procedure.
The analysis of gene variants was performed with the use of the BaseSpace Variant Interpreter, and interpretation was performed according to the American College of Medical Genetics and Genomics (ACMG) and Association for Molecular Pathology (AMP) Standards and Guidelines (Richards et al. 2015). The initial variation filtering included the following: (1) all coding consequences (stop gain or loss, splice site, indels, missense, and protein altering); (2) Gno-mAD frequency value less than 2% for all populations; and (3) small variant QC metrics with value > 25% for variant read frequency. To investigate the functional predictions of the variants by several in silico online programs, PolyPhen-2 (http:// genet ics. bwh. harva rd. edu/ pph2/), SIFT (https:// sift. bii.a-star. edu. sg/), and MutationTaster (http:// www. mutat ionta ster. org/) were used. The analysis and interpretation of the clinical significance of the DNA variants were determined by using ClinVar, the Leiden Open Variation Database (LOVD), and the Human Gene Mutation Database (HGMD), and the variant frequency in populations was assessed by Genome Aggregation Database (gnomAD). The assessment of the quality of the NGS data was performed by using the genome visualization tool Integrative Genomics Viewer (IGV) and included only variants that had ≥ 15 reads.

Sanger sequencing
All pathogenic or likely pathogenic variants identified by NGS were confirmed using Sanger sequencing on an ABI 3130 genetic analyzer. The available additional (affected and healthy) family members were tested for the segregation of identified pathogenic/likely pathogenic variants and variants of uncertain significance (VUS).

General characteristics of the patients
The 29 patients enrolled in the study presented progressive ataxia as the main clinical manifestation. Adult onset was observed in 48.3% of cases (14/29), whereas the remaining 37.9% (11/29) were young adult onset (18-29 years) and 13.8% (4/29) were childhood onset (≤ 18 years). The mean and median age of the study participants at disease onset was 32.7 (range between 7 and 70) and 26.0 years, respectively. The cohort comprised 14 male (48.3%) and 15 female individuals (51.7%).

NGS parameters
The mean coverage depth of each region was estimated to be 110.0x , with the highest score of 153.4x and the lowest of 68.1x. The mean proportion of the sequence achieving 30x sequencing coverage was 95.5% and ranged from 89.6 to 97.2%.
One of the patients presented an ultrarare variant in the ANO10 gene due to consanguinity. A homozygous splice site pathogenic mutation in ANO10, c.1218+1G>C, was identified in a 33-year-old man (patient HA). Sanger sequencing was performed for the affected proband and both asymptomatic parents, who were heterozygous carriers for this mutation (Fig. 1). The affected patient clinically presented with gait and limb ataxia, dysarthria, intention tremor, and age of onset of 20 years. Remarkably, he also manifested other neurological conditions, such as increased deep tendon reflexes, extensor plantar responses, dysdiadochokinesia, and spasticity. Brain MRI showed bilateral cortical atrophy and cerebellar vermis.

Variants of uncertain significance
In the remaining patients (13/29), more than one variant of uncertain significance (VUS) per case was identified. Interpretation of these variants is difficult and requires a more detailed analysis, including clinical and familial interpretation, for the proper identification of the most likelihood disease-causing mutations (Table 3). Among them, 2 patients had a VUS with a dominant mode of inheritance (AFG3L2, CCDC88C), and 3 had a single VUS in a recessive manner (CAPN1, SACS, CLCN2).
In 5 individuals, the interpretation of genetic findings remained unknown because of the complexity of genotype-phenotype correlation, and in 3 cases, it ultimately remained undiagnosed.

Discussion
Application of the designed panel in ataxia patients revealed that the overall mutation detection rate amounted to 55.2% and varied from 33.3% in those with a familial history and autosomal dominant mode of inheritance (3/9 cases) to 53.3% among sporadic cases (8/15 cases). Among the latter cases, we identified variants in genes associated with disorders that are inherited in both autosomal recessive and autosomal dominant traits. The obtained results present significant molecular effectiveness in comparison with other NGS-based studies. For example, Németh et al. reported definite molecular diagnosis by gene panel sequencing in 18.0% of 50 probands with ataxia, although the detection rate varied from 8.3 to 40.0% depending on the age of symptoms onset (adult-, childhood-, or adolescent-onset disease) (Németh et al. 2013). Other studies using NGS exome sequencing revealed a diagnostic yield of 21.0% in a cohort of 76 patients with chronic progressive cerebellar ataxias and potential additional diagnoses in 40.0% (Fogel et al. 2014). A study of 319 cerebellar ataxia cases showed a yield of 22.6%, with possible additional diagnoses in 5.9% (Coutelier et al. 2018). Due to comprehensive diagnostic investigations that involve clinical assessment and differentiate ataxia from acquired, primary cause, or secondary cause by examining the clinical findings, such as blood tests, neuroimaging tests, and genetic tests, a definitive molecular diagnosis of patients with suspected clinically hereditary ataxia often takes years and is highly complex.
In our study, the most frequently pathogenic mutations were located in the POLG, CACNA1A, SACS, and SLC33A1 genes. Other variants were detected in the following genes: STUB1, SPTBN2, TGM6, SETX, ANO10, and SPAST. The most common type of mutation was missense mutation.
In the POLG gene, we identified the two known pathogenic variants most frequently present in Caucasians: c.1399G>A (p.Ala467Thr) and c.2243G>C (p.Trp748Ser). These variants usually occur in the homozygous state and have high variability in their clinical presentation (Neeve et al. 2012;Van Goethem et al. 2004), which may be caused by genetic, epigenetic, and environmental factors (Neeve et al. 2012). Notably, our tested individual with a heterozygous mutation of c.1399G>A in the POLG gene clinically presented with gait and limb ataxia, motor and sensory neuropathy, positive Romberg sign, decreased sensory nerve conduction velocity, impaired vibration sense in the right foot, cardiomyopathy, and age of onset of 70 years (patient RM). However, it was previously assumed that heterozygous carriers were unaffected cases (Neeve et al. 2012; Van Goethem et al. The gold standard in clinical practice should be the use of a focused approach for both NGS and Sanger sequencing. As an example, the homozygous splice site pathogenic mutation c.1218+1G>C in ANO10 was confirmed by Sanger sequencing in the proband and both parents in a consanguineous family. In addition, the analysis of cosegregation of the variants with the disease in families should always be performed when possible. Interestingly, genetic analysis of two unrelated patients with suspected hereditary ataxia identified the heterozygous mutation c.1559T>C (p.Ile520Thr) in the SLC33A1 gene, which is related to autosomal dominant spastic paraplegia type 42 (Lin et al. 2008) or recessive congenital cataracts, hearing loss, and neurodegeneration (Huppke et al. 2012). To the best of our knowledge, the heterozygous missense mutation c.339T>G (p.Ser113Arg) in the SLC33A1 gene was identified in one Chinese family with autosomal dominant pure HSP (Lin et al. 2008). In addition, no mutations in SLC33A1 were detected among 220 Caucasian patients with autosomal dominant hereditary spastic paraplegias and negative for mutations in the SPAST gene (Schlipf et al. 2010). Our results may suggest that mutations in the SLC33A1 gene can be associated with spinocerebellar ataxia, hereditary spastic paraplegia, or both. However, further segregation analysis of gene variants in SLC33A1 and functional assays or analysis of a large group of ataxia patients should be performed.
Additionally, we assume that further reanalysis of the most likely variants, e.g., c.2062C>G (p.Pro688Ala) in the AFG3L2 gene, which ranges from being of unknown significance to pathogenic, may increase the diagnostic yield. Therefore, eventually, it may be extremely important to check for updates of the databases, i.e., HGMD, OMIM, ClinVar, and LOVD, related to gene variants that were previously classified as VUS. At present, these diagnoses cannot be unequivocally settled due to the lack of samples from parents.
The analysis of cosegregation of the variant c.146A>G in the STUB1 gene for the affected mother and cousin of the proband showed the presence of a heterozygous mutation in both, confirming that this variant cosegregates with SCA48. These genetic findings enable the renaming of this variant from VUS to pathogenic. Moreover, 128 affected patients tested for NGS did not reveal the presence of this mutation, which could be strong evidence for the presumed pathogenic variant. These findings together account for the pathogenicity of this variant in the STUB1 gene. Therefore, we propose that the most suspect VUS should be reported in the scientific literature because, with the growing knowledge in online databases, some gene variants can be reclassified from VUS to pathogenic.
The gene-specific NGS approach is subject to some limitations. The first limitation is the overlapping of neurological phenotypes and the presence of pathogenic mutations in genes encompassing other neurodegenerative disorders. The second limitation is the identification of more than one VUS per single case. The third limitation is the increasing information about novel variants in databases and the renaming of VUS as pathogenic variants. Thus, we do not exclude the possibility of a genetic cause of the disorder in patients with no known pathogenic variants detected in the SCA-SPG genes to date. In one such case, we were able to detect a deleterious heterozygous missense mutation of c.305C>T (p.Pro102Leu) in the PRNP gene that was associated with autosomal dominant Gerstmann-Straussler disease. A 32-year-old man (patient BW) presented with gait ataxia, progressive lower limb weakness, imbalance and incoordination, clumsiness, cerebellar dysarthria, scanning speech, rigidity, postural tremor, dysmetria, bilateral extensor plantar responses, hypertonia, dysdiadochokinesia, excessive reflexes of lower limbs, and axonal polyneuropathy, with an age at onset of 27 years old. MRI neuroimaging showed cerebellar and brainstem atrophy and thinning of the corpus callosum. The mutation was found through the analysis of different NGS panels that included 118 genes related to neurodegenerative and dementia disorders, which included the PRNP gene. However, we believe that the high detection rate confirmed that our strategy of using a targeted NGS approach that focuses on genes associated with both hereditary ataxias and hereditary spastic paraplegias was appropriate. This establishment is associated with the clinical and genetic overlapping of these two diseases, which is termed the ataxia-spasticity spectrum (Synofzik and Schüle 2017;Elert-Dobkowska et al. 2019).
Furthermore, together with single nucleotide variants, rare cerebellar ataxias are also caused by different types of mutations undetectable by NGS. Copy number variants (CNVs) in GRID2 are the main cause of SCAR18 (Ceylan et al. 2020). In addition, Friedreich's ataxia is caused by noncoding GAA repeat expansion or ultrarare point mutations in the FXN gene. Recently, Cortese et al. (2019) identified a biallelic intronic pentanucleotide AAGGG repeat expansion in the RFC1 gene that is associated with cerebellar ataxia, neuropathy, and vestibular areflexia syndrome (CANVAS) (Cortese et al. 2019). This finding expands the known molecular genetic basis of autosomal recessive cerebellar ataxia. All these examples can explain the likelihood of a lack of a definite molecular diagnosis in some ataxic patients after the application of only selected genetic tests.
In conclusion, we demonstrated that an approach based on a targeted NGS panel can be a highly effective and