Clinical and genetic analysis of ATP13A2 in hereditary spastic paraplegia expands the phenotype

Abstract Background Hereditary spastic paraplegias (HSP) are neurodegenerative disorders characterized by lower limb spasticity and weakness, with or without additional symptoms. Mutations in ATP13A2, known to cause Kufor–Rakeb syndrome (KRS), have been recently implicated in HSP. Methods Whole‐exome sequencing was done in a Canada‐wide HSP cohort. Results Three additional patients with homozygous ATP13A2 mutations were identified, representing 0.7% of all HSP families. Spastic paraplegia was the predominant feature, all patients suffered from psychiatric symptoms, and one patient had developed seizures. Of the identified mutations, c.2126G>C;(p.[Arg709Thr]) is novel, c.2158G>T;(p.[Gly720Trp]) has not been reported in ATP13A2‐related diseases, and c.2473_2474insAAdelC;p.[Leu825Asnfs*32]) has been previously reported in KRS but not in HSP. Structural analysis of the mutations suggested a disruptive effect, and enrichment analysis suggested the potential involvement of specific pathways. Conclusion Our study suggests that in HSP patients with psychiatric symptoms, ATP13A2 mutations should be suspected, especially if they also have extrapyramidal symptoms.

Herein, we report three additional HSP patients from three different families with homozygous ATP13A2 mutations. Long-term follow-up, genetic analysis, protein structure, and network analyses were done to explore the clinical and genetic spectrum of ATP13A2-related disease.

| Population
HSP patients (n = 696) from 431 families were recruited across Canada, and data on diagnosis, recruitment, and the cohort were previously published (Chrestian et al., 2017). Of those, 383 HSP genetically undiagnosed patients went through whole-exome sequencing (WES). All participants have signed an informed consent form and the study protocol was approved by the institutional review board.

| Genetic analysis
Whole-exome capture, sequencing, alignment, annotation, and variant calling was performed as previously described (Chrestian et al., 2017). Nonsynonymous, frameshift, splicesite, and stop variants with allele frequencies <0.005 in the Exome Aggregation Consortium (ExAC) database were filtered-in, and segregation analysis was performed. The potential pathogenicity of variants was estimated based on their frequency of in gnomAD and ExAC, and by in silico tools: MutationTaster, combined annotation-dependent depletion (CADD), genomic-evolutionary rate profiling (GERP++), sorting intolerant from tolerant (SIFT), and PolyPhen-2.

| In silico analysis of ATP13A2
Genic intolerance of ATP13A2 was assessed using the residual variation intolerance score (RVIS) tool. Pathways enrichment and interaction network were analyzed using GeneMANIA, g:Profiler, and STRING, and networks were visualized by Cytoscape. Clustal Omega program was used for protein sequence alignment of multiple species. A 3D atomic model of human ATP13A2 was built using the automated I-TASSER server. The steric clashes induced by each mutation were evaluated using the mutagenesis toolbox in  Figure 1a) demonstrated that the ATP13A2 protein closely interacts with other HSP-related proteins. Pathway enrichment analysis of genes which are known to be involved in HSP, ALS, and Parkinsonism showed enrichment (FDR p < .05, Table S1) of genes involved in copper ion binding, vesiclemediated transport cellular response to oxidative stress, among others. Both p.   Table 1 details the clinical characteristics of previously published ATP13A2-related HSP patients and the three patients identified in this study. The description of the patients below will detail only the main characteristics. | 3 of 7 ESTIAR ET Al.

| Patient A
The patient, a 44-year-old woman of Inuit-Canadian origin, was initially evaluated at age 31 due to gait dysfunction. She was found to have bilateral lower extremity spasticity, weakness, hyperreflexia, nonsustained bilateral ankle clonus and speech difficulties. On evaluation at age 40, she was laughing excessively and seemingly had an inappropriate affect. Minor Parkinsonian tremor and action tremor were noted. At age 43, the patient was agitated and verbally and physically aggressive. Fine movements were decreased, and spinocerebellar ataxia and prominent spastic paraplegia were present. Some of her parkinsonian symptoms may be attributed to her treatment with haloperidol. Brain and spine MRI demonstrated diffuse cerebellar atrophy and normal spine ( Figure  2a). WES revealed a p.(Leu825Asnfs*32) mutation which results in a truncated peptide of 857 a.a and deletion of six C-terminally located transmembrane alpha-helixes. The mutation has not been reported in gnomAD and ExAC, but was previously reported (also as homozygous) in a patient with KRS from a Greenlandic Inuit family (Eiberg et al., 2012). As our patient is of Inuit-Canadian family, this might suggest that this is an old, founder Inuit mutation.

| Patient B
During childhood, this patient from an Armenian-Lebanese consanguineous family had impairment of fine motor movements, and experienced learning difficulties and delayed mental development noticeable at the age of 6 years. On evaluation at age 18, the patient presented with increased  muscle tone especially in the lower extremities and gait was spastic. A Levodopa trial did not result in any improvement. At age 24, the patient had a seizure for the first time. MRI demonstrated moderate diffuse cerebral and cerebellar atrophy ( Figure 2b). EEG demonstrated mild, slow biposterior dysfunction but without epileptiform patterns. On evaluation at the age of 31 years, the patient started to develop ideas of reference and delusions. WES was performed and identified a novel homozygous missense ATP13A2 mutation, p.(Arg709Thr), in exon 19 within the hydrolase domain. The mutation is predicted to be deleterious by CADD (25), Polyphen-2 (0.99), MutationTaster (1), and was located in a highly conserved amino acid with GERP++ score of 5.

| Patient C
At age 6, after normal development, this male patient of French-Canadian origin was reported to have learning difficulties that became more pronounced through high school. At age 12, the patient started abusing alcohol and drugs, and throughout his teenage years he had two psychotic episodes and paranoid delusions. On evaluation at age 32, the patient had presented with spasticity and ataxia, spastic and mildly magnetic gait with frequent falls. Brain MRI done at the age of 29 showed cortical and cerebellar atrophy (images are not available). Metabolic workup, EEG, EMG, nerve conduction studies, and an abdominal ultrasound were normal. Clinical WES identified a homozygous ATP13A2 missense mutation in exon 20, p.(Gly720Trp), predicted to be deleterious by SIFT (0), Polyphen-2 (1), and CADD (31).

| DISCUSSION
We describe three unrelated patients with predominant spastic paraplegia features, harboring homozygous ATP13A2 mutations that are either novel or were not previously reported in HSP. ATP13A2-HSP is rare, responsible for 0.4% of all HSP patients and 0.7% of all HSP families in CanHSP. Interestingly, all three patients suffered from psychiatric symptoms, which were previously reported in only one SPG78 patient (Estrada-Cuzcano et al., 2017). One of the patients has developed seizures, which have not been previously reported in SPG78. Mild extrapyramidal symptoms/ signs were present in patient A and bradykinesia in patient B. MRI in all three patients demonstrated cerebellar and/or cerebral atrophy, consistent with previous reports on ATP13A2-HSP (Table 1).
Interestingly, the p.(Leu825Asnfs*32) mutation in patient A resulted in HSP-predominant phenotype, while in previous patients reported with the same mutation it was Parkinsonismdominant phenotype (Eiberg et al., 2012). This may suggest that the clinical presentation may be affected by other genetic F I G U R E 2 MRI images, pedigrees, and Sanger sequencing chromatograms. (a) Patient A's MRI showed diffuse cerebellar atrophy (arrow).
(b) Patient B's MRI showed moderate diffuse cerebral and cerebellar atrophy (arrows). (c) MRI and DNA for sanger sequencing were not available for patient C and/or environmental factors. We also identified a novel missense ATP13A2 variant, p.(Arg709Thr) in a highly conserved amino acid located within the hydrolase domain which is critical for the catalytic activity of ATP13A2. Furthermore, this variant is affecting the last nucleotide of the exon, which may also affect splicing and possibly result in nonsense mediated decay. This possibility needs to be studied preferably in neuronal models with the variant. The mutation in patient C, p.(Gl-y720Trp) changes Glycine to Tryptophan at codon 720, which could unfold the N-domain of ATP13A2 (Figure 1e). Our pathway enrichment analysis may suggest that copper ion binding is involved in the pathogenesis of specific forms of HSP, and further studies are required to examine this possibility.
This study has several limitations. Since DNA was not available for segregation analysis, and since our genetic data include only in silico prediction tools and structural models, we could not prove with full confidence that the detected variants in ATP13A2 are disease causing. However, one of the mutations was previously described in a patient, and it is unlikely that by chance alone two extremely rare biallelic variants in a gene that is already known as disease causing will be found in two HSP patients. Therefore, it is probable that these variants are disease causing. An additional limitation is the lack of available DNA for patient C, therefore the variant reported by the clinical lab could not be independently confirmed.
Our study expands the genetic and phenotypic spectrum of ATP13A2-related HSP. The different phenotypes observed in carriers of ATP13A2 mutations imply that genetic and nongenetic modifiers exist. Our findings may also suggest that in HSP patients with psychiatric symptoms, mutations in ATP13A2 should be suspected, especially if mild parkinsonian symptoms are also present.