Neurology Publish Ahead of Print The Phenotypic Continuum of ATPLA3-Related Disorders

the may be corrected during these


INTRODUCTION
Throughout the last 20 years, pathogenic variants in ATP1A3 have been discovered to cause an ever-expanding range of rare neurological phenotypes, affecting both children and adults.
ATP1A3 encodes the α3 subunit of a Sodium-Potassium-ATPase (NKA) present in excitable (neuronal and cardiac) cells. The α3-subunit has a relatively low Na+ affinity coupled with a high affinity to ATP 1,2 . Consequently, the NKA carrying the α3-subunit is ideally configured to clear high intraneuronal sodium concentrations occurring after intense neuronal firing, by being able to utilise the low concentration of ATP that will occur near the neuronal membrane shortly after an energy-demanding task.
In this study, we aim to describe the phenotypic features of a cohort of previously undiagnosed individuals with developmental delay and a neurological presentation, carrying a pathogenic/likely pathogenic ATP1A3 variant and examine where they fit within the current spectrum of ATP1A3-related disorders. We also perform a literature review of all ATP1A3 variants published thus far in association with human neurological disease. Our work clearly demonstrates the heterogeneous clinical spectrum associated with ATP1A3 variants, as well as phenotypic overlap between patients, that will streamline the diagnostic process.

SUBJECT COHORT
An application was made to the Deciphering Developmental Disorders (DDD) study 17

for a
Complementary Analysis Project (CAP), allowing access to anonymised details of individuals with ATP1A3 variants identified through this study (https://www.ddduk.org/). If variant analysis and phenotypic details made pathogenicity likely, responsible clinicians were contacted to invite patients and their families to study recruitment. Some of the contacted clinicians had further individuals with ATP1A3 variants in their care, which they put forward as potential participants. Genomic diagnosis was reached through trio whole exome sequencing (WES) for the DDD participants and through either WES or diagnostic gene panels for the other study participants. Phenotypic details were collected using a standardized clinical proforma covering all symptoms previously reported in ATP1A3-related disorders, as well as MRI features.
We used the UpSetR package in R to visualize intersections of signs and symptoms, trying to identify common phenotypes amongst individuals. As the clinical proforma included a long list of symptoms and signs due to the phenotypic variability of ATP1A3-related conditions, we decided to also group symptoms into broader categories where possible. We formed 4 Spanish language (where phenotypic and genomic information were included in the English abstract) reporting individuals carrying a heterozygous ATP1A3 variant considered to be pathogenic. One publication in Japanese, one in Russian and 3 in Chinese were not included.
Only publications with sufficient details about the variant (nucleotide change and/or amino acid change and gene transcript) and patient phenotype were included. All individuals published were counted, unless clearly stated that they had already been published elsewhere, in which case they were only counted once. However, it is possible that cohorts overlap.
Although ATP1A3 variants have been reported in different gene transcripts, in this article, all variant nomenclature adhere to transcript NM_152296 (isoform 1). Variants that were inconsistent with all available transcripts were presumed to have been reported incorrectly and excluded.
We calculated Combined Annotation Dependent Depletion (CADD) scores 21 for all missense ATP1A3 variants collected from the literature, as well as for all missense variants reported within ClinVar as likely benign and benign and compared them. Unlike other genomic annotations, that tend to exploit a single information type (i.e. conservation), CADD is a framework that objectively integrates many diverse annotations into a single, quantitative score. The integrated annotations include conservation metrics, functional genomic data, transcript information and protein-level scores (Grantham, SIFT, PolyPhen). CADD calculates a raw score and a 'PHRED-scaled' score. 'PHRED-scaled' scores are normalized to all potential ∼9 billion SNVs, thus providing a comparable unit for analysis. So, a 'PHRED-scaled' score of >=10 indicates a raw score in the top 10% of all possible reference genome SNVs, a score of >= 20 or greater indicates a raw score in the top 1%, etc 23 . The developers of CADD do not suggest a rigid cutoff to suggest pathogenicity; however, looking at various HGMD molecular categories of 174,183 disease-associated deleterious mutations, Itan et al. 24 found mean CADD scores for pathogenic missense variants to be above 20. The tool is freely available on cadd.gs.washington.edu.

Constraint Analysis
Constraint analysis 25 was performed on all missense pathogenic variants in the study cohort, as well as for published cases and compared to reported benign missense variants in the population database GnomAD. The number of benign missense variants present within every 10 amino acid residues was plotted across the length of the gene and from this, the missense constraint heat map was generated using the following parameters: Dark green = >20

SUBJECT COHORT
27 individuals with ATP1A3 variants in the DDD cohort (nearly 14000 children recruited with their parents) were identified. In seven individuals, the ATP1A3 variant was classified as class 2 (likely benign) due to either an incongruent phenotype or high prevalence in healthy populations. For the remaining 20 individuals, two clinicians did not report back, two clinicians declined participation in the study and three individuals did not consent to study participation. As a result, 13 individuals from the DDD cohort were included in this study.
Another 11 individuals with class 4 and 5 (likely pathogenic/pathogenic) ATP1A3 variants were volunteered by collaborating clinicians and included as they fulfilled the study criteria.
Three were family members of two separate DDD patients. Three individuals have been previously published in the literature (Patients 3, 5 and 14). Table 2 summarises the ATP1A3 variants present in our cohort, including 13 variants that have not been published previously.
Motor delay was reported in 20 (83.3%) individuals, with walking age ranging from normal at 13 months to some individuals not having learnt to walk by 18 years. Language delay was also reported in 20 (83.3%) individuals. Communication skills were very varied, ranging from starting to communicate at nine months to not having aquired language at age 20 years. Looking at the phenotypes of individuals sharing the same ATP1A3 variant, it seems that, whilst some genotypes are strongly associated with specific phenotypes, there are others that result in more phenotypic variability, such as those associated with c.2116G>A (p.Gly706Arg) and c.2839G>T (p.Gly947Trp). However, this is an observation based on a very small number of individuals and it might be that given the opportunity to look at larger cohorts of the rarer ATP1A3 variants, mutation-specific phenotypes will arise, as they have for c.2452G>A (p.Glu818Lys) and variants at amino acid residue 756. This is important information to gather as it may help clinicians provide families with more accurate prognosis after diagnosis.
Fifty percent of our patients were reported to have a history of epileptic seizures. In half of these, epileptiform features were seen on EEG. The epilepsy phenotype varied amongst patients with some having focal seizures, whilst others had generalised epilepsy. This variability in epilepsy phenotype has been described previously in a study of 51 patients with an AHC phenotype 35  Earlier studies of patient cohorts with a clinical phenotype of AHC mostly reported normal MRIs 37 . More recently, however, as the diverse phenotypes associated with ATP1A3 are evolving, several reports of abnormal neuroimaging have also been published including cerebellar atrophy 38 and polymicrogyria [8][9][10] . Eleven of our patients also had abnormal MRI, most commonly with cerebellar atrophy, in one case proven to be progressive. Traditionally, patients with an ATP1A3-related phenotype, such as AHC or RDP, have been diagnosed by utilising clinical diagnostic criteria 41,42 . In recent years with the association of ATP1A3 variants with a broadening clinical spectrum, this approach is not feasible for all patients, as many do not fulfil classic phenotypic criteria. Also, with broad genetic testing (gene panels, WES, whole genome sequencing (WGS)) being brought into the diagnostic process at a much earlier stage, clinicians are often faced with an ATP1A3 variant in an undiagnosed patient, trying to decide whether it is responsible for the phenotype, rather than having already reached a clinical diagnosis and using genetic investigations to confirm or inform it. We compared the phenotypic characteristics of our cohort to the diagnostic criteria published for AHC, RDP, CAPOS, D-DEMØ (Table 1) Overall, we found that most of our patients cannot be grouped into any of the existing described phenotypes. Rosewich et al. 44 published major and minor criteria to support a diagnosis of an ATP1A3-related condition. The authors identified five major and five minor criteria for patients with infantile and early childhood onset, six major and six minor criteria for patients with childhood and adult onset, and seven major and seven minor criteria applicable for patients presenting at any age. So, for early onset there are 12 major and 12 minor criteria overall, whilst for late onset there are 13 major and 13 minor criteria. No cutoff is given as to how many criteria should be fulfilled to establish an ATP1A3-related condition diagnosis. In our cohort, 22 individuals had an onset in infancy/ early childhood, and two had an onset in later childhood or adulthood. All patients met at least three minor criteria. The only criteria met by all individuals in this cohort were cognitive impairment and negative family history or history suggesting autosomal dominant inheritance (both minor criteria). On average, patients met 3.4 major and 5 minor criteria. This approach of defining a spectrum of associated symptoms seems to be preferable for patients with ATP1A3-related disorders, if the threshold for number of criteria needing to be fulfilled to prompt testing is kept low.
In addition to this, we found that in our cohort looking for a combination of broad symptom categories, namely paroxysmal symptoms, hyperkinetic symptoms, neuropsychiatric symptoms, and cognitive impairment, rather than specific symptom combinations was more likely to identify patients with ATP1A3-related disorder. A CADD-score above 20 and a variant located within the mutation clusters in regions of constraint further support diagnosis of an ATP1A3-related disorder.
There are limitations to our study. The phenotypic information was collected in retrospect through patient interview or case note review, rather than prospective evaluation. The phenotypic information available for published cases is variable and sometimes limited; patients are reported at different ages, clinical information is collected retrospectively, and different authors focus on different symptoms.