Functional impact of the ATP1A3-p.A813V variant: insights into a calcium-driven hyperexcitability cascade in rapid-onset dystonia-Parkinsonism

Abstract

Background

Mutations in the neuronal Na⁺/K⁺-ATPase subunit ATP1A3 are linked to a spectrum of neurological disorders, including rapid-onset dystonia-parkinsonism (RDP), yet their pathogenic mechanisms remain incompletely understood. We describe the complex clinical characteristics of a patient with early-onset movement disorders and a likely pathogenic de novo variant in ATP1A3(c·2438C>T, p.A813V).

Methods

We identified a de novo heterozygous ATP1A3 p.A813V variant in a patient with clinically confirmed RDP and employed an integrative approach combining molecular dynamics (MD) simulations, zebrafish models, and patient-derived induced neurons (iNeurons) to delineate its pathogenesis.

Results

MD simulations revealed that the p.A813V substitution structurally distorts transmembrane helix packing, reduces structural stability, and diminishes water accessibility at the cation-binding site, predicting impaired Na⁺/K⁺-ATPase function. In vivo, atp1a3b knockout zebrafish developed pronounced neuronal hyperexcitability—reflected by elevated c-fos and pERK expression—that emerged before overt neurodegeneration, motor axonopathy, and neuromuscular junction defects. Complementarily, neurons expressing ATP1A3-p.A813V displayed significantly prolonged calcium transient decay times, suggesting a potential mechanism of altered Ca²⁺ handling and delayed clearance mechanisms compatible with ATP1A3 dysfunction. Consistent with these findings, patient-derived iNeurons exhibited markedly reduced ATP1A3 protein abundance and Na⁺/K⁺-ATPase activity.

Conclusions

Together, these findings lead us to propose a mechanistic model in which ATP1A3 dysfunction disrupts Ca²⁺ homeostasis, triggers neuronal hyperexcitability, and culminates in progressive neurodegeneration. This work provides a molecular and functional framework for targeting ionic and calcium homeostasis as a promising therapeutic strategy for ATP1A3-related disorders.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-08203-0.

Introduction

Na⁺/K⁺-ATPase is a membrane transporter critical for establishing the electrochemical gradient necessary for neuronal signaling [1–4]. While the α₁ Na⁺/K⁺-ATPase encoded by ATP1A1 serves as a neuronal housekeeping gene to maintain basal intracellular Na⁺ concentrations, the neuronal α3 subunit, encoded by the ATP1A3 gene, is uniquely responsible for responding to transient, elevated Na⁺ loads that exceed the saturation level of the α1 subunit [5, 6]. This specialized function positions ATP1A3 as a vital regulator of neuronal excitability [7–11]. Consequently, mutations in ATP1A3 are linked to a spectrum of severe neurological conditions, including alternating hemiplegia of childhood, cerebellar ataxia with other symptoms (areflexia, pes cavus, optic atrophy, sensorineural hearing loss), childhood-onset schizophrenia, and rapid-onset dystonia-Parkinsonism (RDP) [1, 2]. RDP is a rare disorder characterized by sudden asymmetric dystonia and Parkinsonism, often triggered by fever, emotional stress, or infections [1]. Symptoms typically stabilize with minimal response to levodopa treatment, with onset occurring between 9 months and 55 years of age [3]. Heterogeneous clinical manifestations across affected families lead to frequent diagnostic delays, underscoring the need for a deeper understanding of the varied profiles of RDP symptoms to enable early diagnosis and effective treatment [12].

Functional analysis of ATP1A3 variants in RDP often focuses narrowly on reduced protein levels or ATPase activity, typically relying on case reports without the functional depth provided by patient-derived cells or integrated in vivo animal models. Here, we present the complex clinical characteristics of a patient with early-onset movement disorders and a likely pathogenic de novo variant in ATP1A3(c.2438C>T, p.A813V). Given the patient’s presentation, including spastic gait and hyperreflexia, which suggest upper motor neuron involvement and potential structural degeneration, a deeper mechanistic understanding is urgently required. To address this gap, we performed a comprehensive, multi-modal analysis. We first used molecular dynamics (MD) simulations to analyze structural changes in the ATP1A3-p.A813V mutation compared to the wild type (WT). Based on these structural predictions, we then utilized a zebrafish model to investigate the physiological and circuit-level consequences in vivo, specifically focusing on calcium dynamics, neuronal hyperexcitability, and the progression toward neurodegeneration. Subsequently, we investigated the resulting loss-of-function using patient-derived induced neurons (iNeurons) to validate the molecular impact on protein expression and enzymatic activity within a human genetic background. By integrating these cross-species findings, we propose a mechanistic cascade that explains how the p.A813V mutation leads to the complex clinical phenotype observed in RDP.

Methods

Clinical phenotypic profile

We describe the complex clinical characteristics of a patient with early-onset movement disorders and a likely pathogenic de novo variant in ATP1A3(c.2438C>T, p.A813V). A 25-year-old Korean female experienced progressive gait disturbance for 8 months. She reported a spastic gait with intermittent dystonic posture in the right hand during the initiation of gait, after symptom onset. Neurological examination revealed gaze fixation instability in lateral gaze, spastic dysarthria, spasticity, and increased deep tendon reflexes in the lower limbs. Additionally, intermittent dystonic posture was observed during gait initiation. The sensory and cerebellar functions were normal. She had a history of febrile seizures since birth. No family history of neurological diseases was reported.

Standard laboratory tests for thyroid disease, autoimmune disease, heavy metals, paraproteinemia, vasculitis, and cerebrospinal fluid (CSF) were performed and yielded negative results. Magnetic resonance imaging of the brain showed no remarkable findings. Genetic tests were done to identify the spectrum of genetic causes of motor neuron disease and various types of dystonia.

Whole-exome sequencing and Sanger sequencing

Informed consent was obtained from the proband and her parents for the Trio study using whole-exome sequencing to identify the genetic cause of the motor neuron disease and dystonia. Genomic DNA was extracted using the Wizard Genomic DNA Purification Kit according to the manufacturer’s instructions (Promega, Madison, Wisconsin). Library preparation was performed using SureSelect Human All Exon V5 (Agilent Technologies, Santa Clara, California), and sequencing was conducted on the Illumina NextSeq500 platform (Illumina Inc., San Diego, California). Alignment, reference genome indexing (hg19), and variant calling were performed using the GATK Best Practices-based pipeline. Variants with allele frequencies > 0.01 in public databases (gnomAD, https://gnomad.broadinstitute.org, Korean Reference Genome Database) were filtered out. Known MND and dystonia-causing genes and related genes were screened. A de novo variant was identified and confirmed by Sanger sequencing in proband DNA. Sanger sequencing revealed that the ATP1A3 p.A813V pathogenic variant was only present in the proband but absent in the asymptomatic younger sister and parents.

Modeling and simulations

The initial structure of the WT, featuring a K⁺-occluded state (PDB ID: 8D3X), as determined via cryo-EM, served as our starting point [13]. Because this structure lacks potassium ions, their positions were inferred from the crystal structure of the Squalus acanthias sodium-potassium pump (PDB ID: 2ZXE) [14]. Using UCSF Chimera for structure superimposition, the ions were placed accordingly [15]. The mutant structure was built by adopting the highest probability of Dunbrack’s backbone-dependent rotamer library for valine [16]. Potassium ions were placed in the mutant structure in the same manner as in the WT structure.

A 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine bilayer was generated with dimensions of 150 Å along the x- and y-axes. Both the WT and mutant structures were embedded within this bilayer using the membrane builder module [17] of the CHARMM-GUI web server [18]. Cubic boxes were constructed to enclose the protein structures by extending 10 Å along the z-axis at each end. TIP3P water molecules were then added to fill the boxes, followed by Na⁺ and Cl^- ions to neutralize the systems. Protonation of five residues (E324, E776, D805, D923, and E951) was conducted following established protocols at pH 7.4 [19].

The MD simulations of the entire system were performed using AMBER20 with a CHARMM36m force field [20]. Initially, the systems were minimized using a 10 000-step steepest descent algorithm, followed by 40 000 steps of the conjugate gradient method. Subsequently, they were equilibrated for 15 ns with an NVT ensemble while gradually reducing the constraints, as recommended by the CHARMM-GUI web server. The production steps of 200 ns trajectories producing 2 000 snapshots were then carried out and subsequently analyzed using CPPTRAJ [21].

Conversion of human skin fibroblasts to iNeurons

Forearm skin punch biopsy was used to obtain fibroblasts cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 20% FBS, non-essential amino acids (all from Gibco, Grand Island, New York), sodium bicarbonate (Sigma-Aldrich, St. Louis, Missouri), and 1% (v/v) penicillin/streptomycin/fungizone (Cellgro, Manassas, Virginia). To minimize potential confounding factors from the genetic background, two healthy control fibroblast lines matched for sex, passage (passages 3–5), and similar age were utilized. Direct conversion of human fibroblasts into induced neurons (iNeurons) was achieved by lentiviral transduction of shRNAs against human PTBP1 and PTBP2 (Genenmed Inc.), according to our previously described protocol [22–24]. Based on these established methodologies, this specific conversion method is known to yield a highly pure population (>90%) of functional glutamatergic neurons [22–24]. Selection with 2 μg/ml puromycin was performed for 2 days. Selected cells were replaced for 3 days in N3 media (DMEM/F12, Gibco, Grand Island, New York) supplemented with 25 μg/ml insulin, 50 μg/ml apo-transferrin, 20 nM progesterone, 100 nM putrescine, and 30 nM sodium selenite (all Sigma-Aldrich, St. Louis, Missouri); 10 ng/ml bFGF (Gibco, Grand Island, New York), supplemented with BDNF, CNTF, GDNF, and NT3 (all PeproTech, Cranbury, New Jersey), as previously described [22–24]. From day 4 to the day of analysis, the cells were maintained in N3 medium supplemented with 2% FBS.

Immunocytochemistry and confocal microscopy

iNeurons underwent fixation with 4% paraformaldehyde, permeabilized with 0.3% Triton X-100, and blocked with 5% normal goat serum (VectorLabs). The cells were incubated with mouse monoclonal antibodies against ATP1A3 (Cat.No.AB182571, Abcam, Cambridge, United Kingdom), Alexa Fluor 488-conjugated mouse antibodies (Gibco, Grand Island, New York), and rhodamine phalloidin (Cat.No.R415, ThermoFisher Scientific, Carlsbad, California). Nuclei were stained with 4′,6-diamidino-2-phenylindole, and imaging was conducted using a Leica TCS SP8 confocal microscope (n = 3 independent experiments).

Immunoblot analysis of ATP1A3 levels in patient-derived iNeurons

For western blotting, iNeurons from patients carrying the ATP1A3-p.A813V mutants and healthy controls were harvested on day 20 of differentiation. Cells were lysed using the N-PER Neuronal Protein Extraction Reagent (Thermo Scientific, Carlsbad, California) and a protease inhibitor cocktail (Sigma-Aldrich, St. Louis, Missouri). Equal amounts of protein from each sample were separated using 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to a PVDF membrane (GE Healthcare). Membranes were blocked with 5% skim milk. The primary antibodies used were mouse monoclonal antibodies against ATP1A3 (Cat.No.AB182571, Abcam, Cambridge, United Kingdom) and rabbit polyclonal antibodies against GAPDH (Cat.No.SC32233, Santa Cruz Biotechnology, Dallas, Texas). The membranes were probed with horseradish peroxidase-conjugated secondary antibodies (Santa Cruz Biotechnology, Dallas, Texas) and developed using a West-Q Chemiluminescent Substrate Plus Kit (GenDEPOT, Barker, Texas).

ATPase assay

Lysates from patient-derived neurons carrying the ATP1A3-p.A813V mutation and from two healthy controls were harvested on day 20 of differentiation. Cells were lysed using the N-PER Neuronal Protein Extraction Reagent (Thermo Scientific, Carlsbad, California) and a protease inhibitor cocktail (Sigma-Aldrich, St. Louis, Missouri). The lysates were then incubated with 250 μM of ATP, 40 mM NaCl, 25 mM KCl, 3 mM MgCl₂, and 1 mM ethylene glycol tetra-acetic acid at 37 °C for 1 hour. ADP-Glo Kinase Assay (Cat.No.V6930, Promega, Madison, Wisconsin) detected a synthesized ADP by ATPase reaction, which measures total Na⁺/K⁺ ATPase activity. Luminescence signals were measured using an Infinite 200 PRO plate reader and the i-control software (Tecan Life Sciences, Mannedorf, Switzerland).

Zebrafish lines

WT, Tg(Olig2:dsRed) [25], Tg(Hb9:mCherry) [26], Tg(Huc:gal4vp16) [27], Tg(uas:egfp) [28], nacre (mitfab692) [29], and Tg(uas:GCaMP6s) [30] were used in this study. All experimental procedures were approved by the Korea University Institutional Animal Care & Use Committee (No. KOREA-2020–0206) and were performed in accordance with the animal experiment guidelines of the Korea National Veterinary Research and Quarantine Service.

Generation of knockout zebrafish

To generate atp1a3b knockout zebrafish, we injected one-cell stage embryos with a single guide RNA (target sequence: 5ʹ-ACCACCTACAACACCCGAGTGGG-3ʹ) and Cas9 protein (ToolGen, Seoul, South Korea; 50 pg). Genomic DNA was extracted from injected embryos, and genome editing was confirmed using a T7 endonuclease I assay [31] from PCR product using the following genotyping PCR primer (Forward: 5ʹ-GGTTTAACCAATGCAAAGGCC-3ʹ, Reverse: 5ʹ-CCACGGCAGACAGGACGATTC-3ʹ). The injected embryos were raised to adulthood (F0) and outcrossed with WT embryos to generate F1 zebrafish. F1 zebrafish carrying mutations were screened using the T7 endonuclease I assay, and short genomic DNA, including the target site, was cloned to identify the mutations. To visualize motor neurons and axons, F1 knockout zebrafish were crossed with Tg(Olig2:dsRed) zebrafish.

Immunohistochemistry and imaging

Zebrafish were fixed with 4% paraformaldehyde overnight at 4 °C. Fixed samples were embedded in 1.5% agar blocks containing 5% sucrose for freezing. Frozen blocks were sectioned into 10-μm slices using a cryostat microtome (Microm HM525, Thermo Fisher Scientific, Carlsbad, California). Mouse anti-HuC/D (Cat.No.A-21271, Thermo Fisher Scientific, Carlsbad, California), rabbit anti-caspase3 (Cat.No.C8487, Sigma-Aldrich, St. Louis, Missouri) and rabbit anti-phospho p44/42 MAP Kinase (Thr202/Tyr204) (pERK) (Cat.No.4377S, Cell signaling, Danvers, Massachusetts) primary antibodies and Alexa 488, 568-conjugated secondary antibody (1:500, Invitrogen, Carlsbad, California) were used. For neuromuscular junction analysis, zebrafish larvae were rinsed three times with PBS/0.02% Triton X-100 and incubated in 2% BSA with 5% sheep serum for 1 hour at 25 °C for blocking. After blocking, embryos were incubated with mouse anti-SV2 (Developmental Hybridoma Bank, Iowa City, IA, USA) and Alexa 647-conjugated α-bungarotoxin (Cat.No.B35450, ThermoFisher Scientific, Carlsbad, California, USA) to mark presynaptic terminals and postsynaptic terminals, respectively. Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was performed using an in situ Cell Death Detection Kit (Cat.No.12156792910, Roche, Mannheim, Germany) according to the manufacturer’s specifications. Fluorescent images were obtained using an A1si laser-scanning confocal microscope (Nikon, Tokyo, Japan) and analyzed using NIS-Elements AR Analysis 4.30 software (Nikon, Tokyo, Japan).

Behavior analysis

Startle responses were measured to analyze locomotor responses to mechanical stimuli. All the zebrafish larvae were placed individually in 48-well plates and allowed to acclimate for 1 hour. Mechanical tapping stimuli and video images were obtained using DanioVision (Noldus, Wageningen, Netherlands) and Ethovision XT12 (Noldus, Wageningen, Netherlands) programs. We measured the peak distance moved within 1 s of the tapping stimulus.

In situ hybridization

The zebrafish c-fos gene (NM_205569) was cloned from 5 days post-fertilization (DPF) cDNA into the pGEM-T easy vector (Promega, Madison, Wisconsin). The linearized vector was transcribed using a digoxigenin-labeled mixture (Roche, Mannheim, Germany). Whole-mount in situ RNA hybridization was performed following previously established protocols [32]. Images were captured using an Ni-U microscope (Nikon, Tokyo, Japan).

GCaMP6s imaging and analysis

To compare the neural activity between neurons expressing the WT form of ATP1A3 and variant ATP1A3 (p.A813V), we transcribed using the mMESSAGE mMACHINE SP6 kit (Cat.No.AM1340, Invitrogen, Carlsbad, California). One microgram of each messenger RNA (mRNA) was injected into one-cell nacre Tg(Huc:Gal4;uas:GCaMP6s);nacre embryos. For immobilized larvae, the caudal fin was slightly cut and incubated with 100 μM D-tubocurarine chloride hydrate (Cat.No.T2379, Sigma-Aldrich, St. Louis, Missouri), which blocked the neuromuscular junction for 10 minutes. Then, larvae were mounted laterally on 2% low-melting agarose in a confocal imaging dish (VWR, Radnor, Pennsylvania, USA). Time-lapses of GCaMP6s fluorescence were acquired under 100-ms exposure for 2 minutes using a spinning disk confocal microscope (CSU-X1, Nikon, Tokyo, Japan). Individual neurons were designated for each region of interest using the NIS-Elements AR Analysis 4.30 software (Nikon, Tokyo, Japan). ΔF/F0 of each neuron was calculated as (F-F0)/F0, in which F0 is the baseline fluorescence level averaged over 2 seconds. The number of GCaMP transients were counted as single transients exceeding 40% ΔF/F0. The rise and decay times were measured as the time between the baseline and the peak and between the peak and the decay back to the baseline.

Quantification and statistical analysis

All data are presented as mean ± standard deviation. One-way analysis of variance (ANOVA) was used to analyze ATPase activity in the two controls and patient-derived iNeurons. Graphs and statistical analyses were performed using Prism 9 software (GraphPad Software Inc., La Jolla, CA). The two groups were compared using unpaired t-tests when data were normally distributed and the Mann–Whitney U test when data were not normally distributed. To compare more than two groups, ANOVA was used in combination with Dunn’s (non-parametric) multiple comparison test (Kruskal–Wallis test for unpaired data). Statistical significance was set at p < 0.05. Statistical analyses were performed using Prism 9.0 (GraphPad).

Results

Clinical characteristics and structural changes of the ATP1A3-p.A813V mutation

A 25-year-old female presented with dystonic posture, spastic gait, gaze instability, and dysarthria, leading to a diagnosis of RDP. Trio-based whole-exome sequencing identified a de novo heterozygous missense variant (Chr19:42480315C > T, GRCh38; NM_152296.5: c.2438C > T, p.A813V) in ATP1A3 gene (Fig. 1a, b), resulting in an A813V substitution within exon 18, a known disease hotspot [33]. According to the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) guidelines [34], the variant was classified as likely pathogenic based on the following criteria: its de novo occurrence with confirmed maternity and paternity (PS2); its absence from the large-scale population database gnomAD v4.1.0 (PM2); and multiple lines of computational evidence (including PolyPhen-2, SIFT, MutationTaster, MetaSVM, MetaLR, and AlphaMissense) supporting a deleterious effect (PP3). Given its location at the C-terminus of transmembrane (TM) helix 6, near the intracellular surface of the bilayer, we sought to determine how this mutation affects ATP1A3 structure and function.

Fig. 1.

Family pedigree and genetic analysis and structural changes of the ATP1A3-p.A813V mutation. (a) Genetic lineage of a rapid-onset dystonia-parkinsonism (RDP) patient family. The inheritance pattern, showing the heterozygous mutation carrier alongside unaffected parents. (b) DNA sequence electropherograms (father, mother, proband) revealing wild-type sequence of the parents and the de novo mutation c.2438C>T, A813V (p.Ala813Val) in the proband. (c-d) wild-type (WT) ATP1A3 structure illustrating stable interactions of A813 with C333 (TM4) and N761 (TM5), which preserves TM helix packing. (e-f) ATP1A3-p.A813V mutant structure (red) exhibits increased Cα distances between V813 and interacting residues (A813V-C333: 4.34 Å → 6.59 Å; A813V-N761: 4.76 Å → 6.66 Å) compared with the wild type (blue), suggesting destabilization due to the bulkier valine residue. (g) Root-mean-square distance (RMSD) analysis shows that while the WT remains stable, the mutant deviates after 100 ns, indicating progressive structural instability. (h) Superimposition of the initial structure (light gray color) and the representative structure of the most populated cluster of A813V mutant MD trajectory. ATP1A3, ATP1B1, FXYD, and V813 are colored pink, light blue, light green, and blue, respectively. Number of water molecules in the first shell of CBS (i) and TM6/9 pathway (j). Blue and red lines are WT and mutant, respectively

Molecular modeling reveals structural destabilization and functional impairment

To examine the structural and functional consequences of the missense mutation identified in the patient, we performed MD simulations. The mutation altered the packing of the TM helices. In the WT protein, TM helices remained closely clustered, with A813 forming stable interactions with C333 (TM4) and N761 (TM5) (Fig. 1c, d). However, in the mutant protein, the increased distances between A813V and these residues suggested an altered helical arrangement. The average Cα distances between A813V and C333 increased from 3.4 Å in the WT to 6.59 Å in the mutant (Fig. 1e), while the distance between A813V and N761 increased from 4.76 Å to 6.66 Å (Fig. 1f). The substitution of alanine with valine, a bulkier residue, likely disrupts compact TM packing, promoting conformational perturbations in the channel.

To further characterize these structural alterations, we analyzed the backbone root-mean-square distance (RMSD) relative to the initial structure during a 200-ns MD simulation (Fig. 1g). Although WT and mutant proteins initially displayed similar RMSD values, the mutant protein exhibited substantially greater deviation after 100 ns, indicating progressive structural drift. The largest cluster from the k-means clustering analysis (k = 10) of the mutant MD trajectories identified a major structural conformation in which TM6 unwinds slightly at its C-terminal end, leading to an outward shift in multiple helices, including ATP1B1 and FXYD6 (Fig. 1h). This helical rearrangement may significantly affect protein function by altering Na⁺/K⁺ ion transport.

Next, we assessed how these conformational changes influenced hydration around critical ion-binding regions, given that perturbed water dynamics can induce ion leakage and ATPase dysfunction. Using the “watershell” module in CPPTRAJ, we quantified water molecules within a 3.4-Å shell surrounding the cation-binding site (CBS), composed of residues from TM4, TM5, and TM6 during the MD simulations. The hydration of proteins, particularly CBS, has been demonstrated to facilitate ion leakage, consistent with MD studies of NavAB channels by Qiu et al. [35], which demonstrated that Na+ remains partially hydrated in sodium channels, whereas K+ adopts a nearly dehydrated state in potassium channels. The analysis showed that the mutant ATP1A3 exhibited decreased hydration in the CBS (Fig. 1i), suggesting a potential loss of ion coordination, which is essential for ATPase activity. As hydration at the CBS directly governs ion binding and transport, we also examined water accessibility in the TM6/9 water pathway implicated in Na⁺/K⁺ transport (Additional File 1), as identified by Cechova et al. [36]. Mutations in this pathway disrupt ATPase function by altering the water-mediated ion movement [37]. In ATP1A3-p.A813V, the water molecule density within this pathway was significantly reduced compared with that of the WT, particularly after 150 ns of simulation (Fig. 1j). This dehydration effect, coupled with increased helical distance, suggests that the mutation disrupts normal ATPase cycling, potentially leading to ion imbalance and neuronal dysfunction. Collectively, our findings highlight how the ATP1A3-p.A813V mutation alters transmembrane helix interactions, reduces protein stability, and impairs ion transport by modifying hydration dynamics [35–37].

ATP1A3-p.A813V prolongs calcium signaling and induces apoptosis via impaired Ca²⁺ clearance

To determine the functional consequences of ion transport defects arising from these structural perturbations, we analyzed calcium dynamics in zebrafish neurons overexpressing ATP1A3-WT or the p.A813V mutant by injecting mRNA into embryos of Tg(HuC:Gal4vp16;uas:GCaMP6s);nacre^-/- zebrafish, which express the calcium indicator GCaMP6s in neurons. Neuronal activity was quantified using fluorescence imaging of calcium levels (Fig. 2a). The average amplitude (△F/F), maximum amplitude (max△F/F), and number of GCaMP6s transients showed no significant differences between neurons overexpressing ATP1A3-WT and ATP1A3-p.A813V (Fig. 2b–d). However, a distinct finding emerged: the duration of GCaMP6s transients was notably prolonged in neurons overexpressing ATP1A3-p.A813V compared with those overexpressing ATP1A3-WT (Fig. 2e). Further analysis of the rise and decay times of GCaMP6s transients revealed comparable rise times between neurons overexpressing ATP1A3-p.A813V and ATP1A3-WT cells (Fig. 2f). In contrast, the decay times of GCaMP6s transients were significantly extended in neurons overexpressing ATP1A3-p.A813V compared with those overexpressing ATP1A3-WT (Fig. 2g). The prolonged decay time of GCaMP6s transients, without changes in rise time, indicates an extended duration of calcium signaling, which we interpret as a proxy for altered calcium homeostasis due to reduced ATP1A3 activity. Such prolonged calcium elevation is expected to promote neuronal hyperexcitability and subsequent cell death via sustained intracellular Ca^{2 +} accumulation. Consistent with this interpretation, we observed significantly increased caspase-3–positive cell death in HB9⁺ motor neurons expressing ATP1A3-p.A813V compared with those expressing ATP1A3-WT (Fig. 2h–j). Collectively, these results demonstrate that the p.A813V mutation induces structural alterations that compromise intracellular calcium homeostasis, leading to neuronal vulnerability and cell death.

Fig. 2.

ATP1A3-p.A813V mutation induces prolonged Ca²⁺ transients, leading to neuronal apoptosis. (a) Representative image of the spinal cord in Tg(Huc:Gal4;uas:GCaMP6s);nacre zebrafish at 3 days post-fertilization (DPF). Regions of interest (ROIs) designated 10 neurons per zebrafish. Scale bar: 50 μm. Individual traces depict GCaMP6s fluorescence signals from spinal neurons overexpressing ATP1A3 (blue) and those overexpressing ATP1A3-p.A813V (pink). (b–g) Quantification of various parameters of GCaMPs transients in neurons. (b) The average amplitude of GCaMPs transients measured in neurons. Mann–Whitney test; n = 18 and n = 16, respectively. (c) Quantification of the maximum amplitude of GCaMPs transients. Unpaired t-test; n = 18 and n = 16, respectively. (d) Quantification of the number of GCaMPs transients. Mann–Whitney test; n = 19, n = 18, and n = 16, respectively. (e) Quantification of duration of GCaMPs transients. Unpaired t-test; n = 18 and n = 16, respectively. (f) Quantification of the rise time of GCaMPs transients. Mann–Whitney test; n = 18 and n = 16, respectively. (g) Quantification of the decay time of GCaMPs transients. Mann–Whitney test; n = 18 and n = 16, respectively. (h, i) representative transverse section images of the spinal cord in zebrafish overexpressing ATP1A3 (WT) (h) and ATP1A3 (p.A813V) (i) at 5 DPF. Immunohistochemistry labeling with anti-active caspase3 antibody (green) in Tg(Hb9:mCherry) zebrafish revealing neuronal cell death. Scale bar: 10 μm. (j) Quantification of Caspase3+ Hb9+ neurons per section at 5 DPF. Mann–Whitney test. analysis conducted on n = 8 zebrafish per group. Statistical significance denoted as not significant (n.s.), *p < 0.05, **p < 0.01, ***p < 0.001

atp1a3b loss-of-function triggers neuronal hyperexcitability in vivo

To investigate the physiological consequences of ATP1A3 loss-of-function, we generated atp1a3b knockout zebrafish using the CRISPR/Cas9 system (Additional File 2). Neuronal hyperexcitability often precedes neuronal cell death in various neurodegenerative conditions, including amyotrophic lateral sclerosis [38–40]. Therefore, we hypothesized that ATP1A3 deficiency induces early hyperexcitability prior to overt neurodegeneration in atp1a3b knockout zebrafish. Neuronal activity was assessed at 2 DPF, a time point where apoptotic neuronal cell death was minimally evident. atp1a3b knockout larvae exhibited markedly elevated c-fos expression, a marker of neuronal activity, in the brain and spinal cord compared with WT counterparts (Fig. 3a–d). This robust activation was further substantiated by increased pERK expression observed specifically in Hu⁺ neurons, confirming hyperactivation (Fig. 3e, f, g). Additionally, ectopic c-fos expression was noted in the spinal muscles of atp1a3b knockout zebrafish, suggesting that neuronal hyperactivation may induce depolarization-mediated c-fos expression in the skeletal muscle (Fig. 3b, d) [41]. These findings suggest that ATP1A3 loss-of-function is linked to increased neuronal activity markers compatible with hyperexcitability prior to detectable neurodegeneration.

Fig. 3.

ATP1A3 loss-of-function elicits neuronal hyperexcitability. (a–d) expression of c-fos messenger RNA (a marker of neuronal activity) at 2 days post-fertilization (DPF) in atp1a3b^+/+ (a, c) and atp1a3b^-/- (b, d) zebrafish. Dorsal views of the brain (a, b) and lateral views of the spinal cord (c, d) are presented. Scale bar: 250 μm (a, b), 500 μm (c, d). (e, f) representative transverse section images of the spinal cord in 2 DPF atp1a3b^+/+ (e) and atp1a3b^-/- (f) zebrafish. Immunohistochemistry labeling with anti-Hu antibody (green) and anti-pERK (red) highlights neuronal populations and phosphorylated ERK. Scale bar: 10 μm. (g) Quantification of pERK+ Hu+ neurons per section. Unpaired t-test; n = 8

Neuronal hyperexcitability due to ATP1A3 loss-of-function leads to neurodegeneration and motor deficits

Following the observation of early neuronal hyperactivation (Fig. 3), we next examined whether this hyperexcitability directly contributes to neurodegeneration in atp1a3b knockout zebrafish. At 2 DPF, immunostaining with a neuronal anti-Hu antibody showed normal neuronal development in atp1a3b knockout zebrafish, similar to that in WT zebrafish (Fig. 4a, b, e). However, at 5 DPF, the neuronal population significantly declined and neuronal cell death increased in atp1a3b knockout zebrafish (Fig. 4c, d, f, g). Motor neurons displayed reduced axonal length alterations in atp1a3b knockout zebrafish, leading to motor axonopathy and altered neuromuscular structure (Fig. 4h, i). Quantification confirmed a significant reduction in Olig2⁺ motor axon length (Fig. 4l). Furthermore, the structure of the neuromuscular junction (NMJ) was impaired. Using staining for anti-SV2 (presynaptic) and α-bungarotoxin (postsynaptic), we observed defects in NMJ morphology (Fig. 4j, k). Quantification demonstrated a marked decrease in NMJ colocalization compared with WT controls (Fig. 4m). These structural defects translated directly into impaired motor function. Upon a tapping stimulus, atp1a3b knockout zebrafish exhibited a clear defect in their startle response (Fig. 4n, o). Quantification confirmed a significantly decreased locomotor response, evidenced by a reduction in the peak distance moved within 1 second of the stimulus (Fig. 4p). Together, these findings establish that ATP1A3 loss-of-function induces neuronal hyperexcitability, leading to progressive neurodegeneration and severe motor dysfunction.

Fig. 4.

Neuronal hyperexcitability drives neurodegeneration, motor axonopathy, and behavioral deficits. (a-d) Representative transverse section images of the spinal cord from atp1a3b+/+ (a, c) and atp1a3b-/- (b, d) zebrafish. Immunohistochemistry labeling with anti-Hu antibody (green) and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining (red) highlights neuronal populations and apoptotic cells. Scale bar: 10 μm. (e, f) Quantification of Hu+ neurons per section at 2 (e) and (f) 5 days post-fertilization (DPF). Unpaired t-tests were performed for statistical analysis. N = 8 at 2 DPF and n = 12 at 5 DPF. (g) Quantification of TUNEL+ cells per section at 5 DPF between atp1a3b+/+ and atp1a3b-/- zebrafish. Unpaired t-test, n = 8. (h, i) Representative lateral views of the spinal cord from Tg(Olig2:dsRed) atp1a3b+/+ (h) and atp1a3b-/- (i) at 5 DPF. Dsred fluorescence shows motor axons, neuromuscular structure, and motor neurons. Scale bar: 100 μm. (j, k) Representative lateral views of the spinal cord of atp1a3b+/+ (j) and atp1a3b-/- (k) zebrafish at 5 DPF. Immunohistochemistry labeling with anti-SV2 antibody (red) and α-bungarotoxin staining (blue) marking presynaptic and postsynaptic terminals. Scale bar: 50 μm. (l) Quantification of Olig2+ motor axon length at 5 DPF. Unpaired t-test; 32 axons from eight zebrafish per group. (m) Quantification of neuromuscular junction (NMJ) colocalization at 5 DPF. Mann–Whitney test; n = 12. (n, o) The representative images of startle behavior upon tapping stimulus of atp1a3b+/+ (n) and atp1a3b-/- (o). (p) Quantification of the peak distance moved. Mann–Whitney test; n = 19. All data are expressed as medians with interquartile ranges represented in violin plots. Statistical significance denoted as not significant (n.s.), **p < 0.01, ****p < 0.0001

ATP1A3-p.A813V mutation reduces protein abundance and impairs ATPase function in patient-derived iNeurons

To determine whether the structural alterations observed in MD simulations and the functional deficits in zebrafish models are recapitulated in a human context, we generated induced neurons (iNeurons) from skin fibroblasts of a patient carrying the p.A813V mutation and two healthy controls. Direct neuronal conversion was achieved using shRNAs targeting PTBP1 (PTB) and PTBP2 (nPTB), a protocol known to yield a highly pure population of glutamatergic neurons. After 20 days of differentiation, confocal microscopy showed significantly reduced ATP1A3 expression in patient-derived iNeurons compared with healthy controls (Fig. 5a). Western blot analysis corroborated this reduction, demonstrating a substantial loss of ATP1A3 protein in patient neurons (Fig. 5b). Additionally, patient-derived iNeurons exhibited significantly reduced Na⁺ /K⁺-ATPase activity compared with healthy controls (Fig. 5c), indicating that the p.A813V mutation disrupts Na⁺ /K⁺ pump function. While this assay measures global Na⁺ /K⁺-ATPase activity rather than an isoform-specific contribution, the observed reduction is highly consistent with the decreased expression of α3. The substantial loss of ATP1A3 protein observed in patient neurons suggests that the structural perturbations identified in our MD simulations may compromise protein stability, potentially leading to enhanced cellular degradation of the mutant protein. Collectively, these findings suggest that the ATP1A3-p.A813V mutation induces profound structural and functional deficits that compromise neuronal stability. While functional phenotypes such as calcium dysregulation were primarily modeled in vivo using zebrafish, the marked enzymatic deficit observed in iNeurons provides a human cellular basis for the proposed mechanistic cascade.

Fig. 5.

Reduced ATP1A3 and ATPase activity in patient-derived iNeurons expressing ATP1A3 (p.A813V). (a) Confocal images displaying reduced ATP1A3 expression (green) with phalloidin (red) in patient-derived induced neurons, fixed at day 20 post-differentiation from fibroblasts. Representative images from n = 3 experiments, each representing an independent differentiation. Scale bars: 25 µm. (b) Western blotting reveals reduced ATP1A3 protein levels in patient-derived iNeurons at day 20 compared with healthy controls. Data representative of n = 3 independent experiments. GAPDH serves as the loading control. (c) Quantification of Na⁺/K⁺-ATPase activity in two controls (CTRL1, CTRL2) and patient-derived iNeurons from three independent experiments, confirming significant functional impairment. Bars depict mean ± standard error of the mean. Statistical analysis was performed using one-way analysis of variance. *p < 0.025

Discussion

Our findings revealed that the p.A813V substitution distorts TM helix packing, compromising the Na⁺/K⁺-ATPase function through reduced hydration dynamics essential for ion transport. Critically, using a zebrafish atp1a3b knockout model, we established a key pathological sequence: ATP1A3 loss-of-function triggers aberrant neural activity and hyperexcitability that precede observable neuronal degeneration. This phenotype was mirrored in ATP1A3-p.A813V overexpressing neurons, which showed prolonged calcium transients indicative of repolarization failure. Our comprehensive analysis using patient-derived neurons and zebrafish models enhances our understanding of ATP1A3-related disorders and highlights potential therapeutic targets.

ATP1A3-related disorders encompass a wide spectrum of severe neurological phenotypes; however, the precise molecular mechanisms underlying these disorders remain insufficiently characterized. Previous research has focused on the clinical characterization of ATP1A3 variants through case reports, familial analyses, and meta-analytical studies, whereas investigations into their molecular and cellular pathophysiology remain limited. In the present study, we provide convergent structural, cellular, and organismal evidence demonstrating that the p.A813V variant induces profound functional impairment of the Na⁺ /K⁺-ATPase activity, driving neuronal hyperexcitability, calcium dysregulation, and neurodegeneration.

Clinically, ATP1A3 variants are associated with a spectrum of rare, often early-onset neurologic syndromes, with overlapping but heterogeneous presentation [1]. Our patient displayed both functional abnormalities, such as gaze instability and intermittent dystonia, and upper motor neuron signs, including spastic gait and hyperreflexia, potentially indicative of structural neurodegeneration. These clinical features are aligned with previously reported individuals harboring the same variant [33]. Notably, earlier reports classified c.2438C > T (p.A813V) as likely pathogenic based solely on in silico prediction tools. However, functional validation was lacking. By identifying an additional patient and experimentally characterizing the effects of the variant across molecular, cellular, and in vivo levels, our study provides definitive mechanistic confirmation of its pathogenicity.

Using MD simulations, we showed that the p.A813V substitution destabilizes transmembrane helix interactions, reduces structural compactness, and disrupts hydration at the cation-binding site—an essential determinant of ion selectivity and pump cycling. These structural perturbations predict impaired ion transport, consistent with reduced ATP1A3 protein abundance and Na⁺ /K⁺-ATPase activity observed in patient-derived iNeurons. Together, these findings suggest that the p.A813V variant is a loss-of-function mutation with significant impact on ATPase performance and potential implications for ionic homeostasis. Specifically, the structural instability and helix packing defects observed in our MD trajectories likely render the p.A813V mutant a substrate for the endoplasmic reticulum-associated degradation (ERAD) pathway, a quality control mechanism known to eliminate misfolded ATP1A3 variants [42–44]. This provides a plausible explanation for the significantly reduced ATP1A3 abundance in our iNeuron model. While the reduced activity in iNeurons is consistent with reduced ATP1A3 expression, we acknowledge that this assay lacks isoform-specific resolution, measuring global Na⁺ /K⁺-ATPase activity that encompasses both α₁ and α₃ contributions.

Nevertheless, a key insight from our work is the proposal of a mechanistic model linking ATP1A3 dysfunction to aberrant calcium handling. Disruption of this hydration, as observed for ATP1A3-p.A813V (Fig. 1), is predicted to compromise sodium extrusion. Intracellular Na⁺ accumulation is predicted to impair the sodium-calcium exchanger (NCX), reducing calcium efflux and perturbing calcium homeostasis [45]. This model is consistent with the prolonged calcium transient decay times in neurons expressing ATP1A3-p.A813V and the increased neural activity (c-fos and pERK) observed in atp1a3b knockout zebrafish, although Na⁺ and NCX activity were not directly measured in this study (Figs. 2 and 3). Notably, the decay phase of calcium transients may reflect the integration of multiple regulatory processes beyond Na⁺/Ca²⁺ exchange alone. Alterations in cytosolic buffering proteins, Ca^{2 +} sequestration in the endoplasmic reticulum via SERCA and release via IP₃R or ryanodine receptors, mitochondrial Ca^{2 +} uptake and extrusion, and plasma membrane Ca²⁺-ATPase activity can all shape calcium decay kinetics. Therefore, while ATP1A3 dysfunction is expected to impact calcium homeostasis, the prolonged decay observed in ATP1A3-p.A813V–expressing neurons should be interpreted as evidence of disturbed calcium regulation rather than as a uniquely defined defect in a single clearance pathway.

Recent studies further support a role for ATP1A3 in regulating neuronal calcium homeostasis. Fujii et al. [46] reported that iPSC-derived neurons carrying the ATP1A3 p.R756C variant exhibit altered calcium dynamics characterized by reduced calcium influx, indicating that ATP1A3 mutations can directly influence calcium entry pathways. Although their study focused primarily on impaired calcium influx, whereas our data highlight prolonged calcium decay kinetics in ATP1A3-p.A813V–expressing neurons, these observations are not mutually exclusive. Instead, together they suggest that ATP1A3 dysfunction perturbs neuronal calcium regulation at multiple levels, affecting both calcium entry and clearance in a variant- and context-dependent manner. The apparent differences between reduced calcium influx (p.R756C) and delayed calcium clearance (p.A813V) may reflect mutation-specific effects on Na⁺ /K⁺-ATPase structure, ionic coupling, and downstream signaling environments. Thus, rather than a single uniform mechanism, ATP1A3-associated disorders likely involve heterogeneous disruptions of calcium handling that converge on altered neuronal activity and vulnerability. Integrating our zebrafish and cellular findings with prior iPSC-based studies strengthens the view that ATP1A3 plays a central role in maintaining calcium homeostasis necessary for stable neuronal excitability.

Importantly, our zebrafish studies reveal that neuronal hyperexcitability precedes neurodegeneration. Elevated c-fos and pERK expressions at 2 DPF occurs at a developmental stage when neuronal loss is minimal, indicating that hyperexcitability is an early pathogenic event rather than a secondary consequence of neuronal degeneration. Motor axonopathy, NMJ disruption, and reduced locomotor performance emerge later, supporting a temporal model in which impaired ATP1A3 function initiates aberrant excitability that leads to neuronal and neuromuscular dysfunction. This is consistent with prior observations in atp1a3 knockout mice, where hyperexcitability and seizure activity precede degenerative changes [47]. Calcium signaling is central to numerous cellular processes, including neuronal excitability and muscle contraction, and its dysregulation has been implicated in movement disorders, such as dystonia and ataxia [48]. Intriguingly, we also observed ectopic c-fos induction in skeletal muscle of atp1a3b mutants. Although atp1a3b expression in zebrafish is predominantly neuronal, this ectopic muscular c-fos activation may arise secondary excitation resulting from hyperactive motor circuits, consistent with reports that depolarization-triggered calcium influx induces c-fos expression in muscle cells [41]. These cellular alterations recapitulate the clinical features in our patient, reinforcing the mechanistic association among ATP1A3 dysfunction, calcium imbalance, and neurodegeneration.

Currently, treatment options for ATP1A3-related disorders are limited to symptomatic management. The mechanistic framework established here highlights calcium overload and hyperexcitability as central contributors to disease progression, suggesting that interventions targeting calcium handling, NCX function, or excitability modulators may offer therapeutic benefit. A prior case study described symptomatic improvement in a 13-year-old patient with RDP treated with flunarizine, a selective calcium entry blocker [49], supporting the concept that calcium-regulating therapies may mitigate disease severity. Our results provide experimental justification for pursuing such targeted strategies.

Nevertheless, this study has certain limitations. Although the patient carries a heterozygous ATP1A3 variant, our zebrafish experiments employed both atp1a3b knockout and variant overexpression approaches, which might not fully reproduce the allelic dosage of the human condition. Furthermore, the absence of gene-edited isogenic control lines in our iNeuron model means that the influence of background genetic modifiers remains to be fully elucidated. Future research utilizing heterozygous knock-in human embryonic stem cell (hESC) models will likely be necessary to further refine dosage-dependent mechanisms and the specific impact of the patient’s genetic background. Additionally, while we were not able to directly characterize the neuronal subtype composition in the present study, this necessitates a cautious interpretation of cellular phenotypes as ATP1A3 expression and function may vary across cell types. To complement these human cellular findings and provide a more comprehensive physiological context, we utilized the zebrafish system as a parallel model. This approach allows for the evaluation of the functional consequences of the mutation within an intact, physiologically complex nervous system, thereby addressing the inherent limitations of isolated in vitro models. The high degree of consistency observed between the molecular deficits in human iNeurons and the functional impairments in our genetically controlled zebrafish model provides supportive, rather than direct, convergent evidence that the p.A813V variant is a potential primary driver of the observed hyperexcitability cascade. Furthermore, neuronal hyperactivity observed in this study was inferred from activity-dependent markers and calcium dynamics rather than direct electrophysiological measurements; thus, the proposed links between ATP1A3 dysfunction, ionic and calcium dysregulation, and increased neuronal activity should be interpreted as a mechanistic model consistent with our data and prior literature. Future electrophysiological and ion-transport-focused studies will be required to define these mechanisms more precisely.

Conclusions

Our multi-level investigation provides a mechanistic framework for understanding the pathogenicity of the ATP1A3-p.A813V variant. By integrating structural simulations with vertebrate and human cellular models, we propose a pathological cascade wherein impaired ATPase function leads to calcium-driven hyperexcitability and neurodegeneration, offering potential insights for future targeted therapeutic interventions.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Abbreviations

RDP

Rapid-onset dystonia-parkinsonism

MD

Molecular dynamics

WT

Wild type

iNeurons

Patient-derived induced neurons

CSF

Cerebrospinal fluid

DMEM

Dulbecco’s Modified Eagle Medium

pERK

phospho-p44/42 MAP Kinase

TUNEL

Terminal deoxynucleotidyl transferase dUTP nick end labeling

DPF

Days post-fertilization

mRNA

messenger RNA

ANOVA

Analysis of variance

TM

Transmembrane

RMSD

Root-mean-square distance

CBS

Cation-binding site

NMJ

Neuromuscular junction

NCX

sodium–calcium exchanger

Author contributions

Conceptualization: S. M. Lim, S. Kim, H.-C. Park, S. H. Kim. Data curation: S. M. Lim, S. Kim, J. Park, Y. E. Kim, and W.-H. Shin. Clinical profiling: J. Park and K. W. Oh, S. H. Kim. Whole-exome and Sanger sequencing: Y. E. Kim and C. S. Ki. Modeling and simulations: W.-H. Shin. Cell culture, enzyme activity, and immunolabeling: S. M. Lim and O. C. Na. Zebrafish modeling, behavior & GCaMP6s imaging: S. Kim. Confocal imaging: S. M. Lim and S. Kim. RT-qPCR: M.-Y. Noh, M. Nahm. Funding: S. Kim and H. C. Park, S. H. Kim. Writing: S. M. Lim, S. Kim, J. Park, Y. E. Kim, and W. -H. Shin, H.-C. Park, S. H. Kim. All authors contributed to and reviewed the final manuscript.

Funding

This research was supported by the K-Brain Project of the National Research Foundation, funded by the Korean government (MSIT) (RS-2023-00265515, RS-2022-NR071990, RS-2026-25472618, and RS-2026-25494588), and by a grant from the Korea Dementia Research Project through the Korea Dementia Research Center, funded by the Ministry of Health and Welfare and the Ministry of Science and ICT, Republic of Korea (RS-2024-00348451).

Data availability

Knockout zebrafish generated in this study will be made available on request, but we may require a payment and/or a completed materials transfer agreement if there is potential for commercial application. All data are contained within the manuscript. Raw data is included in the supplementary materials.

Declarations

Ethics approval and consent to participate

All experimental procedures were approved by the Korea University Institutional Animal Care & Use Committee (No. KOREA-2020–0206) and were performed in accordance with the animal experiment guidelines of the Korea National Veterinary Research and Quarantine Service.

Consent for publication

Informed consent was obtained from the patient and her parents for publication.

Competing interests

The authors declare that they have no competing interests.