ATP1A3 – this is what you need to know

ATP1A3. This is the Epilepsiome page for ATP1A3, the gene that encodes the alpha-3 subunit for Na+-K+ ATPase. This ATPase is mainly present in neurons and is critical for the maintenance of the electrochemical gradient, which in turn allows for action potential propagation and proper functioning of voltage-gated ion channels. The phenotypic spectrum of ATP1A3-related disorders is very broad, involving multiple types of neurological disorder including stroke-like syndromes, epilepsy, neurodevelopmental disorders, and movement disorders, and ranging in severity, as will be described below.

Here are the most recent blog posts that mention ATP1A3:

In a nutshell. The Na+-K+ ATPases utilize adenosine triphosphate (ATP), manufactured by our cells’ mitochondria, to actively transport Na+ out of the cell and K+ into the cell. This constant cycling is essential to the maintenance of the intra/extracellular electrochemical gradient, which in turn underlies critical cellular functions in every tissue in our bodies. Expressed primarily in the brain and the heart, ATP1A3 codes for the α3 subunit of the Na+-K+ ATPase. In the brain, the electrochemical gradient is critical for neuronal cell signaling and for maintenance of electrochemical stability, enabling cell excitation and action potential propagation. Both ATP1A3 and one of its counterparts, ATP1A2, are expressed in neurons during embryonal brain development, and ATP1A3 is also thought to contribute to regulation of non-ionic neuronal transporters and receptors. However, whereas ATP1A2 is primarily expressed in glial cells postnatally and into adulthood, ATP1A3 continues to be expressed primarily in neurons (PMID: 34161264).

Phenotypes | Genetics | Mechanism | Community

Figure 1. The gray figure represents an axon terminal and demonstrates the function of ATP1A3, which uses ATP to shuttle Na+ out of and K+ into the axon terminal, thereby maintaining an electrochemical gradient that allows the propagation of action potentials and ultimately the influx of Ca++ to result in neurotransmitter release into the synapse. The purple figure represents an astrocyte, with another Na+/K+ ATPase (encoded by ATP1A2) embedded in the membrane. This channel also plays a role in maintaining the electrochemical gradient, as well as serving other downstream functions. Figure created with


Summary. Pathogenic variants in ATP1A3 cause a spectrum of neurological phenotypes, described previously as rapid-onset dystonia-parkinsonism (RDP), alternating hemiplegia of childhood (AHC), cerebellar ataxia, areflexia, pes cavus, optic atrophy, and sensorineural hearing loss (CAPOS). While it is helpful to consider ATP1A3-related conditions through the lens of these broad classifications, it is likely that most individuals with ATP1A3-related conditions exist along a spectrum of these conditions. Furthermore, there have been hints of variable expressivity within families – both in our own practice and in the literature, and there have been affected children who have inherited the variant from reportedly asymptomatic parents (PMID: 33868146). We are continuing to learn about the functional effect of the different ATP1A3 variants that have been discovered, as well as the genotype-phenotype correlations. Below, we describe the various phenotypes that have been described in association with disease-causing ATP1A3 variants, keeping in mind that many patients may have overlap between these general categories. Of note, further phenotypic categories have been described, though given substantial overlap between phenotypes, it is more useful to consider ATP1A3-related disorders within a smaller subset of disease categories, keeping in mind that certain individuals may exhibit symptoms related to different phenotypic subsets. Overall, ATP1A3-related disorders seem to be characterized by paroxysmal episodes of neurologic dysfunction, sometimes triggered by illness, at other times by innocuous triggers or no clear trigger at all, and sometimes accompanied by progressive symptomatology related to ataxia or movement disorder. This is not true for all variants, however, where some have no evident paroxysmal feature at all.

Alternating hemiplegia of childhood. Individuals with alternating hemiplegia of childhood (AHC) typically present in infancy or early childhood, usually before the age of 18 months. The classic AHC phenotype has been described as episodic neurologic symptoms that may include hemiplegia, dystonia, seizures, and eye movement abnormalities, affecting either side of the body at different times, and resolving with sleep – though symptoms may return as soon as 10-20 minutes after waking (PMID: 20301294). These episodes may occur abruptly, without any forewarning, and may be triggered by excitement, illness, temperature changes, or lack a trigger altogether. The term “alternating” comes from the tendency for the affected side to alternate; in some cases, the hemiplegia may affect one side and then the other within the same episode (PMID: 35945798). Abnormal eye movements including nystagmus or intermittent esotropia/exotropia have also been reported (PMID: 35945798). The natural history of AHC has been examined by Bourgeous and colleagues (PMID: 8496742). However, there is still much to be learned – these early studies suggest that while paroxysmal episodes of hemiplegia may improve with age, the movement disorder is persistent, and those with epilepsy may not outgrow their seizure disorder (PMID: 35945798). Many individuals with AHC may have an associated neurodevelopmental or neuropsychological disorder including cognitive impairment, developmental delay, or intellectual disability (PMID: 35945798). While little is known regarding genotype-phenotype correlation, there have been postulations that AHC phenotypes tend to cluster before amino acid position 400 or after amino acid position 700, and the majority of AHC-causing variants are transmembrane (PMID: 33868146, 24739246). AHC-causing variants are thought to be primarily gain-of-function.

Epilepsy. Up to 50% of patients with AHC may have epilepsy. On the other hand, there have been reports of children with ATP1A3 variants whose phenotype is not entirely consistent with AHC, but who do have early life epilepsy and episodic apnea, where the latter may be ictal or non-ictal (PMID: 25656163, 33880529). On the most severe end of the spectrum are neonates with early infantile epileptic encephalopathy or epilepsy of infancy with migrating focal seizures. Similar to AHC, the individuals with more severe epilepsy have variants either before 400 or after 700 (PMID: 33880529).

Dystonia and other movement disorders: Dystonia in ATP1A3 typically involves one limb or one hemibody but may be generalized. Distinguishing features include rostro-caudal pattern, prominent bulbar symptoms and upper extremity involvement, and long duration; the episodes may also be painful (PMID: 32443735, 32280259). There has been a report of 4 children with ATP1A3 variants presenting with paroxysmal dystonia with onset in childhood, without episodes of alternating hemiplegia. These cases were phenotypically variable, where three of the four described cases had progressive movement disorders associated with developmental delay, and one of the four was otherwise neurotypical between episodes (PMID: 31061839). There have also been rare cases reported of recurrent ataxia, particularly associated with febrile illness, with much rarer cases of progressive ataxia, along the spectrum of the CAPOS phenotype (PMID: 33868146). These disease phenotypes are seen in domains spread throughout the gene.

Cerebellar ataxia, areflexia, pes cavus, optic atrophy, and sensorineural hearing loss (CAPOS). This clinical syndrome is characterized by paroxysmal, recurrent episodes of ataxia with symptom onset between 1 and 5 years, with additional progressive optic atrophy and sensorineural hearing loss, which may be abrupt in onset then progressive. Episodes are typically triggered by fever and are characterized by encephalopathy, hypotonia, weakness, abnormal eye movements, and ataxia (PMID: 35945798, 33868146). There also may be association with interictal slowly progressive ataxia. Seizures are not associated with the pure subtype of this syndrome.

Structural brain abnormalities: Presumably because of its role in early brain development, ATP1A3 has also been seen in association with structural brain abnormalities, including polymicrogyria (PMID: 33880529) and hydrocephalus (PMID: 33868146). Per Smith and colleagues (PMID: 34161264), variants associated with polymicrogyria cluster between transmembrane regions 7 and 8.

Neurodevelopmental phenotype: Some individuals with ATP1A3-related conditions may also have associated developmental delays, intellectual disability, or autism spectrum disorder. On the most severe end of the spectrum, there have been individuals with early onset epileptic encephalopathy reported, as described above (PMID: 33880529). Recently, Calame and colleagues (PMID: 37043503) have reported on a recurrent variant (Pro775Leu) associated with spasticity and intellectual disability, where some of the reported individuals did not have any paroxysmal symptoms.

Cardiac phenotype. Importantly, ATP1A3 is also highly expressed in cardiac cells. A recent publication by Balestrini and colleagues (PMID: 32913013) showed that EKG abnormalities were seen in over 50% of individuals with the AHC phenotype, with dynamic ECG changes noted in some of these patients; of their cohort of 98 patients with AHC, 3 (3%) required cardiac intervention (one of whom did not carry the ATP1A3 variant). For this reason and echoing the conclusions of this study, we recommend screening cardiac evaluation for all patients with ATP1A3-related conditions; if patients have had syncopal episodes, we recommend implantation of a loop recording device (PMID: 32913013).


Genotype versus phenotype

Many of the conditions reported in association with ATP1A3 variants do not fit any previously reported pattern or broad phenotypic category; furthermore, there may be a wide phenotypic spectrum even in individuals with the same variant (PMID: 34161264). This variability, without a clear or predictable pattern emerging, again emphasizes the broad spectrum of clinical features associated with this gene. The broad phenotypic spectrum on its own does not make ATP1A3 unique – as we have seen, the more we discover about genetics, the more we are forced to acknowledge the wide spectrum of disease associated with the majority of the genes. However, unlike some of its compatriots in the channelopathies, including CACNA1A, it is challenging even to draw conclusions on clinical phenotype based on the domain inhabited by the ATP1A3 variant. Different studies have shown, however, that individuals with an AHC phenotype tend to have variants either below amino acid position 400 or above 700, and variants tend to inhabit transmembrane domains; other studies have shown a clustering of AHC phenotypes in transmembrane domains 2-9. In their study on ATP1A3-related DEE and structural abnormalities, Vetro and colleagues found enrichment of variants in transmembrane domains M3-M10, clustering around ion binding sites.

It is likely that these different clinical symptoms are related to different degrees of alteration of function of the Na+-K+ ATPase, including dominant negative functional effects in the most severe cases, as well as downstream effects of these functional changes.

Mutational spectrum. The majority of reported variants in ATP1A3 have been missense variants, though there have been nonsense, frameshift, and del/ins variants reported as well. The abundance of missense variants lends credence to the idea that the mechanism of disease is unlikely to be a simple haploinsufficiency model, but rather relates to alterations in function of the channel, leading ultimately to an overall loss or gain of function (see below).

Recurrent variants. There have been several recurrent missense variants reported, including the following, which have been seen in association with AHC: p.Asp801Asn, p.Glu815Lys and p.Gly947Arg (PMID: 32280259). The p.Glu818Lys has been recurrent in individuals with CAPOS syndrome (PMID: 33868146, 24468074).

Segregation. There have been reports of affected patients having inherited variants from seemingly unaffected parents, though the majority of severe cases of ATP1A3-related disorders appear to be de novo.


The precise function of the Na+-K+ ATPase relies on the presence of ATP, extracellular K+, and the ability of the channel to undergo complex conformational changes which allows for the binding and release of Na+ and K+.  As in other channelopathy genes (SCN1A, CACNA1A), the functional consequences of different disease-causing variants in ATP1A3 may be broadly categorized into two subtypes: loss-of-function, where the mechanism of disease is typically haploinsufficiency, and gain-of-function, where the mechanism of disease is related to pathological hyperactivity of the channel or a dominant negative functional effect. While it is useful to consider functional changes as a binary loss versus gain, this does not truly capture the diversity of functional effects induced by different variants. Furthermore, the functional effects in question may be variable in themselves, depending on brain region or cell type.

Loss of function. While protein-truncating variants clearly lead to a loss-of-function effect via haploinsufficiency, there have been a number of missense variants reported to have loss-of-function effect, due to a more complicated process. For example, a variant may result in loss of function due to reduced trafficking of the channel to the cell membrane; however, if the malformed channel does not make it to its preordained destination, it may clog up intracellular machinery, thereby having a toxic effect in addition to simple haploinsufficiency. This is the presumed functional effect of the Leu924Pro variant (PMID: 34161264). Other loss-of-function effects include more simple effects such as reduced α3 expression or increased endoplasmic reticulum retention of the protein, while still others may have no effect on the quantity of channels or trafficking, but may alter the function of the channel in such a way as to reduce its function (i.e. changing Na+ or K+ affinity, impaired conformational changes making active transport more challenging, change in voltage dependence, et cetera (PMID: 34161264). Other variants may disrupt the motif critical for ATP binding (PMID: 33880529).

Gain of function.  In our review of the literature, the majority of reported gain-of-function effects have been in the form of a dominant negative effect. All of the variants examined by Vetro and colleagues, which were associated with DEE and/or severe structural brain abnormalities, were likely dominant negative, including D801N, E815K, and G947R. The D801N variant is known to impair transport of Na+ and K+ directly as D801 is a binding site for Na+ and K+ (PMID: 33880529). A study by Paciorkowski and colleagues (2015) similarly found that two patients with severe phenotypes, one with severe early onset epilepsy and the other with epilepsy and recurrent apneas, had loss-of-function variants in ATP1A3 (Gly358Val and Ile363Asn), though again, it is likely that these effects were not true loss-of-function but rather dominant negative effects (PMID: 25656163). The dominant negative functional effect may be due to competition of the wild-type and the variant for binding to the b subunit of the Na+-K+ ATPase (as you recall, ATP1A3 encodes the α3 subunit); this might occur if the abnormal protein retains binding ability to the b subunit (or perhaps has increased affinity for the b subunit). It has been proposed that some of the variation in phenotype may be due to differential levels of competition, where in less severe cases, the wildtype wins out (PMID: 31425744). Arystarkhova and colleagues further propose a dominant negative/cytotoxic mechanism wherein the accumulation of misfolded proteins may trigger neuronal cell death.  On the other hand, one recent publication, by Calame and colleagues (PMID: 37043503), explores the recurrent variant Pro775Leu and proposes a cation leak – in the opposite direction of active transport – as the principal mechanism of disease. This would be characterized as a gain-of-function because it represents a clear alteration of function from typical, even though the movement of cations (Na+ and H+ in this case) is actually in reverse to active transport.

The Clinical Perspective

My patient has a mutation in ATP1A3 – what does this mean?

Assessing ATP1A3 variants is difficult in many cases. This gene is included in many gene panels and clinicians may be faced with the problem to interpret milder symptoms in transmitting carriers correctly. Here are three criteria that may help you interpret ATP1A3 variants in your patient.

1 – Variant. Both PTVs and missense variants have been described in association with ATP1A3-disorders. Though there have been some examples of variable penetrance, we can assume that all PTVs are pathogenic. Regarding missense variants, it is helpful to search ClinGen or GeneMatcher to see if a variant has been previously reported as pathogenic or likely pathogenic. If this is not helpful, you may look at the structure of the protein itself in order to determine whether the variant is located in a highly-conserved domain such as a binding site or transmembrane region, using structure to make inferences on pathogenicity.

2 – Segregation. ATP1A3-associated disorders are inherited in a dominant fashion and segregation data (testing parents and families) may help figure out whether a variant is pathogenic. Severe, early onset childhood phenotypes in otherwise unaffected families may be de novo. If the patient has a severe phenotype and the variant is found to be inherited, it is unlikely that the variant is explanatory of the patient’s phenotype. Alternatively, the variant can be inherited from a parent with a milder phenotype, and there have been infrequent cases of inherited variants where the parent appears to be unaffected.

3 – Phenotype. The phenotypic spectrum of ATP1A3 is large. The question may arise as to which phenotype would be considered incompatible with ATP1A3 as the underlying gene. As yet, we would be reticent to disregard a pathogenic variant in ATP1A3 based on phenotype alone, as we have only been adding to the complexity of described phenotypes over time.


While flunarizine has been touted as a preventive medication for AHC, there is no consistent evidence to demonstrate its efficacy. Other options include calcium-channel blockers (again without proven efficacy) or acetazolamide. For prevention of seizures, there are no particular medications that stand apart, though given possible improvement of AHC with acetazolamide, we might recommend trialing topiramate (for its similar action as a carbonic anhydrase inhibitor, as well as additional mechanisms for seizure prevention) and ketogenic diet, among other options.


Community advocacy:

Research advocacy: