Episodic Ataxias: Primary and Secondary Etiologies, Treatment, and Classification Approaches

Background: Episodic ataxia (EA), characterized by recurrent attacks of cerebellar dysfunction, is the manifestation of a group of rare autosomal dominant inherited disorders. EA1 and EA2 are most frequently encountered, caused by mutations in KCNA1 and CACNA1A. EA3–8 are reported in rare families. Advances in genetic testing have broadened the KCNA1 and CACNA1A phenotypes, and detected EA as an unusual presentation of several other genetic disorders. Additionally, there are various secondary causes of EA and mimicking disorders. Together, these can pose diagnostic challenges for neurologists. Methods: A systematic literature review was performed in October 2022 for ‘episodic ataxia’ and ‘paroxysmal ataxia’, restricted to publications in the last 10 years to focus on recent clinical advances. Clinical, genetic, and treatment characteristics were summarized. Results: EA1 and EA2 phenotypes have further broadened. In particular, EA2 may be accompanied by other paroxysmal disorders of childhood with chronic neuropsychiatric features. New treatments for EA2 include dalfampridine and fampridine, in addition to 4-aminopyridine and acetazolamide. There are recent proposals for EA9–10. EA may also be caused by gene mutations associated with chronic ataxias (SCA-14, SCA-27, SCA-42, AOA2, CAPOS), epilepsy syndromes (KCNA2, SCN2A, PRRT2), GLUT-1, mitochondrial disorders (PDHA1, PDHX, ACO2), metabolic disorders (Maple syrup urine disease, Hartnup disease, type I citrullinemia, thiamine and biotin metabolism defects), and others. Secondary causes of EA are more commonly encountered than primary EA (vascular, inflammatory, toxic-metabolic). EA can be misdiagnosed as migraine, peripheral vestibular disorders, anxiety, and functional symptoms. Primary and secondary EA are frequently treatable which should prompt a search for the cause. Discussion: EA may be overlooked or misdiagnosed for a variety of reasons, including phenotype-genotype variability and clinical overlap between primary and secondary causes. EA is highly treatable, so it is important to consider in the differential diagnosis of paroxysmal disorders. Classical EA1 and EA2 phenotypes prompt single gene test and treatment pathways. For atypical phenotypes, next generation genetic testing can aid diagnosis and guide treatment. Updated classification systems for EA are discussed which may assist diagnosis and management.


INTRODUCTION
The term 'episodic ataxia' originally refers to a small group of rare autosomal dominant inherited disorders [1]. These are characterized by discrete attacks of cerebellar dysfunction (ataxia) of variable duration and frequency, often accompanied by other ictal and interictal symptoms. The incidence is likely less than 1 per 100,000, but may be underestimated due to restricted genetic testing and unidentified genes [2]. The group comprises eight subtypes (EA [1][2][3][4][5][6][7][8]. EA1 and EA2 are the most common subtypes, caused by mutations in KCNA1 and CACNA1A respectively. They are classical neurological channelopathy disorders. They have welldefined phenotypes and are reported in multiple families of different ethnicity. In contrast, EA 3-8 are reported in rare families. Genes for EA5 (CACNB4), EA6 (SLC1A3) and EA8 (UBR4) have been identified, while causative genes are inconclusive for EA3, EA4 and EA7 (either mapped to chromosomal location or unknown) [3].
With increased availability of genetic testing, the phenotypic spectrum of EA1 and EA2 has broadened. Moreover, an increasing number of reports have surfaced of EA as a rare manifestation of other genetic disorders (e.g. epilepsies, paroxysmal dyskinesias, metabolic disorders) [4]. There are also secondary causes of EA that may be encountered, most commonly vascular, multiple sclerosis, or inflammatory disorders. These may be suggested by onset after adolescence, negative family history, greater attack variability, and accompanied by abnormal laboratory and imaging findings [4]. However some clinical features can overlap with primary EAs. There are also a variety of EA mimickers much more commonly encountered, such as migraine or vestibular disorders. Hence there are increasing diagnostic challenges for physicians encountering patients with EA.
This review will focus on providing an update for neurologists and movement disorders specialists regarding clinical and genetic classifications of EA, and a diagnostic and management approach.

METHODS
A systematic literature search of PubMed was performed in October 2022 using the search term 'episodic ataxia' (656 articles) and 'paroxysmal ataxia' (535 articles). (Figure 1). Restricting the search to manuscripts published within the last 10 years, English language, human subjects, and removal of duplications, yielded 330 articles. After screening titles and abstracts, nonrelevant articles were excluded. Of these, 157 reviews, case reports, case series, and literature reviews including systematic reviews were evaluated. Additional articles identified from bibliographic review of screened articles (35 additional articles) resulted in a total of 192 articles reviewed. Of these, 154 were referenced in this review. The author undertook a descriptive analysis, where episodic ataxia was discussed according to the subtype, clinical description, genetics, and treatment modalities.

PRIMARY EA
These still formally comprise 8 subtypes, with recent proposals for EA9 and EA10. (Table 1) Amongst EA1-8, there are 5 known genes, and at least 8 loci. All identified genes, except UBR4 (EA8), encode ion channel proteins, and are important in excitatory neurotransmission [2]. Both EA1 and EA2 are wellestablished classical channelopathies, with numerous cases/families reported, a known gene comprising numerous mutations, and a narrow classical phenotype. Both EA1 and EA2 have broadened their phenotypes considerably in recent years. However, no fixed genotypephenotype relationship is identified, and there can be marked clinical variability between family members with the same mutation. Amongst EA3-8, additional cases of EA6 (SLC1A3) and EA8 (UBR4) were recently found. However, EA3, EA4, EA5 and EA7 remain elusive with no additional cases identified in the past 15-20 years since these were first reported, despite the increased availability of genetic testing. EA 3-8 generally resemble EA1 or EA2 with a few clinical differences, including variability with age of onset (infancy to late adulthood), attack duration (seconds to days), and associated ictal and interictal symptoms.

Clinical
The classical description, first defined in 1975, is brief attacks of ataxia and vertigo [5]. Constant myokymia of the face or limb muscles, detected clinically or via electromyography, affects almost all patients [6]. Onset is typically in childhood, on average at age 7.8 years [6,7]. Attack triggers are numerous; most commonly physiological stressors (exercise, emotional stress, environmental heat, fever, menstruation), caffeine, or alcohol. Sudden movement (kinesogenic trigger), startle, and spontaneous onset are also reported [6,8]. Attacks typically last seconds to minutes, but can last hours or days [5,6,9]. The attack frequency can range from multiple daily attacks to monthly attacks [6,8]. During the attack, gait impairment may range from mild dysfunction to complete inability to walk [6].
Genetics EA1 is mostly familial, although de novo mutations occur [10]. The KCNA1 gene, discovered in 1994, encodes the fast voltagegated potassium channel Kv1.1, and mutations result in a potassium channelopathy [3,11]. Kv1.1 is a critical regulator of neuronal excitability in the central and peripheral nervous system, reflecting the neurologic manifestations of EA1. Each Kv1.1 channel is composed of four αsubunits forming a functional transmembrane pore [3]. Each αsubunit has six transmembrane spanning segments (helices S1-S6) and intracellular N and C terminal domains. Helices S1-S4 form the voltagesensing domain, with S4-S5 helical linker to the channel pore. The S5-S6 segments forms the pore region which allows ion flux. RNA editing of KCNA1 transcripts is important to control protein function. The channels have a low threshold for activation, and a hyperpolarizing effect on membrane potential which limits neuronal excitability. Dysfunction of the channel leads  to excessive excitability and increased duration of action potentials, which may cause excessive neurotransmitter release. Kv1.1. is abundant in the brain, mainly in cerebellum, hippocampus, neocortex and peripheral nerves [10]. In EA1, Kv1.1 dysfunction is thought to cause hyperexcitability of cerebellar interneurons, resulting in excessive inhibition of Purkinje cells, which then reduces cerebellar inhibitory output, with subsequent cerebellar deficits [3,5]. To date, 63 KCNA1 pathogenic mutations are reported on OMIM and most are missense mutations. Functional studies have correlated mutations with Kv1.1 lossoffunction by various mechanisms [3]. Some mutations have a dominant negative effect, meaning that the mutated αsubunit adversely affects the other subunits in the tetrameric structure of the K channel. Other mutations affect Kv1.1 expression or function in other ways [3].

Expanded description
Diplopia, dysarthria, nausea, or headache may accompany attacks. Neuromyotonia of variable severity is common, suggested by muscular stiffening, painful contractures, muscle cramps, twitching, or muscle hypertrophy [2]. Other neuromuscular features may include cataplexy, dystonia, distal weakness, and malignant hyperthermia [5,6,9,[11][12][13]. Sweating, hot flushes, palpitations, paroxysmal dyspnea, or sensory symptoms are rare features [6,9,14]. Most patients have normal cerebellar function between attacks and normal MRI brain imaging. However longer disease duration is correlated with permanent cerebellar signs and cerebellar atrophy [6]. Epilepsy is more common in EA1 patients than the general population implicating KCNA1 as a cause of epilepsy [15,16]. There may be comorbid cognitive disability or deafness [8]. Quality of life is impaired, with mental health the worst affected domain [6].
No single phenotypegenotype correlation reported. The same KCNA1 mutation can show marked clinical variability within the same family or even in twins. This suggests other genetic modifiers or environmental factors influence clinical severity [10,17]. There may be other genes responsible for EA1, as phenocopies (i.e. KCNA1negative cases of EA1) have been identified. The KCNAnegative phenocopies have male predominance and longer attacks versus KCNA1 positive cases [6]. Over half of KCNA1 variants result in EA1, either with or without epilepsy. Other KCNA1 variants occur without EA, and instead present with epilepsy, epileptic encephalopathy, hypomagnesemia, muscle cramps, myokymia, cataplexy, dystonia or paroxysmal kinesogenic dyskinesia [3,10,18]. In an analysis of 47 pathogenic KCNA1 mutations, EA1 associated variants occur along the whole length of the protein, whereas epilepsyrelated variants tend to cluster in the S1/S2 transmembrane domains and pore region of Kv1.1 [3]. Research into small molecules that selectively open Kv1.1 channels may permit a future treatment strategy for treating KCNA1 [19]. In a rat model of focal neocortical epilepsy gene, Kv1.1 overexpression was effective in controlling seizures, although this has not yet been studied for EA1 [20].

Management
The diagnosis is based on clinical findings, electrophysiology studies, and genetic confirmation of KCNA1 mutation. Many patients do not seek treatment because attacks are brief and improve with age [6]. A variety of antiseizure medications can diminish attacks, including carbamazepine, phenytoin, and lamotrigine [1,5,6]. Carbamazepine improved the severity of myokymia, in addition to ataxia, in a patient with a novel KCNA1 mutation [11]. Acetazolamide and benzodiazepines are helpful in rare cases [10,21]. However medication response is highly variable, and severe drug resistant phenotypes are encountered [10].

EA2
Clinical This is the most common hereditary episodic ataxia. The classical description, first published in 1946, is intermittent spells of ataxia and dysarthria lasting several hours, possibly up to 2-3 days [2,22]. There is interictal nystagmus between attacks, a useful clinical clue. This may be primary position downbeat, gazeevoked or rebound nystagmus [2,22,23]. The attack triggers include emotional or physiological stress, exercise, alcohol and caffeine [22]. Onset in childhood is most common but it can occur in the sixth decade [2,24]. The dysarthriaataxia spells may be isolated, or accompanied by other brainstem symptoms (vertigo, diplopia, tinnitus) or other features (migraine, abdominal pain, seizures, dystonia, cognitive impairment) [22,23,25,26]. While initially episodic, some patients develop a progressive cerebellar ataxia syndrome and cerebellar midline atrophy, similar to EA1 [27]. Migraine is reported in up to 50% of cases, and may be hemiplegic migraine [2,22].

Genetics
CACNA1A was identified as the genetic cause in 1996, and causes a calcium channelopathy [28]. CACNA1A encodes the alpha 1A subunit of the P/Qtype voltagegated calcium channel (Cav2.1). The P/Q channel is expressed throughout the CNS, most densely in cerebellar Purkinje cells and granule layer neurons. It is mainly found on presynaptic terminals and is important for synaptic transmission. CACNA1A gene mutations for EA2 have high but incomplete penetrance at 80-90%. Over 100 pathologic mutations are reported to date on OMIM, typically nonsense or frameshift mutations leading to a premature stop in protein transcription. These result in loss of P/Q channel function in the cerebellum. Similar to EA1, the mutation may exert a dominant negative effect [29].
In addition to EA2, CACNA1A mutations can result in two other autosomal dominant disorders: familial hemiplegic migraine type 1 (FHM1) and spinocerebellar ataxia type 6 (SCA6). While EA2 is associated with lossoffunction mutations, FHM1 is caused by gainoffunction mutations of the alpha subunit of the Cav 2.1 channel, while SCA6 is caused by a polyglutamine repeat expansion in the alpha subunit [1]. Clinical overlap between EA2, FHM1 and SCA6 is reported. Most FHM1 patients have cerebellar signs and symptoms, while half of EA2 patients have migraine, and SCA6 cases can present with fluctuating ataxia at onset before the progressive cerebellar dysfunction evolves [1]. Acetazolamide can reduce or completely abolish attacks, and is a hallmark of the disease. It is a carbonic anhydrase inhibitor but the mode of action is not well understood for EA [58]. About 50-75% patients report improvement with 250 mg to 1000 mg daily. However side effects of nephrolithiasis, paresthesia, and fatigue may limit tolerance [58]. The potassium channel blocker 4aminopyridine (4AP) is also effective, reducing the number of attacks and improving quality of life in an RCT [59]. Dalfampridine, a slow release formulation of 4AP, is also effective for EA2 [60]. In a recent headtohead trial, both 4AP 5 mg TID and fampridine (a newer slow release version of 4AP) 20 mg daily significantly reduced the number of attacks in patients with EA2 and related disorders, in comparison to placebo [61]. Fampridine had fewer side effects than acetazolamide. The combined use of topiramate and 4AP was effective in a patient with EA2 with migraine that was refractory to acetazolamide [62]. A mouse model of EA2 suggests early treatment with 4AP may be neuroprotective for secondary progressive ataxia [63]. Levetiracetam is also reported to be beneficial in EA2 [64,65]. Brief naps alleviated attacks in one case of EA2 with migraine, but the attack persisted if they remained awake, suggesting a sleepneuromodulation effect [66].

EA3: (gene unknown)
This was reported in a single large Canadian family in 2001. It resembles EA1, with short attacks with vertigo and tinnitus and interictal myokymia. It may respond to acetazolamide [67]. It is distinguished by EA1 and EA2 by vertigo and tinnitus accompanying the attacks, absent interictal nystagmus, and shorter attacks. Linkage studies excluded KCNA1 and CACNA1A as a cause, and mapped the gene 1q42 with a high LOD score, but only after adapting linkage parameters [68]. Some have questioned the reliability of this finding.

EA4: (gene unknown)
This is also termed 'periodic vestibulocerebellar ataxia' and 'North Carolina autosomal dominant ataxia'. It is reported in 2 families from North Carolina, suggesting a common single founder. It is characterized by ataxia, vertigo, episodic impaired smooth pursuit, gazeevoked nystagmus, and diplopia [69]. The onset is between 30-60 years, and symptoms worsen over time. It resembles EA2 but without interictal nystagmus, and it does not respond to acetazolamide. Gabapentin may relieve vertigo symptoms in EA4 [70]. Linkage studies in 1996 excluded autosomal dominant ataxias with known chromosomal localization at that time, including KCNA1, CACNA1A, SCAs 1-5 and DRPLA [71]. Autopsy findings in a 91year old EA4 patient showed polyglutamine repeats in Purkinje and granule cells, without intranuclear inclusions, similar to SCA6 brains [72]. This was of interest as SCA6 can present as a fluctuating ataxia.

EA5: (CACNB4)
This was reported in a single FrenchCanadian family 20 years ago, with a mutation in CACNB4 gene, coding for the beta4 auxiliary subunit of voltagegated calcium channels (Cav2.1) [73]. It is late onset and responds to acetazolamide. There were attacks of vertigo and ataxia lasting for several hours, although 1 family member had a single attack lasting for weeks. Interictal examination revealed spontaneous downbeat and gazeevoked nystagmus, mild dysarthria and truncal ataxia. However no additional cases have subsequently been identified, despite frequent screening of CACNB4 mutation in EA patients [74,75]. Meanwhile, the same mutation was reported in a family with epilepsy [20]. Other CACNB4 mutations have been associated with epilepsies [76]. Some researchers have questioned whether there is sufficient data to support EA5 [7].

EA6: (SLC1A3)
EA6 has been reported in several Caucasian and Korean families to date, and associated with SLC1A3 mutations in all cases [77][78][79][80][81][82]. The phenotype resembles EA2 with long duration attacks, interictal nystagmus, similar triggers, and acetazolamide response [79]. Migraine, alternating hemiplegia, progressive ataxia and epilepsy may cooccur. Childhood and adult onset is reported, and it appears to have reduced penetrance [74]. The SLC1A3 gene codes for EAAT1 (excitatory amino acid transporter), a glial Na+dependent glutamate transporter and ion channel [1,81]. Functional studies suggest SLC1A3 mutations impair EAAT1 by altering transport function in various ways via reduced or enhanced glutamate uptake and/or anion currents [78,83]. There are differing clinical phenotypes, according to the glutamate reuptake capability [7,74,80]. A SLC1A3 mutation underlying EA6 has also been reported in a family with adultonset progressive ataxia [81]. SCL1A3 mutations have also been found in migraine, ADHD, autism and Tourette syndrome. In a patient with familial migraine, the mutation impaired K+ binding to the EAAT1 channel and completely disrupted glutamate transport [84].

EA7: (unknown gene)
This was reported in 7 members of a 4generation family in 2007. It is similar to EA2 but without interictal nystagmus. This was mapped to 19q13 with a LOD score slightly above significance cutoff. Sequencing of 2 candidate genes in this region (KCNC3, SLC17A7) did not identify a mutation [85].

EA8: (UBR4)
This was first reported in 2016 in an Irish 3generation family [86]. This presented by age 2, much earlier than EAs 1-7. There are episodic attacks with impaired balance, dysarthria, and generalized weakness. Attacks can be triggered by physical fatigue or stress. Interictal examination can show intention tremor, eyelid myokymia, and impaired tandem gait. Attacks vary in duration from minutes to hours, and frequency ranges from daily to every few months. Migraine with aura may cooccur. Attacks respond to clonazepam and are not improved with acetazolamide [86]. The gene has been mapped to a large region on 1p36.13-p34.3 with a LOD score near to cutoff. Exome sequencing revealed variants in 2 genes, SPG2 and UBR4. UBR4 had a greater likelihood of pathogenicity than SPG2, as UBR4 is ubiquitin ligase protein that interacts with calmodulin and may potentially disrupt calcium sensor in neurons as hypothesis for ataxia [86]. A Korean study reported 2 patients with mutations in both UBR4 and CACNA1A, and suggested UBR4 may act as a genetic modifier with synergic effects on abnormal CACNA1A activity [74]. No functional analysis studies have been performed as yet for UBR4.

EPISODIC ATAXIAS ASSOCIATED WITH OTHER GENETIC DISORDERS
There are a growing number of genetic disorders that can present with EA either alone or embedded in a complex syndrome. These include chronic ataxia disorders (SCA-14, Some of these might explain prior classical familial EA cases that were negative for EA1 and EA2 and genetic loci of other EAs [1,6,87]. Three of them (SCA27/FGF14, SCA42/CACNA1G, and SCN2A) have been proposed to be categorized as EA9 or EA10.

SCA-14 (PRKCG)
SCA14 is a dominantly inherited slowly progressive ataxia, sometimes accompanied by parkinsonism, dystonia, myoclonus and cognitive impairment. It is caused by mutations in PRKCG gene encoding protein kinase C gamma (PRKγ). PRKCG mutations may also present with adultonset episodic ataxia, with a frequency of 1/14 PRKCGpositive patients [88].

SCA-27 (FGF14)
SCA27 is a lateonset progressive ataxia with parkin sonism, postural tremor and titubation; 20% have coexistent episodic ataxia [89]. It is caused by mutations in FGF14 gene which encodes Fibroblast Growth Factor 14. This protein is highly expressed in the brain, especially Purkinje cells, where it interacts with voltagegated Na+ channels to regulate neuronal excitability [89]. Isolated EA is also reported to be caused by heterozygous FGF14 gene mutations [75,[89][90][91][92]. Onset age ranges widely from early childhood to adulthood. Attacks may be accompanied by vertigo, dizziness, unsteadiness, with interictal nystagmus and tremor. Attacks are highly variable, lasting seconds up to several days. There are a variety of triggers; a fever trigger with a prolonged attack in a young child can mimic febrile cerebellitis. Attacks may respond to acetazolamide, and may improve with age. Developmental delay and paroxysmal dyskinesia have also been observed [90]. Some authors suggested designating this EA9.

SCA-42/epilepsy (CACNA1G)
CACNA1G encodes the poreforming α1G subunit of Ttype voltage gated calcium channel (VGCC). Mutations in CACNA1G cause generalized absence epilepsy and SCA42. A single family is reported with episodic vestibulocerebellar ataxia associated with a mutation in the CACNA1G gene [93]. There were attacks of dizziness, unsteadiness, headache and facial numbness, and headmovement induced vertigo. Attacks lasted up to several months in duration. Interictal examination showed cerebellar findings and bilateral vestibulopathy. The attack duration and absence of myokymia or tinnitus distinguishes this from EA3. Attacks were worsened by acetazolamide, and suppressed by carbamazepine. The authors proposed this be designated EA10.

AOA2 (SETX)
Mutations in SETX (senataxin) account for two separate clinical syndromes. Oculomotor apraxia type 2 (AOA2), an autosomal recessive spinocerebellar ataxia, with adolescent or early adult onset progressive ataxia, oculomotor apraxia, neuropathy, cerebellar atrophy, and elevated alpha fetoprotein levels. Autosomal dominant juvenileonset motor neuron disease (ALS4) is also characterized. There is a single case report of a 4 year old boy presenting with isolated severe EA attacks lasting 20-30 minutes, and intermittent mild impaired tandem gait between attacks [94]. Genetic testing excluded EAs 1,2,5,6. Whole exome sequencing (WES) identified a heterozygous deletion in SETX gene, possibly explaining the milder phenotype compared to homozygous mutations in AOA2.

GLUT-1 (SLC2A1)
The SLC2A1 gene encodes glucose transporter protein type 1 (GLUT1) which facilitates glucose transport across the bloodbrain barrier, and is critical for brain energy. The GLUT1 deficiency syndrome is a result of inadequate brain glucose transport. The main phenotype is a severe chronic neurologic disorder (microcephaly, developmental delay, early infantile seizures, ataxia) [98]. About 10% do not have this phenotype, and instead have milder paroxysmal variants often provoked by fasting or exercise, such as EA or paroxysmal exerciseinduced dyskinesia (PED) [99]. Amongst 25 SLC2A1 carriers, 1 had EA [100]. The GLUT 1 spectrum disorder has grown to encompass other paroxysmal dyskinesias (PKD, PNKD), myotonia, migraine, hemiplegic migraine and episodic eye movements [100,101]. EA may be pure or with additional neurological findings [99,102,103]. A diagnostic test to assess for GLUT1 deficiency is spinal tap showing a low CSF to serum glucose ratio (hypoglycorrhachia). Treatment for GLUT1 deficiency is avoidance of triggers, or ketogenic diet to provide an alternative energy substrate for brain energy metabolism [98,104,105]. However EA attacks may also respond to acetazolamide, which may lead to misdiagnosis as EA2 [99,102].

Epilepsy spectrum disorder (KCNA2)
A spectrum of neurological disorders may be caused by mutations in the KCNA2 gene, which encodes voltagegated potassium channel Kv1.2. Early onset developmental and epileptic encephalopathy, intellectual disability, and ataxia are recognized. A milder phenotype of episodic ataxia, epilepsy, and complicated hereditary spastic paraplegia is reported [106,107].

Epilepsy spectrum disorder (SCN2A)
Mutations in SCN2A are associated with a spectrum of neurological disorders from benign to severe epilepsies, autism spectrum disorder and intellectual disability [108,109]. SCN2A encodes the alpha subunit of voltage gated neuronal Nav1.2 channel. Lossoffunction mutations result in severe epilepsy, intellectual disability and autism, whereas gainoffunction mutations cause benign familial neonate infantile seizures (BFNIS) with or without EA [110]. Patients with BFNIS have seizures before age 3 months which resolve in early life, and may later develop EA between age 10 months to age 14 years [1,108,111]. Cognitive outcome is mostly favorable in these cases [108]. Amongst cases with a more severe intellectual disability phenotype, EA appears to be uncommon [112]. Phenotypes may vary in family members, e.g. EA in infant and episodic hemiplegia in parent [113]. EA may be triggered by vaccinations, minor head trauma, and sleep deprivation [114,115]. Some but not all have a favorable response to acetazolamide [108]. Seizures, but not EA attacks, respond to Nachannel blockers (phenytoin, carbamazepine) suggesting different pathophysiologic mechanisms. Some authors have designated this EA9 [116].

PKD/epilepsy (PRRT2)
PRRT2 (prolinerich transmembrane protein 2) mutations are responsible for a spectrum of paroxysmal neurological disorders. The 3 main phenotypes are paroxysmal kinesogenic dyskinesia (PKD), benign familial infantile convulsions (BFIC) and infantile convulsions and choreoathosis (ICCA) [117][118][119]. Other phenotypes include migraine, FHM and epilepsy [100]. EA appears to be a rare manifestation of PPRT2 mutations. In 1 large study of 374 PRRT2positive patients, episodic ataxia was only occasionally reported [120]. In another study of 182 EA patients, only one case was attributed to a PRRT2 gene mutation [121]. MRI brain imaging performed in a PRRT2 patient during an EA attack showed cerebellar diffusion restriction [122]. Most PRRT2 mutations respond exquisitely to carbamazepine [123] so EA attacks due to PRRT2 may cause diagnostic confusion with EA1. The PRRT2 protein interacts with SNAP25 and may play a role in synaptic transmission. Most mutations are lossoffunction, which may result in disrupted synaptic transmission and neuronal hyperexcitability. This does not explain phenotype variation, and no specific phenotype genotype correlation is identified [124]. A handful of cases are reported with homozygous PRRT2 mutations and a severe phenotype with episodic ataxia, intellectual disability, and infantile seizures [125][126][127].

Epilepsy spectrum/DOORS (TBC1D24)
Mutations in the TBC1D24 presynaptic protein are associated with a neurological spectrum of epilepsy, chronic encephalopathy, DOORS (deafness, onychodystrophy, osteodystrophy, mental retardation and seizures), hearing loss, and myoclonus. Biallelic mutations of TBC1D24 were found in an infant with EA and myoclonus, with a later finding of cerebellar atrophy in adolescence [128].

Mitochondrial disorders
EA cases are reported in mitochondrial disorders, such as pyruvate dehydrogenase complex deficiency, (PDHx, PDHA1), TPK1, DARS2, MTATP6, ACO2 genes [4]. EA can be isolated or occur with other neurological abnormalities. Diagnostic clues are the presence of serum and CSF lactic acidosis. A mild presentation of fevertriggered EA attacks lasting 2 to 7 days, and a normal interictal exam, was observed in a young child with PDH deficiency [129]. Tests showed elevated serum and CSF lactate, and MRI brain showed dentate nucleus hyperintensity. Attacks responded to thiamine, levocarnitine, and alphalipoic acid. In comparison, homozygous or compound heterozygous mutations in ACO2 (encodes mitochondrial aconitase 2 that catalyzes citrate to isocitrate) can cause a spectrum of disorders with often severe neurologic impairment. A recent report of 2 siblings with ACO2 mutations had EA plus mild developmental delay and neuropsychiatric symptoms [130].

Unknown genes
A "lateonset EA" was reported in 2009 of 4 cases in a single 2generation family but the gene is not known [87]. Onset was in the fifth or sixth decade. Phenotype severity was variable, with more severe cases exhibiting daily attacks with slowly progressive ataxia and poor acetazolamide response. Screening excluded KCNA1 (EA1), CACNA1A, (EA2), and locus for EA2, EA5, EA6 and EA7.

SECONDARY (ACQUIRED) EPISODIC ATAXIA
There is a broad differential diagnosis for acuteonset recurring ataxia [131]. (Table 2). Secondary or acquired EA may resemble primary EA with regards to onsetage, attack variability, and interictal cerebellar findings, but are more likely to have abnormal laboratory and MRI imaging [132]. Many secondary causes are treatable, and collectively more common that primary EAs, so they are important to consider.
The most common secondary disorders are transient ischemic attacks or stroke, multiple sclerosis or other immunemediated disorders [133]. "Paroxysmal dysarthria and ataxia" (PDA) is a wellrecognized phenomenon in multiple sclerosis, with stereotyped multiple daily episodes of sudden ataxia lasting seconds to minutes, attributed to ephaptic transmission [134]. This PDA syndrome can mimic genetic EA, and is also reported in immunemediated diseases such as antiphospholipid syndrome, Bickerstaff's/ Bickerstafflike encephalitis, certain autoimmune ataxias, and ischemic stroke [133,135,136]. This has been attributed to a lesion in the midbrain, near or in the red nucleus [134]. Other inflammatory disorders (postinfectious cerebellitis, MillerFisher syndrome) and vascular disorders (Behcet's disease with brainstem and red nuclei involvement, Kawasaki disease) can present with prolonged attacks of acute ataxia. Structural lesions in the posterior fossa or cerebellum such as a tumor or occult neuroblastoma can present with recurrent ataxia.
Epileptic pseudoataxia may transiently occur after a seizure. Hypothyroidism can present with recurrent ataxic episodes, and responds to thyroxine. Toxins (e.g. alcohol, antiseizure medications, lead) can present with reversible acute ataxia. Metabolic disorders (e.g. maple syrup urine disease, pyruvate dehydrogenase deficiency, ornithine transcarbamylase deficiency, biotinidase deficiency, Hartnup disease, argininosuccinic aciduria, citrullinemia, thiamine pyrophosphate deficiency) causing EA usually present in childhood with severe neurologic symptoms, but may present in adults with much milder features [4]. Thiamine pyrophosphate deficiency has been reported in a small number of patients with EA, delayed development and dystonia, and may respond to thiamine supplementation [137]. Maple syrup urine disease may also have significant clinical or biochemical improvement with thiamine supplementation [137]. Citrullinemia is a rare recessive urea cycle disorder due to mutations in the ASS1 (type I citrullinemia) gene which cause deficiency of arginosuccinate synthetase enzyme, necessary for catalyzing the formation of arginosuccinic acid from citrulline and aspartic acid. A typical presentation is a neonate with toxic hyperammonemia and progressive encephalopathy. Mild lateonset childhood or adultonset forms with intermittent symptoms (ataxia, headache, stroke, intellectual disability, or encephalopathy) are reported. A case of citrullinemia presented in late childhood with brief EA attacks with fever, a normal interictal neurological exam, cerebellar atrophy, and elevated citrulline and ammonia blood levels [138].
Autoimmune ataxias are usually chronic, but three types to date may manifest with EA. CASPR2 (VGKC complex) can present with episodic ataxia and dysarthria,  seizures and cognitive dysfunction. MRI brain may be normal or show medial temporal hyperintensity, with elevated CSF protein and positive CASPR2IgG in serum and CSF [139][140][141]. It responds to immunotherapy. Anti NMDA receptor autoimmunity can present with paroxysmal dysarthriaataxia syndrome [142]. AntiHu (ANNA1) associated paraneoplastic limbic encephalitis presented in a child as episodic ataxia and progressive behavioral changes evolving to intractable epilepsy [143]. Iatrogenic intermittent ataxia may be provoked during deep brain stimulation programming [144]. Finally, functional ataxia may be suggested by incongruent examination findings, distractibility, and the presence of other functional signs.

DIFFERENTIAL DIAGNOSIS FOR EPISODIC ATAXIA
EA can be mimicked by other paroxysmal disorders with stereotyped attacks of central or peripheral origin (e.g. vestibular migraine, migraine with brainstem aura, seizures, paroxysmal dyskinesias or benign paroxysmal positional vertigo). Patients with EA have been misdiagnosed with migraine, seizures, functional or anxiety disorders, resulting in premature diagnostic closure [22, 55, 57]. A personal or family history of epilepsy or migraine may have suggested these alternative more common diagnoses, rather than EA. It may also be difficult to distinguish chronic ataxia with stepwise exacerbations or stepwise decline (e.g. SCA 6), from EA with persistent cerebellar dysfunction. Suspicion for EA should be heightened with acetazolamide responsive attacks.

APPROACH TO EPISODIC MOVEMENT DISORDERS IN THE CLINIC: (SEE TABLE 3)
Clinical assessment EA can be readily misdiagnosed or overlooked. In order to recognize, it is important to routinely include it in the differential diagnosis of spells, whether these are movement or nonmovement based. A detailed history should include: onset age, triggers, duration, frequency, aura, baseline between spells, and response to treatment trials. Patient descriptors may pose challenges (e.g. episodic stiffening due to EA1 versus PKD attack, or EA with episodic cognitive impairment due to EA2 versus seizure). Events during early childhood development (e.g. infantile paroxysmal torticollis, episodic oculomotor dysfunction, BFNIS) may provide diagnostic clues for EA2. The interictal examination can offer clues to primary EAs when findings are present (e.g. nystagmus EA2, myokymia EA1). Ictal examination or a video of the attack can help reconstruct the phenomenology. A 3generation family history is important to look for other paroxysmal neurologic disorders because of considerable phenotypic variability in families. Family history may appear negative with de novo mutations, false paternity, early death, or estrangement from biological family, deceptively pointing away from a genetic cause. Secondary EAs or mimicking conditions can be suggested by history, examination, and imaging findings. While functional features may suggest a nongenetic EA, functional embellishment of primary EA may lead to diagnostic uncertainty regarding the predominant etiology

Treatment
For cases that resemble EA1 or EA2, one may proceed directly to firstline treatment (antiseizure medication or acetazolamide, respectively), without requiring genetic confirmation [145]. If unsuccessful, consider second line treatment trials. Ultimately, genetic testing is gold standard for the diagnosis to guide appropriate treatment and longterm management. Many of the primary EAs, other genetic causes of EA, and secondary forms of EA are treatable. (See Table 4).  [116]. This suggests a shared pathophysiological basis, and advances into the underpinnings of EA may translate into better understanding for these other paroxysmal disorders.

DISCUSSION
The phenotype is often not accurately predicted by the underlying genotype. Within the same family there can be large variability in attack frequency, disease severity, and treatment response, despite the same genotype. It is presumed that the phenotype must therefore be modulated by environmental factors, modifier genes, or agedependent expression [146]. This seems plausible as these are episodic (not fixed) disorders, and environmental modifiers are already illustrated by the presence of attack triggers. A study suggested UBR4 and SLC1A3 may act as genetic modifiers with a synergistic effect on CACNA1A mutation [74]. Plausibly, these modifier genes could be developed as a therapeutic target or a new precision therapy [116]. For example, several genes have identified in EA1 mice that modify the epilepsy phenotype in EA1 mice, so potentially this could be adapted for EA treatment [3]. Moreover, the agedependent expression observed with some of these disorders, such as CACNA1A, may simply reflect properties of neurologic channelopathies, where different phenotypes can arise at different ages, and the adult phenotype may differ considerably from the childhood syndrome [147]. It is also transpiring that the infantile CACNA1A paroxysmal phenotypes are not so benign as their names would suggest, given their increasing association with chronic neuropsychiatric impairment.
Improved understanding of genotypephenotype relations using molecular and electrophysiological study in animal models and patients may result in better precision medicine. A machinelearning method was recently applied to 47 patients with 33 unique variants in CACNA1A (pathogenic or likely pathogenic) to predict LOF or GOF mutations [148]. The severity score was  significantly higher for GOF variants, S5/S6 helices variants and pVal1392 Met variant. This was interpreted as demonstrating broad disease severity in CACNA1A disease and that clinical phenotypes likely reflect diverse molecular phenotypes [148]. A recent study used gene interaction networks to investigate common gene signatures associated with paroxysmal phenotypes of ataxia, migraine, epilepsy and other movement disorders [149]. Nineteen candidate genes were used to create an interaction network, which further revealed 39 associated genes (including KCNA1, SCN2A, CACNA1A, and CACNB4).
The metaregression analysis showed the strongest association of SCN2A with genes in neurodevelopmental disorders, and KCNMA1 as a common gene signature with a link to epilepsy, movement disorders and wide paroxysmal neurologic presentations. Identifying gene interactions may help future drug targets [149].
Using advanced genetic testing may be the crucial step for undiagnosed hereditary EAs, although this can create its own challenges. Genetic testing may be restricted by methodology or techniques, such that the pathogenic gene was omitted, or mutations may not be adequately detected (e.g. repeat expansion, microdeletion). If mutations are detected, there may be additional challenges because of broad phenotype variability, poor genotypephenotype correlations, or a large number of VUS identified (e.g. CACNA1A) [120]. Frequently, no functional study of a mutated protein is performed so we cannot be certain of its pathogenicity [1]. This could be improved by accessible functional readouts, particularly for atypical cases or with cheaper or more readily available nextgeneration testing [40]. On a global scale, there may be underdiagnosis of genetic EAs in resourcepoor settings. Apart from EA1 and EA2, most other EA reports reflect Caucasian cases. A recent Korean study found genetic heterogeneity in 33/39 EA patients, when examining a range of suspected pathogenic mutations in CACNA1A, SLC1A3, UBR4, SCNA1, TTBK2, TGM6, FGF14 and KCND3 [74]. However more studies are needed to update global genetic differences of EA, similar to the SCAs and genetic parkinsonisms.
Prior to advances in genetic testing, all EAs were thought to be channelopathies. Genetic mutations in KCNA1, CACNA1A, CACNB4, SLC1A3, SCN8A, KCNMA1, and ATP1A3 genes that encode ion channels lend support to the channelopathy theory [150]. Moreover, the overlap of movement disorders, migraine and epilepsy is often described in channelopathy disorders [151]. However other EA genes do not encode ion channels, suggesting alternative mechanisms [152]. There is evidence to suggest that the presynaptic terminal is involved, as PRRT2 and SLC1A3 likely act on the presynaptic terminal, and both KCNA1 and CACNA1A are presynaptic [100]. KCNA1 and CACNA1A have the highest levels of expression in the cerebellum, and in frontal, temporal and occipital cortices, compared with GLUT1/SLC2A1, so the regional effect of vesicle release could explain phenotypic differences. This may be why KCNA1 and CACNA1A are more likely to present with ataxia than GLUT1. This regional effect may not explain their other phenotypes such as migraine. Instead, they might be attributed to the consequences of dysregulated presynaptic terminals. Mice models of migraine with single gene mutations have shown increased glutamatergic neurotransmission and cerebral hyperexcitability, which may reflect abnormal neurotransmitter release from the presynaptic terminal [100,153]. The current classification system and diagnostic algorithm for EA frequently designates EA1, EA2 and others. This seems too simplistic given the current number of genes identified. Moreover, clinical prediction for the underlying gene is unreliable, as even classical EA1 and EA2 phenotypes can be KCNA-1 and CACNA1A negative respectively. Empiric treatment may also result in misdiagnosis, e.g. acetazolamide responsive GLUT1. It is likely time to reconsider the nosology for EA.
The simplest solution might be to ascribe EA numbers to all the genes identified to date for EA. This would be similar to the SCAs, which currently number 50. The caveat is that some genes are more commonly associated with nonEA syndromes. Another suggestion is to classify EA by its mutation. This has been proposed for the PKDs e.g. PKD PRRT2 or PKDSCN8A [150]. This could be readily used for EA, e.g. EAPRRT2, EA SLC2A1, EASLC1A3, etc. However, a limitation is that this could only be used if a causative gene is identified. Instead, we could consider diagnostic algorithms proposed for the paroxysmal dyskinesias, another episodic disorder. One suggestion uses a 2axes system. Axis 1 is clinical classification by trigger to establish PKD, PNKD, or PED, and Axis 2 classification is the presence or absence of 4 causative genes (PRRT2, MR1, KCNMA1, and SLC2A1); if negative, further testing can be pursued [152]. However genetic EAs share multiple triggers, may not be reliably clinically distinguished by a single clinical feature alone, and have a wider genetic spectrum.
Therefore, turning to the dystonia classification system may provide a better model for EA classification. This also combines 2 axes: clinical characteristics and etiology, with the goal of helping guide diagnosis and treatment [154]. Adapting this model for EA, Axis 1 clinical characteristics could include age at onset, attack duration, simple (dysarthria ataxia) or complex attack, interictal exam (normal or abnormal), and other neurologic comorbidities (e.g. epilepsy, intellectual impairment) and Axis 2 etiology could include nervous system pathology, and whether inherited, acquired or unknown. (See Tables 5 and 6). An additional category to consider in Axis 1 etiology is empiric treatment response e.g. acetazolamide responsiveness in many EAs, FGF14, and ATP1A3 but absent in others. These combined clinical aspects may suggest EA syndromes to help guide genetic diagnosis and treatment. For example, childhoodonset short duration simple EA attacks, with a normal interictal exam and normal imaging could suggest KCNA-1 or PRRT2, whereas the presence of abnormal imaging could suggest PDHx, or interictal ataxia could suggest SETX. Adultonset long duration EA attacks may suggest CACNA1A, SLC1A3, or FGF14. Adultonset short duration simple EA attacks, with abnormal interictal exam and imaging could be MS or autoimmune ataxia. Future research into analysis of this proposal would be of interest to assess if it could improve clinical diagnosis and genetic prediction.
A final alternative strategy, also borrowing from PKDs, is to consider grouping EAs into categories by presumed pathogenic mechanism: i.e. channelopathies, neurotransmission syna ptopathies, brain energy transportopathies, to create a new classification system [150].
Most EAs are treatable or even curable, so it is important to correctly diagnose them. There are now four effective treatments for EA2 include longacting formulations of 4aminopyridine (dalfampridine and fampridine) in addition to acetazolamide and 4AP. Novel observations of sleep alleviated EA2 attacks may suggest innovative treatment modulators. Many genetic EAs and GLUT1 respond to acetazolamide. A trial of thiamine supplementation could be considered in cases suspicious for disorders of thiamine metabolism and PDH complex disorders.