Clinical and molecular characterization of KCNT1 -related severe early-onset epilepsy

1Autosomal dominant pathogenic variants in KCNT1, encoding the sodium-activated potassium channel, are identified in a wide spectrum of epileptic disorders with variable age at onset and cognitive outcome. These include severe early-onset epileptic encephalopathies such as Ohtahara and West syndromes, and epilepsy of infancy with migrating focal seizures (EIMFS), as well as autosomal dominant and sporadic severe nocturnal frontal lobe epilepsies (ADNFLE and NFLE),,, but the genotype-phenotype relationship appears to be unclear. We undertook detailed clinical, molecular genetic, and functional characterization of a cohort of patients with KCNT1-related epilepsy.

2We recruited patients with EIMFS (n = 31) to a research study investigating the genetic basis of early-onset epileptic encephalopathy (EOEE) between 2011 and 2016, following an earlier national surveillance study. Inclusion criteria were epilepsy with onset at <2 years and unknown etiology. Diagnostic criteria for EIMFS were as described in the previous study. Patients were recruited at Great Ormond Street Hospital, London, UK, and by referral from other centers in the United Kingdom and internationally. Two patients who had routine local diagnostic multiple gene panel testing revealing KCNT1 variants were also included.

3We obtained written informed consent from families in whom research genetic investigations were undertaken. The study was approved by the National Research Ethics Service (London-Bloomsbury, Research Ethics Committee reference 13/LO/0168, Integrated Research Application System project identifier 95005). We collected anonymized data from patients tested on the diagnostic next-generation sequencing panel (n = 3) as part of an approved case note review project (Great Ormond Street Hospital Research and Development Department, 16NM11).

4We used a variety of different methods (table e-1, http://links.lww.com/WNL/A6), including direct Sanger sequencing, multiple gene panel testing with the TruSeq Custom Amplicon panel and SureSelect panel, exome sequencing (e-Methods, http://links.lww.com/WNL/A8; tables e-1 and e-2, http://links.lww.com/WNL/A6), and diagnostic chromosomal microarray.

5HMMscan against Pfam (database of sequence-based domain families) identified 2 domains in the sequence of human KCNT1 (isoform 1): ion channel (PF07885, at position 278–346) and calcium-activated BK potassium channel alpha-subunit family (PF03493, at position 495–598) (e-Methods).

6We introduced variants into a wild-type (WT) human KCNT1 expression construct using QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA). cDNAs were transcribed in vitro (mMessage mMachine; Ambion, Austin, TX). Oocytes were prepared, and 2-electrode voltage clamp recording was performed after 14 to 24 hours of expression. We also recorded currents before and after the application of Quinidine (e-Methods).

7We identified pathogenic variants in KCNT1 in 12 patients, 5 through direct Sanger sequencing, 2 from whole-exome sequencing, and 5 from the Great Ormond Street Hospital diagnostic panel (5 of 800 tested patients with EOEE/developmental delay).

8Clinical features are summarized in table 1. Median age at seizure onset was 3.5 weeks (range 1 day–6 months). Most patients developed seizures consistent with EIMFS. Two patients (patients 3 and 11) presented with severe, early-onset NFLE, characterized by asymmetric tonic posturing and later fencing posture. We noted similar frontal seizure semiology in patients with EIMFS (e.g., patient 12). All patients developed axial hypotonia, and upper motor neuron signs emerged in 3 patients. Four patients had a choreiform movement disorder (onset 14–24 months); 1 patient developed generalized dystonia at 18 months. Onset of hyperkinesia was not related to medication (including vigabatrin) nor triggered by intercurrent illness. Initial age at presentation, disease course, response to medication, brain MRI, and EEG findings were similar for both KCNT1 pathogenic variant–positive and –negative patients from the cohort. However, 5 of 12 pathogenic variant–positive patients with EIFMS presented with a severe movement disorder compared to 2 of 19 KCNT1-negative cases. Most had extensive uninformative laboratory metabolic and genetic investigations. Abnormal muscle respiratory chain enzyme activity for complex I and/or II was detected in patients 4 and 8 of uncertain significance (table e-3, http://links.lww.com/WNL/A6). For patient 4, the muscle biopsy was taken during an intercurrent illness and repeated after clinical recovery, revealing a more borderline result. In patient 8, borderline abnormalities in complex I and II ratios were found. Neither patient had other systemic, biochemical, or radiologic features of mitochondrial disease or concurrent sodium valproate treatment.

9In general, neurodevelopmental outcome was markedly impaired in all patients with EIMFS. All patients had a trial of at least 5 different medications. Response to treatment was, in general, poor (table 1). Three of 8 patients who received the ketogenic diet in combination with other antiepileptic drugs responded with ≈75% seizure reduction. Three patients were treated with quinidine. Patient 2 received 40 mg/kg/d without adverse events but with no effect on seizure burden. Patient 5 was treated with quinidine at 40 mg/kg/d, leading to a marked reduction in seizure frequency. Patient 6 showed some initial transient reduction in seizure frequency at 30 mg/kg/d. For this patient, the unexpected development of a severe proliferative pulmonary and mediastinal vasculopathy resulted in life-threatening pulmonary hemorrhage. Investigations failed to identify an underlying vasculitis, and quinidine was subsequently withdrawn. The patient later died despite initial successful pulmonary embolization.

10All patients with an EIMFS phenotype had a “migrating” ictal focus with discrete ictal involvement of differing cortical areas within the same EEG (table e-4, http://links.lww.com/WNL/A6). Although not always evident at initial presentation, it developed by 7 months of age in most patients. Periods of EEG suppression or burst suppression were noted in 8 of 12 patients; 6 of these patients had seizure onset in the first 4 weeks of life. Further atypical EEG features included a generalized electrodecremental response in 5 patients and hypsarrhythmia in 1 patient.

11Neuroimaging was available for review in 11 of 12 patients. The majority developed predominantly frontal cerebral atrophy by 3 years of age (figures e-1A and e-1B, http://links.lww.com/WNL/A7; table e-5, http://links.lww.com/WNL/A6). Cerebellar atrophy was also evident in 4 patients (figure e-1C). We noted an open operculum in the first 6 months of life in patient 4 (figure e-1B). Delayed myelination was evident in 9 of 11 patients who had imaging after 3 months of age. In some patients, early brain imaging was normal. Magnetic resonance spectroscopy was abnormal with a relatively reduced N-acetylcholine peak in 3 of 4 patients.

12We identified 12 patients with pathogenic variants in KCNT1 (table 1 and table e-6, http://links.lww.com/WNL/A6); 8 have been previously reported and 4 are unpublished.,,,, Eight of 12 patients had C-terminus variants, of whom 5 had the commonly reported variant A934T. We have identified 4 (including 2 unpublished) pathogenic variants causing EIMFS, namely V271F, L274I, G288S, and F346L located in or between transmembranes 5 and 6. All are missense variants that are predicted to be pathogenic (table e-6), affecting highly conserved amino acid residues (figure e-2, http://links.lww.com/WNL/A7), and are not reported in 1000 Genomes, the ExAC database, or the Exome Variant Server. For 9 of 12 cases, variants occurred de novo. Parental DNA was not available for patient 1. In patient 5, we found the same KCNT1 variant in an asymptomatic mother and her affected child. We noted a lower heterozygous peak on Sanger sequencing of both salivary and blood-derived maternal genomic DNA (figure e-3), which may reflect somatic mosaicism. In patient 7, the variant was inherited from the unaffected father with no difference in peak size on Sanger sequencing (figure e-4). The recurrent A934T variant was identified in 5 patients, 4 with an EIMFS presentation and 1 (patient 11) with an NFLE phenotype.

13Homology modeling was performed for 2 novel mutations: F346L, located in the ion channel domain (residues 270–353), and F502V, located in the gating region (residues 373–1174, although residues 1,045–1,174 could not be modeled). F346 is located on the inner helix of the transmembrane pore (figure 1, A and B). It is part of the hydrophobic cavity, which mediates interactions between the inner-membrane helices of 2 adjacent subunits (figure 1C) and is thus responsible for maintaining the stability of the open conformation. In the modeled closed-state conformation, the helix containing F346 and the inner helix from the other protomer undergo conformational changes (figure e-5, http://links.lww.com/WNL/A7). Therefore, mutation to leucine (F346L) is likely to destabilize the open state by perturbing the hydrophobic interactions because the side chain of leucine is smaller (figure 1D), affecting the equilibrium between the closed and open states. In addition, the packing arrangement in the K+ channels involving the pore and the inner helix is known to be critical for the stability of the tetrameric assembly, ion conduction function, and cation selectivity. Thus, F346L might be detrimental to these functions.

14Within each protomer of the KCNT1 gating region, there are 2 tandem RCK domains (RCK1 and RCK2) that serve as regulators of potassium conductance (figure 1E). These form flexible intrasubunit and intersubunit (figure 1E) interfaces that facilitate functional tetramer formation. F502 is located in RCK1 and predicted to form a pi-pi interaction with W476 from αD (figure 1F). F502 is also surrounded by a number of hydrophobic residues (I472, L473, A475, V500, and A503), which may play a role in stabilizing the gating ring (figure 1F). The amino acid substitution F502V is predicted to result in destabilization of these hydrophobic interactions, given the smaller valine side chain (figure 1G), and abolition of potential pi-stacking with resultant disruption of the stable assembly interface.

15We evaluated the 4 previously unpublished variants and V271F, which we previously described and was recently studied in a xenopus oocyte system. All mutations resulted in an increased current magnitude compared to WT (figure 2A). We noted that for variants V271F and F346L, the rate of activation was slowed at higher voltages compared to WT, and in others (M896K, F502V and L274I), the activation rates were generally faster than WT (figure 2A). Investigation of the current-voltage relationship showed that mutant channels were very weakly voltage dependent, and in some cases, voltage dependence of steady-state activation was essentially absent (figure 2, B and C) with only a residual Goldman-Hodgkin-Katz rectification. Assessment of average peak currents at 10 mV revealed a significant difference between both individual mutant channels and summated data compared to WT (figure 2, D and E).

16Quinidine 300 µmol/L had variable current-blocking effects in different mutant channels. For F346L, peak current was completely insensitive to quinidine, although it had some effect on activation kinetics (figure 3A). The differential sensitivity of KCNT1 mutants to quinidine was clearly shown in the current-voltage relationship (figure 3B) and percentage of inhibition at maximum current, 80 mV (figure 3C). There is some correlation between the in vitro studies and clinical response in patient 5 (figure 3). M896K had the most marked in vitro blockade by quinidine, and patient 6 showed some initial clinical response. F346L showed no quinidine response at all, and the patient harboring this mutation was not treated with quinidine.

17We report a cohort of patients with early-onset epilepsy associated with pathogenic variants in KCNT1, which encodes the sodium-activated potassium channel KCa4.1 (sequence like a calcium-dependent potassium channel [SLACK], Slo2.2). KCNT1 is widely expressed throughout the brain, as well as in the dorsal root ganglia, kidney, and heart, and is responsible for slow hyperpolarization after bursts of action potentials., KCNT1 also has direct interactions with Fragile X-related protein. Compared with other potassium channels, KCNT1 is involved in a highly extensive protein network, suggesting a putative role in cognitive developmental processes.,,,

18To date, KCNT1 variants have been reported in a wide range of epilepsies (table e-7, http://links.lww.com/WNL/A6).,, We identified patients with the same variant associated with varying electroclinical phenotypes (table 1). Phenotypic variability has been reported within single families in which different individuals may present with either ADNFLE or EIMFS. Such intrafamilial variation in phenotype is also described in SCN1A kindreds; Dravet syndrome, febrile seizures, and a variety of other generalized epilepsies may be reported in the same family. Furthermore, while the majority of variants in our cohort occurred de novo, 2 patients inherited variants from an unaffected parent. The mechanisms underlying phenotypic variability and true/apparent nonpenetrance are unclear but may be related to somatic mosaicism, variant type, other genetic/epigenetic factors, or differential expression of alternative KCNT1 transcripts.,,,,

19The majority of patients with pathogenic KCNT1 variants in our cohort had electroclinical EIFMS, although this is likely to reflect ascertainment bias. Indeed, 2 of the 5 KCNT1-positive patients identified by the diagnostic panel from a larger cohort of 800 patients with EOEE/developmental delay had an NFLE-like presentation. Although movement disorders are increasingly reported in other severe early-onset genetic epilepsies, they appear to be rare in KCNT1 epilepsy. We describe several atypical EEG features. Generalized electrodecrement and hypsarrhythmia, more classically associated with infantile spasms, have been previously described in EIMFS.,,,, EEG suppression, classically seen in Ohtahara syndrome, has been only rarely described in EIMFS., Extensive diagnostic investigations undertaken in patients with KCNT1 mutations were unyielding other than abnormal respiratory chain enzyme analysis of muscle tissue in 2 patients. The relevance of these findings is not clear, but secondary mitochondrial effects may be evident in KCNT1 epilepsy, as often reported in other severe drug-resistant epilepsies. Other genetic and environmental influences on mitochondrial function may also play a role.

20KCNT1 tetramers form a transmembrane sodium-activated potassium channel. Each subunit consists of 6 transmembrane domains with an extended cytoplasmic carboxy (C-) terminus (figure 4). The majority of reported pathogenic variants (table e-7, http://links.lww.com/WNL/A6), as seen in this study, are located in the C-terminus with clustering around the RCK and nicotinamide adenine dinucleotide–binding domains (figure 4). More recently, several variants have been identified within transmembrane domain 5 and in the pore-forming regions between transmembrane domains 4 and 5,, (table e-7), and this study also demonstrates epilepsy-associated mutations in transmembrane domains.

21To date, different model systems have been used to determine the functional effects of KCNT1 variants.,,,,,, Our protein homology structural modeling data predict abnormal gating or protein instability within the pore-forming region as a putative disease mechanism. In silico modeling of G288S has predicted similar detrimental effects, while Y775H is predicted to affect sodium sensitivity of the channel. KCNT1 variants may therefore alter structural properties of the protein, contributing to altered channel function. Consistent with previous reports,,,, (table e-7, http://links.lww.com/WNL/A6), our xenopus oocyte model demonstrated that KCNT1 pathogenic variants display a gain-of-function effect with increased current amplitude (figure 2). Previous studies have sought to correlate disease severity with the degree of gain of function., However, in keeping with more recent studies, such correlation was not evident in our study. KCNT1 variants result in an increased Po (probability of the channel being open), which may be due to increased mutant channel cooperativity or altered sodium sensitivity., In a recent study, sodium removal from the pipette solution had a less negative effect on G288S channel amplitude than WT, suggesting reduced sodium sensitivity in the mutant. Heterotetramer formation may be of importance in vivo. In 1 study, mutant KCNT1 homomers revealed a more marked gain of function than mutant WT heteromers. A significant remaining question is how KCNT1 gain-of-function variants with predicted effects on neuronal hyperpolarization result in epilepsy. Altered voltage sensitivity may result in KCNT1 channels opening at more depolarized potentials, allowing a persistent hyperpolarizing current, with resultant interneuronal disinhibition as reported in SCN1A-related epilepsy. Conversely, increased repolarization permitting more frequent and rapid action potentials may also play a role.,

22Recently, quinidine has been identified as a novel therapy for patients with KCNT1-related epilepsy. In in vitro models, quinidine has been shown to reduce the abnormal increase in mutant KCNT1 channel amplitude. For 1 patient with EIMFS with the KCNT1 variant R428Q, in vitro testing showed quinidine sensitivity, and treatment resulted in a dramatic improvement in seizure control with neurodevelopmental gains., However, in more recent studies, patient response has been variable and not always as predicted by in vitro studies. Indeed, another patient with the same variant (R428Q) but different epilepsy phenotype (unclassified EOEE) failed to respond to quinidine, albeit at a later stage in the disease course. Most recently, a patient with West syndrome had a good response but only with a higher dose of 60 mg/kg/d. Clinical response may possibly be determined by the specific variant, other genetic factors, epilepsy phenotype, and drug timing within a therapeutic window. In our series, we treated 3 patients with quinidine, and 1 patient showed a clinical response. One patient developed a severe pulmonary vasculopathy, after which quinidine was discontinued. Systemic vasculitis has been reported with quinidine treatment. While investigations in our patient did not reveal overt evidence of vasculitis, the observed pulmonary dysfunction may represent an adverse drug-related event. The precise mechanism of KCNT1 blockade by quinidine is unclear, and it is possible that the disease mechanism for F346L, perhaps involving abnormal channel-opening dynamics as suggested by the modeling data, is not modifiable by quinidine. Our data suggest that quinidine should be considered as a therapeutic option for patients with KCNT1 variants, but used with caution. Larger studies will provide further guidance about clinical utility, patient selection, optimum age at administration, and dose. Other KCNT1 modulators, including bepridil and clofililum, have been identified as possible alternative therapies. Like quinidine, bepridil has been shown in vitro to reversibly block mutant KCNT1 channels at a lower concentration than WT channels. However, similar to quinidine, potential cardiac effects and lack of specificity may limit use in patients.

23Pathogenic variants in KCNT1 cause a wide spectrum of severe epilepsies typically associated with impaired neurologic development and significant disease burden. As demonstrated, in vitro model systems may be useful to validate putative variants and to confirm pathogenicity, although genotype-phenotype correlations remain unclear. Evaluation of new therapies, including KCNT1-specific blockers, remains a research priority for this devastating pharmacoresistant group of epilepsies.