Infantile spasms and encephalopathy without preceding neonatal seizures caused by KCNQ2 R198Q, a gain-of-function variant

SUMMARY

Variants in KCNQ2 encoding for Kv7.2 neuronal K⁺ channel subunits lead to a spectrum of neonatal-onset epilepsies ranging from self-limiting forms to severe epileptic encephalopathy. Most KCNQ2 pathogenic variants cause loss-of-function, whereas few increase channel activity (gain-of-function). We herein provide evidence for a new phenotypic and functional profile in KCNQ2 related epilepsy: infantile spasms without prior neonatal seizures associated to a gain-of-function gene variant. Via an international registry, we identified four unrelated patients with de novo heterozygous KCNQ2 c.593G>A, p.R198Q variants. All were born at term and discharged home without seizures or concern of encephalopathy, but developed infantile spasms with hypsarrhythmia (or modified hypsarrhythmia) between the ages of 4-6 months. At last follow up (ages 3-11 years), all patients were seizure-free and had severe developmental delay. In vitro experiments showed that Kv7.2 R198Q subunits shifted current activation gating to hyperpolarized potentials, indicative of gain-of-function; in neurons, Kv7.2 and Kv7.2 R198Q subunits similarly populated the axon initial segment, suggesting that gating changes rather than altered sub-cellular distribution contribute to disease molecular pathogenesis. We conclude that KCNQ2 R198Q is a model for a new subclass of KCNQ2 variants causing infantile spasms and encephalopathy, without preceding neonatal seizures.

Variants in the KCNQ2 gene are responsible for a wide phenotypic spectrum of epileptic diseases, ranging from benign familial neonatal epilepsy (BFNE; OMIM: 121200) to neonatal-onset epileptic encephalopathy (EIEE7, OMIM: 613720)¹. In autosomal-dominant KCNQ2-BFNE, seizures start in otherwise healthy infants around the first week of life and spontaneously disappear after few months, with mostly normal interictal electroencephalogram (EEG), neuroimaging, and psychomotor development. By contrast, EIEE7 is nearly always caused by de novo KCNQ2 missense variants, and characterized by multiple, pharmacoresistant seizures beginning in the first few days of life, burst-suppression pattern or multifocal epileptiform activity at the EEG, and moderate to severe developmental impairment.² KCNQ2 pathogenic variants are among the most common cause for neonatal-onset EE, overlapping with Ohtahara syndrome and Early Myoclonic Encephalopathy.¹

KCNQ2 encodes for Kv7.2 K⁺ channel subunits underlying the neuronal voltage-gated M-current (I_KM).³ Most KCNQ2 pathogenic variants decrease channel function, with stronger functional impairment often causing more severe epileptic phenotypes.^4,5 In addition, a few EIEE7 de novo variants enhance channel function.^{6, 7}

Genotype-phenotype correlations are difficult to establish in rare disorders due to inter- and intra-familial phenotypic variability; comparing histories of patients carrying the same KCNQ2 pathogenic variant provides the unique opportunity to correlate the specific variant with the disease clinical course and severity. Here, we describe the case histories of four unrelated patients carrying the same de novo KCNQ2 missense variant (c.593G>A) leading to the p.R198Q substitution. All patients exhibited similar clinical courses, characterized by infantile spasms (IS) without previous neonatal seizures. In vitro, when compared to wild-type, mutant Kv7.2 R198Q channels activated at less depolarizing potentials, thus showing a gain-of-function (GOF) effect.

METHODS

Clinical database

The Rational Intervention for KCNQ2/3 Epileptic Encephalopathy (RIKEE) database is currently edited and maintained at Baylor College of Medicine under an IRB approved research protocol. RIKEE collects information about patients where genetic tests show KCNQ2 and KCNQ3 sequence variants, their phenotypes, and related lab studies; summaries are disclosed on a website (www.rikee.org). For each patient reported here, the registering parent optionally consented to be re-contacted by study staff. A study neurologist (ECC, JJM) reviewed proband and parental CLIA-certified genetics lab reports to confirm genotypes, reviewed medical charts, and interviewed a parent to resolve uncertainties in the medical record.

Heterologous expression and Whole-cell electrophysiology

Variants were engineered in human KCNQ2 cDNA; channel subunits were expressed in Chinese Hamster Ovary (CHO) cells by transient transfection, as described⁴ (Data S1). Currents were recorded under whole-cell patch-clamp at room temperature (20-22°C) 1-2 days after transfection.

Multistate structural modelling

Three-dimensional models of K_v7.2 subunits were generated by using as templates the coordinates of six different states of K_v1.2/2.1 paddle chimera (PDB accession number 2R9R; 29% of sequence identity with K_v7.2) obtained in molecular dynamics simulations,⁸ as previously described (Data S1).⁴.

Neuronal cell transfection and immunocytochemistry

The subcellular distribution of Kv7.2 subunits was investigated in primary cultures of rat hippocampal neurons transfected with HA/EGFP-tagged Kv7.2 (wild-type or mutant) and Kv7.3 cDNAs. The axon initial segment (AIS)/soma and AIS/dendrites ratios were calculated frm confocal images as described in the Supplementary methods section (Data S1), as reported.⁹

Statistics

Data are expressed as the mean±SEM. The Student's t-test was used for statistical comparison (p<0.05).

RESULTS

Clinicogenetic data

In all four patients, genetic testing performed as part of clinical care revealed an heterozygous c.593G>A, p.R198Q variant in the KCNQ2 gene (Table I); this variant is not found in healthy subjects (http://exac.broadinstitute.org/gene/ENSG00000075043). All parents were found negative for the KCNQ2 R198Q variant. All four patients in the RIKEE database carrying the KCNQ2 c.593G>A, p.R198Q variants were included.

Table 1.

Clinical characteristics

Patient	1	2	3	4
Sex	F	M	F	F
Age	3 y	3 y	5 y	11 y
Diagnosis	Encephalopathy	Encephalopathy	Encephalopathy	Encephalopathy
KCNQ2 Pathogenic Variant	c.593 G>A p. R198Q de novo	c.593 G>A p. R198Q de novo	c.593 G>A p. R198Q de novo	c.593 G>A p. R198Q de novo
Variant detection method	Targeted NGS (53 genes panel)	Targeted NGS (70 genes panel)	WES	WES
Pregnancy duration, Birth mode	39 weeks, C-section	38 weeks, C-section	40 weeks, vaginal delivery	40 weeks, C-section
Initial symptoms (age)	Developmental regression; loss of smiling; irritable (5 m)	Infantile spasms (4 m)	Developmental regression (4 m)	Developmental delay; poor head control (4 m)
Initial seizure type (age)	Infantile spasms (6 m)	Infantile spasms (4 m)	Infantile spasms (5 m)	Infantile spasms (6 m)
Transition to other seizures (frequency)	Tonic (last seizure 2 y)	None (last seizure 1 y)	None (last seizure 1 y)	Status epilepticus at 2 y and 3 y (last seizure 6 y)
Seizures (last f/u)	Seizure-free (3 y)	Seizure-free (3 y)	Seizure-free (5 y)	Seizure-free (11 y)
Development (last f/u)	Severe delay (22 m)	Moderate delay (5 m)	Severe delay (5 y)	Severe delay (11 y)
Initial treatment	TPM	PRD	ACTH	LVT
All Treatments	TPM; LVT; PRD; EZG; OXC; CLB; CBM	PRD; ZNS; LVT	ACTH; TPM; ZNS; VGB; KD	LVT
Current Therapy	CBM	LVT	none	LVT
Initial EEG (age)	Modified hypsarrhythmia (6 m)	Hypsarrhythmia (4 m)	Hypsarrhythmia (5 m)	Hypsarrhythmia (6 m)
Transition to other EEG findings (age)	Diffuse slowing (18 m)	Diffuse slowing and multifocal epileptiform discharges (6 m)	Diffuse slowing (22 m)	Diffuse slowing (9 y)
MRI brain	Mild cerebral volume loss	Normal	Mild cerebral volume loss	Mild cerebral volume loss; increase T2 signal in periventricular white matter

ACTH, adrenocorticotropin; CBM, carbamazepine; LVT, levetiracetam; TPM, topiramate; PRD, prednisolone; EZG, ezogabine; OXC, oxcarbazepine; CLB, clobazam; ZNS, zonisamide; VGB, vigabatrin; KD, ketogenic diet; NGS, next generation sequencing; WES, whole exome sequencing.

In addition to the KCNQ2 c.593G>A, p R198Q variant, all four patients had additional variants of uncertain significance in other epilepsy-related genes that were not considered clinically significant based on the currently available information. These were: TSC2 (c.599+3 A>G; heterozygous), CLN5 (c.127 G>A; heterozygous), and PRICKLE1 (c.425 C>T; heterozygous) for patient 1, PCDH19 (c.1321 G>C; hemizygous) and PPT1 (c.904 A>G; heterozygous) for patient 2, NDUFAF2 (c.221 G>A; heterozygous; paternally inherited; autosomal recessive) for patient 3, and RET (c.2556 C>G; heterozygous; maternally inherited) for patient 4.

Patient Characteristics

All four patients were born at term, and discharged home from the nursery without seizures or concern of encephalopathy. Onset of IS with hypsarrhythmia (or modified hypsarrhythmia) occurred between the ages of 4-6 months (Table 1).

Patient 1. At 5 months, parents noticed regression marked by decreased smiling and increased irritability. At 6 months, clusters of IS characterized by neck flexion and limb extension began. Initially, about 2 clusters per day (25 spasms per cluster) occurred. EEG obtained at the local hospital showed a modified hypsarrhythmia pattern. Initial treatment (topiramate and levetiracetam) was incompletely effective. Video EEG at age 6 months captured numerous clusters of flexor IS and a modified hypsarrhythmia pattern consisting of generalized background slowing and multifocal epileptiform discharges (Fig. S1A). For ongoing spasms, oral prednisone was given, producing a complete but temporary remission. Clobazam was started while topiramate was titrated down. She was diagnosed with KCNQ2 encephalopathy at 8 months old and ezogabine was added to oxcarbazepine, clobazam, and levetiracetam. Ezogabine was titrated to 13 mg/kg/day that resulted in an initial improvement in seizures and alertness that was not sustained. Due to recurrence of refractory IS, combinations of ezogabine, levetiracetam, oxcarbazepine, carbamazepine, and clobazam were tried. EEG background at 18 months old was significant for diffuse background slowing (Fig. S1B). By 2 years old, seizures stopped on a combination of carbamazepine, ezogabine, clobazam, and levetiracetam. Anticonvulsants were serially weaned off. At last follow up (3 years old), she remains seizure-free on carbamazepine monotherapy. She has severe developmental delay and is unable to speak or sit.

Patient 2. At 4 months, parents noted episodes of trunk flexion that occurred multiple times per day and were associated with irritability. Video EEG showed background activity consistent with hypsarrhythmia and captured electroclinical IS. Oral prednisolone was started. Spasms stopped immediately and hypsarrhythmia resolved within 1 month. He was started on zonisamide, but was subsequently changed to levetiracetam due to side effects. Last available EEG showed a continuous background with diffuse slowing and multifocal epileptiform discharges. At last follow up (3 years old), he remains seizure-free on levetiracetam monotherapy. Development was slow for the first 2 years of life and at 3 years old he is able to babble and sit unassisted.

Patient 3. At 4 months there was parental concern for developmental regression, including loss of babbling, smiling, and rolling. Shortly thereafter, parents noted clusters of bilateral arm extension associated with irritability. EEG was performed, showing hypsarrhythmia. The patient was treated with ACTH producing a temporary remission. However, spasms recurred. Subsequent treatments included topiramate, zonisamide, levetiracetam, and vigabatrin. Ultimately, spasms ceased after ketogenic diet was started at 10 months old. All anticonvulsants were weaned off after several months of seizure-freedom. Ketogenic diet was discontinued 4 months later and she remained seizure-free off anticonvulsants since then. At age 5 years, she has global developmental delay, is nonverbal, and cannot sit unassisted.

Patient 4. At 3-4 months, parents noted developmental regression in the form of loss of head control, and paroxysmal events characterized by blank staring and unresponsiveness for 10-15 seconds associated with upward eye deviation and slight hand movements. At 6 months, an initial EEG performed for “jerks” was significant for a severely abnormal background that was discontinuous with asynchronous bursts of high amplitude epileptiform discharges. Seizures were treated with levetiracetam and the parent reported a good response. After 2 episodes of prolonged seizures in early childhood, she has remained seizure-free on continued levetiracetam since age 6. EEGs at 7-9 years old showed diffuse slowing. At age 11 years, she has global developmental delay, is attentive, but cannot speak or sit unassisted.

Functional properties and neuronal subcellular localization of channels incorporating Kv7.2 R198Q subunits

The c.593G>A variant substitutes the highly conserved, most extracellular, positively-charged Arginine (R1) with a neutral Glutamine (Q) (p.R198Q, Fig. 1A) along the fourth transmembrane segment S4 (Fig. 1B). In CHO cells, Kv7.2 channels generated K⁺ selective and voltage-dependent currents that activated around −40 mV (Fig. 1C). At the holding voltage of −80 mV, homomeric Kv7.2 channels are closed; therefore, the ratio between the current measured at the beginning of the depolarizing step (I_inst) and that at the end of the 0 mV depolarization (I_steady-state), was negligible (0.02±0.01). Homomeric Kv7.2 R198Q channels displayed a 30 mV hyperpolarizing shift in their activation gating (Figs. 1C, 1D) and the I_inst/I_steady-state ratio was 0.26±0.02 due to the presence of a large fraction of channels already open at the holding potential of −80 mV, as described.¹⁰ Charge reversal at the same position (R198E, R198D) completely abolished voltage-dependent gating, leading to I_inst/I_steady-state ratios close to unity (Fig. S2; Table S1). When compared to wild-type, Kv7.2 R198Q channels displayed slower deactivation kinetics, a result consistent with a destabilization of the closed state (Figs. 1E, 1F). Multistate molecular modeling (Fig. S3) revealed that the R198 residue is involved in electrostatic interactions with several partially negatively charged residues, including a critical cysteine (C106) in S1 in the voltage sensor resting and closeby (early activated and late deactivated) gating states, whereas no interactions seem to occur in the fully activated state,¹¹ suggesting the involvement of R198 in resting state VSD stabilization. No difference in ion selectivity¹⁰ and tetraethylammonium (TEA) sensitivity (Table S1) was observed between wild-type and mutant homomeric channels. Cells transfected with Kv7.2+Kv7.2 R198Q+Kv7.3 cDNAs (0.5/0.5/1 cDNA ratio), an experimental strategy mimicking the genetic balance of the affected patients and reproducing I_KM composition³ generated heteromeric K⁺ currents retaining a significant 10 mV hyperpolarizing shift in their activation gating when compared to Kv7.2+Kv7.3 channels (Fig. 1G; Table S1). Finally, in cultured rat hippocampal neurons, Kv7.2+Kv7.3 and Kv7. 2 R198Q+Kv7.3 heteromers were similarly expressed at the ankyrin-G-identified axon initial segments (AISs), showing identical AIS/dendrite and AIS/soma ratios (Fig. S4).

Figure 1. Functional properties of channels incorporating Kv7.2 R198Q subunits.

(A) Topological representation of a Kv7.2 subunit, showing the six transmembrane segments (S₁-S₆) and the intracellular N- and C-termini. Arrow indicates the location of the variant investigated. (B) Sequence alignment of S₄ segments of the indicated K⁺ channel subunits (www.ebi.ac.uk/Tools/psa/). (C) Macroscopic currents from Kv7.2 and Kv7.2 R198Q channels. Current scale, 200 pA; time scale, 200 ms. (D) Conductance/voltage curves; continuous lines are Boltzmann fits of the experimental data. (E) Normalized and superimposed current traces from the indicated channels. (F) Time constants for ionic current activation (filled symbols) and deactivation (empty symbols) (n=6-9) for Kv7.2 and Kv7.2 R198Q channels. (G) Conductance/voltage curves for Kv7.2+Kv7.3 and Kv7.2+Kv7.2 R198Q+Kv7.3 heteromeric channels. Continuous lines are Boltzmann fits of the experimental data.

DISCUSSION

We describe four patients with unremarkable neurodevelopmental status around birth, in particular, without histories of neonatal-onset seizures or encephalopathy. Each presented between ages 4 and 6 months with IS in clusters, with hypsarrhythmia at the time of EEG, accompanied by varying levels of developmental delay. Remarkably, all four patients carried the same de novo KCNQ2 variant (c. 593G>A; p.R198Q). The similarity within our small cohort strongly suggests that the shared R198Q variant, never previously reported in neonatal-onset KCNQ2 related epilepsy, is pathogenic for IS, an infantile onset EE.

A KCNQ2 variant (R144Q) was previously detected by WES in 1 of 149 patients with no apparent history of epilepsy before the onset of IS;¹² in another cohort of 73 patients with IS, no KCNQ2 mutation was found.¹³ The KCNQ2 R144Q variant was shown to exhibit GOF in heterologous expression studies.⁷ Here, functional studies reveal that the Kv7.2 R198Q subunit, expressed alone or in combination with Kv7.2 and/or Kv7.3 subunits, also increases currents activated by partial membrane depolarization, without major changes in pore properties, heteromerization, maximal current density, and neuronal sub-cellular localization. KCNQ2 GOF is an atypical functional profile, recently described in rare patients with severe neonatal-onset epileptic encephalopathy carrying KCNQ2 R201C, and R201H^12,14 or KCNQ3 R230C^12,15 variants. Thus, to date, two phenotypes have been associated with KCNQ2 GOF variants: a severe neonatal-onset EE (paradigmatically exhibited by R201C-carrying patients), and IS in the absence of neonatal seizures (as in R198Q patients, and the single reported R144Q patient^7,12). Notably, expressed as homotetramers, R201C channels display significantly more dramatic GOF effects compared to R198Q or R144Q channels^7,10; making this in vitro assay a potential prognostic tool. Very recently, a different KCNQ2 variant affecting the pore residue A306 has also been reported in a patient with sporadic IS syndrome and no apparent history of neonatal seizures;¹⁶ further studies, possibly including transgenic animal models,¹⁷ are needed to understand such complex genotype-phenotype correlations.

The limitations of retrospective chart review, limited patient numbers, the natural fluctuations in seizure frequency, and concurrent modification of multiple anticonvulsants make it difficult to draw conclusions regarding the effect, either positive or negative, of ezogabine or other anticonvulsants in patients with the R198Q variant. Although there was a perceived early positive response to ezogabine in patient 1, this was not sustained or complete and additional anticonvulsant modifications were necessary. Similarly, response to sodium channel blockers could not be assessed.¹⁸

Although recognized only recently, KCNQ2 encephalopathy already exhibits a diversity of apparently correlated molecular mechanisms and phenotypes. Here, identification of 4 unrelated patients enabled clearer conclusions to be drawn about R198Q. These finding highlight the benefits of transnational networks and multidisciplinary-curated gene-specific databases for hastening progress in phenotypic characterization, patient stratification, and individual tailoring of therapeutic approaches.¹⁹

Biography

One sentence line describing the first author, Dr, John J. Millichap: Assistant Professor of Pediatrics and Neurology, Northwestern University Feinberg School of Medicine