KCNJ4 variants disrupt inward‐rectifier potassium channel function and cause refractory epilepsy

Abstract

Objective

Epilepsy is a common neurological disorder with a strong genetic basis, most frequently arising from ion channel dysfunction. Although multiple inwardly rectifying potassium (Kir) channels have been implicated in epileptogenesis, the contribution of KCNJ4, which encodes the Kir2.3 channel, has not previously been established in human epilepsy. The present study aimed to identify pathogenic KCNJ4 variants and to elucidate their functional consequences in the context of epilepsy.

Methods

Trio whole exome sequencing was performed in four unrelated individuals with refractory epilepsy and neurodevelopmental abnormalities. Identified KCNJ4 variants were evaluated for rarity and inheritance patterns. Functional consequences were assessed using two‐electrode voltage‐clamp recordings in Xenopus laevis oocytes coexpressing wild‐type or mutant Kir2.3 together with Kir2.1. Protein expression levels were examined by Western blot analysis to exclude effects attributable to altered channel expression or trafficking.

Results

We identified four rare heterozygous missense variants in KCNJ4 (Gly136Ser, Val206Met, Met293Lys, and Glu384Lys), all of which were absent from public population databases. Clinically, affected individuals exhibited a broad phenotypic spectrum ranging from isolated epilepsy to severe developmental and epileptic encephalopathy. Electrophysiological analyses revealed variant‐specific functional alterations; the Gly136Ser and Glu384Lys variants significantly increased inwardly rectifying potassium currents, consistent with gain‐of‐function effects, whereas the Val206Met and Met293Lys variants markedly reduced current amplitudes, indicating loss of function. These functional changes were independent of channel protein expression levels.

Significance

Our findings establish KCNJ4 as a novel epilepsy‐associated gene and demonstrate that both gain‐ and loss‐of‐function mechanisms of Kir2.3 can contribute to epileptogenesis. This study expands the genetic landscape of epilepsy and highlights the critical role of inward‐rectifier potassium channel regulation in neuronal excitability, with potential implications for mechanism‐based therapeutic strategies.

Key points.

• Novel KCNJ4 variants were identified in four unrelated patients with refractory epilepsy and variable neurodevelopmental phenotypes.

• Kir2.3 channel dysfunction is implicated in epilepsy, confirming KCNJ4 as a newly established epilepsy‐associated gene.

• Both gain‐ and loss‐of‐function mechanisms were observed, with different KCNJ4 variants altering inward‐rectifier potassium currents.

• Electrophysiological studies showed variant‐specific increases or decreases in Kir2.3‐mediated currents independent of protein expression.

1. INTRODUCTION

Epilepsy is a prevalent neurological disorder characterized by unprovoked seizures resulting from aberrant neuronal hyperexcitability and synchronized electrical discharges. ¹ This condition exhibits marked phenotypic heterogeneity, attributable to diverse etiologies involving complex gene–environment interactions. With an estimated 50 million affected individuals worldwide, epilepsy remains one of the most common neurological diseases. ² Emerging evidence suggests a strong genetic contribution, with 70%–80% of cases demonstrating heritable components. ³ , ⁴ Both inherited and de novo variants contribute to disease pathogenesis, further underscoring the intricate genetic architecture of epilepsy. These genetic epilepsies encompass a broad phenotypic spectrum, ranging from benign, self‐limiting syndromes to severe, refractory forms. Advances in genomic research have identified nearly 1000 genes implicated in epileptogenesis, highlighting the disorder's genetic complexity. A significant subset of genetic epilepsies are associated with developmental epileptic encephalopathy (DEE), a severe and often treatment‐resistant subgroup. DEE is characterized by rare, age‐specific electroclinical syndromes with distinct seizure patterns and electroencephalographic (EEG) abnormalities. ⁵ More than 50 genes have been definitively associated with DEEs, and with the widespread adoption of next generation sequencing, this number continues to expand rapidly. ⁶ Notably, pathogenic variants in ion channel‐associated genes—including potassium (K⁺), sodium (Na⁺), and other ion channels—account for approximately 25% of inherited epilepsies. ⁷

K⁺ channels are a class of ion channels that selectively facilitate potassium ion flux through a sophisticated gating mechanism. ⁸ Genetic variants affecting K⁺ channel‐encoding genes can induce structural and functional alterations, ultimately contributing to epilepsy pathogenesis. Inward‐rectifier potassium (Kir) channels represent a distinct class within the K⁺ channel superfamily, including voltage‐gated (Kv), two‐pore domain (K2P), and calcium‐activated cyclic nucleotide‐gated (HCN/CNG) channels. ⁹ , ¹⁰ These channels exhibit unique rectification properties, functioning as biological diodes that preferentially conduct inward K⁺ currents at hyperpolarized membrane potentials while limiting outward K⁺ flux during depolarization. Kir channels play a pivotal role in maintaining the resting membrane potential and modulating cellular excitability across both excitable and nonexcitable tissues. ¹¹ To date, approximately 20 members have been identified, which assemble as homotetramers or heterotetramers. Based on sequence homology and functional characteristics, these subtypes are classified into seven families (Kir1–Kir7). Notably, several Kir channels—including Kir2.1, Kir3, Kir4, and Kir6—have been implicated in epileptic phenotypes, underscoring their significance in neuronal hyperexcitability and seizure disorders. ¹²

The KCNJ4 gene encodes Kir2.3, an inwardly rectifying potassium channel highly expressed in the forebrain. Unlike other Kir2.x family members, Kir2.3 demonstrates significantly smaller ionic conductance while maintaining strong inward rectification properties critical for regulating neuronal resting membrane potential and firing threshold. ¹³ Current research indicates the involvement of Kir2.3 channels in epileptogenesis. Postmortem analyses have revealed a significant downregulation of Kir2.3 expression in hippocampal tissue from chronic temporal lobe epilepsy patients, potentially explaining the increased propensity for spontaneous seizure generation in this condition. ¹⁴ Additionally, the Kir2.3‐specific agonist tenidap has shown efficacy in attenuating seizure activity and neuronal hyperexcitability in pentylenetetrazole‐induced acute seizure models. ¹⁵ Despite these associations between Kir2.3 and epilepsy, no pathogenic variants in KCNJ4 have been reported to date in human epilepsy or other neurological disorders.

Here, we report four unrelated individuals with refractory epilepsy carrying rare missense variants in the KCNJ4 gene. Three patients exhibited global developmental, language, and motor delays, whereas one presented with isolated epilepsy. Functional analysis identified two distinct pathogenic mechanisms in KCNJ4; Gly136Ser and Glu384Lys acted as gain‐of‐function (GOF) variants, whereas Val206Met and Met293Lys acted as loss‐of‐function (LOF) variants, both disrupting channel conductance. Our findings identify pathogenic KCNJ4 variants as a novel genetic cause of epilepsy. These results underscore an evolutionarily conserved role for Kir2.3 channels in neuronal development and the regulation of excitability.

2. MATERIALS AND METHODS

2.1. Diagnosis and human genetics

This study received approval from the institutional review board of Hunan Provincial Maternal and Child Health Care Hospital and adhered to the principles outlined in the Declaration of Helsinki. Informed consent for participation was obtained by clinical practitioners responsible for patient care. The probands from Families 1 and 2 underwent thorough clinical evaluations, including medical histories and physical examinations. Genomic DNA was extracted from peripheral blood samples using Aidlab's Genomic DNA extraction kit. Whole exome sequencing (WES) was conducted on patients and their parents, except for I:2 in Family 1, who had passed away, leaving only the proband and I:1 for testing. Sequencing was performed on the Illumina HiSeq6000 platform, achieving an average depth of 100×.

Initial data processing involved the removal of adapter sequences using cutadapt (v1.15) ¹⁶ and aligning sequences to the human reference genome (UCSC hg19) with BWA (v0.7.15). ¹⁷ Picard tools (v2.4.1) were used to identify duplicate reads, and Qualimap (v2.2.1) ¹⁸ assessed quality control metrics, including base quality, genome mapping efficiency, and target region coverage. Further data refinement included base quality recalibration, indel realignment, and variant identification (single nucleotide variants and InDels) using GATK (v3.8) ¹⁹ , ²⁰ following best practice guidelines. Variant filtration was performed using a customized in‐house algorithm, with subsequent annotation through Pubvar and VEP (release 88). ²¹

As for Family 4, exome analysis was done in 2025 strictly following manufacturer's instructions and standard operating procedures. Library preparation was done using Nextera DNA Flex Pre‐Enrichment LibraryPrep with Illumina Nextera DNA UD Indexes by Illumina. Target enrichment was achieved by using Human Core Exome hybridization probes from Twist Bioscience. Paired‐end next generation sequencing was then performed on a NovaSeq 6000 Instrument using an S1 Reagent Kit (300 cycles) by Illumina. Analysis of the raw data was performed using the software Varfeed (Limbus), and the variants were annotated and prioritized using the software Varvis (Limbus).

2.2. Western blot

Forty‐eight hours after channel RNA injection, proteins (40 mg/sample) were extracted from the oocytes and quantified according to the standard protocol of the Cell Membrane and Cytosol Protein Extraction Kit and Enhanced BCA Protein Assay Kit (P0033 and P0010s, Beyotime) and transferred onto polyvinylidene difluoride membranes. The membranes were incubated at −4°C overnight with the primary rabbit antibody for KCNJ2 (Proteintech 19 965‐1‐AP, 1:1000) and KCNJ4 (Abmart PS18310S, 1:1000), respectively. Subsequently, the membranes were washed and incubated with horseradish peroxidase (HRP)‐conjugated antirabbit IgG antibody (1:2000, A0216/A0208, Beyotime). Also, the membranes were stripped and incubated with anti‐Na‐K‐adenosine triphosphatase (ATPase) antibody (Abbkine, ABL1141, 1:1000) following incubation with HRP‐conjugated antirabbit IgG antibody (1:2000, A0216/A0208, Beyotime). The chemiluminescent detection was performed using the Bio‐Rad Laboratories Quantity One system.

2.3. Bioinformatic analysis

Genetic variants were evaluated under both dominant and recessive inheritance models. Variants were selected for validation if they met the following criteria ²² : (1) heterozygous variants with a minor allele frequency (MAF) < .1% in public databases, including the 1000 Genomes Project, NHLBI‐ESP, gnomAD, and ExAC; (2) homozygous or compound heterozygous variants with an MAF < 1% in the aforementioned databases; and (3) variants predicted to have functional consequences, such as stop‐gain/loss, frameshift, splice‐site alterations, or missense variants. Candidate variants were validated by Sanger sequencing in probands and their biological parents.

The three‐dimensional protein structure model was generated using AlphaFold2. ²³ Analysis and visualization of amino acid residue interactions were performed with PyMOL (v3.1.3.1). Sequences of various species and Kir2.3 proteins were retrieved from the NCBI database, and multiple sequence alignment was conducted using ClustalW in MEGA11 with default parameters.

2.4. Electrophysiology

cRNAs for human potassium channels KCNJ2 and KCNJ4 (including the mutants Gly136Ser, Val206Met, Met293Lys, and Glu384Lys in KCNJ4) were synthesized in vitro using T7 RNA polymerase from constructs cloned into the pGEMSh vector. All plasmids were verified by sequencing (Genewiz). For expression, cRNAs were diluted to 200 ng/μL, and a 1:1 (wt/wt) mixture of KCNJ4 (or mutant) and KCNJ2 cRNA was microinjected into Xenopus laevis oocytes. Two‐electrode voltage clamp (TEVC) recordings were performed 30 h postinjection at room temperature (20°C–25°C) using an OC‐725C amplifier and WinWCP software for data acquisition. The standard extracellular bath solution contained (in millimolars): 96 KCl, 1 NaCl, 1 MgCl₂, 1.8 CaCl₂, and 5 hydroxyethylpiperazine ethane sulfonic acid (pH adjusted to 7.4 with NaOH). Microelectrodes were filled with 3 mol·L⁻¹ KCl containing 1.5% agarose. After confirming functional expression with a voltage ramp from −100 to +100 mV, currents were elicited by 70‐ms test pulses from −150 to +50 mV in 10‐mV increments. Data were analyzed using pCLAMP software to generate current–voltage (I–V) curves. Electrophysiological recordings were performed in two independent experimental batches on different days, using separately prepared oocytes and cRNA injections. Data from both batches were pooled for analysis. To account for variability in expression levels, macroscopic current amplitudes were normalized to relative protein expression (Figure S1C,D), and normalized I–V relationships were used for comparative analyses.

2.5. Statistics

Statistical analyses were conducted using Prism 8 (GraphPad Software). An unpaired t‐test was employed for comparisons between two variables.

3. RESULTS

3.1. Identification of heterozygous variants in KCNJ4 in four unrelated children with epilepsy

3.1.1. Family 1

The proband, a 13‐year‐old Chinese girl, was the first child born to nonconsanguineous parents with severe epilepsy. Born at full term without perinatal complications, she achieved normal developmental milestones during early childhood. At 12 years 3 months of age, she began experiencing recurrent epileptic episodes, predominantly tonic–clonic seizures occurring during noontime sleep. These episodes were characterized by limb rigidity, tremors, foaming at the mouth, upward eye deviation, and trismus, lasting approximately 5 min and resolving spontaneously, followed by postictal confusion and drowsiness. During the interictal period, no neurological deficits were observed. Despite oxcarbazepine therapy, seizure frequency escalated markedly from once monthly to more than 10 episodes per month, peaking at three daily events. Notably, the proband's mother had a history of epilepsy onset at age 5 years (with undocumented seizure semiology due to retrospective reporting) and died at age 30 years due to status epilepticus, although cognitive function remained intact throughout her life.

Physical examination showed normal anthropometric parameters (height 158 cm, weight 45 kg, head circumference 54 cm) and intact cognitive and motor functions, without dysmorphic features, hypertonia, or pyramidal signs. Routine laboratory tests, such as complete blood count, renal function, and electrolyte panels, were unremarkable. Electrocardiogram and echocardiogram revealed no structural or functional abnormalities. EEG at age 13 years, she demonstrated right frontotemporal‐predominant epileptiform activity, including synchronous and asynchronous spikes, sharp waves, polyspike–wave complexes, and slow waves. Brain magnetic resonance imaging (MRI) showed no structural abnormalities except for an incidental right temporo‐occipital arachnoid cyst. Trio WES revealed a novel heterozygous missense variant in the KCNJ4 gene (NM_004981.2:c.616G>A:p.Val206Met; Table 1). This variant is absent from public databases, including gnomAD, ExAC, and the 1000 Genomes Project (Figure 1A and Table S1).

TABLE 1.

Clinical features of four patients with KCNJ4 variants (NM_004981.2).

Patients	Family 1 (III:1)	Family 2 (II:1)	Family 3 (II:1)	Family 4
Sex	Female	Male	Male	Female
HGVS	c.616G>A: p.(Val206Met)	c.878T>A: p.(Met293Lys)	c.1150G>A: p.(Glu384Lys)	c.406G>A: p.(Gly136Ser)
Age at presentation	13 years	8 months	6 years	59 years
Family history	Yes	No	No	Mother with alcohol abuse, parents deceased
Height/length, cm	158	73	110	155
Weight, kg	45	8.5	18.5	52.3
OFC, cm	54	44	50
DD/ID	No	Yes	Yes	Yes
Speech problems	No	Nonspecific vocalizations	Nonspecific vocalizations	Nonspecific vocalizations
Motor problems	No	Minimal gross motor function	Minimal gross motor function	Minimal gross motor function
Epilepsy	Limb rigidity, tremors, foaming at the mouth, upward eye deviation, and trismus, lasting approximately 5 min and resolving spontaneously	Generalized tonic seizures, limb convulsions, and repetitive eye blinking	Episodes of limb rigidity, tremors, occasional rightward eye deviation, and cyanosis	Focal epilepsy with myoclonic–atonic seizures and tonic seizures with falls
Age of first symptom	12 years	4 months	Neonatal period	No information available
Average monthly frequency	More than 10	Almost daily occurrences	–	Multiple per year
Postictal confusion	Yes	Yes	Yes	No
Muscle tone	Normal	Increased upper limb muscle tone	–	Normal
Abnormal behavior	No	No	No	Yes
Other neuromuscular features	No	No	No	No
EEG	Synchronous and asynchronous spikes, sharp waves, polyspike waves, and slow waves, predominantly affecting the right frontal, temporal, and sphenoidal electrodes, with the most prominent activity in the temporal regions	Posterior‐predominant atypical hypsarrhythmia with intermittent low‐amplitude paroxysmal bursts	–	Left frontotemporal cerebral dysfunction with rare interictal epileptiform discharges (maximum T1)
Imaging results	Right temporo‐occipital arachnoid cysts	Thin corpus callosum	Bilateral lateral and third ventricular dilatation, reduced basal ganglia volume, increased thalamic density, and decreased cerebral white matter volume	Cerebral white matter atrophy
Additional symptoms	No	No	No	Microcephaly, dysphagia, low‐set ears, narrowed philtrum, narrow red lips

Abbreviations: DD, developmental delay; EEG, electroencephalography; HGVS, Human Genome Variation Society; ID, intellectual disability; OFC, occipitofrontal circumference.

FIGURE 1.

Pedigrees, genetic variants, and structural characterization of KCNJ4 variants. (A–C) Pedigrees of three families (Families 1–3) with corresponding Sanger sequencing chromatograms illustrating heterozygous KCNJ4 variants. In Family 1 (A), the shaded symbol (I:2) represents a symptomatic individual for whom genetic data are unavailable. Arrows indicate probands. (D) Multiple sequence alignment highlighting the evolutionary conservation of the affected amino acids across species. Mutated residues are marked in red for emphasis. (E) Schematic representation of the KCNJ4 protein's secondary structure. (F) Tructural mapping of KCNJ4 variants onto the predicted Kri2.3 channel model.

3.1.2. Family 2

The proband was an 8‐month‐old Chinese male infant born to nonconsanguineous parents with no family history of epilepsy. The mother had gestational hypertension and diabetes mellitus during pregnancy, with prenatal monitoring revealing transient fetal arrhythmia. Delivery at 38 weeks 3 days gestation was uncomplicated, with normal anthropometric parameters (birth weight 3.28 kg, length 52 cm, head circumference 33.6 cm). Developmental milestones were delayed, with head control achieved only at 3 months of age. At 4 months of age, the patient developed recurrent seizures characterized by eye closure, fist clenching, and synchronous nodding with embracing movements, followed by developmental regression including loss of acquired head control. Antiepileptic therapy with valproic acid proved ineffective, with daily seizures and multiple forms, including generalized tonic seizures, limb convulsions, and repetitive eye blinking. Each episode lasted approximately 3 min, resolving spontaneously, although occasional episodes exceeding 10 min required diazepam for termination. Combination therapy with levetiracetam, vigabatrin, and valproate subsequently reduced seizure frequency.

Current examination at 8 months revealed appropriate growth parameters (weight 8.5 kg, length 73 cm, head circumference 44 cm) but significant neurodevelopmental impairment. The patient demonstrated minimal gross motor function, only producing nonspecific vocalizations, and showed responsiveness exclusively to loud auditory stimuli. Neurological examination revealed increased upper limb muscle tone. Electrocardiography showed no abnormalities. EEG demonstrated posterior‐predominant atypical hypsarrhythmia with intermittent low‐amplitude paroxysmal bursts. Brain MRI revealed a thin corpus callosum. Trio WES discovered a de novo heterozygous missense variant in the KCNJ4 gene (NM_004981.2:c.878T>A:p.Met293Lys), which has not been documented in any public databases (Figure 1B and Table S1).

3.1.3. Family 3

The proband, a 6‐year‐old Chinese boy, was admitted to the Pediatric Intensive Care Unit due to intractable seizures and severe cough, with unremarkable family history for genetic or congenital disorders. He was delivered vaginally at 32 weeks of gestation with a birth weight of 3 kg (normal range), requiring intubation for resuscitation shortly after birth. Episodes of limb rigidity, tremors, occasional rightward eye deviation, and cyanosis were notable in the neonatal period. Although the epilepsy showed partial response to levetiracetam and clonazepam treatment, his coughing symptoms progressively worsened. The patient ultimately developed respiratory failure secondary to severe pneumonia, and care was withdrawn at parental request, resulting in death at home.

The proband exhibited profound global developmental delays, including the inability to sit independently, walk, or feed himself. He also lacked language communication. His facial features appeared normal. Cranial computed tomography (CT) revealed bilateral lateral and third ventricular dilatation, reduced basal ganglia volume, increased thalamic density, and decreased cerebral white matter volume. Trio WES identified a missense heterozygous variant c.1150G>A (p.Glu384Lys) occurring de novo in the KCNJ4 gene that has not been previously reported, which was confirmed by Sanger sequencing (Figure 1C and Table S1).

3.1.4. Family 4

The proband was a 59‐year‐old female; her prenatal history was not available, but suspicion of fetal alcohol syndrome was raised given the maternal history of alcohol abuse. Early developmental delay was significantly delayed; at the age of 3 years, her developmental level corresponded to less than a 1‐year‐old child. Severe intellectual disability was confirmed in subsequent assessments (a formal intelligence quotient test was not performed), with no receptive language and only occasional vocalizations or single words. She was placed in institutional care at the age of 2 years and has since required 24‐h supervision.

At 43 years of age, she was hospitalized for recurrent seizures and frequent falls. Clinical examination revealed microcephaly, severe intellectual disability, minimal expressive speech, and mild gait ataxia. Muscle tone was normal, but behavioral abnormalities such as self‐injury and psychomotor agitation were present. Brain CT showed symmetric cerebellar atrophy and supratentorial ventricular enlargement. Seizures were characterized as myoclonic–atonic and tonic seizures, occurring multiple times per year with frequent falls and injuries. Valproate therapy led to partial reduction of seizure frequency.

At 52 years, she was admitted after a fall with dental fractures. Her neurological status had worsened, with a broad‐based unsteady gait, increased urinary incontinence, hypersalivation, and reduced nutritional intake. Clinical examination demonstrated mild generalized hypotonia, severe impairment of speech (limited to vocalizations), and scoliosis with right trunk deviation. EEG revealed left frontotemporal cerebral dysfunction with rare interictal epileptiform discharges (maximum T1). Electroneurography confirmed an axonal‐demyelinating lesion of the right radial nerve due to reduced arm use.

At 59 years, her condition was marked by profound intellectual disability, atypical autism, behavioral disturbances, and further decline in mobility with fatigue and gait worsening. Dysmorphic features included low‐set ears, smooth philtrum, and thin vermilion border. She remained nonverbal, limited to vocalizations. MRI revealed cerebral atrophy with gliotic changes in the white matter, and repeated CT confirmed global atrophy. Until currently, the patient was treated with phenytoin, phenobarbital, valproate, lamotrigine, carbamazepine, and clobazam. Under clobazam her gait worsened, and lamotrigine led to an increase of falls. With a combination of valproate and levetiracetam, the frequency of her myoclonic–atonic and tonic seizures lessened, but an encephalopathy was suspected. After the therapy with valproate was terminated, there was a significant increase in seizure frequency, so a good effect of valproate in the proband was proven. Genetic investigations, including cytogenetics and exome sequencing, identified no pathogenic variants in known disease‐associated genes. Targeted analysis of the RNU4‐2 gene was unremarkable. Research analysis revealed a heterozygous variant in KCNJ4 (NM_152868.3:c.406G>A, p.Gly136Ser).

3.2. Identification of functional KCNJ4 variants effects on Kir channel activity

Coexpression of KCNJ4 (wild‐type or mutant) with KCNJ2 in Xenopus oocytes resulted in inwardly rectifying potassium currents upon membrane hyperpolarization. These currents exhibited a characteristic weak inward rectification, with the rectification property being most pronounced at depolarizing potentials between approximately +60 and +80 mV (Figure 2A,B). The rectification effect weakened at more positive voltages. Robust currents were successfully recorded for the wild‐type KCNJ4/KCNJ2 channel (n = 7) and four mutants, Gly136Ser/KCNJ2 (n = 8), Val206Met/KCNJ2 (n = 8), Met293Lys/KCNJ2 (n = 8), and Glu384Lys/KCNJ2 (n = 8). To address the concern that differences in current amplitudes might arise from variable expression levels, we performed Western blot analysis of membrane proteins from Xenopus oocytes coexpressing Kir2.1 with either wild‐type or mutant Kir2.3 at equal ratios. Quantification revealed that the membrane expression levels of both Kir2.1 and Kir2.3 were comparable among all groups, with only minor variations. Current amplitudes were normalized to relative protein expression levels prior to quantitative comparison; Na⁺/K⁺‐ATPase was used as the internal control (Figure S1C,D). These data indicate that the observed differences in current amplitudes are not primarily attributable to altered channel expression or trafficking (Figure 2I). The current amplitudes of both the Gly136Ser and Glu384Lys mutants were significantly larger than those of the wild‐type channel (p = .0007 and p = .003, respectively; Figure 2I). Conversely, the Val206Met and Met293Lys mutants displayed significantly smaller inwardly rectifying currents compared to the wild‐type (p = .001 and p = .003, respectively; Figure 2I).

FIGURE 2.

Voltage protocols and representative current traces for wild‐type (WT) and mutant KCNJ4/KCNJ2 channels expressed in Xenopus oocytes. (A) Voltage protocol for eliciting inward‐rectifier currents, with membrane potential ramped from −100 to 100 mV. (B) Voltage step protocol, stepping membrane potential to −150 mV. (C) Representative current trace of WT KCNJ4/KCNJ2 channel, showing inward‐rectifier current upon hyperpolarization. (D) Family of current traces for WT KCNJ4/KCNJ2 channel in response to a series of voltage commands. (E) Representative current trace of Gly136Ser/KCNJ2 mutant channel, with larger current amplitude compared to WT. (F) Family of current traces for Gly136Ser/KCNJ2 mutant. (G) Representative current trace of Val206Met/KCNJ2 mutant channel, exhibiting smaller current amplitude than WT. (H) Family of current traces for Val206Met/KCNJ2 mutant. (I) Current–voltage relationships for WT and four KCNJ4 mutants (Gly136Ser, Val206Met, Met293Lys, Glu384Lys) coexpressed with KCNJ2, illustrating differences in current amplitudes and rectification properties. Data are presented as mean ± SEM (n = 7 for WT; n = 8 for each mutant). Current amplitudes are expressed in microamperes (µA) and presented as normalized values.

4. DISCUSSION

In this study, we report for the first time the identification of heterozygous KCNJ4 variants as a cause of severe epilepsy with global developmental delay/intellectual disability. Our functional analyses revealed a dual pathogenic mechanism; the Gly136Ser and Glu384Lys variants significantly increased potassium current amplitudes, suggesting a GOF effect, whereas the Val206Met and Met293Lys variants markedly reduced inwardly rectifying currents, indicative of an LOF mechanism. This bidirectional dysregulation of channel activity, encompassing both GOF and LOF, underscores the critical and complex role of Kir channel function in the pathogenesis of epilepsy.

Kir2.3 typically forms homotetramers or heterotetramers with other Kir2.x subunits to function properly. Its activity is influenced by factors such as pH, phosphorylation, and other posttranslational modifications, as well as proteostasis processes like endoplasmic reticulum–Golgi trafficking, membrane targeting, and endocytic recycling. ²⁴ , ²⁵ The C‐terminus of Kir2.3 contains signals that direct the channel to the basolateral membrane during the biosynthetic pathway and includes important regulatory sites, such as the PDZ‐binding motif. ²⁶ Additionally, the CO₂/pH sensitivities of Kir2.3 channels are abolished or compromised when alterations occur in the N and C termini. ²⁷ , ²⁸

KCNJ4 has previously been proposed as an epilepsy‐associated gene. This is supported by the observed downregulation of Kir2.3 expression in the hippocampal tissue of patients with chronic temporal lobe epilepsy ¹⁴ and the identification of the Kir2.3‐specific agonist tenidap as a potential antiepileptic therapy. ¹⁵ These findings are consistent with electrophysiological data showing that the Val206Met and Met293Lys variants lead to heteromeric Kir channels with markedly reduced current amplitudes, indicating the formation of nonfunctional channels.

Despite sharing a common electrophysiological defect, the two variants resulted in distinct clinical manifestations. The Val206Met variant was associated with isolated epilepsy, featuring later onset and no significant intellectual impairment. In contrast, the Met293Lys variant was linked to refractory epilepsy accompanied by global developmental delay. The difference in cognitive outcomes may be partly explained by the later onset of symptoms in the Val206Met case. Furthermore, ion channelopathies frequently exhibit considerable clinical heterogeneity, where even the same variant can lead to diverse phenotypes across individuals. ²⁹

The patients from Family 3 and Family 4 presented with refractory epilepsy and global developmental delay/intellectual disability. Electrophysiological analysis revealed that the Glu384Lys and Gly136Ser variants identified in these patients exhibited enhanced inward rectification compared to the wild‐type channel, which may alter the neuronal resting membrane potential. This disruption is likely to affect neural network dynamics—for example, by reducing the activity of intermediate inhibitory neurons—thereby contributing to dysregulated neuronal signaling. ³⁰ , ³¹ In support of this mechanism, studies have reported increased inwardly rectifying potassium currents with properties of Kir2 channels in epileptic granule cells, along with a substantial upregulation of Kir2 channel expression, suggesting that Kir2 overexpression underlies the enhanced Kir currents observed in these cells. ³² These findings suggest that GOF mutations leading to dysregulation of neural network activity may represent another mechanism by which KCNJ4 mutations contribute to epilepsy. Such cases, in which both LOF and GOF mutations result in cellular electrophysiological disturbances and subsequent disease, are not uncommon among genes encoding Kir channels. For example, GOF mutations in KCNJ3 are associated with bradycardia and atrial fibrillation, ³³ whereas LOF mutations in the same gene have been linked to increased neuronal excitability and epileptogenesis. ³⁴ Similarly, mutations in KCNJ2 can cause life‐threatening ventricular arrhythmias and sudden cardiac death; LOF mutations lead to Andersen–Tawil syndrome, whereas GOF mutations are responsible for short QT syndrome 3. ³⁵ Notably, when current alterations were considered in the context of reduced protein expression, the disproportionate increase in current amplitude observed for GOF variants supports a primary channel‐intrinsic functional effect rather than an artifact of altered expression levels. At the same time, the limited number of affected individuals precludes definitive genotype–phenotype correlations, highlighting the intrinsic clinical heterogeneity of KCNJ4‐related epilepsy.

We acknowledge several limitations in our experimental approach. We also acknowledge that prenatal history was unavailable for the proband in Family 4 and that long‐term longitudinal follow‐up data, including dynamic treatment responses, were incomplete for some individuals due to the retrospective nature of data collection. Despite these limitations, the consistency of genetic findings and functional validation support the pathogenic relevance of the identified variants. Specifically, Kir2.1 and Kir2.3 RNAs were injected into Xenopus oocytes at a 1:1 mass ratio, which does not reflect the physiological condition in neurons, where Kir2.1 is the predominantly expressed subunit. Previous studies have shown that the incorporation of Kir2.3 subunits into heterotetrameric channels can alter their sensitivity to protons. ²⁷ This suggests that our current experimental system may not fully replicate the complex subunit interactions or functional consequences of these variants in a native neuronal context. Genotype–phenotype correlations, however, remain challenging to establish in KCNJ4‐related disorders. Given the limited number of patients, the GOF or LOF classification should be interpreted as a functional description at the channel level rather than a strict determinant of clinical phenotype. In this context, our TEVC data provide a reliable and internally consistent functional readout, and the observed electrophysiological changes are biologically coherent with Kir2.3 channel properties. Future studies incorporating patient‐derived neuronal models or in vivo systems will help clarify how GOF and LOF alterations translate into distinct clinical outcomes.

In conclusion, our study identifies KCNJ4 as a novel gene linked to a broad spectrum of epilepsy disorders, underscoring the critical role of Kir channels in epileptogenesis and their potential as targets for therapeutic interventions. Our findings suggest that further identification and analysis of affected individuals will be essential to further define the molecular and phenotypic spectrum of this disorder.

AUTHOR CONTRIBUTIONS

Hu Pan: Writing—original draft; investigation; funding acquisition; data curation. Deng Liu: Writing—review & editing; resources; data curation. Wuhen Xu: Writing—review & editing; data curation. Yang Li: Writing—review & editing; supervision; funding acquisition. Juanli Hu: Methodology; writing—review & editing. Mei He: Methodology; data curation. Yuanlu Chen: Investigation; formal analysis. Shengnan Wu: Writing—review & editing. Xiaohuan Dong: Writing—review & editing. Zhen Liu: Methodology; formal analysis. Yiqiao Chen: Methodology. Pengchao Wang: Methodology. Oppermann Henry: Editing. Alexander Fuchs: Editing. Rami Abou Jamra: Data curation. Weiyue Gu: Validation; supervision. Han Jing: Writing—review & editing; methodology. Xiao Mao: Funding acquisition. Ya‐Jie Wang: Writing—review & editing; supervision; resources. Yabing Tang: Writing—review & editing; supervision; methodology. Neng Xiao: Writing—original draft & editing; project administration; visualization; funding acquisition.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no conflict of interest. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Supporting information

FIGURE S1 Electrophysiological recording protocols and membrane expression of Kir2.1 and Kir2.3 channels. (A) Representative voltage step protocol (I–V step) and corresponding whole‐cell current traces recorded from Xenopus oocytes expressing Kir2.1 together with wild‐type (WT) or mutant Kir2.3. The inset shows an expanded view of the current traces. Scale bars: 50 μA, 10 ms (inset, 5 ms). (B) Voltage ramp protocol and representative current responses used to evaluate the current–voltage (I–V) relationship and rectification properties of Kir2.1/Kir2.3 channel complexes. Scale bars: 20 μA, 20 ms. (C) Western blot analysis of membrane proteins from Xenopus oocytes coexpressing Kir2.1 with WT or mutant Kir2.3 (Gly136Ser, Val206Met, Met293Lys, and Glu384Lys). Kir2.1 and Kir2.3 were detected using specific antibodies, and Na⁺/K⁺‐adenosine triphosphatase (ATPase) was used as a membrane protein loading control. (D) Quantification of Kir2.3 (upper panel) and Kir2.1 (lower panel) membrane expression levels normalized to Na⁺/K⁺‐ATPase. Data are presented as mean ± SEM from independent experiments. No marked differences in membrane expression levels were observed among WT and mutant groups, indicating that the electrophysiological differences are not primarily due to altered membrane expression.

TABLE S1 Summary of KCNJ4 variants identified in the study: genomic coordinates, protein consequences, in silico functional predictions, and population allele frequencies.

DATA AVAILABILITY STATEMENT

The datasets used and/or analyzed in the current study are available from the corresponding author (N.X.) on reasonable request.