This study received the ethical approval by the Institutional Review Board of the Fondazione IRCCS Istituto Neurologico Carlo Besta of Milan, Italy. Written informed consent for genetic analysis and use of anonymized clinical data for research purposes was obtained from all patients or guardians of participants. This research was performed in accordance with GCP and the ethical standards laid down in the 1964 Declaration of Helsinki.

In this retrospective observational study, we selected patients with pediatric onset drug-resistant epilepsies caused by variants in potassium channel-encoding genes, consecutively recruited from a large cohort subjected to NGS genetic screening of target genes for epilepsy at the Fondazione IRCCS Istituto Neurologico Carlo Besta of Milan, Italy (Castellotti et al., 2024). Diagnosis of epilepsy was established according to the most recent criteria from the International League Against Epilepsy (ILAE) (Scheffer et al., 2017; Wirrell et al., 2022).

Following acquisition of study consent, genomic DNA was extracted from peripheral blood lymphocytes, according to standard procedures. Genetic analysis was assessed with a NGS approach as previously reported (Castellotti et al., 2024). Briefly, we designed target panels using Illumina design studio software¹ and Agilent’s Sure Design tool.² All the panels the following potassium channel coding genes: KCNA1, KCNA2, KCNB1, KCNC1, KCND2, KCNE2, KCNH1, KCNH5, KCNH8, KCNJ10, KCNJ6, KCNK4, KCNMA1, KCNQ2, KCNQ3, KCNQ5, KCNT1, KCNT2, KCNV2, KCTD7, KDM6A, KIF1A, KIF2A, KIF5A, KIF5B, and KMT2D.

According to ACMG criteria (Richards et al., 2015), the genetic variants of potassium channels passing the filtering process were classified as “pathogenic” (class V) or “likely pathogenic” (class IV); segregation analysis was performed by Sanger sequencing in both parents.

We consulted medical records from a pseudo-anonymized database (available exclusively to the study researchers) and extracted the following information: genetic data [gene mutated, pathogenic variant(s), characteristics of variant(s), inheritance, ACMG class, functional effect]; clinical aspects [gender, age at diagnosis (months), neurological examination, neurodevelopmental delay/intellectual disability, behavioral features, other clinical aspects, brain MRI, age at seizure onset, epileptic syndrome, anti-seizure medications (ASMs), drug resistance, status epilepticus, EEG organization/background, epileptic activity].

We also collected the following information regarding pharmacological treatment with in vitro selected compounds (IVSCs): ASMs at IVSC start, IVSC compound, age of IVSC start, frequency of seizures at IVSC start, IVSC dose (mg/kg), IVSC in use, side effects of IVSC, frequency of seizure during IVSC treatment, improvement of other features, ASMs modification during IVSC treatment.

Variants herein reported were engineered by QuikChange site-directed mutagenesis (Agilent Technologies, Milan, Italy), as previously described (Miceli et al., 2015). In particular, the Kv7.2 D563E variant (numbering is based on the longest Kv7.2 transcript isoform a; accession number NM 172107.2; 872 amino acids) was introduced in a human KCNQ2 cDNA at the codon corresponding to the D535 residue in transcript isoform c (accession number: NM 004518.4; 844 amino acids), while the K_Na1.1 R262Q variant was inserted in a plasmid containing the cDNA for a myc-DDK-tagged human isoform 2 (Q5JUK3–2) of KCNT1 (RC214820; Origene, Rockville, MD, USA).

Previously-studied variants in Kv7.2 [G310S (Soldovieri et al., 2020)], Kv7.3 [D542N and V359L (Ambrosino et al., 2018)], Kv3.1 [V425M (Ambrosino et al., 2023)], and K_Na1.1 [S937G (Mosca et al., 2024)] were engineered with a similar strategy in plasmids containing the cDNA sequence for the corresponding wild-type Kv channels.

Chinese Hamster Ovary (CHO) cells were grown in 100-mm plastic Petri dishes in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum (FBS), 2 mM L-glutamine, penicillin (50 U/mL) and streptomycin (50 μg/mL) in a humidified atmosphere at 37°C with 5% CO₂.

Channel subunits were expressed in CHO cells by transient transfection using Lipofectamine (Thermofisher, Milan, Italy; 3.6 μg cDNA). An additional plasmid encoding for enhanced green fluorescent protein (EGFP; Clontech, Palo Alto, CA, USA; 0.4 μg cDNA) was used to identify transfected cells.

Patch-clamp recordings in the whole-cell configuration from transiently-transfected CHO cells, as well as data processing and analysis, were performed at RT 24 h after transfection as reported (Soldovieri et al., 2007; Rizzo et al., 2016). Briefly, for Kv7 currents cells were held at −80 mV and then depolarized for 1.5 s from −80 to +40 mV using an incremental pulse of 10 mV, followed by an isopotential pulse at 0 mV of 0.8 s; for K_Na1.1 currents, cells were held at −80 mV and then depolarized for 0.6 s from −90 to +60 mV using an incremental pulse of 10 mV.

Current values recorded at the beginning of the 0 mV isopotential step (to record Kv7 currents) or at the end of the incremental pulse (to record K_Na1.1 currents) were measured, normalized, and expressed as a function of each test voltage. The data were then fit to a Boltzmann distribution of the following form: y = max/[1 + exp(V_½ − V)/k], where V is the test potential, V_½ is the half-activation potential, and k is the slope factor.

To analyze activation kinetics, current traces recorded in response to incremental voltage steps were fitted to a single- or a double-exponential function of the following forms: y = amp exp(−t/τ) or y = amp_FAST exp(−t/τ_FAST) + amp_SLOW exp(−t/τ_SLOW), respectively, where amp_FAST and amp_SLOW indicate the amplitude of the fast and slow exponential components, respectively, whereas τ_FAST and τ_SLOW indicate the time constants of these components. The relative contribution of the components in the total current was reported as amp_RATIO = amp_FAST/(amp_FAST + amp_SLOW).

Gabapentin (Sigma, Milan, Italy) was dissolved in distilled water, while fluoxetine (Sigma, Milan, Italy) in 0.01% DMSO. Both drugs were tested using ramp protocols from −80 to 0 mV in 3 s (for Kv7 currents) or from −90 to +60 mV in 1 s (for K_Na1.1 currents), and perfused using a fast solution exchange system (<1 s), including a cFlow 8 flow controller attached to a cF-8VS 8-valve switching apparatus (Cell MicroControls, Norfolk, VA, USA), as previously described (Ambrosino et al., 2014). Drug-induced effects were calculated upon quantification of the area under the curve (AUC) of the current trace before (control) or after 1–3 minutes after drug application.

Each data point shown in Figures or in the text is the mean ± SEM of at least five distinct determinations, obtained in five distinct cells. Statistically significant differences were evaluated with the Student’s t test or with the ANOVA followed by the Student–Newman–Keuls test, when multiple groups were compared, with significance considered when p < 0.05.

We identified six patients (five females) with drug-resistant epilepsy caused by pathogenic variants in genes coding for different potassium channels. Their genetic and clinical characteristics are summarized in Table 1.

In five patients (#1–4 and #6), the clinical picture was characterized by developmental and epileptic encephalopathy, with epilepsy beginning in the neonatal period or shortly thereafter. During the evolution of the disease, all patients developed moderate to severe intellectual disability with spastic tetraparesis, and experienced recurrent seizures refractory to several ASMs. The remaining case (#5) reported focal seizures with onset at the age of 3 years; the clinical course was characterized by recurrent focal seizures, refractory to ASMs, and evidence of mild intellectual disability, aggressiveness and mood disorder.

To investigate the functional properties of currents expressed by Kv channels incorporating the variants herein described, we performed electrophysiological in vitro experiments in mammalian cells transiently expressing distinct Kv channels.

Previous works showed that the Kv7.2 variant G310S (identified in patient #1), as well as those found in compound heterozygosity in Kv7.3 (V359L and D542N, identified in patient #3), prompt severe LoF effects, as these mutant subunits were non-functional (Soldovieri et al., 2020; Ambrosino et al., 2018; Mosca et al., 2022). Heteromeric Kv7.2/Kv7.3 channels mainly recapitulate the native M-current in adult neurons (Shah et al., 2002; Stewart et al., 2012); thus, the effects of these variants were also tested upon co-expression of mutant subunits with wild-type Kv7.2 and/or Kv7.3 subunits. The results obtained showed that, while Kv7.2 G310S subunits prompted dominant-negative effects on Kv7.2/Kv7.3 currents, haploinsufficiency was instead observed upon co-expression of either Kv7.3 variant with wild-type Kv7.2 subunits.

Similarly, currents from Kv7.2 channels incorporating the novel D535E variant identified in patient #2 were undistinguishable from background (Figures 1A,B). When mutant subunits were co-expressed with wild-type channels (transfection ratio 1:1, 1.8 + 1.8 μg), a significant decrease of maximal currents was measured when compared to cells expressing only wild-type subunits (transfection ratio Kv7.2 channels+empty vector 1:1, 1.8 + 1.8 μg): in fact, current densities measured at 0 mV were 36.5 ± 5.5 or 16.9 ± 2.2 pA/pF in cells expressing Kv7.2 or Kv7.2 + Kv7.2 D535E subunits, respectively (n = 11–20; p < 0.05), thus suggesting dominant-negative effects prompted by mutant on wild-type Kv7.2 subunits. Furthermore, when Kv7.2 D535E subunits were co-expressed in vitro with Kv7.2 and Kv7.3 subunits at a transfection ratio (Kv7.2 + Kv7.2 D535E + Kv7.3,0.5:0.5:1) reproducing the genetic balance of the proband who was heterozygous for the KCNQ2 D535E variant, maximal currents were significantly reduced when compared to those measured in cells expressing Kv7.2 + Kv7.3 subunits, an experimental condition reproducing the genetic balance of a healthy control (Kv7.2 + Kv7.3, 1:1; Figures 1A,B). These results confirm the LoF effects prompted by the D535E variant. Moreover, currents measured in cells expressing Kv7.2 + Kv7.2 D535E + Kv7.3 subunits were identical to those observed in cells expressing Kv7.2 + Kv7.3 subunits (0.5:1 transfection ratio; Figures 1A,B), suggesting that in this heteromeric configuration mutant subunits prompt only haploinsufficiency (not dominant negative) effects. Finally, mutant subunits appear to alter activation kinetics of Kv7.2 + Kv7.3 currents: in fact, at −30 mV (where the maximal differences between Kv7.2 + Kv7.3 and Kv7.2 + Kv7.2 D535E + Kv7.3 channels were measured), τ_FAST were 116.3 ± 9.5 or 193.7 ± 21.6 ms (n = 18–50; p < 0.05), τ_SLOW were 535.2 ± 75.6 or 855.0 ± 108.3 ms (n = 11–28; p < 0.05), and amp_RATIO were 0.97 ± 0.02 or 0.81 ± 0.04 (n = 14–31; p < 0.05), for Kv7.2 + Kv7.3 or Kv7.2 + Kv7.2 D535E + Kv7.3 channels, respectively. These results suggest that heteromeric Kv7.2 + Kv7.3 channels activate at a significant slower rate in the presence of mutant subunits, thus confirming variant-induced LoF effects.

These results are similar to those previously reported for Kv7.3 V359L and Kv7.3 D542N variants carried in compound heterozygosity in patient #3 (Ambrosino et al., 2018): while homomeric Kv7.3 V359L or Kv7.3 D542N channels were non-functional per se, currents recoded in cells co-expressing Kv7.2 + Kv7.3 V359L + Kv7.3 D542N were significantly reduced when compared to wild-type Kv7.2 + Kv7.3 subunits, confirming variant-induced LoF effects. Also in this case, currents measured in cells expressing Kv7.2 + Kv7.3 V359L + Kv7.3 D542N subunits (transfection ratio 1:0.5:0.5) were identical to those measured upon co-expression of wild-type Kv7.2 + Kv7.3 subunits (1:0.5 transfection ratio), thus suggesting only haploinsufficiency effects in the presence of these mutant subunits.

By contrast, the herein reported K_Na1.1 R262Q variant prompts a strong increase in maximal current density compared to wild-type channels (n = 15–34; p < 0.05; Figures 1C,D), as well as a leftward shift in their voltage-dependence of current activation (V_½ were 1.9 ± 2.3 mV or − 39.2 ± 2.5 mV for wild-type or mutant channels, respectively; n = 15–37; p < 0.05; k values were 21.3 ± 1.3 or 19.0 ± 1.8 mV/efold, respectively; n = 15–37; p > 0.05; Figure 1E). In addition, currents expressed by K_Na1.1 R262Q channels showed a significant increase in the % of instantaneously-activated component (I_INST) versus the total current (I_STEADY-STATE or I_SS) compared to wild-type channels (n = 15–35; p < 0.05; Figure 1F); by contrast, the time-dependent component of maximal currents appeared slower in K_Na1.1 R262Q mutant when compared to wild-type channels (τ values measured at +60 mV were 134.8 ± 6.6 ms or 226.9 ± 39.8 ms for wild-type or mutant channels, respectively; n = 12–28; p < 0.05). When mutant subunits were co-expressed with wild-type channels (transfection ratio 1:1), to reproduce the heterozygous state of the proband, voltage-dependence of current activation (V_½ was −11.5 ± 3.1 mV; n = 17; p < 0.05; Figure 1E) and the % of instantaneously-activated component (n = 16; p < 0.05 versus K_Na1.1 and K_Na1.1 R262Q homomeric channels; Figure 1F) resulted intermediate between homomeric wild-type and K_Na1.1 R262Q mutant channels; by contrast, current density at +60 mV (n = 16; p > 0.05 versus K_Na1.1 and K_Na1.1 R262Q homomeric channels; Figures 1C,D) and the slope of the voltage-dependence activation curve (k value was 23.6 ± 1.9 mV/efold; n = 17; p > 0.05 versus K_Na1.1 and K_Na1.1 R262Q homomeric channels) were not significantly changed. Finally, current activation kinetics of the time-dependent component of maximal currents resulted indistinguishable from wild-type channels (τ value at +60 mV was 116.7 ± 7.8, n = 16; p > 0.05 versus K_Na1.1; p < 0.05 versus K_Na1.1 R262Q; Figure 1C): altogether, these results suggest that R262Q variant induces mainly GoF effects on wild-type channels, both in the homomeric and in heteromeric configurations.

Similarly, both previously-reported variants identified in Kv3.1 (V425M; Ambrosino et al., 2023) or K_Na1.1 (S937G; Mosca et al., 2024) subunits prompted GoF effects when compared to the corresponding wild-type channels, in terms of increased current densities, leftwardly-shifted voltage-dependence of activation, absence of inactivation and/or slower deactivation kinetics. Qualitatively similar effects, although quantitatively smaller, were measured when wild-type and mutant subunits were simultaneously expressed to mimic the genetic balance of the proband (Ambrosino et al., 2023; Mosca et al., 2024).

To counteract the distinct functional alterations prompted by the investigated Kv variants and considering that all reported patients had drug-resistant epilepsy, we searched for alternative treatments by studying in vitro the pharmacological sensitivity of Kv channels incorporating each variant to distinct drugs already approved for clinical use which were able to revert variant-induced changes in channel function.

According to the LoF effects described above in the presence of Kv7.2 or Kv7.3 variants, we performed additional patch-clamp recordings upon exposure to the anti-epileptic drug gabapentin (GBP), recently shown as a potent in vitro Kv7 activator (Manville and Abbott, 2018) and as an effective pharmacological treatment in vivo of patient #1 carrying the LoF-causing Kv7.2 G310S variant (Soldovieri et al., 2020). In vitro exposure to 10 μM GBP prompted a similar current increase in total currents (measured as the integral of the entire current trace, AUC) in cells expressing either wild-type Kv7.2 + Kv7.3, Kv7.2 + Kv7.2 D535E + Kv7.3, or Kv7.2 + Kv7.3 V359L + Kv7.3 D542N (Figures 2A–C), thus suggesting this drug as a rationale pharmacological alternative for probands carrying each of these variants.

On the contrary, since Kv3.1 V425M and K_Na1.1 S937G variants both prompted GoF effects, a possible blocker of these currents was tested to counteract variant-induced functional alterations. As reported (Ambrosino et al., 2023; Mosca et al., 2024), the anti-depressant drug fluoxetine (FLX) showed the ability to block currents expressed by both Kv3.1 and K_Na1.1 channels (also when carrying the Kv3.1 V425M and K_Na1.1 S937G variants, respectively), with similar potency, thus suggesting the use of this specific drug in both probands carrying these variants.

Based on this evidence and according to GoF effects shown by K_Na1.1 channels carrying the novel R262Q variant, 10 μM FLX was tested, revealing the ability of this drug to reduce current AUCs in CHO cells expressing wild-type K_Na1.1 and K_Na1.1 R262Q channels (Figures 2D–F; n = 12–17; p > 0.05).

The presence of pathogenic variants in different potassium channel-encoding genes advocated the chance for targeted treatments. Different compounds were selected based on the patients’ specific gene variant causing distinct changes in channel function when studied in cellular models (Table 2). Based on the results of our functional studies, gabapentin was tested in three patients harboring LoF variants in KCNQ2 (patient #1 and #2) and KCNQ3 (patient #3), while fluoxetine was tested in those with GoF variants in KCNC1 (patient #4) and K_Na1.1 (patients #5 and #6).

We previously demonstrated that the exposure to the Kv7 activator gabapentin can potentiate channel function in the presence of the Kv7.2 G310S variant carried by patient #1 (Soldovieri et al., 2020). Gabapentin was introduced and titrated up to 28 mg/kg/daily, reaching a plasma level of 3.4 mg/L. This treatment resulted significantly and persistently effective. During the subsequent six years of follow-up, gabapentin was well tolerated (daily dose is currently 26 mg/kg, with a drug plasma level of 5.7 mg/L), seizures did not recur, and the girl significantly improved in interaction and communication. Moreover, the EEG showed a marked decrease of epileptiform discharges.

Based on the successful treatment of patient #1, together with in vitro evidence showing the ability of gabapentin in recovering the functional deficit due to the Kv7.2 D535E variant (Figures 2A–C), we also tested the effect of this compound in patient #2. Unfortunately, treatment with gabapentin up to 9.6 mg/kg/daily induced worsening of seizure frequency, and the drug had to be withdrawn. Carbamazepine, known to be effective in patients carrying Kv7.2 variants (Falsaperla et al., 2023; Dilena et al., 2022; Pisano et al., 2015; Sands et al., 2016), was than titrated up to 18 mg/kg/daily (plasma level 7 μg/mL), with full control of seizures. During the following three years of follow up, topiramate and clonazepam were withdrawn, no seizure recurred and we observed a mild improvement in attention and participation.

According to the in vitro results reported in (Figures 2A–C), gabapentin was also applied in patient #3, carrying the two LoF variants Kv7.3 V359L and Kv7.3 D542N. This drug, gradually increased up to 28 mg/kg/daily reaching plasma level 3.4 mg/L, led to the rapid disappearance of tonic spasms, and allowed withdrawing vigabatrin four months later. The EEG showed a marked decrease of epileptic activity, and no seizure recurred in the following three years of follow-up. Furthermore, postural control of head and trunk slightly improved, so that the child is now able to drive her own wheelchair.

On the other hand, for the treatment of patients #4, #5 and #6, carrying GoF variants in Kv3.1 or K_Na1.1 potassium channels, we tested in vivo the effect of the potassium channel-blocker fluoxetine.

In particular, patient #4 did not respond to numerous ASMs and developed drug-resistant epilepsy. In vitro incubation with fluoxetine of cells expressing the Kv3.1 V425M variant resulted in the reduction of the current produced by the mutated channels (Ambrosino et al., 2023). Treatment with this compound at 0.26 mg/kg (plasma level 171.3 ng/mL) resulted in the complete control of seizures, and allowed to suspend valproate. During the five and a half years of follow-up (current dose 0.24 mg/kg, plasma level 137 ng/mL), the patient did not experience seizures, and a noticeable improvement in behavioral aspects with greater participation in school activities and social interaction was observed. Moreover, serial EEG traces showed significant reduction of epileptic activity.

Fluoxetine was also used for the treatment of patient #5, carrying the K_Na1.1 S937G variant. This drug at the dosage of 1.2 mg/kg (plasma level 412.6 ng/mL) led to the complete control of epileptic seizures, along with significant improvement of the EEG pattern (Mosca et al., 2024). The patient has been under treatment with fluoxetine for over three years, no seizures have occurred during follow-up, and the patient has an active social life. However, despite the complete control of seizures, the girl refused to reduce the ASMs oxcarbazepine and clonazepam.

According to in vitro data demonstrating the ability of fluoxetine in recovering the functional defect prompted by the K_Na1.1 R262Q variant (Figures 2D–F), we treated with this drug also patient #6. Fluoxetine was gradually increased up to 0.8 mg/kg, reaching a plasma level of 447.5 ng/mL, leading to disappearance of seizures in wakefulness and their reduction during sleep. However, along with improvement of the epileptic symptoms, we observed a progressive worsening of behavior disorder, including marked agitation, akathisia, and the emergence of compulsions and mannerisms. Suspecting a possible side effect of fluoxetine, this was discontinued after ten months, leading to the regression of behavioral disorders, but with reappearance of high-frequency seizures during wakefulness.

The main limitation of this work derives from the small number of patients recruited. However, these are extremely rare conditions and patients come from a single center. Notably, we included in this study only cases with drug-resistant epilepsy caused by KCN variants who required therapeutic approaches selected on the basis of drug’s ability to counteract mutation-induced channel dysfunction in vitro, while we excluded those well controlled by conventional ASMs.