Epilepsy and neurobehavioral abnormalities in mice with a dominant-negative KCNB1 pathogenic variant.

Developmental and epileptic encephalopathies (DEE) are a group of severe epilepsies that usually present with intractable seizures, developmental delay, and often have elevated risk for premature mortality. Numerous genes have been identified as a monogenic cause of DEE, including KCNB1. The voltage-gated potassium channel KV2.1, encoded by KCNB1, is primarily responsible for delayed rectifier potassium currents that are important regulators of excitability in electrically excitable cells, including neurons. In addition to its canonical role as a voltage-gated potassium conductance, KV2.1 also serves a highly conserved structural function organizing endoplasmic reticulum-plasma membrane junctions clustered in the soma and proximal dendrites of neurons. The de novo pathogenic variant KCNB1-p.G379R was identified in an infant with epileptic spasms, and atonic, focal and tonic-clonic seizures that were refractory to treatment with standard antiepileptic drugs. Previous work demonstrated deficits in potassium conductance, but did not assess non-conducting functions. To determine if the G379R variant affected KV2.1 clustering at endoplasmic reticulum-plasma membrane junctions, KV2.1-G379R was expressed in HEK293T cells. KV2.1-G379R expression did not induce formation of endoplasmic reticulum-plasma membrane junctions, and co-expression of KV2.1-G379R with KV2.1-wild-type lowered induction of these structures relative to KV2.1-WT alone, consistent with a dominant negative effect. To model this variant in vivo, we introduced Kcnb1G379R into mice using CRISPR/Cas9 genome editing. We characterized neuronal expression, neurological and neurobehavioral phenotypes of Kcnb1G379R/+ (Kcnb1R/+) and Kcnb1G379R/G379R (Kcnb1R/R) mice. Immunohistochemistry studies on brains from Kcnb1+/+, Kcnb1R/+ and Kcnb1R/R mice revealed genotype-dependent differences in the expression levels of KV2.1 protein, as well as associated KV2.2 and AMIGO-1 proteins. Kcnb1R/+ and Kcnb1R/R mice displayed profound hyperactivity, repetitive behaviors, impulsivity and reduced anxiety. Spontaneous seizures were observed in Kcnb1R/R mice, as well as seizures induced by exposure to novel environments and/or handling. Both Kcnb1R/+ and Kcnb1R/R mutants were more susceptible to proconvulsant-induced seizures. In addition, both Kcnb1R/+ and Kcnb1R/R mice exhibited abnormal interictal EEG activity, including isolated spike and slow waves. Overall, the Kcnb1G379R mice recapitulate many features observed in individuals with DEE due to pathogenic variants in KCNB1. This new mouse model of KCNB1-associated DEE will be valuable for improving the understanding of the underlying pathophysiology and will provide a valuable tool for the development of therapies to treat this pharmacoresistant DEE.

Introduction

Developmental and epileptic encephalopathies (DEE) are a group of severe disorders that present early in life and are characterized by paroxysmal activity on EEG, multiple seizure types that are often medically intractable, and serious cognitive, behavioral and neurological deficits (Berg et al., 2010; Scheffer et al., 2017). In 30–50% of infants diagnosed with an epileptic encephalopathy, a causative mutation has been identified in a known epilepsy gene (McTague et al., 2016). Previous work has identified heterozygous de novo KCNB1 pathogenic variants in individuals with DEE (Torkamani et al., 2014; Saitsu et al., 2015; Thiffault et al., 2015; Allen et al., 2016; Calhoun et al., 2017; de Kovel et al., 2017; Latypova et al., 2017; Marini et al., 2017; Miao et al., 2017; Bar, 2020; Kang et al., 2019). In addition to seizures, individuals with KCNB1 variants display comorbidities that include developmental delay, intellectual disability, features of autism spectrum disorder, ADHD and aggression, as well as borderline long QT syndrome in some cases (Calhoun et al., 2017; de Kovel et al., 2017; Marini et al., 2017; Bar, 2020). Since the initial reports of KCNB1 variants in DEE type 26 (DEE26), the phenotype has expanded to include cases with less severe and/or later onset epilepsy or no discrete seizures, and thus the term KCNB1 encephalopathy encompasses this broader phenotype spectrum.

The KCNB1-p.G379R variant was identified in a male child with seizures commencing at 8 months of age (Torkamani et al., 2014). Glycine 379 is located in the Kv2.1 channel ion selectivity filter and is an evolutionarily invariant residue critical for potassium selectivity, supporting pathogenicity of the G379R missense variant (Fig. 1A) (Kuang et al., 2015). The epilepsy phenotype included infantile spasms, atonic, focal, and tonic-clonic seizures that were refractory to various antiepileptic therapies, including topiramate, valproic acid, pyridoxine, ACTH, and the ketogenic diet (Torkamani et al., 2014). At 15 months of age, hypsarrhythmia was present on EEG and later in childhood, monitoring revealed multifocal and diffuse polyspikes, polyspike and slow waves, and diffuse polyspike bursts (Torkamani et al., 2014). The individual also had severe motor and language delays, autism spectrum disorder, and strabismus (Torkamani et al., 2014; de Kovel et al., 2017; Srivastava et al., 2018).

Fig. 1.

Generation and molecular characterization of Kcnb1 Mice. A) Location of glycine-379 in the Kv2.1 tetrameric potassium channel subunit (PDB 29R9 Kv2.1/Kv1.2 chimera). G379 (red) is one of the critical ‘GYG’ residues that determines potassium selectivity and resides in the central pore where backbone carbonyls coordinate the K+ ions as they traverse the selectivity filter (black circle). B) Sequencing chromatograms of Kcnb1 genomic PCR products with the first nucleotide of the G379 codon highlighted in black. Top chromatogram was from a heterozygous Kcnb1 mouse (G379R) showing heterozygosity for the 3 nucleotides changes introduced by CRISPR/Cas9 editing and homology directed repair. Two nucleotide changes in codon 379 resulted in substitution of arginine for glycine, and a silent substitution in codon 377 disrupted an adjacent PAM site. Bottom chromatogram was from a heterozygous Kcnb1 mouse (G379Vfs) showing the 2 nucleotide deletion (yellow highlighted box) introduced by CRISPR/Cas9 editing and non-homologous end joining. This resulted in the frameshift G379VfsTer6. C) Multiplex ddPCR-based genotyping using hydrolysis probes for the WT and G379R alleles distinguishes between the three possible genotypes, WT, Kcnb1 and Kcnb1. Representative two-dimensional scatterplots of each genotype are shown along with a no template control (blank). Each scatterplot represents >10,000 partitioned reactions. Within a scatterplot, each point represents a droplet with a given fluorescence level and droplet colors indicate target amplification about threshold (magenta lines). Droplet colour code: green = positive for WT; blue = positive for G379R; orange = positive for both; black = negative.

KCNB1 encodes the Kv2.1 voltage-gated potassium channel subunit that is primarily responsible for delayed rectifier potassium current, an important regulator of neuronal excitability (Murakoshi and Trimmer, 1999; Guan et al., 2007; Liu and Bean, 2014; Honigsperger et al., 2017; Palacio et al., 2017). Kv2.1 is widely and robustly expressed throughout the nervous system. Within neurons, the majority of Kv2.1 is localized to high-density clusters on the soma, proximal dendrites and axon initial segment (Trimmer, 1991; Du et al., 1998; Sarmiere et al., 2008; Bishop et al., 2015). Kv2.1 clusters align with and are components of endoplasmic reticulum (ER) and plasma membrane (PM) junctions (ER-PM junctions), membrane microdomains important in calcium and lipid signaling and homeostasis (Du et al., 1998; Mandikian et al., 2014; Bishop et al., 2015; Henne et al., 2015; Gallo et al., 2016; Dickson, 2017; Wu et al., 2017; Bishop et al., 2018). Kv2.1 serves a structural role at ER-PM junctions independent of its potassium conducting function by binding to the ER proteins VAPA and VAPB, resulting in the recruitment of the ER to the PM (Fox et al., 2015; Johnson et al., 2018; Kirmiz et al., 2018). At these sites in brain neurons, Kv2.1 facilitates the spatial coupling of PM L-type calcium channels with ryanodine receptor ER calcium release channels, forming compartmentalized calcium signaling microdomains (Vierra et al., 2019). In addition to neuronal expression, Kv2.1 is also expressed in rodent heart within high-density clusters in atrial myocytes, as well as non-clustered Kv2.1 in ventricular transverse-axial tubules and sarcolemma (O’Connell et al., 2008), and in high density clusters on pancreatic β cells (Fu et al., 2017).

Several KCNB1 variants have been functionally characterized in vitro and shown to exhibit reduced Kv2.1 conductance, as well as altered ion selectivity, expression and localization (Torkamani et al., 2014; Saitsu et al., 2015; Thiffault et al., 2015; Kang et al., 2019). KV2.1-G379R channels exhibited loss of voltage-sensitivity and ion-selectivity, resulting in a non-specific cation leak current (Torkamani et al., 2014). When co-expressed with wild-type (WT) Kv2.1 subunits, Kv2.1-G379R exerted dominant-negative effects on potassium currents (Torkamani et al., 2014). However, the effect of Kv2.1-G379R on expression and subcellular distribution of Kv2.1 channels was not examined.

In the present study, we evaluated the effect of the DEE-associated KCNB1-p.G379R variant on Kv2.1 expression and clustering in HEK293T cells and observed a dominant negative effect on ER-PM clustering. To model this variant in vivo, we used CRISPR/Cas9 genome editing to generate a mouse model carrying the KCNB1-p.G379R pathogenic variant and characterized Kv2.1 expression and localization, as well as neurological, neurobehavioral, and cardiac phenotypes. Kcnb1G379R mice exhibited spontaneous and handling-induced seizures, abnormal EEG patterns, profound hyperactivity, inattention/impulsivity, and reduced anxiety-like behavior. The whole animal phenotype was more severe than that of a previously described global knockout allele (Speca et al., 2014). These mice recapitulate core features of KCNB1-associated DEE and will be a useful tool for understanding disease pathophysiology and evaluating potential therapies.

Materials and methods

HEK293T cell culture, immunocytochemistry, and epifluorescence and TIRF imaging

HEK293T cells were transfected with WT rat Kv2.1 in pRBG4, rat Kv2.1 S586A in pCGN, HA-tagged WT human Kv2.1, or HA-tagged WT human Kv2.1 with the G379R mutation (Shi et al., 1994; Lim et al., 2000; Kang et al., 2019). Cells were transiently transfected using Lipofectamine 2000 following the manufacturer’s protocol within 18 h of plating. For immunocytochemistry, fixation was performed as previously described (Dickson et al., 2016; Kirmiz et al., 2018). Briefly, HEK293T cells were fixed in 3.2% formaldehyde, blocked and permeabilized. Cells were incubated in anti-HA epitope tag IgG2a mouse monocloncal antibody (mAb) 12CA5 (RRID:AB_ 2,532,070, pure, 5 μg/mL) and monoclonal anti-VAPA antibody N479/24 (RRID:AB_2,722,709, tissue culture supernatant 1:5). Coverslips were immunolabeled with Alexa-conjugated fluorescent IgG subclass-specific secondary antibodies (ThermoFisher; 1:1500) and Hoechst 33258 dye (ThermoFisher H1399; 200 ng/mL).

Epifluorescence and TIRF imaging of fixed cells and image analysis was performed essentially as described (Bishop et al., 2018; Kirmiz et al., 2018; Vierra et al., 2019). Images were obtained with an Andor iXon EMCCD camera installed on a TIRF/widefield equipped Nikon Eclipse Ti microscope using a Nikon LUA4 laser launch with 405, 488, 561 and 647 nm lasers and a 100× PlanApo TIRF/1.49 NA objective run with NIS Elements software (Nikon). Images were collected within NIS Elements as ND2 images.

Mice

Kcnb1 (abbreviated Kcnb1) mice on the C57BL/6J inbred strain were generated using CRISPR/Cas9 to introduce the modification of glycine 379 (same codon number in human and mouse) by homology directed repair. Three nucleotide changes were introduced, including two nucleotide changes in codon 379 (GGT > CGC) to maintain similar codon usage, and a silent change to disrupt an adjacent PAM site (C > A) (Fig. 1B). A single guide RNA (TAGATGTCTCCGTAACCAA NGG) with good targeting efficiency in Neuro2A cells and low predicted off-target effects, and a 200 bp repair oligo (5ʹ-TCTCCAGCCTGGTCTTCTTTGCCGAGAAGGATGAGGATGACACCAAGTTCAAAAGCATCCCCGCCTCTTTCTGGTGGGCTACCATCACCATGACGACAGTTCGCTACGGAGACATCTACCCTAAGACTCTCCTGGGGAAAATCGTGGGGGGCCTCTGTTGCATTGCCGGTGTCCTGGTGATTGCCCTCCCCATTCCAATT) were micro-injected into C57BL/6J embryos by the Northwestern University Transgenic and Targeted Mutagenesis Laboratory.

Potential founders were screened by PCR of genomic DNA using primers outside the repair oligo homology region (Table 1), and the region of interest was TOPO-cloned and Sanger sequenced (n = 12–22 clones per founder). The mosaic G379R founder was backcrossed to C57BL/6J mice (Jackson Labs, #000664, Bar Harbor, ME) to generate N1 offspring. An additional founder heterozygous for a 2 bp deletion resulting in a frameshift (G379VfsTer6; abbreviated as Kcnb1) was also bred similarly (Fig. 1B). N1 offspring were genotyped by Sanger sequencing to confirm transmission of the G379R editing event or the 2 bp deletion, and were screened for off-target events by Sanger sequencing of all potential sites with <3 mismatches. N1 mice with the confirmed on-target event and without off-target events were bred with C57BL/6J females to establish the lines, Kcnb1 (official allele designation: Kcnb1) and Kcnb1 (2 bp deletion frameshift; official designation: Kcnb1), which are maintained by continual back-crossing of heterozygous mice with inbred C57BL/6J mice. Male and female Kcnb1 mice were intercrossed to generate Kcnb1+/+ (WT), heterozygous Kcnb1, and homozygous Kcnb1 mice for experiments. Male and female Kcnb1 mice were intercrossed to generated WT, Kcnb1, and Kcnb1 mice; or crossed to C57BL/6J mice to generate WT and heterozygous Kcnb1 mice for neurobehavioral experiments. For immunohistochemistry experiments, Kcnb1 knockout mice (Kcnb1) that we previously characterized were included for comparison (Speca et al., 2014).

Table 1

List of primers and probes.

Assay	Description	Sequence
Founder Screening PCR	Primer 1	5ʹ-TGGGCTTGCTCATCCTCTTC
	Primer 2	5ʹ-CTCCGCTTGATGGCTTTCTC
Genotyping Kcnb1G379R	Primer 1	5ʹ-GCATCCCCGCCTCTTTCT
	Primer 2	5ʹ-CAGGAGAGTCTTAGGGTAGATGTCT
	Probe 1 (WT allele)	5ʹ-VIC-CACCATGACGACCGTT-NFQ
	Probe 2 (mutant allele)	5ʹ-FAM-CACCATGACGACAGTT-NFQ
Genotyping Kcnb1fs	Primer 1	5ʹ-TCTTTGCCGAGAAGGATGAG
	Primer 2	5ʹ-CCAGGAGAGTCTTAGGGTAGAT
	Probe 1 (common allele)	5ʹ-VIC-TAGCCCACCAGAAAGAGGC-NFQ
	Probe 2 (WT allele)	5ʹ-FAM-ATGACGACCGTTGGTTACGG-NFQ

Abbreviations: FAM, FAM reporter dye; VIC, VIC reporter dye; NFQ, nonfluorescent quencher.

Mice were maintained in a Specific Pathogen Free (SPF) barrier facility with a 14-h light/10-h dark cycle and access to food and water ad libitum. Both female and male mice were used for all experiments. All animal care and experimental procedures were approved by the Northwestern University and UC Davis Animal Care and Use Committees in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The principles outlined in the ARRIVE (Animal Research: Reporting of in vivo Experiments) guideline and Basel declaration (including the 3R concept) were considered when planning experiments.

Genotyping

DNA was isolated from tail biopsies using the Gentra Puregene Mouse Tail Kit according to the manufacturer’s instructions (Qiagen, Valencia, CA). Genomic DNA was digested with BAMH1-HF (R3136, New England Biolabs, Ipswich, MA) at 37 °C for 1 h. Digested DNA was then diluted 1:1 with nuclease-free water and used for template for digital droplet PCR (ddPCR) using ddPCR Supermix for Probes (No dUTP) (Bio-Rad, Hercules, CA, USA) and custom TaqMan SNP Genotyping Assays (Life Technologies, Carlsbad, CA) to detect the mutations (Table 1). Reactions were partitioned into droplets in a QX200 droplet generator (Bio-Rad). PCR conditions were 95 °C for 10 min, then 44 cycles of 95 °C for 30 s and 60 °C for 1 min (ramp rate of 2 °C/s) and a final inactivation step of 98 °C for 5 min. Following amplification, droplets were analyzed with a QX200 droplet reader with QuantaSoft vl.6.6 software (Bio-Rad) (e.g., Fig. 1C).

Transcript analysis

Kcnb1 expression increases over the first 3 weeks of life and then stabilizes (Cardoso-Moreira et al., 2019). Whole brain RNA was extracted from 4 to 15-week-old WT, Kcnb1, Kcnb1, Kcnb1 and Kcnb1 mice. Total RNA was isolated using TRIzol reagent according to the manufacturer’s instructions (Invitrogen, Carlsbad, CA, USA). First-strand cDNA was synthesized from 2 to 4 micrograms of RNA using oligo(dT) primer and Superscript IV reverse transcriptase (RT) according to the manufacturer’s instructions (Life Technologies). First-strand cDNA samples were diluted 1:5 and 5 μL was used as template. Quantitative ddPCR was performed using ddPCR Supermix for Probes (No dUTP) (Bio-Rad) and TaqMan Gene Expression Assays (Life Technologies) for mouse Kcnb1 (FAM-MGB-Mm00492791_m1) and Tbp (VIC-MGB-Mm00446971_m1) as a normalization standard. Reactions were partitioned into a QX200 droplet generator (Bio-Rad). Thermocycling conditions and analysis were performed as described for genotyping. Both assays lacked detectable signal in genomic, no-RT and no template controls. Relative transcript levels were expressed as a ratio of Kcnb1 concentration to Tbp concentration, with 6–10 biological replicates per genotype. Statistical comparison between groups was made using ANOVA with Tukey’s post-hoc comparisons (GraphPad Prism, San Diego, CA).

Immunoblotting

Whole brain membrane proteins were isolated from 4 to 15-week-old WT, Kcnb1, Kcnb1, Kcnb1 and Kcnb1 mice. Membrane fractions (Kcnb1 samples for Kv2.1 blots: 50 μg/lane; Kcnb1 samples for Kv2.1 blots: 100 μg/lane; All samples for Kv2.2 and AMIGO-1 blots: 100 μg/lane) were separated on a 7.5% SDS-PAGE gel and transferred to nitrocellulose. Blots were probed with anti-Kv2.1 mAbs (Kcnb1 samples - NeuroMab K89/34, RRID:10673392, pure, 2 μg/mL; Kcnb1 samples - NeuroMab K39/25, RRID:2131649, pure, 10 μg/mL), anti-Kv2.2 mAb (NeuroMab N372B/1.1, RRID:2315869, tissue culture supernatant, diluted 1:2) or anti-AMIGO-1 mAb (NeuroMab L86/36.6, RRID: AB_2315799, tissue culture supernatant, diluted 1:2) and anti-mortalin/GRP75 mAb (NeuroMab N52A/42, RRID:2120479, pure, 1 μg/mL) which served as a normalization control. Alexa Fluor 790 and 680 goat anti-mouse antibodies (Jackson ImmunoResearch, 1:20,000) were used to detect signal on an Odyssey imaging system (Li-COR). Relative protein levels were determined by densitometry as previously described with ImageJ (NIH) (Rasband 1997-2018) or Image Studio (Li-COR) and expressed as a ratio of Kv2.1, Kv2.2 or AMIGO-1 to GRP75, with 5-7 biological replicates per genotype. Line scan analysis of Kv2.1 immunoblots was performed in ImageJ and expressed as the average ± SEM. Statistical comparison between groups was made using one-way ANOVA with Tukey’s post-hoc comparisons (GraphPad Prism).

Immunohistochemistry

P24-25 WT, Kcnb1 and Kcnb1 mice were used for immunohistochemistry, while Kcnb1 mice were not included in this analysis because epitopes for the immunolabeling-suitable Kv2.1 antibodies are all downstream of G379VfsTer6 frameshift. Animals were deeply anesthetized with pentobarbital and transcardially perfused with 4% formaldehyde prepared from paraformaldehyde, in 0.1 M sodium phosphate buffer pH 7.4 (0.1 M PB). Sagittal brain sections (30μm thick) were prepared and immunolabeled using free-floating methods as previously described (Rhodes et al., 2004; Speca et al., 2014; Palacio et al., 2017). Briefly, free floating sections were blocked and permeabilized. Sections were incubated with primary antibodies (see Table 2 for antibody details) and then immunolabeled with Alexa-conjugated fluorescent IgG subclass-specific secondary antibodies and Hoechst 33258 DNA stain at 200 ng/mL (ThermoFisher Cat# H21491). Images were taken using the same exposure time to compare the signal intensity directly. Images were identically processed in Photoshop to maintain consistency between samples. For further details, see supplemental materials and methods.

Table 2

List of antibodies.

Antibody (target)	Immunogen	Manufacturer information	Concentration/dilution used	Figures
12CA5 (HA epitope)	Influenza hemagglutinin HA epitope tag	Mouse IgG2b mAb, In-house (Trimmer Laboratory) RRID:AB_2532070	Pure, 5 μg/mL	2
N479/24 (VAPA)	Fusion protein aa 1–219 of rat VAPA	Mouse IgG2a mAb, In-house (Trimmer Laboratory) RRID:AB_2722709	Tissue culture supernatant, 1:5	2
N52A/42 GRP75/mortalin)	Identified as off-target mAb in screen for anti-SALM2 mAbs	Mouse IgG1 mAb, NeuroMab RRID:AB_2120479	Pure, 1 μg/mL	3, 4
K89/34 (Kv2.1)	Synthetic peptide aa 837–853 of rat Kv2.1	Mouse IgG1 mAb, NeuroMab RRID:AB_RRID:10673392	Pure, 2–5 μg/mL	3,4, S3
K39/25 (Kv2.1)	Synthetic peptide aa 211–220 of human Kv2.1	Mouse IgG2a mAb, NeuroMab RRID:AB:2131649	Pure, 10 μg/mL	3
L86/33 (AMIGO-1)	Fusion protein aa 394-492 of mouse AMIGO-1	Mouse IgG2a mAb, Neuro Mab RRID:AB_2315798	Tissue culture supernatant, 1:5	3,4
Kv2.2C (Kv2.2)	Fusion protein aa 717–907 of rat Kv2.2 long isoform	Rabbit pAb, In-house (Trimmer Laboratory), RRID:AB_2801484	Affinity purified, 1:100	3,4
ab32454 (MAP2)	Synthetic peptide within Rat MAP2 aa 1–100 (N terminal)	Rabbit pAb, Abeam RRID:AB_776174	Affinity purified 1:1000	S3
D3/71R (Kv2.1)	Fusion protein aa 506-533 of rat Kv2.1	Mouse IgG2a recombinant mAb, In-house (Trimmer Laboratory) RRID: AB_2750651	Tissue culture supernatant, 1:5	S4
L80/21 (Kv2.1)	Synthetic peptide aa 837-853 of rat Kv2.1	Mouse IgG3 mAb, Neuro Mab RRID:AB_2315862	Tissue culture supernatant, 1:5	S4
L105/31 (Kv2.1)	Synthetic peptide aa 596-616 of rat Kv2.1	Mouse IgG1 mAb, In-house (Trimmer Laboratory) RRID:AB_2801485	Tissue culture supernatant, 1:5	S4
L114/3 (Parvalbumin)	Fusion protein amino acids 1–110 of rat Parvalbumin	Mouse IgG2a mAb, NeuroMab RRID:AB_2629420	Tissue culture supernatant, 1:5	S4
L122/6 (Calretinin)	Fusion protein amino acids 1–178 of human Calretinin	Mouse IgG2b mAb, NeuroMab RRID:AB _2716256	Tissue culture supernatant, 1:5	S4

Hippocampal neuron culture and immunocytochemistry

Immunostaining was performed on cultured hippocampal neurons isolated from WT, Kcnb1 and Kcnb1 mice. Detailed methods and images are presented in the supplement.

Seizure threshold testing

Male and female mice were tested between 9 and 12 weeks of age by experimenters blinded to genotype. Separate cohorts of mice were used for each inducing agent.

Flurothyl seizure induction

Susceptibility to seizures induced by the chemoconvulsant flurothyl (Bis(2,2,2-trifluoroethyl) ether, Sigma-Aldrich, St. Louis, MO) was assessed in WT, Kcnb1R/+, Kcnb1R/R, Kcnb1 and Kcnb1 mice at P71-84. Flurothyl was introduced into a clear, plexiglass chamber (2.2 L) by a syringe pump at a rate of 20 μL/min and allowed to volatilize. Latencies to first myoclonic jerk, generalized tonic-clonic seizure (GTCS) with loss of posture, and time interval between these phases were recorded (n = 27-32 per genotype). Groups were compared using one-way ANOVA with Tukey’s post-hoc comparisons for parametric data unless otherwise indicated (Table 3) (GraphPad Prism).

Table 3

Statistical comparisons.

Figure	Comparison	Test	Value	Post hoc
2E	PCC (Kv2.1:VAPA)	One-way ANOVA	F(5,43) = 62.453, p < 0.0001	Tukey
2F	Kv2.1 Puncta Size	One-way ANOVA	F(4,36) = 6.763, p = 0.0004	Tukey
2G	VAPA Puncta Size	One-way ANOVA	F(5,42) = 5.39, p = 0.0006	Tukey
3A	Kcnb1G379R Expression	One-way ANOVA	F(2,19) = 0.3806, p > 0.68	n/a
3C	Kv2.1-G379R Expression	One-way ANOVA	F(2,15) = 139.9, p < 0.0001	Tukey
3E	Kcnb1fs Expression	One-way ANOVA	F(2,22) = 11.48, p = 0.0004	Tukey
3G	Kv2.1-fs Expression	One-way ANOVA	F(2,13) = 68.6, p < 0.0001	Tukey
4E	Kv2.2 Expression	One-way ANOVA	F(2,18) = 9.525, p = 0.0015	Tukey
4G	AMIGO-1 Expression	One-way ANOVA	F(2,18) = 7.168, p = 0.0051	Tukey
5A	G379R GTCS Flurothyl	Kruskal-Wallis	H(2) = 38.92, p < 0.0001	Dunn’s
5A	fs GTCS Flurothyl	One-way ANOVA	F(2,57) = 0.8234, p = 0.4441	Tukey
5B	G379R Myoclonic Jerk Flurothyl	Kruskal-Wallis	H(2) = 2.556, p = 0.2786	Dunn’s
5B	G379R Time for Seizure Progression	Kruskal-Wallis	H(2) = 12.02, p = 0.0025	Dunn’s
5B	fs Myoclonic Jerk Flurothyl	One-way ANOVA	F(2,50) = 2.583, p = 0.0856	Tukey
5B	fs Time for Seizure Progression	Kruskal-Wallis	H(2) = 0.2239, p = 0.8941	Dunn’s
5C	6 Hz Seizure Induction	Log-Probit	p < 0.0001	n/a
5D	Seizure Incidence, WT vs. Kcnb1R/R	Fisher’s Exact	p < 0.001	n/a
8A	Neurological exam	One-way ANOVA	F(3,114) = 25.82, p < 0.0001	Tukey
8B	Open Field Assay-Distance	One-way ANOVA	F(3,100) = 28.31, p < 0.0001	Tukey
8C	Open Field Assay-Center Time	One-way ANOVA	F(3,102) = 8.435, p < 0.0001	Tukey
8D	Zero Maze- Open Arm Time	One-way ANOVA	F(3,97) = 35.51, p < 0.0001	Tukey
8E	Zero Maze-Distance	One-way ANOVA	F(3,96) = 19.96, p < 0.0001	Tukey
8F	Cliff Avoidance WT vs. Kcnb1R/+	LogRank Mantel-Cox	p = 0.0088	n/a
8F	Cliff Avoidance WT vs. Kcnb1R/R	LogRank Mantel-Cox	p < 0.0001	n/a
8F	Cliff Avoidance Kcnb1R/+ vs. Kcnb1R/R	LogRank Mantel-Cox	p = 0.0052	n/a
8F	Cliff AvoidanceKcnb1fs/+ vs. Kcnb1WT	LogRank Mantel-Cox	p = 0.7263	n/a
8F	Cliff Avoidance Kcnb1fs/+ vs. Kcnb1R/+	LogRank Mantel-Cox	p = 0.0719	n/a
8F	Cliff Avoidance Kcnb1fs/+ vs. Kcnb1R/R	LogRank Mantel-Cox	p = 0.0006	n/a
8G	Marble Burying	Two-way repeated measures ANOVA	F(2,41) = 3.518, p = 0.0389	Sidak

6 Hz seizure induction

Susceptibility to 6 Hz psychomotor seizures was assessed in WT and Kcnb1R/+ littermates at P67-89. Individual mice were subjected to 6 Hz of auricular stimulation (0.2 ms pulse width, 3 s duration) using an electroconvulsive unit (Ugo Basile, Gemonio (VA) Italy). Mice that did not exhibit a seizure on the first stimulation were subjected to a second stimulation (≥ 2 mA higher) separated by a delay of ≥15 min. Mice were scored for presence or absence of psychomotor seizure activity, defined as forelimb clonus, rearing, paddling or loss of posture. CC50 curves were calculated with n = 2–20 subjects per current stimulus. Stimulus response curves, CC50, and respective confidence intervals (CI) were determined for each sex and genotype using log-probit analysis.

Video-EEG monitoring

Between P23-25, male and female WT, Kcnb1R/+ and Kcnb1R/R mice were implanted with prefabricated 3-channel EEG headmounts (Pinnacle Technology, Lawrence, KS, USA). Briefly, mice were anesthetized with isoflurane or ketamine/xylazine and placed in a stereotaxic frame. Headmounts with four stainless steel screws that served as cortical surface electrodes were affixed to the skull with glass ionomer cement. Anterior screw electrodes were 0.5–1 mm anterior to bregma and 1mm lateral from the midline. Posterior screws were 4.5–5 mm posterior to bregma and 1 mm lateral from the midline. EEG1 represents recordings from right posterior to left posterior (interelectrode distance ≈2 mm). EEG2 represents recordings from right anterior to left posterior (interelectrode distance ≈5 mm). The left anterior screw served as the ground connection. Following at least 48 h of recovery, tethered EEG and video data were continuously collected from freely moving mice with Sirenia acquisition software (Pinnacle Technology) as previously described (Hawkins et al., 2016). At least 216 h of EEG data were acquired from each subject (Range: WT: 216–504 h/mouse (n = 5 mice); Kcnb1R/+: 216–504 h/mouse (n = 4 mice); Kcnb1R/R: 360–32 h/mouse (n = 3 mice)). Raw data was notch filtered with a 1 Hz window around 60 and 120 Hz prior to analysis. Video-EEG records were analyzed with Persyst13 software (Persyst, Solana Beach, CA, USA), MATLAB (Math-Works, Massachusetts) and EEGLAB (Swartz Center for Computational Neuroscience, California) by a reviewer blinded to genotype. Epileptiform discharges were defined as isolated events with a spike and slow wave morphology, an amplitude of ≥3 times baseline, duration of 150–500 ms, and with increased power in frequencies >20 Hz compared to baseline. Samples with high baseline artifact were excluded from analysis.

Neurobehavioral assays

Male and female WT, Kcnb1R/+, Kcnb1R/R and Kcnb1 mice were tested between 11 and 15 weeks of age. Male and female mice were tested separately with at least a one-hour delay between sessions. For all experiments, mice were acclimated in the behavior suite with white noise for 1 h prior to behavioral testing. At the end of each procedure, mice were placed into a clean cage with their original littermates. Behavioral testing was performed by experimenters blinded to genotype. Evaluation occurred over 4 consecutive days: Day 1- neurological exam; Day 2- open field (OF); Day 3- zero maze; and Day 4-cliff avoidance. The marble burying assay was performed on a separate cohort of mice. Statistical comparison between groups were made using one-way ANOVA with Tukey’s post-hoc comparisons for parametric data or two-way repeated measures ANOVA with Sidak’s post-hoc comparisons, unless otherwise indicated (Table 3).

Neurological exam

Neurological assessment was based on a modified Irwin screen to evaluate baseline behavior (Irwin, 1968). Individual mice were placed in a small static cage for 3 min and observed for transfer behavior, body position, spontaneous activity, tremor, gait, pelvic elevation, tail elevation, palpebral closure, piloerection, air puff startle response, trunk curl, limb grasping, Preyer reflex, provoked salivation, and provoked biting. Each parameter was scored as shown in Table S1 and then summed for a total exam for each mouse. Total exam scores are shown with 23–32 mice per genotype.

Open field

Mice were evaluated for baseline activity in an open field. Individual mice were placed in the center of an open field arena (46 cm × 46 cm) and monitored for 10 min. Limelight software (Actimetrics, Wilmette, IL, USA) was used to video record each trial, track the position of the mouse, and calculate distance traveled and relative position in the arena (n = 21–30 mice per genotype).

Zero maze

Mice were evaluated for anxiety-related behavior in an elevated zero maze, a variant on the elevated plus maze that eliminates the ambiguous center region (Shepherd et al., 1994). Individual mice were placed near the closed arm of the maze and allowed to freely explore for 5 min. Limelight software was used to video record each trial, track the position of the mouse, and calculate time spent in closed or open arms and total distance traveled (n = 19–28 mice per genotype). Trials where mice jumped off the maze were excluded from the analysis.

Cliff Avoidance

Mice were evaluated for impulsive behavior using a cliff avoidance test. Individual mice were placed on an elevated platform (16 cm diameter, 25 cm height) for 7 min. Number of peering events (defined as the entire head extending over the edge of the platform) and fall/jumping events (defined as the entire body leaving the platform) were recorded. Jumping events were compared using time-to-event analysis, p-values were determined by LogRank Mantel-Cox tests (n = 15–29 per genotype).

Marble burying

Marble movement and burying were evaluated in WT, Kcnb1 and Kcnb1 mice to assess phenotypes related to anxiety- and obsessive-compulsive behavior. Individual mice were placed into a static rat cage with 5 cm of corncob bedding and acclimated for 15 min. Mice were then briefly removed from the cage while bedding was flattened and 20 marbles were evenly placed across the cage in 5 rows with 4 marbles each, with a small open space at the front of the cage. A baseline image of the cage was taken prior to reintroduction of the mouse and reimaged after the 30-min trial. The two images were compared for the number of marbles buried, defined by at least ½ of the marble being submerged under the bedding. Total number of marbles moved or buried were compared between groups by two-way repeated measures ANOVA with Sidak’s multiple comparisons test (n = 9–25 mice per genotype).

Cardiac function

Preliminary screening of cardiac function in WT, Kcnb1 and Kcnb1 mice was performed with electrocardiography and echocardiography. Detailed methods and initial results are presented in the supplement (Supplemental Materials and Methods; Supplemental Figs. S1 and S2).

Statistical analysis

Table 3 summarizes statistical tests used for all comparison along with computed values. Values for post-hoc comparisons are reported in the results text and figure legends, and group n’s are reported in figure legends. There were no significant differences between sexes on any measurements except 6 Hz seizure threshold. Thus, groups were collapsed across sex for all variables, except the 6 Hz seizure threshold.

Results

Kcnb1-p.G379R affects subcellular localization of Kv2.1 in HEK293T cells

Previous reports demonstrated that KCNB1 pathogenic variants have altered expression and localization of Kv2.1 in CHO-K1 and COS-1 cell lines (Torkamani et al., 2014; Thiffault et al., 2015). Additionally, it was shown that rat Kv2.1 (rKv2.1) exogenously expressed in HEK293T cells is specifically localized in large PM clusters similar to those seen in neurons (Trimmer, 1991; Scannevin et al., 1996; Mohapatra and Trimmer, 2006; Bishop et al., 2015; Fox et al., 2015; Johnson et al., 2018; Kirmiz et al., 2018). These clusters are located at and participate in ER-PM junctions that result from the interaction of Kv2.1 with ER proteins VAPA and VAPB (Fox et al., 2015; Johnson et al., 2018; Kirmiz et al., 2018). The clustering of Kv2.1 and recruitment of VAP proteins at ER-PM junctions is a non-conducting function of Kv2.1, which instead is dependent upon serine phosphorylation within the PRC domain, a small motif in the Kv2.1C-terminus (Lim et al., 2000; Johnson et al., 2018; Kirmiz et al., 2018). To determine whether the Kcnb1-p.C379R variant affected this non-conducting function of Kv2.1, we investigated human Kv2.1 (hKv2.1) and hKv2.1-G379R expression in HEK293T cells, which lack endogenous Kv2 channel expression even though they have attributes of a neuronal lineage (Yu and Kerchner, 1998; Shaw et al., 2002).

We first determined whether hKv2.1 formed PM clusters and recruited VAPA proteins to ER-PM junctions in HEK293T cells. We used two imaging modalities to assess subcellular localization: conventional epifluorescence microscopy to evaluate overall expression pattern, and Total Internal Reflection Fluorescence (TIRF) microscopy to interrogate fluorescent signals within 100–200 nm of the cover slip. The “TIRF field” corresponded to the first ≈1% of the distance into the cell and primarily comprised the PM and PM-associated structures. Expression of hKv2.1 in HEK293T cells localized to PM clusters in both epifluorescence and TIRF images, with the majority of hKv2.1 present in the clusters (Fig 2A). Expression of hKv2.1 led to a significant reorganization in the subcellular localization of VAPA relative to that seen in untransfected cells, such that VAPA was now enriched at Kv2.1-containing ER-PM junctions (Fig 2A). This colocalization of hKv2.1 and VAPA was indicated by a relatively high Pearson’s Correlation Coefficient (PCC) of 0.7 (Fig. 2E). Expression of hKv2.1-G379R in HEK293T cells suggested a primarily intracellular localization, with lack of concordance between epifluorescence images showing the entire cell and TIRF images showing the region at/near the PM (Fig. 2C). The hKv2.1-G379R PCC value for colocalization with VAPA of 0.3 was lower than for hKv2.1 (p < 0.0001) (Fig. 2E). Additionally cells expressing hKv2.1-G379R had smaller average VAPA cluster size of 0.23 ± 0.02 μm2 compared to cells expressing WT hKv2.1 (0.63 ± 0.13 μm2; p = 0.0028) (Fig. 2C,G).

Fig. 2.

G379R mutation disrupts Kv2.1-mediated recruitment of VAPA to ER-PM junctions in HEK293T cells. A–D) Representative wide-field epifluorescence (“Epi”) and TIRF images of HEK293T cells expressing hKv2.1 WT (A), rKv2.1 S586A (B), hKv2.1 G379R (C) and hKv2.1 G379R + rKv2.1 (D). Red- Kv2.1, Green- VAPA and Blue- Hoechst. Scale bar = 10 μm. E) Scatter plot of Pearson’s correlation Coefficient (PCC) values for colocalization of Kv2.1 and VAPA for each transfection condition. PCC values were higher for hKv2.1 WT (0.699 ± 0.02), rKv2.1 (0.69 ± 0.04), and hKv2.1 WT + rKv2.1 (0.84 ± 0.02) compared to rKv2.1 S586A (0.28 ± 0.04), hKv2.1 G379R (0.3 ± 0.03), and hKv2.1 G379R + rKv2.1 (0.4 ± 0.05) (F(5,43) = 62.453, p < 0.0001; one-way ANOVA; n = 4–12 cells). ****p < 0.0001. F) Scatter plot of Kv2.1 puncta size for each transfection condition. hKv2.1 G379R (0.284 ± 0.2 μm2) and hKv2.1 G379R + rKv2.1 (0.34 ± 0.4 μm2) had smaller Kv2.1 puncta size compared to hKv2.1 WT (0.82 ± 0.14 μm2) (F(4,36) = 6.763, p = 0.0004; one-way ANOVA; n = 4–12 cells). ND: Not determined, p = ***0.0006, *p = 0.0130. G) Scatter plot of VAPA puncta size for each transfection condition. hKv2.1 G379R (0.23 ± 0.02 μm2), hKv2.1 G379R T rKv2.1 (0.20 ± 0.02 μm2) and rKv2.1 S586A (0.181 ± 0.02 μm2) had smaller VAPA puncta size compared to hKv2.1 WT (0.63 ± 0.13 μm2) (F(5,42) = 5.39, p = 0.0006; one-way ANOVA). n = 4–12 cells. **p = 0.0028, *p = 0.0181, **p = 0.0014. Error bars represent SEM.

The rKv2.1-S586A mutation was previously shown to not recruit VAPA to ER-PM junctions (Kirmiz et al., 2018). As a control, we compared Kv2.1:VAPA co-localization and VAPA cluster size of rKv2.1-S586A with WT hKv2.1 or hKv2.1-G379R. Consistent with our previous report, expression of rKv2.1-S586A resulted in a lower PCC value of 0.28 (p < 0.0001) and smaller average VAPA cluster size of 0.18 ± 0.02 pm2 (p < 0.003) compared to WT hKv2.1 cluster size of 0.63 ± 0.13 μm2 (Fig. 2B, E, G). Co-localization PCC values and VAPA cluster sizes were not different between hKv2.1 and rKv2.1, or hKv2.1-G379R and rKv2.1-S586A mutants (Fig. 2E, G). As an additional control, rKv2.1 and hKv2.1-G379R were co-expressed and found to have similar localization and size properties of the hKv2.1-G379R mutation alone, suggesting G379R may exert a dominant negative effect on co-expressed Kv2.1 (Fig. 2E–G). This is consistent with the previously reported dominant negative effect of G379R on Kv2.1 conducting functions (Torkamani et al., 2014). Based on the complexity of the G379R dominant-negative cellular phenotypes, it is difficult to predict effects on neurophysiology; therefore, we generated a mouse model.

Generation and initial characterization of Kcnb1G379R mice

To model the DEE-associated KCNB1-p.G379R pathogenic variant in vivo, we used CRISPR/Cas9 genome editing to introduce the G379R missense variant in the mouse Kcnb1 gene by homology directed repair. Kcnb1 heterozygous and Kcnb1 homozygous mutants (abbreviated as Kcnb1 and Kcnb1, respectively) were born at the expected Mendelian ratios. As a by-product of genome editing, we also obtained a 2 bp deletion allele resulting in a frameshift. Heterozygous and homozygous mice carrying the G379VfsTer6 frameshift allele (abbreviated Kcnb1 and Kcnb1, respectively) were also born at the expected Mendelian ratios. Droplet digital RT-PCR (RT-ddPCR) evaluating whole brain Kcnb1 transcript indicated no expression difference between WT, Kcnb1R/+ and Kcnb1R/R mice (Fig. 3A). There was a modest elevation in transcript for Kcnb1 (p = 0.004) and Kcnb1 (p = 0.0004) mice relative to WT littermates (Fig. 3E).

Fig. 3.

Expression of Kcnb1 transcript and Kv2.1 protein in Kcnb1 and Kcnb1 mice. A) Relative expression of whole brain Kcnb1 transcript in WT, Kcnb1 and Kcnb1 assayed by quantitative RT-ddPCR. There was no difference in transcript expression between genotypes (F(2,19) = 0.3806, p > 0.68; one-way ANOVA; n = 6–8 mice per genotype). B) Representative Kv2.1 immunoblot of Kcnb1 brain membrane protein using the anti-Kv2.1 K89/34 mAb. The anti-GRP75 mAb is used as a loading control. C) Quantification of the Kv2.1/GRP75 ratio on immunoblots showed ≈15% lower Kv2.1 expression in Kcnb1 and ≈67% lower in Kcnb1 relative to WT (F(2,15) = 139.9, p < 0.0001; one-way ANOVA; n = 6 mice per genotype; **p = 0.0049, ****p < 0.0001). D) Line scan analysis of immunoblots showing a genotype-dependent shift in post-translational modification state of Kv2.1 from a heterogeneous pool (Mr ≈ 100–130 kDa) in WT and Kcnb1 samples to an Mr more closely approximating the predicted molecular weight of 95 kDa in Kcnb1 samples. Lines are the average of 6 samples per genotype and SEM is represented by the shading. E) Relative expression of whole brain Kcnb1 transcript in WT, Kcnb1 and Kcnb1 assayed by quantitative RT-ddPCR. There was a modest elevation of Kcnb1 transcript expression compared to WT (F(2,22) = 11.48, p = 0.0004; one-way ANOVA; n = 7–10 mice per genotype; **p = 0.004, ***p = 0.0004). F) Representative Kv2.1 immunoblot of Kcnb1 brain membrane protein using the anti-Kv2.1 K39/25 mAb. The anti-GRP75 mAb is used as a loading control. G) Quantification of the Kv2.1/GRP75 ratio on immunoblots showed ≈45% lower Kv2.1 expression in Kcnb1 and ≈89% lower in Kcnb1 relative to WT littermates. Kcnb1 control sample (purple circles) had similar expression of ≈85% compared to WT, suggesting that residual signal is non-specific and KcnbP line is comparable to a knockout (F(2,13) = 68.6, p < 0.0001; n = 2–6 mice per genotype; **p = 0.0003, ***p < 0.0002). For panels A, C, E and G, symbols represent samples from individual mice and error bars represent SEM.

Lower Kv2.1 channel expression in Kcnb1G379R mice

To evaluate the effect on Kv2.1 protein expression, we first performed immunoblotting using whole brain membrane preparations from Kcnb1 and Kcnb1 lines. These analyses revealed that whole brain expression of Kv2.1 was ≈15% lower in Kcnb1 (p < 0.005) heterozygotes and ≈67% lower in Kcnb1 (p < 0.0001) homozygous mice compared to WT littermates (Fig. 3B–C). There was also an evident shift in the relative electrophoretic mobility (Mr) from the fully post-translationally modified form (≈125 kDa) of Kv2.1 toward an Mr of ≈96 kDa, suggesting a deficit in the processing of the Kcnb1 variant into the mature, fully post-translationally modified species (Fig. 3B, D) (Murakoshi et al., 1997; Misonou et al., 2004). Immunoblotting analysis of the Kcnb1 line with an antibody targeting an epitope upstream of the frameshift (Table 2) revealed that Kv2.1 expression was ≈45% lower in Kcnb1 (p = 0.0001) and ≈89% lower in Kcnb1 (p < 0.0001) mice compared to WT littermates (Fig. 3F,G). This was similar to the level of expression observed with a Kcnb1−/− knockout control, which had ≈85% lower expression compared to WT. This suggests that the residual signal is non-specific and that the Kcnb1 line is comparable to a knockout line.

To assess the impact of the Kcnb1 variant on neuroanatomy and Kv2.1 expression, we performed immunolabeling of brain sections from WT, Kcnb1 and Kcnb1 mice. The Kcnb1 variant had no overt effect on gross hippocampal neuroanatomy based on labeling with the DNA-specific dye Hoechst 33258 (data not shown), which was similar to previous observations in Kcnb1−/− mice (Speca et al., 2014). Kv2.1 immunolabeling (green) in hippocampal, subicular and neocortical neurons in Kcnb1 mice was lower in intensity compared to WT (Fig. 4A–C). Kcnb1 mice had little detectable Kv2.1 immunolabeling (Fig. 4A–C), and substantially reduced immunolabeling for Kv2.2 and AMIGO-1. This was qualitatively consistent with immunoblotting results (Fig. 3B, C) and results obtained with cultured hippocampal neurons (Supplemental Fig. S3). To ensure specificity of the Kv2.1 antibody, we also immunolabeled Kcnb1brain sections, which lacked detectable Kv2.1 signal (Fig. 4C). As previously reported (Speca et al., 2014), the pattern and intensity of Kv2.2 immunolabeling in the Kcnb1−/− brain sections was comparable to that in WT sections (Fig. 4C). Notably, when we enhanced the intensity of Kcnb1 images, the remaining signal was found to be predominantly Kv2.2 and AMIGO-1 (Fig. 4C) such that the enhanced images resembled the non-enhanced image from the Kcnb1−/− brain sections (Fig. 4C’). Additional validation of lower Kv2.1 expression due to the Kcnb1 variant was performed by immunolabeling with three additional Kv2.1 antibodies each recognizing a distinct region of Kv2.1, D3/71R, L105/31 and L80/21 (Table 2), each of which detected a similarly low level of Kv2.1 immunolabeling (Supplemental Fig. S4).

Fig. 4.

Immunolabeling of Kv2 channel complexes in Kcnb1 brain sections and membrane preps. A) Representative images of somatosensoiy cortex and hippocampus from littermate WT, Kcnb1 and Kcnb1 mice. Signal intensities for Kv2.1 (green), Kv2.2 (red), and AMIGO-1 (blue) immunolabeling are lower in Kcnb1R/R mice. Scale bar = 50 μm. B) Representative high magnification images of layer 5b of somatosensory cortex from littermate WT, Kcnb1 and Kcnb1 mice. Inset panel is 3x size reduced image with enhanced signal intensity to show that the immunolabeling remaining in the sample from the Kcnb1 mouse appears to be primarily intracellular compared to WT and Kcnb1 mice. Scale bar = 20 μm. C) Image of bottom left panel of A with enhanced signal intensity, showing that the remaining signal is predominantly Kv2.2 (red) and to a lesser extent AMIGO-1 (blue). CO Representative image of somatosensoiy cortex and hippocampus from Kcnb1−/− at the same exposure as panels A and B showing a pattern of immunolabeling similar to the pattern observed in the Kcnb1 section after enhancement of signal intensity as shown in panel C. D) Representative Kv2.2 immunoblot. The anti-GRP75 mAb is used as a loading control. E) Quantification of the Kv2.2/GRP75 ratio on immunoblots showed ≈30% lower expression in Kcnb1 and ≈50% lower in Kcnb1 relative to WT (F(2,18) = 9.525, p = 0.0015; one-way ANOVA; n = 7 mice per genotype; *p < 0.05, **p < 0.002). F) Representative AMIGO-1 immunoblot. The anti-GRP75 mAb is used as a loading control. G) Quantification of AMIGO-1/GRP75 ratio on immunoblots showed ≈25% lower expression in Kcnb1 and ≈47% lower in Kcnb1 relative to WT (F(2,18) = 7.168, p = 0.0051; one-way ANOVA; n = 7 mice per genotype; **p < 0.004). For panels E and G, symbols represent samples from individual mice and error bars represent SEM.

Kcnb1 also affected expression of Kv2.2 and the Kv2 channel auxiliary subunit AMIGO-1. Immunolabeling experiments performed on WT, Kcnb1 and Kcnb1 brain sections revealed lower Kv2.2 and AMIGO-1 expression in Kcnb1 versus WT mice (Fig. 4A–C). To quantitate the effect of the Kcnb1 variant on Kv2.2 and AMIGO-1 protein expression, immunoblotting was performed on whole brain membrane preparations (Fig. 4D–G). Immunoblotting showed that whole brain expression of Kv2.2 was ≈30% lower in Kcnb1 (p < 0.05) and ≈50% lower in Kcnb1 mice (p < 0.002) compared to WT littermates (Fig. 4D, E). Immunoblotting for AMIGO-1 whole brain expression revealed ≈25% lower expression in Kcnb1 and ≈47% lower expression in Kcnb1 mice (p < 0.004) compared to WT littermates (Fig. 4F, G). This is consistent with a dominant negative effect on Kv2 channel complexes.

Lower threshold for induced seizures in Kcnb1G379R mice

In order to examine seizure sensitivity, we evaluated response of Kcnb1 mice to the volatile chemoconvulsant flurothyl, a GABAA antagonist (van Vliet et al., 2017). Latency to the first myoclonic jerk did not differ between genotypes, while latency to the first generalized tonic-clonic seizure (GTCS) with loss of posture was affected by genotype (Table 3). GTCS latency was lower in Kcnb1 compared to either WT (p < 0.0001) or Kcnb1 mice (p = 0.0004). Kcnb1 mice also had a lower GTCS threshold than WT mice (p = 0.0477). Average latency was 112 ± 5 s for Kcnb1 mice, 153 ± 8 s for Kcnb1 mice, and 178 ± 8 s for WT littermates (Fig. 5A). Time to progress from the first myoclonic jerk to GTCS was affected by genotype (Table 3), with both Kcnb1 and Kcnb1 mice progressing more rapidly between stages compared to WT (p = 0.0040 and p = 0.0438, respectively) (Fig. 5B). Average progression time was 28 ± 7 s for Kcnb1, 49 ± 9 s for Kcnb1 and 68 ± 8 s for WT. In order to compare flurothyl susceptibility of Kcnb1 and Kcnb1 mice to the heterozygous and homozygous null condition, we ran a concurrent cohort of Kcnb1 mice. Under our experimental conditions, the Kcnb1 alleles did not differ from WT littermates in flurothyl sensitivity measurements, including latencies to MJ, GTCS, or time progression between the two (Table 3) (Fig. 5A–B). Although this differs from the enhanced flurothyl susceptibility previously reported for Kcnb1−/− mice (Speca et al., 2014), the protocol used here had faster introduction of flurothyl, which is less likely to detect small effects on threshold.

Fig. 5.

Kcnb1 and Kcnb1 mice are sensitive to induced seizures. A) Latency to flurothyl-induced GTCS was affected by Kcnb1 genotype (H(2) = 38.92, p < 0.0001, Kruskal-Wallis; n = 28–32 mice per genotype). Kcnb1 mice had a reduced seizure threshold of 153 ± 8 s compared to WT with an average latency of 178 ± 8s. Kcnb1 mice had the shortest average latency of 112 — 5 s, which was lower than WT or Kcnb1 . There were no differences in GTCS latency between WT, Kcnb1fs/+ or Kcnb1fs/fs mice (F(2,57) = 0.8234, p = 0.44; n = 18–22 mice per genotype). Symbols represent individual mice and error bars represent SEM. *p = 0.0412, ***p = 0.0005, ****p < 0.0001 B) Average latencies to first myoclonic jerk (MJ) and GTCS are shown, with the connecting line depicting time of progression between the stages. Although there was no difference in MJ latency for Kcnb1 versus WT mice, progression between the stages was faster for Kcnb1 and Kcnb1 mice relative to WT (p = 0.0040, p = 0.0438, respectively; Dunn’s). There were no differences in average latencies or progression time between the stages for Kcnb1fs/+, Kcnb1fs/fs and WT mice (Table 3). C) Susceptibility to psychomotor seizures induced by 6 Hz stimulation. Kcnb1R/+ mice had lower CC50 values (convulsive current, 50% with seizures) compared to WT (p < 0.0001, log-probit). The CC50 values (95% confidence interval) for male WT and Kcnb1R/+ mice were 23.7 mA (19.5 to 27.9) and 16.4 mA (14.3 to 18.6), respectively. The CC50 values for female WT and Kcnb1R/+ mice were 18.7 mA (16.5 to 20.9) and 16.2 mA (13.3 to 19.1), respectively. Floating box graph represents 95% Cl and CC50 (line). CC50 and CI were determined using log-probit analysis with n = 2–20 per current. D) Handling-induced seizures were observed in Kcnb1 mice at 6–12 weeks of age, while they were not observed in Kcnb1 or WT mice in this age window (p < 0.001, n = 23–24 mice per genotype). Rare handling-induced seizures were observed in Kcnb1 mice at >4 months of age.

We next investigated whether Kcnb1 influenced latency to psychomotor seizures induced by long duration, low frequency stimulation (6 Hz, 3 s), a model of pharmacoresistant focal seizures (van Vliet et al., 2017). Kcnb1 heterozygotes and WT littermates were subjected to 6 Hz stimulation and scored for the presence or absence of seizure activity (forelimb clonus, rearing, paddling or loss of posture). Convulsive current (CC) curves were generated as previously described (Finney, 1971), and population CC50 values (CC at which 50% of mice seized) were determined for each genotype and sex. Both male and female Kcnb1R/+ mice had lower CC50 values compared to sex-matched WT controls (p < 0.0001) (Fig. 5C). The CC50 values (95% confidence interval) for WT and Kcnb1R/+ male mice were 23.7 mA (19.5 to 27.9) and 16.4 mA (14.3 to 18.6), respectively. The CC50 values for WT and Kcnb1R/+ female mice were 18.7 mA (16.5 to 20.9) and 16.2 mA (13.3 to 19.1), respectively.

Handling-induced seizures in Kcnb1G379R mice

We initially observed during the course of routine husbandry that Kcnb1 mice exhibited seizures following handling, as well as overt home cage hyperactivity, repetitive jumping, and rare unexpected deaths in otherwise healthy appearing animals (median age at death P27, range P21-172). Based on these initial observations of seizures, we systematically assessed handling-induced behavioral seizures once per week between 6 and 12 weeks of age by individually transferring mice to clean cages and observing for 1 min. During those observations, 42% (10 of 24) Kcnb1 mice exhibited behavioral seizures with forelimb clonus, rearing and falling, sometimes followed by wild running (Fig. 5D) (Supplemental Video S1). In contrast, Kcnb1 heterozygous mice (n = 23) and WT littermates (n = 23) did not exhibit any seizures during these observations (p < 0.001) (Fig. 5D). Similar handling-induced seizures were observed infrequently in Kcnb1 heterozygotes, but at older ages (>4 months).

EEG abnormalities in Kcnb1G379R mice

The behaviors observed during handling-induced seizures described above were similar to behaviors coinciding with spontaneous generalized tonic-clonic seizures during video-EEG monitoring of Kcnb1 mice at 3 to 6 weeks of age (Fig. 6A; Supplemental Video S2–S2a). Unprovoked spontaneous seizures occurred rarely (one to two times per animal during monitoring period; average frequency of <1 per week) in Kcnb1 mice (Fig. 6A), but were never observed in WT or Kcnb1mice. In addition, homozygous Kcnb1 mice displayed recurrent, brief runs of rhythmic slow spike and wave complexes (1–2 Hz) that lasted up to a minute during wakefulness or up to 15 min during sleep (Fig. 6B–C; Supplemental Videos S3–S4). Several types of interictal EEG abnormalities were obseived (Fig. 7). Both Kcnb1 and Kcnb1 mice exhibited isolated spike and slow wave complexes (Fig. 7B–D; Supplemental Video S5) that have an increase in power across low and high frequencies up to 170 Hz (Fig. 7E). The occurrence of the isolated spike and slow wave complexes was elevated relative to WT littermates (Fig. 7F) when quantified over a 24-h period.

Fig. 6.

Ictal EEG abnormalities in Kcnb1G379R mice. A) Representative EEG traces from Kcnb1R/R mouse exhibiting a spontaneous tonic-clonic seizure (see Supplemental Video S2 and S2a for the corresponding videos). The top line corresponds to EEG1 (right poster-left posterior) and the second line corresponds to EEG2 (right anterior to left posterior) in each set of traces. The upper two traces display 2 min of EEG and the purple line denotes the one minute expanded in the lower two traces. The scale bar for the 2-min time base are 75 μV and 4 s and the scale bar for the 1-min time base are 75 μV and 1 s. B-C) Representative EEG traces from a Kcnb1R/R mouse with repetitive runs of slow spike-wave discharges. The top line corresponds to EEG1 and the second line corresponds to EEG2. The top two lines display 5 min of EEG and the purple line denotes the 30 s region of expanded time base displayed below. The scale bar for the 5-min time base are 75 μV and 10 s and the scale bar for the 30 s segment are 75 μV and 1 s, respectively. Neither of these runs had apparent clinical correlate with B occurring during wakefulness while the animal was moving around the cage and C occurring during a 15-min run of discharges during sleep.

Fig. 7.

Interietal EEG abnormalities in Kcnb1G379R mice. A–C) Representative EEG traces from WT, Kcnb1R/+, and Kcnb1R/R mice. The top line corresponds to EEG1 (right posterior-left posterior) and the second line corresponds to EEG2 (right anterior to left posterior) in each set of traces. The top two lines display 5 min of EEG and the purple line denotes the 30 s region of expanded time base displayed below. The scale bar for the 5-min time base are 75 μV and 10 s and the scale bar for the 30 s segment are 75 μV and 1 s, respectively. D) Example isolated spike and slow wave discharge corresponding to blue line in B. E) Power spectrum for spike and slow wave discharge in D showing elevated power in decibels across the 1–170 Hz frequency range at the time of the discharge. F) Quantification of isolated spike and slow wave discharges for each genotype (WT n = 5; Kcnb1R/+ n = 4; Kcnb1R/R n = 3) over a 24-h period and displayed as frequency per hour. Symbols represent the mean for an individual animal ± SEM. Horizontal lines represent the mean ± SEM for all animals of that genotype.

Neurological abnormalities in Kcnb1G379R mice

A neurological exam was used to evaluate baseline neurological function in 3 week old WT, Kcnb1R/+ and Kcnb1R/R mice, including analysis of muscle, spinocerebellar, sensory, neuropsychiatric and autonomic functions (Irwin, 1968). Kcnb1 mice were included as control to contrast Kcnb1R/+ to the heterozygous null condition, as heterozygosity is the genotype present in KCNB1 DEE patients. Neurological exam scores were affected by Kcnb1 genotype, with higher scores indicating deficits. Kcnb1 had an average overall exam score of 17.2 ± 0.4, that was elevated (p = 0.0004) compared to the WT control score of 15.1 ± 0.4 (Fig. 8A). Kcnb1 mice had the highest score, with an average of 19.8 ± 0.5 (p < 0.0001 vs WT, Kcnb1 and Kcnb1, respectively) (Fig. 8A). In contrast, Kcnb1 mice had an average score of 16.2 ± 0.3 that did not differ from WT littermates (p = 0.1372). Differences in the overall exam score in Kcnb1 mice were driven largely by differences in activity levels, escape behavior, and trunk curl (Supplemental Table S1). Furthermore, it was noted that upon transfer into the observation cage, 2 of 23 Kcnb1 mice exhibited a tonic-clonic seizure characterized by bilateral forelimb clonus with rearing and falling.

Fig. 8.

Neurological and neurobehavioral phenotypes in Kcnb1 and Kcnb1 mice. A) Modified Irwin neurological exam scores were affected by genotype (F (3,114) = 25.82, p < 0.0001; one-way ANOVA). Kcnb1 mice had an average score of 17.2 ± 0.4, higher compared to the WT score of 15.1 ± 0.4. Kcnb1 mice had the highest score, with an average score of 19.8 ± 0.5. Kcnb1fs/+ mice had an average score of 16.2 ± 0.3 (n = 23–32 mice per genotype; ***p = 0.0004, ****p < 0.0001). B) Distance traveled in an open field assay was affected by genotype (F(3,100) = 28.31, p < 0.0001; one-way ANOVA). Kcnb1 mice traveled an average distance of 59.3 ± 2.2 m, farther compared to WT average distance of 45.6 ± 2.0 m. Kcnb1 mice traveled an average distance of 75.8 ± 3.1 m, farther than both WT and Kcnb1 mice. Kcnb1fs/+ mice traveled an average distance of 52.0 ± 2.2 m (n = 21–29 mice per genotype; ***p = 0.0003, ****p < 0.0001). C) There was a significant effect of genotype on percent of time spent in the center of the open field (F(3,102) = 8.435, p < 0.0001; one-way ANOVA). Kcnb1 mice spent 4.8 ± 0.6% of time in the center, less than WT, Kcnb1 and Kcnb1fs/+ mice (n = 22–30 mice per genotype; **p = 0.0011, ***p < 0.0004). D) Time spent in the open arms of the zero-maze was affected by genotype (F(3,97) = 35.51, p < 0.0001; one-way ANOVA). Kcnb1 mice averaged 38.6 ± 1.3% of the test time in open arms of the maze compared to WT mice that averaged 31.4 ± 1.9% of time. Kcnb1 mice spent 55.7 ± 2.3% of time in the open arms of the maze, more than WT, Kcnb1 and Kcnb1fs/+ (35.4 ± 1.1%) mice (n = 19–28 mice per genotype; *p = 0.01, ****p < 0.0001). E) There was a significant effect of genotype on total distance traveled in the zero maze (F(3,96) = 19.96, p < 0.0001; one-way ANOVA). Kcnb1 mice traveled the farthest distance of 34.3 ± 2.1 m compared to 25.2 ± 0.9 m for Kcnb1, 26.0 ±1.2 for Kcnb1 and 20.4 ± 0.8 m for WT (n = 19–28 mice per genotype; *p = 0.0236, **p = 0.005, ****p < 0.0001). F) Time to step or jump off an elevated platform was evaluated in the cliff avoidance assay. Kcnb1 mice left the platform more frequently (≈52%, 15/29) compared to WT (≈15%, 4/26). Approximately 78% of Kcnb1 mice (18/23) left the platform, more than WT, Kcnb1 and Kcnb1 (≈22%, 3/15) mice. **p < 0.009, ***p < 0.0007, ****p < 0.0001. G) Marble burying was significantly affected by genotype. (F(2,41) = 3.518, p = 0.0389; two-way repeated measures ANOVA). Relative to WT, Kcnb1 mice rarely and Kcnb1 mice never buried marbles (n = 9–25 mice per genotype; *p = 0.0225). H-J) Representative examples of movement paths in the open-field assay for WT, Kcnb1 and Kcnb1 mice. For panels A-E and G, symbols represent individual mice, horizontal lines represent mean and error bars represent SEM.

Kcnb1G379R mice exhibit profound hyperactivity

Kcnb1mice were easily distinguishable from WT littermates in home cages based on their elevated activity level. Baseline activity measured in an open field assay showed that Kcnb1 mice traveled farther than WT controls, with average distances of 59.3 ± 2.2 m and 45.6 ± 2.0 m, respectively (p = 0.0003). Kcnb1 mice traveled 52.0 ± 2.2 m, which did not differ from WT (Fig. 8B). Kcnb1 mice spent the majority of the session moving around the perimeter of the arena, traveling an average distance of 75.8 ± 3.1 m, more than WT controls, Kcnb1 and Kcnb1 mice (p < 0.0001, p < 0.0001, p < 0.0001, respectively; Fig. 8B, C, J). Kcnb1 mice spent <5% of their time in the center of the arena compared to WT, Kcnb1 and Kcnb1 mice (p = 0.0011, p = 0.0001, p = 0.0003, respectively) that all spent —10% of the time in the center (Fig. 8C, H–J).

Impulsivity and diminished anxiety-related behaviors in Kcnb1G379R mice

Along with their noticeably elevated activity, Kcnb1 mice appeared to have altered impulsivity/anxiety-like behavior. To further assess these behavioral abnormalities, we used zero-maze, cliff avoidance and marble burying assays. The zero-maze assay evaluated anxiety-related behavior of WT and Kcnb1 mutants by comparing time spent in open versus closed arms. Kcnb1 mice spent an average of 38.6 ± 1.3% of the test time in open arms compared to WT controls that averaged 31.4 ± 1.9% of time (p = 0.0100) (Fig. 8D). Kcnb1 mice spent more than half their time (55.7 ± 2.3%) in the open arms of the maze, more than WT, Kcnb1 or Kcnb1 mice (p < 0.0001, respectively) (Fig. 8D). Kcnb1 mice spent 35.4 ± 1.1% of their time in the center, which did not differ from WT (p = 0.2749). Distance traveled in the zero-maze differed between Kcnb1 genotypes and mirrored the effects seen in open field distance traveled (Fig. 8B, E). In this assay, distance traveled for Kcnb1 mice was similar to Kcnb1 mice and both were greater than WT (p = 0.005, p < 0.03, respectively) (Fig. 8E).

The cliff avoidance assay takes advantage of the natural tendency of mice to avoid a potential fall from a height and is used to assess inattentive and impulsive behavior (Matsuoka et al., 2005; Yamashita et al., 2013). WT, Kcnb1 and Kcnb1 mice were individually placed on an elevated platform and monitored for jumping or stepping off the platform. Most WT mice ≈85%; 22 of 26) remained on the platform for the duration of the test, and Kcnb1 performed similarly ≈80%; 12 of 15) (Fig. 8F). Fewer Kcnb1 ≈48%; 14 of 29) and Kcnb1 (≈22%; 5 of 23) remained on the platform compared to WT littermates (p = 0.0088 and p < 0.0001, respectively). Furthermore, Kcnb1 mice left the platform more often than Kcnb1 or Kcnb1 mice (p = 0.0052, p = 0.0006, respectively) (Fig. 8F).

The marble burying assay assessed attention, anxiety, and obsessive-compulsive related behaviors of WT and Kcnb1 mutants. Relative to WT controls, Kcnb1 mice showed little interaction with the marbles, moving few and failing to bury any (p = 0.0225) (Fig. 8G).

Discussion

KCNB1-p.G379R was identified as a pathogenic de novo variant in a child with DEE. Electrophysiological studies of Kv2.1 channels with the G379R variant in a heterologous expression system demonstrated lower potassium conductance relative to WT, loss of ion selectivity, and gain of a depolarizing inward cation currents (Torkamani et al., 2014). In addition, mutant subunits exerted a dominant negative effect on potassium currents when co-expressed with WT subunits (Torkamani et al., 2014). We extended in vitro characterization of the Kcnb1-p.C379R variant by demonstrating failure of Kv2.1-G379R to induce ER-PM junctions, as well as dominant negative effects on co-expressed WT Kv2.1 subunits. Together, these results suggest that KCNB1-p.G379R exerts dominant negative effects on both conducting and non onducting Kv2.1 functions. Introduction of this variant in Kcnb1 knock-in mice provides a new model of KCNB1-associated DEE. This model recapitulates several key features of the human disorder, including seizures, interictal epileptiform events on EEG, behavioral hyperactivity, impulsivity/inattention, and attenuated anxiety-related behavior. Additionally, we show that Kcnb1 results in a more severe phenotype than a frameshift allele that had loss of Kv2.1 similar to the global knock-out allele (Speca et al., 2014).

The proband with the KCNB1-p.C379R de novo variant initially presented with infantile spasms at 8 months of age and later developed multiple seizure types including focal dyscognitive, atonic, and generalized tonic-clonic seizures (Torkamani et al., 2014). Seizures were difficult to treat and adequate control was not achieved with ACTH, topiramate, valproic acid, pyridoxine, or the ketogenic diet. EEG findings at different time points included hypsarrhythmia, diffuse polyspikes, diffuse polyspike and slow-waves, right temporal spike and waves, and left occipital spikes. Beyond seizures, there was also developmental delay and an autism spectrum disorder diagnosis consistent with atypical Rett syndrome (Torkamani et al., 2014; Srivastava et al., 2018). The Kcnb1 mouse model shares a number of these key phenotypes and will be a useful platform for evaluation of potential therapies for KCNB1-associated DEE.

Electrographic and behavioral seizures in Kcnb1 mice shared features of seizures seen in the proband (Torkamani et al., 2014). The Kcnb1 mice exhibited generalized tonic-clonic seizures that occurred both spontaneously and with brief handling. Handling-induced seizures are akin to reflex seizures, which are included as equivalent to unprovoked seizures in the ILAE definition of epilepsy (Fisher et al., 2014). Numerous EEG abnormalities were noted in Kcnb1 mice, including isolated spike and slow waves and recurrent runs of rhythmic slow spike and wave complexes lasting up to 15 min without a behavioral correlate. These events and features will be useful biomarkers to evaluate potential therapeutic strategies that can normalize the EEG.

Neurobehavioral abnormalities were prominent in the Kcnb1 mice and share some overlap with features reported in children with KCNB1-associated DEE (Torkamani et al., 2014; Thiffault et al., 2015; Calhoun et al., 2017; de Kovel et al., 2017; Marini et al., 2017; Srivastava et al., 2018; Bar, 2020). Attention-deficit/hyperactivity disorder or hyperactivity with inattention has been reported in numerous cases of KCNB1-associated DEE. Kcnb1 mice exhibit profound hyperactivity both in their home and novel environments, and inattention was suggested by failure to interact with marbles in the marble burying assay. In addition, lack of preference for the closed arms in the zero maze could be due to inattention to surroundings and/or reduced anxiety-like behavior. Failure of the cliff avoidance response in Kcnb1 mice may reflect inattention and/or elevated impulsivity, another behavioral problem reported for individuals with KCNB1-associated DEE. Although visual deficits could explain the observed inattention, there was no deficit in visual placing in Kcnb1 mice and prior characterization of Kcnb1−/− mice included assessment of vision, which detected no impairment (Speca et al., 2014). Autism spectrum disorder has been reported in more than half of KCNB1-associated DEE cases, including KCNB1-p.G379R. Repetitive jumping was another prominent behavior observed in the Kcnb1 mice in their home cages and upon transfer to a novel environment. This perseverative behavior was also previously reported in the Kcnb1−/− mice and may reflect repetitive movements seen in KCNB1 DEE patients with autism spectrum disorder (Speca et al., 2014).

Previous work demonstrated that global homozygous deletion of Kcnb1 in mice resulted in pronounced hyperactivity, reduced anxiety-like behavior, and enhanced susceptibility to flurothyl and pilocarpine-induced seizures (Speca et al., 2014). The neurobehavioral phenotypes of Kcnb1 mice were qualitatively similar to those reported for Kcnb1−/− on the same C57BL/6 J background strain. However, parallel comparison of Kcnb1 and Kcnb1 mice in a subset of neurobehavioral assays suggests that G379R is more severe than the G379VfsTer6 frameshift allele that did not differ from WT. Furthermore, seizure phenotypes are more prominent in Kcnb1 mice relative to Kcnb1 and Kcnb1−/− global knock-out mice (Speca et al., 2014). Seizure activity was reported in ≈10% of homozygous Kcnb1−/− mice during the course of routine handling (Speca et al., 2014). This is similar to the frequency of handling-induced seizure activity observed in heterozygous Kcnb1 mice, while seizures were observed in almost half of homozygous Kcnb1 mice between 6 and 12 weeks of age. Direct comparison of flurothyl sensitivity showed that both Kcnb1 and Kcnb1 mice had lower GTCS thresholds and faster MJ-GTCS transitions relative to WT, while Kcnb1 mice did not differ from WT. This provides further evidence of a more severe phenotype with the Kcnb1 allele. In vitro functional studies of Kcnb1-p.C379R showed change of function effects including loss of voltage-sensitivity and altered ion-selectivity, as well as dominant negative effects when co-expressed with WT subunits (Torkamani et al., 2014). In addition to alterations in the biophysical properties of Kv2.1 channels, the G379R mutation also appeared to exert a dominant negative influence on the expression and subcellular localization of WT Kv2.1, and transdominant negative effects on expression of Kv2.2 and AMIGO-1 subunits. Immunohistochemistry and immunoblotting experiments revealed that immunoreactivity for both Kv2.2 and AMIGO-1 was lower in sections and whole brain membrane preparations from Kcnb1 mice compared to both WT mice and to Kcnb1null mice. Kv2.1 and Kv2.2 are colocalized in some neuron types (e.g., Satb2-positive cells in layer 5 cortex) and form heteromeric channels, with most complexes also containing AMIGO-1 auxiliary subunits (Kihira et al., 2010; Peltola et al., 2011; Bishop et al., 2018). It is possible that effects on heteromeric channels that include the Kcnb1 mutant may contribute to the more severe seizure phenotypes observed in Kcnb1 mice as compared to Kcnb1−/− null mice, which had normal expression of Kv2.2 (Speca et al., 2014). Future studies will investigate the neurophysiological basis of seizure phenotypes in Kcnb1 mice.

Beyond neurological phenotypes, there have been anecdotal reports of borderline long QT syndrome in patients with KCNB1-associated encephalopathy (Kcnb1.org). Additionally, KCNB1 is highly expressed in mouse cardiac tissue and synonymous SNPs in KCNB1 have been associated with long QT syndrome in humans (Iwasa et al., 2001). Our preliminary evaluation of cardiac phenotypes suggested that homozygous Kcnb1 mice had prolonged QTc interval relative to WT at baseline and following an isoproterenol challenge (Supplemental Fig. S1). There was no evidence of structural abnormalities or contractile dysfunction by ECHO (Supplemental Fig. S2 & Table S2). KCNB1 variants could conceivably contribute to spontaneous or induced cardiac arrhythmia (e.g., exposure to proarrythmic drugs), although future studies will be required to further characterize the arrhythmogenic potential of Kcnb1.

In summary, we developed a novel knock-in mouse model of human DEE caused by a missense mutation in Kcnb1 that disrupts both conducting and non-conducting functions of Kv2.1 channels. The Kcnb1 mouse model will be valuable for defining the molecular and neurophysiological consequences of Kcnb1 mutation, understanding disease pathophysiology, and evaluating response to therapeutic interventions.