EEG of diffuse bilateral onset seizure in Gly256Trp/+ individual 1, age 16 day.

The recording is continuous, seizure electrographic evolution is labeled. Settings: LFF 3 Hz, HFF 70 Hz, sensitivity 7uV/mm, 35 sec/panel.

Gly256 is linked to the selectivity filter bridge segment via a hydrogen bond network among residues distinct to KCNQ2.

A, B. Cartoons showing KCNQ2 membrane topology, including transmembrane segments S1-S6 and the P-loop (turret segment, purple; H5 or P-helix, cyan; and selectivity filter segment, yellow). Positions of the K+ selective pore, and the G256W substitution within the turret are indicated. C. Alignment of human KCNQ4 and KCNQ5 sequences with KCNQ2 sequences of major vertebrate groups. Background colors match panels A-B, and the five selectivity filter lining residues are boxed in red. At four aligned positions within the turret and one in the SFB, KCNQ2 substitutions have evolved in amphibians and tetrapods (residues highlighted in red). D. Rendering of the wild type KCNQ2-calmodulin tetrameric structure obtained by cryoEM (PDB 7cr3), highlighting one subunit and the position of the G256W substitution near the channel’s extracellular domain apex. The Trp256 sidechain is at scale but its rotamer is chosen arbitrarily. The subunit closest to the viewer is partially deleted to reveal the highlighted subunit more clearly. E. Ribbon rendering of the extracellular part of the PGD. For clarity, only two opposing side subunits are shown (as schematically in A). A Trp side chain is added at one Gly256 α-carbon. The distance between the G256 α-carbon and Y280 carbonyl oxygen at the selectivity filter mouth is labeled. F. Top down view of the KCNQ2 regions as in panel E, but showing 4 subunits. The Trp rotamer is different from panels D-E. The S5, S6 and P-helices are labeled. G. Hydrogen bonding network of the KCNQ2 turret. All predicted bonds are shown as dashed orange lines. The network extends from the S5 helix (Y251) via the labelled turret residue atoms to bonds involving residues of the SFB. As in C, five residues that diverge in vertebrates are colored red. H, I. The turret peptide region near G256, which is boxed with a grey dashed line in G. The main chain is shown as ball-and-stick; side chains as stick. A tight turn occurs at K255 to N258, stabilized by hydrogen bonding between the G256 carbonyl oxygen and N258 amide. I. The G256-E257 peptide deviates from planarity (ω = +/-180°) by 11.6° (∼2.6 sd). In and out arrows indicate N and C termini, respectively. Abbreviations: mya, million years ago; VSD, voltage-sensor domain; HA-HB, the cytoplasmic helices A and B; CaM, calmodulin.

KCNQ2 G256W co-expression suppresses current in KCNQ2/KCNQ3 heteromeric channels.

A. Cartoon showing the expected combinations of WT and G256W subunits under heterozygosity based on a simple random association model and preferred 2:2 stoichiometry for KCNQ2 and KCNQ3. B-G. In vitro dissection of effects of G256W heterozygosity on currents. B-C. Mean current families are shown for the indicated combinations of expression of KCNQ2 and KCNQ3 prior to and after addition of 10 μM ezogabine (n = 60, 50; 40, 31; 28, 24 for the upper, middle, and lower conditions). D-E. Current/voltage and conductance/voltage relationships for the indicated WT only and G256W/WT cells. F. Current/voltage relationship for WT control and G256W (‘homozygous”) heteromeric channels, studied in parallel. G. Increases in current by ezogabine for WT KCNQ2 only and WT:G256W co-expression.

Immature heterozygous G256W mice exhibit normal development and have infrequent epileptic seizures.

A. Upper, map of the Kcnq2 constructs. Lower, sequence alignments for the region between the middle of exon 5 and the beginning of exon 6. Although the human G256W variant is a single base substitution, Crispr/Cas9 editing introduced two substitutions, since the WT G256 codons differ between mouse (GGT) and human (GGG). Also aligned is the DNA and protein sequences of the frameshift mutation. B. WT and G256W/+ mice showed no difference in weight gain during development. C. WT and G256W/+ mice performed similarly in the developmental milestone assays for negative geotaxis, surface righting, and cliff aversion. D-H. Screenshots of stages of a generalized seizure in a P10 G256W/+ mouse (see also: Figure 4 - supplement 3 - Movie). D. Onset with immobility and myoclonic tail and forelimb shaking. E. Abrupt fall to side with flexion posturing. F. Evolution to hindlimb and tail extension posture. G. Immobility with flaccid appearance, interrupted by brief episodes of tail, individual limb myoclonus or clonus. H. Arouses, quickly regains upright posture, then normal mobility. Labels: time in 15 min source video.

Heterozygous G256W mice have increased CA1 pyramidal cell excitability.

A. Representative voltage responses to increasing current injection steps (step duration 1 sec) in CA1 pyramidal neurons from WT and G256W/+ mice. The resting membrane potential was held at -65 mV. B. Representative voltage responses to decreasing current injections steps (1s) in CA1 pyramidal neurons from WT and G256W/+ mice. C. Time to 1st action potential following step stimulus WT n=16, G256W/+ n=16), D. Summary graph showing the effect of one copy of G256W in the action potential count (WT n=16, G256W/+ n=16, F(12,180)=5.8 P<0.0001). Asterisks indicate statistically significant differences. Data are presented as mean and s.e.m.

Convulsive seizures in adult heterozygous G256W mice show stereotyped electrographic features and reduce survival.

A-B. EEGs of non-fatal and subsequent fatal seizure captured in a P54 male G256W/+ mouse (animal 2, panel C). Electrographic seizures were characterized by fast spiking, high amplitude activity lasting 15-20 s (highlighted in gray). C. Summary showing the sex, ages, duration of recordings and timing of seizures in 8 animals undergoing EEG. Turqoise hashmarks denote a survived seizure, red hashmark denote a fatal seizure. Black bars are periods on EEG; some recording were performed on a 6 hr/day schedule. D. Survival curve of WT vs G256W/+ mice, hashmarks indicate censored mice. G256W/+ mice had signifcant mortality, P = 0.0348 Cox propotional hazards model.

RT-qPCR shows distinct patterns of Kcnq2 and Kcnq3 compensatory change in heterozygous E254fs and G256W mice.

A. Kcnq2 and Kcnq3 mRNA levels in P21 hippocampus and neocortex. Compared to WT, Kcnq2 mRNA levels in E254fs/+ samples were significantly higher than the 50% expected for uncompensated haploinsufficiency (hippocampus: P = 1.28×10-5, neocortex: P = 3.5×10-7, one sample t-test). In G256W/+ mice, Kcnq3 mRNA significantly increased 1.15-fold (+/- 0.10, P = 0.0043) in the hippocampus, and 1.12-fold (+/- 0.09, P = 0.00213) in the neocortex. In E254fs/+ mice, Kcnq3 mRNA significantly increased (1.11 +/- 0.02- fold, P = 0.0245) in the hippocampus only. B. Kcnq2 and Kcnq3 mRNA expression in P100 hippocampus and neocortex. In E254fs/+, Kcnq2 levels were significantly higher than the expected 50% in hippocampus (P = 0.0003) and neocortex: (P = 0.0007, one sample t-test). For G256W/+, Kcnq3 mRNA significantly increased in the hippocampus only (1.16 fold +/- 0.09, P = 0.0026). Symbols are individual animals, n = 3 males and 3 females for adult experiments. For P21 experiments, one G256W/+ sample was removed as an outlier, as determined by the Grubbs test. See supplemental data for computation of one sample t-tests and Grubbs test. For P21 G256W/+ neocortex Kcnq2 expression, n = 1 male, and n = 3 females. For all other P21 experiments, n = 2 males, and n = 3 females. One way ANOVA, * = p<0.05, ** = p<0.005, *** = p<0.0005.

Heterozygous G256W mice show reduced KCNQ2 and KCNQ3 labeling of CA1 pyramidal cell AISs and increased labeling of neuronal somata.

Identically processed age P21 tissue sections of WT (upper) and G256W/+ (lower) mice; area CA1B was imaged under identical settings. Confocal image stacks are shown as maximal intensity projections. In the animation, channels for the indicated markers are allowed to fade into the next, enabling evaluation of colabeling. DAPI marks cell nuclei. AnkG strongly marks AISs and lightly labels somata and proximal apical dendrites. An arrow highlights one stratum oriens interneuron somatically labeled for KCNQ2 only. Labels: DAPI, 4’,6-diamidine-2’-phenylindole; so, Stratum oriens; sp, Stratum pyramidale, sr, Stratum radiatum. Scale: 50 μm. Link to movie F8

The ratios of axonal to somatic KCNQ2 and KCNQ3 labeling are reduced in CA1 and CA3 in heterozygous G256W mice.

A-B. The ratios of AIS to somatic immunofluorescence intensity is significantly reduced for KCNQ2 and KCNQ3 in CA1 (A) and CA3 (B) for G256W/+ but not E254fs/+ mice. C. The ratio of mossy fiber to somatic KCNQ2 and KCNQ3 immunofluorescence intensity is reduced in the CA3 for G256W/+ but not E254fs/+ mice. D. In the dentate gyrus, the ratio between GCL and PML intensity is significantly reduced for KCNQ3 but not KCNQ2 in G256W/+ but not E254fs/+ mice. n=3 per genotype. One way ANOVA, * = P<0.05, ** = P<0.005, *** = P<0.0005.

EEG recording including pre-ictal, post-ictal attenuation and recovery of background of seizure excerpted in Figure 1.

As labeled, onset was preceded by eyeblink and muscle artifact. The interval of uninterrupted voltage attenuation between the end of the high voltage fast activity to the first epileptiform burst was 61 sec. Interburst length progressively shortened in length, over about 3 min. Settings as in Figure 1. Link to movie F1-S1

Awake and sleep EEG background. Settings as in Figure 1, except LFF changed to 1 Hz.

Movie illustrating position of the G256W substitution within the KCNQ2 channel pore turret and its distance to the selectivity filter.

Link to movie F2-S1

The G256W variant affects a divergent neuronal KCNQ turret structure enabling forming a bonding network linked to the ion selective pore.

A-C. Aligned structural models of the extracellular portions of PGDs of KCNQ1, KCNQ4 and KcsA with that of KCNQ2. Single subunits are shown. D. Cartoon of structural model of turret region of KCNQ4 highlighting the predicted hydrogen bonding network. Several bonds are conserved between KCNQ2 and KCNQ4, but the KCNQ4 network has fewer bonds (compare with Figure 2G). E. Cladogram summarizing evolutionary relationships among several voltage-gated potassium channel genes. Gene duplication(s) are indicated by red circles, and are labeled by a common ancestor (or their extant descendant) possessing both duplicate genes. F. Sequence alignments of KCNQ1-5 P loops and flanking S5 and S6 regions reveal relative conservation of KCNQ5, KCNQ4, and KCNQ2 (as in Figure 2C), and divergence of KCNQ1 and KCNQ3.

Movie illustrating locations of residues contributing to a non-covalent bonding network extending from S5 to the selectivity filter.

Link to movie F2-S3

KCNQ2 G256W co-expression suppresses current in KCNQ2/ KCNQ3 heteromeric channels recorded by manual patch-clamp.

A. Representative current families for the indicated ratios of subunits. Note currents are larger than in Figure 3. B-C. Current/ voltage and conductance/voltage relationships for the indicated WT only and G256W/WT cells.

KCNQ2 G256W co-expression suppresses current in KCNQ2 homomeric channels recorded by manual patch-clamp.

A. Representative current families for the indicated ratios of subunits. Homomeric currents are smaller than in Figure 3 - supplement 1. B-C. Current/voltage and conductance/voltage relationships for the indicated WT only and G256W/ WT cells. D. Cartoon showing the expected combinations of WT and G256W subunits under heterozygosity based on a simple random association model. Mutant subunits are included in 15/16 of channel tetramers.

DNA, RNA, and predicted protein consequences of the G256W and E254fs*16 mutations.

Upper, DNA and predicted protein alignment of the frameshift mutation. Lower, Sanger sequence for cDNA from hippocampal mRNA of an E254fs/+ mouse. Splicing occurs at the WT junction, resulting in the predicted in-frame stop codon.

Western blotting reveals a lower molecular weight band (Mr ∼ 28 kDa) that is equally detected in WT and heterozygous E254fs, but no evidence of the E254fs truncated protein product.

A. Western blot of WT and E254fs/+ cortical homogenates (3 biological replicates per genotype, all males), probed with KCNQ2 N-terminal antibody. Black arrow indicates the monomer (Mr ∼90kDa), red arrow indicates the estimated relative mobility (∼29.73 kDa), of the truncated protein product made from the E254fs allele. Red asterisk indicates a ∼28 kDa band equally detected in both WT and E254fs/+. B. Same blot as in A but windowed to show higher molecular weight bands. Bands at ∼180 kDa and ∼360 kDa consistent with predicted mobility of KCNQ2 dimers and tetramers, respectively. A band at ∼250 kDa appears in all immunoblots of whole brain homogenates using our KCNQ2 N-terminal antibody. Nano-LC tandem mass spectrometry of peptides from an in-gel digest of this band showed multiple abundant proteins and few KCNQ2 peptides. C. Quantification of ∼28 kDa band from A.

Generalized seizure in a P10 heterozygous G256W mouse.

This movie includes from 9:01 to 11:05 of a 15:00 min period of open field observation. Animal recovers upright posture at 1:45 in the clip. Link to movie F4-S1

AP biophysical properties that are unchanged

A. RMP (WT n=16, G256W/+ n=16). B. Action potential amplitude (WT n=16, G256W/+ n=16), C. Action potential width (WT n=16, G256W/+ n=16), D. Action potential rise slope (WT n=16, G256W/+ n=16), F. Action potential decay (WT n=16, G256W/+ n=16), and E. input resistance (WT n=16, G256W/+ n=16).

Video of a fatal convulsive seizure in a 4 month old heterozygous G256W mouse.

Seizure onset (not shown) occurred 5-10 sec prior to the start of recording with wild running and jumping, followed by arrest, then resumed (start of video). This was again followed by loss of postural control, followed by sustained forelimb flexor/hindlimb extensor posturing. Attempts to resuscitate the animal were begun immediately and were unsuccessful. Link to movie F6-S1

No significant mortality in heterozygous E254fs mice.

Survival curve of WT vs E254fs/+ mice, hashmarks indicate censored mice. E254fs/+ mice showed no significant mortality, P = 0.452 Cox propotional hazards model.

E254fs and G256W variant transcripts are detectable in cDNA from heterozygous animals.

A. Upper, WT sequence (bases 758-771). Lower, Sanger trace of amplified Kcnq2 cDNA from WT hippocampus. B. Upper, WT sequence; the red line indicates bases deleted in the E254fs allele. Lower, Sanger trace of cDNA from an E254fs/+ mouse. Blue shading highlights the shift following the deletion at positions 761-767. Peaks corresponding to E254fs transcripts are labeled and are smaller than WT peaks. C. Upper, alignment of WT and missense variant DNA sequences, red lines highlight the two base substitutions at codon 256. Below, Sanger trace of amplified cDNA from G256W/+ hippocampus. Double peaks at the mutated bases positions are visible.

In CA1, the KCNQ2 and KCNQ3 cellular and subcellular immunolabeling patterns appear similar for WT and heterozygous E254fs mice.

Ankyrin-G marks position of AISs. KCNQ2 and KCNQ3 strongly label CA1 AISs in E254fs/+ mice, and do not show increased somatic labeling compared to WT. Highlighted by an arrow is one interneuron in stratum pyramidale that was somatically labeled for KCNQ2 only. Scale: 50 μm. Link to movie F8-S1

Heterozygous G256W mice show increased CA3 pyramidal cell somatic labeling and reduced mossy fiber labeling for KCNQ2 and KCNQ3.

Yellow lines demarcate the borders of sp; the sp-sl border is cut obliquely through the tissue section in the G256W/+ sample. PanNav strongly labels the unmylenated axons of the mossy fibers in stratum lucidum of both samples. PanNav also labels the obliquely cut AISs of pyramidal cell neurons, which are mostly located within sp. Scale: 50 µm. Link to movie F8-S2

A heterozygous G256W mouse CA1 Stratum oriens interneuron is somatically labeled for both KCNQ2 and KCNQ3.

A. Wider view of the same G256W/+ CA1 image stack shown in Figure 8 - Movie, showing corpus callosum (cc, partial), alveus (alv), so, sp, and a larger portion of sr. Four interneurons somatically labelled for KCNQ2 but not KCNQ3 are enclosed by red boxes. The interneuron indicated with an arrow in Figure 8 - Movie (KCNQ2 labeled, KCNQ3 unlabeled), is again highlighted with an arrow. The yellow box encloses an interneuron somatically co-labeled for KCNQ2 and KCNQ3. B. Individual laser channels for the interneuron enclosed by yellow box in A. The soma is labeled for both KCNQ2 and KCNQ3. The nearby AIS showing AnkG, KCNQ2, and KCNQ3 labeling likely arises from this cell, but its origin was not verified by higher resolution re-imaging. Scales: A: 50 µm, B: 10 µm.

KCNQ2 protein is reduced in neocortex of P21 heterozygous E254fs and G256W mice.

A-B. Representative western blots for all three genotypes probed for KCNQ2 and KCNQ3. C. Quantified KCNQ2 signal relative to wildtype. D. Quantified KCNQ3 signal relative to wildtype, normalized to protein loaded as assayed by BCA. n=3 per genotype, all males. One way ANOVA, * = P<0.05, ** = P<0.005, *** = P<0.0005.

Images of entire filter used for western blot of lysates, probed for KCNQ2 and KCNQ3.

A. PVDF filter with electrotransferred brain proteins was cut below the 75 kDa marker. Upper portion was probed for KCNQ2; lower portion was probed for tubulin. Black arrow points to KCNQ2 monomer band (Mr ∼90 kDa). Red arrow points to a band at ∼250 kDa. Mass spectrometry of in gel digest of this band showed multiple abundant proteins and few KCNQ2 peptides. Also visible are bands likely representing KCNQ2 dimer formation (Mr ∼180 kDa and higher molecular weight). B. KCNQ3 filter was cut below 75 kDa in order to probe for Tubulin. Black arrow points to predicted KCNQ3 monomer with a Mr ∼100 kDa. The red asterisk indicates a bad sample that was not included in the analysis seen in Figure 9 - supplement 1. C. KCNQ2 band intensity normalized to tubulin. D. KCNQ3 band intensity signal normalized to Tubulin.