An epilepsy-associated KV1.2 charge-transfer-center mutation impairs KV1.2 and KV1.4 trafficking

Significance A child with epilepsy has a previously unreported, heterozygous mutation in KCNA2, the gene encoding KV1.2 proteins. Four KV1.2 assemble into a potassium-selective channel, a protein complex at the neuronal cell surface regulating electrical signaling. KV1.2 subunits assemble with other KV1-family members to form heterotetrameric channels, contributing to neuronal potassium-channel diversity. The most striking consequence of this mutation is preventing KV1.2-subunit trafficking, i.e., their ability to reach the cell surface. Moreover, the mutation is dominant negative, as mutant subunits can assemble with wild-type KV1.2 and KV1.4, trapping them into nontrafficking heterotetramers and decreasing their functional expression. Thus, KV1-family genes’ ability to form heterotetrameric channels is a double-edged sword, rendering KV1-family members vulnerable to dominant-negative mutations in a single member gene.

3 assembled by the ribosome (1); (ii) the mixed-subunit RNA is introduced to the oocyte at precisely the same place and time.
We can challenge this idea by introducing a negative association bias; i.e., wild-type KV1.2 subunits are less likely to co-assemble with KV1.2(F233S). Effectively, this means that the availability of F233S subunits is diminished, which can be calculated by reducing the molar proportion of F233S by a factor <1. Conversely, multiplying by a factor >1 increases its effective concentration. We explore both ideas in Fig.S6. Decreasing F233S effective concentration narrows the spread of the curves, such that KV1.2 data are now better described by the black model (no heteromers / F233S-subunit rescue; Fig.S6B); and the KV1.4 data by the green model (no 2:2 stoichiometry; Fig.S6E). However, both models are contradicted by experimental results from flow cytometry, VCF and concatemer experiments. Increasing F233S effective concentration splays out the curves. For KV1.2, this means that data are best described by the cyan model (2:2 stoichiometry allowed; Fig.S6C); however this is inconsistent with concatemer experiments. No condition with bias >1 could explain the KV1.4 data better than unbiased association (Fig.S6F). To summarize, the "applied transcriptomics data", in combination with other experiments in this work, favor unbiased heteromeric channel assembly.
As discussed, the data go a long way in rejecting most alternatives to specific stoichiometry trafficking. We cannot completely exclude contributions from other mechanisms (such as nonstringent stoichiometric trafficking, open probability / single-channel conductance, biased subunit association), nor is this necessary: the main conclusions of the model are both falsifiable and well supported by our experimental data, all the more impressive as the model carries no free parameters: model predictions are simply overlaid, not fit to the data.

Patient genetic testing
A HIPAA-compliant consent form by Children's Hospital Los Angeles was signed by the patient's parents. DNA was extracted from the peripheral blood of the patient and each of his unaffected parents as comparators using a commercially available kit (Promega Maxwell RSC DNA Extraction Kit). The exome sequencing library was generated for proband DNA using the Agilent SureSelect Human All Exon V6+Mito kit. Captured DNA fragments were then sequenced using the Illumina Nextseq 500 sequencing system, with 2100 basepair (bp) paired-end reads. Single nucleotide variants (SNVs) and small insertions and deletions (<10 bp) were detected by mapping and comparing the DNA sequences with the human reference genome (GRCh37-hg19). Of all the variants identified by exome sequencing, a list of rare variants (minor allele frequency <1%) located within a predefined set of 224 epilepsy associated genes (Table S1) was generated. These variants were further annotated and analyzed using Alissa Interpret (Agilent). The genes included in this panel are known to be associated with primary epilepsy, or associated with syndromes in which epilepsy is a commonly observed feature. The genes (Table S1) were evaluated for single nucleotide substitutions and small indels. Both parents and the patient were tested for the variant by Sanger sequencing (Fig.1D). Variant interpretation was done according to the standards and guidelines of variant interpretation by American College of Medical Genetics and Genomics and Association for Molecular Pathology (2).

Molecular biology
Site-directed mutagenesis (SDM) to introduce F233S, A291C, etc., was performed with a highfidelity Pfu polymerase (Agilent 600850), unless otherwise stated. All molecular biology reagents were provided from New England Biolabs, and all synthetic oligonucleotides from Integrated DNA Technologies, unless stated otherwise. All molecular biology operations were confirmed by sequencing.

Immunocytochemistry constructs
For initial two-color immunocytochemistry studies ( Fig.2B-D), a plasmid containing rat KV1.2 with an N-terminally fused EGFP tag (in the pEGFP-C1 vector) and an extracellular HA site (between transmembrane helices S1 and S2) was used; it was a generous gift from Lily Y. Jan (3,4).
To incorporate a bungarotoxin binding site (BBS; TGG CGG TAC TAC GAG AGC AGC CTG GAG CCC TAC CCC GAC) in KV1.2(HA) and KV1.2(HA,F233S) constructs, the HA tag was excised using BstEII and EcoRI and replaced by a synthetic BBS-containing oligo. The resulting constructs had a BBS at the same position as the excised HA tag. Rat brain KV1.4 in the GW1 plasmid was gifted by James S. Trimmer (5). A hemagglutinin (HA) tag (TAT CCG TAC GAC GTC CCA GAC TAT GCG) was introduced at the S1-S2 extracellular linker after position G346. This was performed by excising a portion of the KV1.4 N-terminus between unique sites NotI and BsmI, and replacing it with a gBlocks synthetic fragment including the HA tag sequence. The new construct was evaluated by cut-open oocyte Vaseline gap electrophysiology (see below): it exhibited fast N-type inactivation, and activated with a voltagedependence similar to the tag-less channel ΔV0.5=−134mV (n=4 per condition).

Electrophysiology constructs
The rat KV1.2 cDNA construct in pMAX oocyte expression vector was a kind gift from Benoît Roux.
To construct concatemeric dimers, a BamHI site was first introduced upstream of KV1.4, to match the 5 untranslated region (UTR) of KV1.2, by SDM. The existing KV1.x (x=2, 2(F233S) or 4) open reading frames (ORFs) would then become the C-terminal domains of the concatemer. Nterminal KV1.x domains, originally generated by PCR amplification, were then introduced, along with a synthetic linker sequence modified from the 5 UTR of the Xenopus β-globin gene (7,8). The N-terminal ORFs lacked a stop codon, and the sequence between the two KV1.x domains was: GCT AGC GAT ACG AAG GAG CGA GGA AAC CTC TTC ACG TCA ACC GGA TCC GCC GCC ACC ATT (NheI and BamHI sites are underlined at the 5 and 3, respectively; bold indicates preserved pMAX sequence preceding the start codon of the C-terminal KV1.x domain), which corresponds to the 20-residue linker: ASDTKERGNLFTSTGSAATI. pMAX plasmids were linearized using PacI, transcribed to cRNA in vitro (T7 mMESSAGE MACHINE, Thermo Fisher Scientific) and stored at −80 C in RNA storage solution (Thermo Fisher Scientific). cRNA was quantified spectrophotometrically and evaluated by gel electrophoresis. Immunocytochemistry / confocal microscopy: Primary (Rat anti-HA; Roche 3F10) and secondary (AlexaFluor 568 conjugated goat anti-rat IgG(H+L); Invitrogen A-11077) antibodies were diluted 1:200 and 1:1000, respectively, in blocking solution. 48 hr post-transfection, COS-7 cells were rinsed with ice-chilled phosphate-buffered saline (PBS) supplemented with 0.9 mM Ca 2+ and 0.5 mM Mg 2+ (Gibco 14040-133). Cells were fixed with 4% paraformaldehyde in PBS (without Ca 2+ or Mg 2+ , 5 min), and washed with PBS (once quickly, then 35 min each) prior to incubation with blocking solution (2 ml of 5% normal goat serum in PBS) for at least 1 hr at room temperature. Blocking solution was replaced with primary antibody solution (0.5 ml) and cells were incubated at room temperature for at least 2 hrs. Cells were washed with PBS prior to incubation with secondary antibody solution (0.5 ml, 1 hr at room temperature). Unbound antibodies were washed off with PBS and a coverslip was mounted over the glass-bottom of each dish using ProLong Glass Antifade Mountant (Invitrogen P36982). Mountant was allowed to set for 18-24 hr at room temperature, then samples stored at 4C. Confocal images were acquired using a Zeiss LSM 800 Microscope with Zen 2.3 software. A z-stack was acquired for each cell (z-section interval = 0.27 μm). Zen Analysis software was used to create an orthogonal projection of 2 sequential z-sections. Final images were prepared for publication using Adobe Photoshop CS6. Identical adjustments to levels, sharpness and pixel sampling were performed on all samples. Epitope Tag Antibody; BioLegend Cat# 682404, RRID:AB_2566616) as previously described (4,9). Cells were detected using a Gallios Flow Cytometer (Beckman Coulter Life Sciences) as previously described (4). The fraction of anti-HA positive cells in the EGFP positive population, normalized to wild type KV1.2(HA) measurements was calculated.

Electrophysiology & voltage-clamp fluorometry
Oocyte preparation: All animal experiments were approved by the Linköping University Animal Care and Use Committee. Defolliculated Xenopus laevis oocytes were either purchased from Ecocyte or isolated and prepared from locally-kept Xenopus laevis (Nasco, WI, USA) as follows: Frogs were anesthetized with 1.4g/L tricaine (ethyl 3-aminobenzoate methanesulfonate, Sigma #A5040). Lobes of ovaries were removed through an abdominal incision and placed into Ca 2+free OR-2 solution (in mM: 82.5 NaCl, 2.5 KCl, 1 MgCl2 and 5 HEPES; pH adjusted to 7.4 by NaOH) for further processing. Frogs were treated with analgesics (5mg/mL Marcain; Astra Zeneca & 2% Xylocain gel; Aspen) and the incision was sutured, prior to being returned to a recovery aquarium for post-surgical monitoring. Ovaries were cut into small clusters of 5-7 oocytes, then enzymatically treated with Liberase TM (7 Units/batch; Roche 05401127001) in ~10 ml of OR-2 with agitation using an orbital shaker for 25-40 min. Liberase was removed by washing with OR-2 solution, then manual agitation for 30-60 min was employed to remove follicular layers. Mature (stage V-VI) defolliculated oocytes were selected and stored at 17°C in SOS (in mM: 100 NaCl, 2 KCl, 1.8 CaCl2, 1 MgCl2 and 5 HEPES; pH adjusted to 7.0 by NaOH).
Parallel two-electrode voltage clamp (TEVC) recordings: Oocytes were injected with 50 nl of cRNA solution using a nanoinjector (UMP3T-1, World Precision Instruments). Specific cRNA amounts used are mentioned in their corresponding experiments. One day after injection, oocytes were recorded using a medium throughput robot capable of eight simultaneous TEVC recordings (OpusXpress 6000A Parallel Oocyte Voltage Clamp; Axon Instruments). Recording electrodes were pulled from TW150F-6 thin-wall capillary glass (World Precision Instruments) on a P-1000 horizontal puller (Sutter Instruments). Only electrodes with resistance ≤1 Ω (I-electrodes), or ≤1.5 Ω (V-electrodes) were used. Electrodes were filled with 3 M KCl. The oocyte recording chambers were perfused continuously with SOS. Once oocytes in all chambers were correctly impaled by both electrodes, oocytes were clamped at a holding potential of −80mV. P/−6 subtraction was used to limit capacitative transients. For KV1.2(WT/F233S) experiments (Figs. 3, 5A, S2A,B), voltage-dependent activation was evaluated using a series of 100ms test-pulses from −80mV to 120mV in 10-mV increments. For experiments including KV1.4 subunits (Figs. 7A-C, S3), voltagedependent activation was evaluated using a series of 100ms test-pulses (from −90mV to 100mV in 10-mV increments). A pre-pulse to −120mV for 800ms preceded each test-pulse to completely relieve N-type inactivation. A 10 s inter-pulse interval was used in all recordings. Current was lowpass filtered to 2 kHz and acquired at 6.25 kHz using the OpusXpress digitizer and OpusXpress 1.1 software (Axon Instruments).
Voltage clamp fluorometry (VCF): VCF was performed by adding epifluorescence and light detection capability to the COVG system (12,13). Prior to mounting on the COVG apparatus, oocytes were stained at room temperature for 5 minutes with 20 μM MTS-TAMRA fluorophore in a depolarizing solution (in mM: 120 K-MeS, 2 Ca(MeS)2, 10 HEPES, pH=7.0), in the dark, to label the introduced Cys (A291C). The oocytes were then rinsed in dye-free SOS prior to being mounted in the recording chamber. Fluorescence emission and ionic current were simultaneously measured from the same area of membrane isolated by the top chamber. The same electrophysiological apparatus and solutions were used as above. The optical setup consisted of an Olympus BX51WI upright microscope with filters (Semrock Brightline) appropriate for rhodamine excitation and emission wavelengths (exciter: FF01-531/40-25; dichroic: FF562-Di02-25x36; emitter: FF01-593/40-25). The light source was a 530 nm, 170 mW LED (M530L3-C1, Thorlabs) driven by a Cyclops LED driver (Open Ephys). The objective (Olympus LUMPLANFL, 40XW, water immersion) had a numerical aperture of 0.8 and a working distance of 3.3 mm, which left sufficient room for the insertion of the microelectrode while fully covering the oocyte domus exposed in the external recording chamber. The emission light from the camera port was focused on a Si photodiode (SM05PD3A, Thorlabs) connected to a current amplifier (DLPCA-200, FEMTO). Its output was filtered to 5 kHz using an in-line low-pass Bessel filter (LPF-8, Warner Instruments) before being acquired at 25 kHz (Digidata 1550B1 and pClamp v.11, Molecular Devices).

Data analysis and modeling
Electrophysiological data analysis: All electrophysiological data analysis was performed by least squares fitting in Microsoft Excel. In experiments with just KV1.2 subunits, steady-state activation was calculated by fitting the macroscopic conductance to a Boltzmann distribution: where Vm was the membrane potential; V0.5 was the half-activation potential; z was the effective valence; F and R the Faraday and Gas constants, respectively; T was temperature (294 K). The maximal macroscopic conductance, G, was calculated by dividing the current (Im) by the driving force: where EK was the equilibrium potential for potassium. In TEVC experiments, the Vm reported by OpusXpress at the same time as the sampled Im was used.
In experiments with KV1.4 subunits, peak current was used. The G-V curve could be better accounted-for with the sum of two Boltzmann distributions: where w is the fractional contribution, or weight, of each component, and w1=1−w2.
The voltage dependence of fluorescence deflections (ΔF; i.e., VSD activation) was estimated by fitting ΔF to a Boltzmann distribution: [4] where ΔFmax and ΔFmin were the maximal and minimal ΔF asymptotes, respectively.

Binomial models of heteromeric channel configurations:
The probability that a tetrameric channel with specific subunit composition will form was calculated starting with the standard binomial distribution: nk k n f n k p p p k −  =−   [5] where n is the number of trials (always 4, for the obligate tetrameric KV channels); k is the number of "successes", i.e., KV1.2(WT) or KV1.4 subunits in the tetramer (k=4 for WT homo-tetramers, k=3 for 3:1 WT:F233S hetero-tetramers, and so on); p is the probability to encounter a KV1.2(WT) or KV1.4 subunit during tetramerization, given by: where m is the relative molar proportion of injected KV1.2(F233S) cRNA, ranging from 0 to 8.
In these experiments (Figs. 5A, 7A), the molar proportion of KV1.2(F233S) cRNA increased, while the amount of KV1.2(WT) or KV1.4 cRNA was kept stable. Therefore, more tetramers could potentially be made with higher F233S proportions due to the increased availability of KV subunits. For this reason, the frequency of tetramers (f, eq. [5]) was scaled by the relative amount of cRNA: ( ) , , ( 1) g f n k p m =  + [7] The black curves in the model (Figs. 5A, 7A) represent g with k=4; the green curves are the sum of g with k=4 and k=3; and so on.
Statistics: All comparative statistics (e.g., "relative Gtotal" or "ΔV0.5") were performed among oocytes or COS-7 cells from the same block of experiments (same batch / passage number, RNA injection / DNA transfection, and experiment day). All significance tests were two-tailed Student's t-tests. Errors are SEM or 95% confidence intervals (C.I.), as stated in the figure legends. relative to oocytes expressing 1 RNA (i.e., 0.5ng/oocyte). This amount is not statistically significant from a sample with mean=9 (i.e., 9-fold increase in conductance) and the same standard deviation and number of observations (p=0.21). This experiment was performed under COVG, to handle the increased conductance, up to 730 μS, with better voltage control; in addition, current is recorded from 20% of the oocyte surface area (12), further decreasing macroscopic currents that could generate clamp errors. It shows that the conductance observed in the presence of increasing F233S cRNA (Figs.5A & 7A) was not under-estimated due to potential oocyte translation issues. (D) Voltage dependence of 1 and 9 WT RNA condition. (E) Parameters for the analysis in panes C and D. Errors are 95% C.I..      (Figs. 4, 6, S1) and VCF (Fig.8B (Fig.5B,C).  (Fig.7D,E).