Reduced axonal surface expression and phosphoinositide sensitivity in K_v7 channels disrupts their function to inhibit neuronal excitability in Kcnq2 epileptic encephalopathy.

Abstract

Neuronal K_v7/KCNQ channels are voltage-gated potassium channels composed of K_v7.2/KCNQ2 and K_v7.3/KCNQ3 subunits. Enriched at the axonal membrane, they potently suppress neuronal excitability. De novo and inherited dominant mutations in K_v7.2 cause early onset epileptic encephalopathy characterized by drug resistant seizures and profound psychomotor delay. However, their precise pathogenic mechanisms remain elusive. Here, we investigated selected epileptic encephalopathy causing mutations in calmodulin (CaM)-binding helices A and B of K_v7.2. We discovered that R333W, K526N, and R532W mutations located peripheral to CaM contact sites decreased axonal surface expression of heteromeric channels although only R333W mutation reduced CaM binding to K_v7.2. These mutations also altered gating modulation by phosphatidylinositol 4,5-bisphosphate (PIP₂), revealing novel PIP₂ binding residues. While these mutations disrupted K_v7 function to suppress excitability, hyperexcitability was observed in neurons expressing K_v7.2-R532W that displayed severe impairment in voltage-dependent activation. The M518V mutation at the CaM contact site in helix B caused most defects in K_v7 channels by severely reducing their CaM binding, K⁺ currents, and axonal surface expression. Interestingly, the M518V mutation induced ubiquitination and accelerated proteasome-dependent degradation of K_v7.2, whereas the presence of K_v7.3 blocked this degradation. Furthermore, expression of K_v7.2-M518V increased neuronal death. Together, our results demonstrate that epileptic encephalopathy mutations in helices A and B of K_v7.2 cause abnormal K_v7 expression and function by disrupting K_v7.2 binding to CaM and/or modulation by PIP₂. We propose that such multiple K_v7 channel defects could exert more severe impacts on neuronal excitability and health, and thus serve as pathogenic mechanisms underlying Kcnq2 epileptic encephalopathy.

INTRODUCTION

Epilepsy is a common neurological disorder that strikes about 1.2% of the US population (www.cdc.gov). It is caused by excessive neuronal excitability characterized by seizures, which are abnormal and uncontrolled discharges of action potentials (Fisher et al., 2005). About 40% of all epilepsies are estimated to be caused by genetic factors including mutations in ion channels and occurs as early as infancy (Guerrini and Noebels, 2014). Among ion channels, neuronal K_v7/KCNQ potassium (K⁺) channels have emerged as critical players in epilepsy because their agonist (ezogabine/retigabine) suppresses seizures in animal models and was clinically used as an anti-epileptic drug until 2017 (Blackburn-Munro et al., 2005; Gunthorpe et al., 2012; Large et al., 2012). Importantly, inherited and de novo mutations in Kcnq2 and Kcnq3 genes encoding K_v7.2 and K_v7.3 subunits are associated with early-onset epileptic disorders including benign familial neonatal epilepsy (BFNE) and epileptic encephalopathy (RIKEE database: www.rikee.org).

Neuronal K_v7 channels are mostly heterotetramers composed of K_v7.2 and K_v7.3 subunits (Shah et al., 2002; Wang et al., 1998), each subunit containing six transmembrane segments (S1–S6), cytoplasmic N- and C-termini, and the pore domain between the S5–S6 segments (Jentsch, 2000). Distributed throughout the brain including the hippocampus and neocortex (Cooper et al., 2001; Geiger et al., 2006; Pan et al., 2006; Weber et al., 2006) and in the peripheral nervous system (Dedek et al., 2001; Schwarz et al., 2006), K_v7 channels generate slowly activating and non-inactivating voltage-dependent K⁺ current (Jentsch, 2000) that potently suppresses burst and repetitive firing of action potentials (APs) and contribute to resting membrane potential (Shah et al., 2008; Shah et al., 2002; Yue and Yaari, 2004; Yue and Yaari, 2006) (Tzingounis and Nicoll, 2008). Their current is also called ‘M-current’ since their inhibition by muscarinic agonists leads to a profound increase in AP firing (Brown and Passmore, 2009; Shah et al., 2002; Wang et al., 1998). Subcellularly, K_v7 channels are preferentially enriched at axonal plasma membrane compared to somatodendritic plasma membrane with highest concentration at the axonal initial segments (AIS) (Cavaretta et al., 2014; Chung et al., 2006), the critical sites for AP initiation and modulation (Clark et al., 2009). Their AIS localization is mediated by their interaction with ankyrin-G (Chung et al., 2006; Pan et al., 2006; Rasmussen et al., 2007), an essential component of the AIS (Bennett and Lorenzo, 2016; Jenkins et al., 2015). Consistent with their axonal enrichment, K_v7 channels regulate spike threshold, frequency, and shape (Gu et al., 2005; Peters et al., 2005; Shah et al., 2008; Soh et al., 2014; Tzingounis and Nicoll, 2008; Yue and Yaari, 2004; Yue and Yaari, 2006).

The first BFNE mutations were mapped to Kcnq2 in 1998 (Biervert et al., 1998; Charlier et al., 1998; Singh et al., 1998). Since then, over 130 mutations in Kcnq2 and Kcnq3 have been associated with dominantly inherited BFNE in humans (RIKEE database: www.rikee.org). Although BFNE is a rare disorder with transient appearance of seizures in neonates, the risk of recurring seizures later in life is 10–13% (Psenka and Holden, 1996). Furthermore, some BFNE mutations also accompany peripheral nerve hyperexcitability, myokymia, drug-refractory seizures, and developmental delay (Borgatti et al., 2004; Dedek et al., 2003; Dedek et al., 2001; Schmitt et al., 2005; Wuttke et al., 2007). Recently, multiple de novo mutations in K_v7.2 have been discovered in patients with early onset epileptic encephalopathy characterized by drug resistant seizures, psychomotor delay, developmental delay, distinct electroencephalogram (EEG) and neuroradiological abnormalities including white matter reduction and enlarged ventricles (Abidi et al., 2015; Kato et al., 2013; Kwong et al., 2015; Milh et al., 2013; Millichap et al., 2016; Numis et al., 2014; Saitsu et al., 2012; Weckhuysen et al., 2013; Weckhuysen et al., 2012; Zhang et al., 2016). Some de novo mutations of K_v7.2 are associated with Ohtahara and West syndromes, which are severe symptomatic drug resistant epilepsy with poor prognosis and profound psychomotor delay (Kato et al., 2013; Samanta et al., 2015). BFNE mutations have been extensively studied for their effects on K_v7 current and surface expression (Borgatti et al., 2004; Cavaretta et al., 2014; Chung et al., 2006; Dedek et al., 2003; Dedek et al., 2001; Maljevic and Lerche, 2014; Schmitt et al., 2005; Wuttke et al., 2007). However, pathogenic mechanisms underlying epileptic encephalopathy mutations of K_v7.2 have not been fully characterized. It is also unclear how BFNE and epileptic encephalopathy mutations of Kcnq2 and Kcnq3 ultimately lead to epilepsy with distinct severity.

Interestingly, one third of epileptic encephalopathy mutations in K_v7.2 are located in helices A-C within the large cytoplasmic C-terminal tail (Abidi et al., 2015; Kato et al., 2013; Kwong et al., 2015; Milh et al., 2013; Millichap et al., 2016; Numis et al., 2014; Saitsu et al., 2012; Weckhuysen et al., 2013; Weckhuysen et al., 2012; Zhang et al., 2016) that mediates channel assembly and interaction with multiple signaling molecules and proteins (Haitin and Attali, 2008) including phosphatidylinositol 4,5-bisphosphate (PIP₂) (Delmas and Brown, 2005; Suh and Hille, 2007; Suh et al., 2006), calmodulin (CaM) (Wen and Levitan, 2002; Yus-Najera et al., 2002), syntaxin-1A (Regev et al., 2009), and A-kinase-anchoring proteins (AKAPs) (Hoshi et al., 2005; Hoshi et al., 2003). PIP₂ binding is critical for proper function of neuronal K_v7 channels (Hernandez et al., 2008; Suh and Hille, 2002), although the precise PIP₂ binding residues in K_v7.2 are unknown. While helix C is responsible for subunit tetramerization (Schwake et al., 2006), CaM binds to helices A and B of all K_v7 subunits (Sachyani et al., 2014; Strulovich et al., 2016; Sun and MacKinnon, 2017; Xu et al., 2013; Yus-Najera et al., 2002). In crystal structure, CaM binds to the antiparallel coiled-coil architecture of helices A and B of K_v7.2 and K_v7.4 with helix A binding to CaM C-lobe, and helix B interacting with the CaM N-lobe (Strulovich et al., 2016; Xu et al., 2013; Yus-Najera et al., 2002). We have recently shown that disruption of CaM binding to K_v7.2 helix A impairs enrichment of heteromeric K_v7.2/ K_v7.3 channels at the axonal surface by promoting their retention in the endoplasmic reticulum (ER) (Cavaretta et al., 2014). Furthermore, decreasing CaM levels or dissociating CaM from K_v7.2 can reduce M-current density and enhance excitability in hippocampal neurons (Shahidullah et al., 2005). However, whether epileptic encephalopathy mutations in helices A-C alter CaM and PIP₂ affinity and subsequently disrupt K_v7 surface expression and function remains unknown.

In this study, we explored molecular pathogenetic effects of selected epileptic encephalopathy mutations (R333W, M518V, K526N, and R532W) in helices A and B of K_v7.2. We chose these specific mutations for several reasons. First, these mutations cause early-onset epileptic encephalopathy with drug-refractory seizures, profound psychomotor delay, and other behavioral comorbidities including autism and speech impairment. Second, we want to compare the effects of mutations buried within CaM contact sites (M518V in helix B) to those located at the periphery (R333W in helix A, K526N in helix B, and R532W in helix B-C linker). We found that the R333W, K526N, and R532W mutations located peripheral to CaM contact sites in helices A and B reduced axonal surface enrichment of heteromeric channels by 30% and disrupted their ability to suppress excitability in hippocampal neurons, although not all of them significantly impaired CaM binding or voltage-dependent activation. However, gating regulation by PIP₂ was impaired by R333W and R532W mutations, revealing these positively charged basic residues as novel PIP₂ modulation sites close to CaM contact sites. In contrast, M518V at the CaM contact site in helix B caused severe impairment in CaM binding, axonal surface expression, and voltage-dependent activation. Interestingly, this mutation also induced de novo ubiquitination and accelerated proteasome-dependent degradation of K_v7.2, whereas the presence of K_v7.3 blocked such degradation. Lastly, expression of K_v7.2-M518V enhanced neuronal injury and death. Taken together, these findings led to a surprising conclusion that pathogenic mechanisms of epileptic encephalopathy mutations in helices A-B of K_v7.2 involves a combination of distinct defects in K_v7 channels as well as neuronal injury depending on the nature of a mutation to disrupt CaM and PIP₂ affinity.

MATERIALS AND METHODS

Materials.

Antibodies used include anti-MAP2 (Millipore, AB5622), anti-ankyrin G (Neuromab, 75–146), anti-phospho IκBα Ser32 (14D4) (Cell Signaling, 2859), anti-GFP (Abcam, ab13970; Cell Signaling 2955S), anti-KCNQ2 (Alomone, APC-050; Neuromab, N26A/23), anti-KCNQ3 (Alomone, APC-051) anti-HA (Covance, MMS-101P, Cell Signaling, 3724 & 2369), anti-CaM (Cell Signaling, 4830), anti-GAPDH (Cell Signaling, 2118), HRP-conjugated secondary antibodies (The Jackson Laboratory, 715-005-151, 715-035-150, 711-035-152, 703-035-155) and Alexa Fluor dye-conjugated secondary antibodies (Invitrogen/Molecular Probes, A21202, A11072, A21076). Reagents used include 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), DL-2-amino-5-phosphonopentanoic (DL-AP5) and bicuculline (Sigma, C239, A5282, and 14343).

DNA Constructs and Mutagenesis.

Plasmids pcDNA3 with Kcnq2 cDNA (Y15065.1) encoding K_v7.2 (CAA 75348.1), HA-K_v7.2, K_v7.2-A343D, Kcnq3 cDNA (NM_004519) encoding K_v7.3, and HA-K_v7.3 have been described (Cavaretta et al., 2014; Chung et al., 2006; Schwake et al., 2000). Kcnq2 mRNA (Y15065.1) is the first Kcnq2 isoform isolated by positional cloning from fetal brain cDNA library (Biervert et al., 1998), and is identical to the reference sequence of Kcnq2 transcript variant 3 (NM_004518.5) that contains a distinct 3’ UTR and lacks two alternate in-frame exons compared to the longest transcript variant 1 (NM_172107.3). In addition to functional characterization of this transcript variant 3 (Biervert et al., 1998), we have previously reported that K_v7 channels composed of this shorter K_v7.2 isoform (CAA 75348.1) and K_v7.3 are preferentially localized to the axonal plasma membrane whereas BFNE mutations block this targeting in cultured hippocampal neurons (Cavaretta et al., 2014; Chung et al., 2006). Plasmid pJPA7 containing wild-type rat CaM was a gift from Dr. John Adelman (Oregon Health and Science University, Portland, OR) (Xia et al., 1998). The rat and human protein sequences of CaM are identical. EYFP-hCaM was a gift from Dr. Emanuel Strehler (Addgene plasmid # 47603). Epileptic encephalopathy mutations (R333W, M518V, R532W, K526N) were generated using the Quik Change II XL Site-Directed Mutagenesis Kit (Agilent) and verified by sequencing the entire cDNA construct.

Structural modeling and visualization.

The functional domains of K_v7.2 (NP_742105.1) from the longest transcript variant 1 (NM_172107.3) (Fig. 1A) were annotated based on multiple sources (Brown and Passmore, 2009; Chung, 2014; Strulovich et al., 2016; Xu et al., 2013) (RIKEE database: www.rikee.org). To model the interaction between calmodulin and K_v7.2 helices A and B, the sequence of human K_v7.2 helix A (amino acid E322-V367) was threaded to the crystal structure of Ca²⁺-CaM bound to a chimeric Q3A-Q2B peptide (PDB 5J03) (Strulovich et al., 2016). The structure was relaxed with Rosetta (Leaver-Fay et al., 2011) using two rounds of sequential rotamer, side chain and backbone minimization, followed by rigid body minimization. The lowest scoring decoy from 20 runs was chosen as the final model. Site-specific mutations were made to the model in Rosetta followed by sequential rotamer, side chain, backbone, and rigid body minimization, and the lowest scoring of 20 decoys is presented in Figure 1D. Structures were visualized using the PyMOL Molecular Graphics System, Version 2.0 (Schrödinger, LLC). Amino acid residues R333, A343, M546, K554 and R560 in K_v7.2 isoform a (NP_742105.1) correspond to R333, A343, M518, K526 and R532 in K_v7.2 isoform c (CAA_75348.1).

Figure 1. Distribution of epileptic encephalopathy mutations in CaM binding helices A and B of K_v7.2.

(A) K_v7.2 protein (NP_742105.1) showing functional domains including transmembrane S1–S6, a pore, an intracellular C-terminal tail containing helices A-D and ankyrin-G binding motif. Missense epileptic encephalopathy mutations are concentrated in helix A (residues 324–350), helix B (residues 532–557) and the linker between helix B and helix C (residues 568–589). The mutations (R333W, M546V, K554N) characterized in this study (red) are shown. (B) Amino acid sequence alignment of K_v7.2 (NP_742105.1) and a shorter K_v7.2 isoform (CAA_75348.1 used in this study) showing helices A and B (blue), CaM-binding consensus IQ motif in helix A (underlined), A343D mutation (purple), and epileptic encephalopathy mutations (red). (C) Amino acids (R333, M518, K526) mutated in epileptic encephalopathy and characterized in this study (red) are shown on a model of K_v7.2 helices A and B (blue) in complex with Ca²⁺-bound CaM (green). A343 (purple) is located in helix A at the CaM C-lobe contact. R532 (not shown) is located distal to the helix B of this structure. (D) The modeled structure of wild type K_v7.2 helix A-B (blue) bound to Ca²⁺-bound CaM (green) is overlaid with the lowest energy model of the mutant K_v7.2 helix A-B (mutations in red) complexed with Ca²⁺-bound CaM (grey).

Immunoprecipitation.

HEK293T cells cultured in 60 mm cell culture dish were transfected with plasmids (total 1.6 μg) containing K_v7.2 or EYFP-hCaM (1:1 ratio), or K_v7.2 and HA-K_v7.3 (1:1 ratio) using FuGENE6 transfection reagent (Promega). At 24 h post transfection, the entire cells were lysed and subjected to immunoprecipitation as described (Cavaretta et al., 2014) with small amount of rabbit anti-K_v7.2 antibody (5 μg) that could immunoprecipitate a fraction but not all of transfected K_v7.2 proteins. This allowed us to immunoprecipitate the equal amount of K_v7.2 proteins and analyze the effects of mutations on the amount of co-immunoprecipitated EYFP-hCaM and HA-K_v7.3 by immunoblotting with anti-GFP and anti-HA antibodies. To analyze K_v7.2 ubiquitination, immunoblotting with anti-HA antibody was performed on immunoprecipitated K_v7.2 proteins from cells transfected with plasmids containing HA-ubiquitin (0.3 μg), K_v7.2 (0.8 μg), with or without K_v7.3 (0.8 μg). For immunoblotting in Fig. S1, HEK293T cells were transfected with plasmids (total 1.6 μg) containing K_v7.2 wild-type or mutants with or without CaM. The lysates were analyzed by immunoblotting with anti-CaM, anti-K_v7.2, and anti-GAPDH antibodies.

Experimental animals and neuronal culture.

All procedures involving animals were reviewed and approved by the Institutional Animal Care and Use Committee at the University of Illinois Urbana-Champaign in accordance with the guidelines of the U.S National Institutes of Health (Protocols 15222). Primary dissociated hippocampal cultures were prepared from 18-day old embryonic rats and transfected with plasmids (total 1.0 μg) at 5 DIV as described (Cavaretta et al., 2014).

Immunocytochemistry.

Primary dissociated hippocampal cultures were prepared from 18-day old embryonic rats, transfected with plasmids (total 1.0 μg) at 5 DIV, and immunostaining for surface and total HA-K_v7.3/ K_v7.2 channels were performed at 48 h post transfection as previously described (Cavaretta et al., 2014). Immunostaining for surface and total K_v7 subunits and AIS markers (ankyrin-G and phospho IκBα-Ser32) in hippocampal neurons and HEK293T cells were performed at 24–48 h post transfection as described (Cavaretta et al., 2014). Fluorescence images were acquired as described (Cavaretta et al., 2014) using Zeiss Axiovert 200M inverted microscope equipped with AxioCam HRm Camera and Axiovert software, and stored with no further modification as ZVI and 16-bit TIFF files. Within one experiment, the images were acquired using the same exposure time to compare the fluorescence intensity of the neurons transfected with different constructs. The analyses of surface and total HA-K_v7.3/ K_v7.2 channels in Figure 3 were performed only from the healthy transfected neurons identified by differential interference contrast (DIC) imaging and the fluorescence imaging of HA-K_v7.3 immunostaining. If the transfected neurons had beaded or broken dendrites or axons, or damaged soma (i.e. dead neurons), then they were excluded from analyses. The background-subtracted mean fluorescence intensity of the soma, the axon within 0–30 μm of the beginning of the axon (AIS), the axon between 50–80 μm from the beginning of the axon (distal axon), and the major primary dendrites were quantified using ImageJ Software (National Institutes of Health) as described (Cavaretta et al., 2014). In Figure 8, the number of all K_v7.2-transfected neurons, the number of K_v7.2-transfected dead neurons, and the number of the K_v7.2-transfected neurons showing prominent immunostaining of endogenous ankyrin-G at the AIS were counted.

Figure 3. Polarized axonal surface expression of K_v7 channels is severely decreased M518V mutation, and modestly reduced by by R333W, K526N, and R532W mutations.

Immunocytochemistry showing surface and total HA-K_v7.3 proteins in hippocampal neurons cotransfected with K_v7.2 WT or K_v7.2 containing epileptic encephalopathy mutations. (A) Representative images of surface HA-K_v7.3 as pseudo-color that display differences in the surface HA intensities from high (red) to low (blue). (B) Representative images of total HA-K_v7.3 (Upper-inverted gray) and the axonal initial segments (AIS) identified by anti-phospho IκBα-Ser32 (14D4) antibodies (Lower-fluorescence) from the same neurons shown in (A). Arrows mark the AIS. Scale bars: 20 μm. (C) Background subtracted mean intensities of surface and total HA fluorescence from healthy neurons that were untransfected (None: n=20) or transfected with HA-K_v7.3 and K_v7.2 WT (n=36), R333W (n=35), M518V (n=22), K526N (n=39), or R532W (n=38) in 3 independent experiments. AU, arbitrary unit. (D) Surface HA intensity ratio at distal axon over dendrite. Data shown represent the Ave ± SEM (*p<0.05, **p<0.01, ***p<0.005).

Figure 8. Expression of K_v7.2 causes cell death in transfected HEK293T cells and hippocampal neurons.

(A) Immunocytochemistry of K_v7.2 in HEK293T cells at 24 h post transfection. Dapi staining shows nuclei. (B-D) Immunocytochemistry of GFP-positive HEK293T cells cotransfected with K_v7.2 WT or M518V alone, or together with HA-K_v7.3. Representative images are shown in Supplemental Fig. S5. (B) Background subtracted mean intensity of total K_v7.2 fluorescence per transfected cell. (C) Quantification of small round transfected cells with average area of 9.9 ± 0.2 μm² as % transfected cells per image. Total number of transfected cells analyzed in n=10–18 images: K_v7.2-WT (n=507), K_v7.2-M518V (n=389), HA-K_v7.3+K_v7.2-WT (n=186), and HA-K_v7.3+K_v7.2-M518V (n=731). The average area of untransfected cells was of 14.5 ± 0.6 μm². (D) Average nucleus area per transfected cell. Total number of transfected cells analyzed: K_v7.2-WT (n=40), K_v7.2-M518V (n=38), HA-K_v7.3+K_v7.2-WT (n=40), and HA-K_v7.3+K_v7.2-M518V (n=39). Data shown represent the Ave ± SEM (*p<0.05, **p<0.01, ***p<0.005). (E, F) Immunocytochemistry of hippocampal neurons transfected with GFP and K_v7.2 WT or K_v7.2-M518V. (E) Representative images. Ankyrin-G immunostaining marked the AIS (arrows). Cotransfected GFP allowed visualization of all neurites. Scale bars: 50 μm. (F) Quantification of the number of healthy transfected neurons containing prominent ankyrin-G immunostaining (Neurons with Ankyrin-G) and transfected neurons that had beaded or broken dendrites or axons, or damaged soma (Dead neurons) as % transfected cells. Total number of transfected neurons analyzed in n=4 independent experiments: GFP (n=300), GFP+K_v7.2-WT (n=252), GFP+K_v7.2-M518V (n=350). (G) Immunocytochemistry of hippocampal neurons transfected with HA-K_v7.3 and K_v7.2 WT or M518V was performed and quantified for live neurons with ankyrin-G and dead neurons as % transfected cells. Total number of transfected neurons analyzed in n=3 independent experiments: HA-K_v7.3+K_v7.2-WT (n=145), HA-K_v7.3+K_v7.2-M518V (n=183). Data shown represent the Ave ± SEM (*p<0.05, ***p<0.005).

Electrophysiology.

Whole cell patch clamp recordings were made at room temperature (20°C–22°C) using a Multiclamp 700B amplifier, acquired with a Digidata 1440A interface, and analyzed with pClamp 10.6 and Clampfit 10.6 (Molecular Devices). Current-clamp recordings were performed from cultured hippocampal neurons, filtered at 2 kHz, and digitized at 10 kHz as previously described (Cavaretta et al., 2014). At 24–48 hr post transfection with plasmids (total 1 μg) including pEGFPN1 and pcDNA3-K_v7.2, spike trains were evoked from untransfected or GFP-positive neurons as described (Cavaretta et al., 2014) by delivering constant somatic current pulses of 500 ms in the range 0 to 200 pA at a holding potential of −60 mV and in the presence of the fast synaptic transmission blockers CNQX (20 μM), DL-AP5 (100 μM) and bicuculline (20 μM).

Voltage clamp recording of K⁺ currents were carried out in Chinese hamster ovary (CHO hm1) cells (kind gifts from Dr. Naoto Hoshi, University of California, Irvine), filtered at 1 kHz, and digitized at 10 kHz as described (Kosenko et al., 2012; Miceli et al., 2015). Cells were plated on 12mm glass coverslips (Warner Instrument, 10⁵ cells per coverslip) coated with Poly D-lysine (0.1 mg/ml) and transfected at 24 h post plating with plasmids (total 1 μg) including pEGFPN1 (0.2 μg) and pcDNA3-K_v7.2 WT or mutant (0.8 μg). At 24–48 h post transfection, recordings from GFP-positive cells were performed in external solution containing (in mM): 138 NaCl, 5.4 KCl, 2 CaCl₂, 1 MgCl₂, 10 D-glucose, and 10 HEPES (pH 7.4, 307–312 mOsm). Recording pipettes had a resistance of 3–5 MΩ filled with internal solution containing (in mM): 140 KCl, 2 MgCl₂, 10 EGTA, 10 HEPES, 5 Mg-ATP, (pH 7.4 with KOH, 290–300mOsm). Resting membrane potential was determined in current clamp after achieving the whole-cell configuration. To record K⁺ current, cells were held at −80 mV and then voltage steps were applied for 1.5 s from −100 mV to 20 mV in 10 mV increments, followed by an isopotential pulse at 0 mV of 300 ms duration. To examine PIP₂ sensitivity of K_v7.2 channels, the recording was repeated with internal solution containing diC8-PIP₂ (100 μM).

All currents were measured under whole-cell voltage clamp with pipette and membrane capacitance cancellation. Leak subtracted currents were computed by subtracting leak currents from peak current at all voltage steps. Leak currents were defined as non-voltage-dependent current from GFP-transfected cells. Current densities (expressed in picoamperes per picofarad) were calculated as leak-subtracted peak K⁺ currents divided by capacitance (C). For Figures 4B and 5B, the leak-subtracted peak current density at each voltage step was measured and plotted as a function of the preceding voltages. For Figures 5C and 5E, the normalized conductance (G / Gmax) was plotted as a function of voltage steps. No corrections were made for liquid junction potentials. To measure the channel kinetics, current traces were fitted with a double exponential function of the following form: y = A_fexp(τ/τ_f) + A_sexp(τ/τ_s), where A_f and A_s indicate the fractions of the fast and slow exponential components of the current, respectively, and τ_f and τ_s indicate the time constants of these components. The time constant representing the weighted average of the fast and slow components of current activation was calculated with the following equation; τ = (τ_fA_f + τ_sA_s)/(A_f+A_s). For Table 2, the leak-subtracted currents were then fit to a Boltzmann distribution of the following form: y = y _max / [1/exp(V_1/2-V)/k], where V is the test potential, V_1/2 the half-activation potential, and k the slope factor.

Figure 4. Voltage-dependent activation of K_v7.2 channels is decreased by R532W mutation and abolished by M518V mutation.

Whole cell voltage clamp recording of macroscopic K⁺ currents from CHO hm1 cells transfected with GFP and K_v7.2 WT or K_v7.2 containing epileptic encephalopathy mutations. The cells were held at −80 mV. Currents were evoked by depolarization for 1.5 s from −100 mV to +20 mV in 10 mV increments, followed by a step to 0 mV for 300 ms. The raw current traces and data are shown in Supplemental Fig. 3. (A) Representative traces of currents from which leak currents were subtracted at all voltage steps. Leak current was defined as non-voltage-dependent current from GFP-transfected cells. (B-D) Average peak current densities (B), normalized currents (C), and normalized conductance (G / Gmax) (D) were quantified using leak subtracted currents at all voltage steps. The number of transfected cells that were analyzed: GFP (n=9), GFP and K_v7.2 WT (n=11), R333W (n=12), A343D (n=12), M518V (n=8), K526N (n=14), or R532W (n=18). (E) Immunoblotting with anti-Kv7.2 N-terminal antibodies on the lysate from CHO hm1 cells expressing Kv7.2 WT (n=4), R333W (n=12), A343D (n=4), M518V (n=4), K526N (n=4), or R532W (n=4). GAPDH served as a loading control. Data shown represent the Ave ± SEM (*p<0.05, **p<0.01, ***p<0.005).

Figure 5. PIP₂ stimulation of K_v7.2 current is decreased by R333W and R532W mutations and increased by K526N mutation.

Whole cell patch clamp recording of K_v7.2 currents from CHO hm1 cells with or without diC8-PIP₂ (100 μM) in the patch pipette solution using the same voltage clamp protocol as described in Figure 4. The raw current traces and data are shown in Supplemental Fig. S4. (A) Representative leak-subtracted current traces. (B-C) Average peak current densities (B) and normalized conductance (G / Gmax) (C) were quantified using leak subtracted currents at all voltage steps. (D) Average leak subtracted peak current densities at + 20 mV in the absence or presence of diC8-PIP₂. The number of GFP-cotransfected cells that were analyzed without diC8-PIP₂: K_v7.2 WT (n=11), R333W (n=12), K526N (n=14), or R532W (n=18). The number of GFP-cotransfected cells that were analyzed with diC8-PIP₂: K_v7.2 WT (n=8), R333W (n=13), K526N (n=14), or R532W (n=11). Data shown represent the Ave ± SEM (#p<0.05, ###p<0.005 compared to WT without diC8-PIP₂, *p<0.05, **p<0.01, ***p<0.005 compared to WT with diC8-PIP₂, and ^^^p<0.005 compared to K526N without diC8-PIP₂)

Table 2.

Biophysical properties of wild-type and mutant K_v7.2 channels in GFP-cotransfected CHO hm1 cells in the presence or absence of diC8-PIP₂.

Transfection	n	Leak substracted peak current (pA)	V1/2 (mV)	k (mV/efold)	Iinstant / Isteady-state	Tau activation at +20 mV (ms)
GFP + Kv7.2 WT	11	321.4 ± 38.5	−8.0 ± 0.6	10.6 ± 0.5	0.14 ± 0.02	79.4 ± 4.8
GFP + Kv7.2 A343D	12	15.6 ± 1.8*	−45.2 ± 3.3*	11.3 ± 2.5	1.00 ± 0.04*	4.9 ± 0.6*
GFP + Kv7.2 R333W	12	241.7 ± 35.5	−8.2 ± 1.1	12.5 ± 1.1	0.26 ± 0.02*	90.3 ± 6.3
GFP + Kv7.2 M518V	8	29.5 ± 8.2*	−32.9 ± 3.6*	10.2 ±1.5	0.89 ± 0.09*	5.2 ± 0.9*
GFP + Kv7.2 K526N	14	430.0 ± 72.2	−7.9 ± 0.9	14.5 ± 0.7*	0.11 ± 0.02	86.3 ± 2.9
GFP + Kv7.2 R532W	18	184.8 ± 30.5	−7.2 ± 1.4	12.2 ± 1.1	0.31 ± 0.04*	112.8 ± 7.2*
GFP + Kv7.2 WT + diC8-PIP2	8	638.4 ± 66.2‡	−6.4 ± 1.2	15.0 ± 0.5†	0.12 ± 0.02	79.7 ± 5.4
GFP + Kv7.2 R333W + diC8-PIP2	13	404.8 ± 54.6†	−7.9 ± 1.2	15.2 ± 1.0‡	0.19 ± 0.02‡	86.2 ± 2.9
GFP + Kv7.2 K526N + diC8-PIP2	14	1276.5 ± 140.8†‡	−5.3 ± 1.3‡	14.2 ± 0.8‡	0.07 ± 0.01‡	58.3 ± 2.4†‡
GFP + Kv7.2 R532W + diC8-PIP2	11	258.0 ± 31.3†	−5.6 ± 0.9	17.2 ± 2.6‡	0.21 ± 0.05†‡	100.3 ± 5.1‡

n, number; Leak subtracted peak current measured at +20 mV; V_1/2, half-activation potential calculated using leak subtracted currents; k, the slope factor calculated using leak subtracted currents; I_instant, leak subtracted current measured at the beginning of the depolarization step; I_steady-state, leak subtracted current measured at the end of the 0 mV depolarization; τ, activation time constant measured at + 20 mV using leak subtracted currents. Ave ± SEM (*p<0.05 for comparing K_v7.2 wild-type (WT) vs. mutants; †p<0.05 for comparing K_v7.2 WT + diC8-PIP₂. vs mutants + diC8-PIP_2; ‡p<0.05 comparing for K_v7.2 wild-type (WT) vs mutants + diC8-PIP₂). Each value represents the Ave ± SEM.

Statistical Analyses.

All analyses are reported as mean ± SEM. Using Origin 9.1 (Origin Lab), the Student t test and one-way ANOVA with post-ANOVA Tukey’s and Fisher’s multiple comparison tests were performed to identify the statistically significant difference with a priori value (p) < 0.05 between 2 groups and for >3 groups, respectively.

RESULTS

K_v7.2 binding to CaM is severely decreased by M518V mutation in CaM contact site and modestly reduced by R333W mutation located peripheral to CaM contact site.

To date, 50 missense mutations in K_v7.2 have been associated with early-onset epileptic encephalopathy. Mapping 21 epileptic encephalopathy mutations to the cytoplasmic C-terminal tail of K_v7.2 revealed their location at 4 distinct domains: helix A that contains consensus IQ motif for CaM binding and interacts with CaM C-lobe (Gamper and Shapiro, 2003; Strulovich et al., 2016; Yus-Najera et al., 2002), helix B that binds to CaM N-lobe and has preferential affinity to Ca²⁺-bound CaM in vitro (Alaimo et al., 2012; Strulovich et al., 2016; Xu et al., 2013), the linker from helix B to helix C, and helix C that mediates subunit interaction (Schwake et al., 2006)(Fig. 1A). Epileptic encephalopathy mutations are clustered in helix A and helix B, but no mutations are found in the linker between helix A and helix B. This interesting trend of mutation clustering at or near CaM binding sites suggests that a molecular pathogenetic mechanism of epileptic encephalopathy involves disruption of K_v7.2 binding to CaM.

To test if epileptic encephalopathy mutations in helices A and B and the helix B-C linker impair K_v7.2 interaction with CaM, we introduced point mutations in helix A (R333W), helix B (M518V, K526N) and helix B-C linker (R532W) in K_v7.2 (protein accession CAA 75348.1) (Fig. 1A, B). We selected these epileptic encephalopathy mutations to compare the effects of mutations buried within CaM contact sites (M518V) to those located at the periphery (R333W, K526N, and R532W) (Fig. 1C, D). The A343D mutation located in the consensus IQ motif of helix A was included in this study as a control (Fig. 1B–D) because this nonpathogenic variant was previously shown to disrupt CaM interaction with K_v7.2 (Alaimo et al., 2009; Cavaretta et al., 2014; Wen and Levitan, 2002). Indeed, our modeling of K_v7.2 helices A-B structure with Ca²⁺-bound CaM predicted that the A343D mutation at the CaM C-lobe contact site brings a desolvated acidic group into a buried hydrophobic interface (Fig. 1D). This creates steric clashes which result in rotamer shifts of surrounding side chains and subsequent destabilization of the interaction between K_v7.2 helix A with CaM C-lobe (Fig. 1D).

We then performed co-immunoprecipitation studies in the presence of the Ca²⁺ chelator EGTA from HEK293T cells transfected with K_v7.2 and CaM N-terminally tagged with yellow fluorescent protein (YFP) (Fig. 2A, B). We first noticed that the expression of K_v7.2-A343D and K_v7.2-M518V was consistently lower compared to wild type K_v7.2 proteins in the presence of YFP-CaM although equal amounts of each plasmid were transfected in HEK293T cells (Fig. 2A, B). To immunoprecipitate the same amount of wild-type and mutant K_v7.2 and compare the effect of mutations on K_v7.2 coimmunoprecipitation with CaM, we therefore used small amount of anti-K_v7.2 N-terminal antibodies that could immunoprecipitate a fraction but not all of transfected K_v7.2 subunits from the large amount of transfected cell lysate.

Figure 2. K_v7.2 expression and binding to apoCaM is impacted by epileptic encephalopathy mutations.

(A, B) Co-immunoprecipitation of YFP-CaM with wild type K_v7.2 (WT) or K_v7.2 containing epileptic encephalopathy mutations from transfected HEK283T cells. (A) Representative immunoblots. (B) Quantification of co-immunoprecipitation and lysate from 7 independent experiments: untransfected (None: n=6), YFP-CaM (n=6), YFP-CaM cotransfection with WT (n=7), R333W (n=7), A343D (n=6), M518V (n=7), K526N (n=6), or R532W (n=5). (C, D) Co-immunoprecipitation of K_v7.3 containing extracellular HA tag (HA-K_v7.3) with K_v7.2 WT or mutant K_v7.2 containing epileptic encephalopathy mutations from transfected HEK283T cells. (C) Representative immunoblots. (D) Quantification of co-immunoprecipitation and lysate from 6 independent experiments: untransfected cells (None: n=6), or cells transfected with HA-K_v7.3 (n=6), HA-K_v7.3 and K_v7.2 WT (n=6), R333W (n=6), A343D (n=4), M518V (n=5), K526N (n=5), or R532W (n=5). GAPDH served as a loading control. Data shown represent the Ave ± SEM (*p<0.05, **p<0.01, ***p<0.005).

Although similar amounts of wild type and mutant K_v7.2 proteins were immunoprecipitated, YFP-CaM co-immunoprecipitated with wild type Kv7.2 but not Kv7.2 containing A343D mutation in the CaM contact site within helix A (Fig. 2A, B), consistent with our previous report (Cavaretta et al., 2014). The M518V mutation is located at the CaM N-lobe contact site within a completely buried hydrophobic region of helix B (Fig. 1C, D). This mutation is expected to create a large void at the interface center with little compensatory conformational changes of the neighboring residues, thereby destabilizing the helix B-CaM interface (Fig. 1D). Consistent with our prediction, M518V mutation severely decreased coimmunoprecipitation of apoCaM with K_v7.2 by 85% (Fig. 2A, B). CaM cotransfection also increased the expression of wild-type K_v7.2 and K_v7.2-M518V but not K_v7.2-A343D (Fig. S1), supporting our coimmunoprecipitation findings that CaM binding to K_v7.2 is severely reduced but not abolished by M518V mutation.

The R333W mutation located proximal to the CaM C-lobe contact site of helix A introduces an indole ring which can pack against hydrophobic residues of CaM (V92, F93, and L113) to make alternative favorable contacts, and thus is expected to have mild impact on K_v7.2 helix A interaction with CaM C-lobe (Fig. 1D). Consistently, the R333W mutation reduced K_v7.2 expression and its interaction with apoCaM by 20% (Fig. 2A, B). The K526N mutation is distal to the CaM N-lobe contact interface in Kv7.2 (Fig. 1C). The K526N mutation is anticipated to modestly affect helix B binding to CaM N-lobe due to the loss of a solvent-exposed electrostatic contact to E48 of CaM (Fig. 1D). Although the K526N mutation decreased K_v7.2 coimmunoprecipitation with apoCaM by 35%, this change did not reach statistical significance (Fig. 2A, B). Consistent with the location of R532W beyond helix B (Fig. 1B, C), the R532W mutation had no effect on K_v7.2 binding to apoCaM (Fig. 2A, B).

To test if epileptic encephalopathy mutations in helices A and B and the helix B-C linker impair K_v7.2 interaction with K_v7.3, we repeated co-immunoprecipitation studies in HEK293T cells transfected with K_v7.2 and HA-K_v7.3 which is extracellularly tagged with a HA epitope (Fig. 2C, D). Although an increasing trend was observed in K_v7.2-M518V coimmunoprecipitation with HA-K_v7.3, none of the tested mutations significantly affected K_v7.2 binding to HA-K_v7.3 (Fig. 2C, D).

Enrichment of K_v7 channels at the axonal surface is severely decreased by M518V mutation, and modestly reduced by R333W, K526N, and R532W mutations.

We have previously shown that A343D mutation in the IQ motif of K_v7.2 helix A disrupts CaM binding to K_v7.2 and blocks preferential targeting of heteromeric K_v7 channels to the axonal membrane in hippocampal neurons by inhibiting their trafficking from the ER to the axons (Cavaretta et al., 2014) (Fig. 2A, B, Fig. S2). These previous findings together suggest that CaM binding to helix A of K_v7.2 is critical for K_v7 targeting to the axonal surface. Therefore, we next tested if epileptic encephalopathy mutations in helices A and B of K_v7.2 affect surface density of heteromeric K_v7 channels in hippocampal neurons.

We performed surface immunostaining of K_v7.3 containing an extracellular HA epitope (HA-K_v7.3) (Cavaretta et al., 2014) since antibodies that recognize extracellular domains of endogenous K_v7.2 and K_v7.3 are not available. HA-K_v7.3 has been shown to display K⁺ current and surface expression in Xenopus oocytes coexpressing K_v7.2 (Schwake et al., 2006; Schwake et al., 2000). We have previously demonstrated in cultured hippocampal neurons that homomeric HA-K_v7.3 channels are not expressed on the plasma membrane, whereas co-transfection of K_v7.2 results in robust axonal surface expression of HA-K_v7.3 (Cavaretta et al., 2014; Chung et al., 2006). Consistent with our previous reports, strong immunostaining of HA-K_v7.3 was detected on the axonal plasma membrane compared to somatodendritic plasma membrane in cultured hippocampal neurons cotransfected with K_v7.2 (Fig. 3A–C), resulting in the surface “Axon / Dendrite” ratio of 3.5 ± 0.4 (Fig. 3D). The highest intensity of HA-K_v7.3/K_v7.2 channels was observed at the AIS (Fig. 3A–C).

In contrast, the M518V mutation, which severely reduced CaM binding to K_v7.2 (Fig. 2A, B), decreased surface expression of heteromeric HA-K_v7.3/K_v7.2 channels at the distal axons to the level of untransfected neurons (Fig. 3A–C), reducing the surface “Axon / Dendrite” ratio to 1.1 ± 0.1 (Fig. 3D). The surface and total expression of HA-K_v7.3/K_v7.2-M518V at the AIS and the soma was also reduced to a third and a half of wild type channels, respectively (Fig. 3A–C). No significant expression of these mutant channels was seen at the dendrites (Fig. 3C). The weak but detectable expression of heteromeric HA-K_v7.3/K_v7.2-M518V mutant channels at the AIS (Fig. 3A–D) is in sharp contrast to the complete absence of heteromeric A343D mutant channels from the AIS and axons (Fig. S2)(Cavaretta et al., 2014).

The R333W and K526N mutations decreased axonal surface expression of heteromeric HA-K_v7.3/K_v7.2 channels (Fig. 3A–C) and the surface “Axon / Dendrite” ratio by 30% (Fig. 3D), consistent with decreasing trend in CaM binding to K_v7.2-R333W and K_v7.2-K526N (Fig. 2A, B). Reduction in total expression was also seen in the soma for heteromeric HA-K_v7.3/K_v7.2 channels harboring R333W and K526N mutation (Fig. 3C). Surprisingly, the R532W mutation also decreased overall surface expression of heteromeric HA-K_v7.3/K_v7.2 channels (Fig. 3A–C) and reduced the surface “Axon / Dendrite” ratio by 30% (Fig. 3D) although this mutation did not affect K_v7.2 binding to CaM (Fig. 2A, B). These data provide strong evidence that all tested epileptic encephalopathy mutations at or near CaM-binding domains reduced preferential enrichment of heteromeric K_v7 channels at the axonal surface.

Voltage-dependent activation of K_v7.2 channels is decreased by R532W mutation in helix B-C linker and abolished by M518V mutation in helix B.

Mutant K_v7.2 channels deficient in CaM binding have been previously reported to be non-functional (Wen and Levitan, 2002), whereas CaM cotransfection has shown to increase current density of K_v7.2 channels (Ambrosino et al., 2015; Soldovieri et al., 2016). To test if epileptic encephalopathy mutations in helices A and B of K_v7.2 affect voltage-dependent activation of K_v7.2 channels, we performed whole-cell patch clamp recording in CHO hm1 cells, a well-established system to measure K⁺ current through exogenously expressed K_v7 channels (Kosenko et al., 2012). The patch pipette solution contained EGTA to sequester free Ca²⁺. In CHO hm1 cells transfected with GFP, the resting membrane potential was −10.2 ± 0.7 mV (Table 1), consistent with very low expression of endogenous K⁺ channels (Gamper et al., 2005). Application of depolarizing voltage steps from −100 to +20 mV in GFP-transfected cells generated very little inward and outward currents (Fig. S3), which displayed no obvious voltage-dependence and reversed around −25.6 ± 1.6 mV (Tables 1, 2).

Table 1.

Passive properties of CHO hm1 cells transfected with GFP and K_v7.2 containing epileptic encephalopathy mutations.

	Without diC8-PIP2				With diC8-PIP2
Transfection	n	Vm (mV)	Cm (pF)	Erev (mV)	n	Vm (mV)	Cm (pF)	Erev (mV)
GFP	9	−10.2 ± 0.7*	11.9 ± 2.1	−25.6 ± 1.6*	ND	ND	ND	ND
GFP + Kv7.2 WT	11	37.0 ± 1.1	14.4 ± 1.5	−37.0 ± 1.1	8	−36.6 ± 1.5	17.2 ± 1.1	−44.6 ± 1.7‡
GFP + Kv7.2 A343D	12	−9.2 ± 0.5*	13.2 ± 0.7	−25.5 ± 2.3*	ND	ND	ND	ND
GFP + Kv7.2 R333W	12	−39.3 ± 1.4	14.7 ± 1.3	−40.5 ± 1.9	13	−37.1 ± 1.4	17.1 ± 1.4	−44.7 ± 2.1‡
GFP + Kv7.2 M518V	8	−28.5 ± 1.6*	22.0 ± 5.9	−29.9 ± 3.7*	ND	ND	ND	ND
GFP + Kv7.2 K526N	14	−37.1 ± 1.4	17.7 ± 3.0	−43.1 ± 0.8*	14	−42.7 ± 1.8†‡	17.8 ± 1.3	−53.1 ± 2.2†‡
GFP + Kv7.2 R532W	18	−35.3 ± 1.3	17.8 ± 1.9	−37.0 ± 1.9	11	−36.3 ± 1.2	17.6 ± 1.2	−45.0 ± 1.2‡

n, number; V_m, resting membrane potential; C_m, whole cell membrane capacitance; E_rev, reversal potential. Ave ± SEM (*p<0.05 compared to K_v7.2 wild-type (WT); †p<0.05 for K_v7.2 WT + diC8-PIP₂. vs mutants + diC8-PIP_2; ‡p<0.05 for K_v7.2 WT vs mutants + diC8-PIP₂). ND, not determined. Each value represents the Ave ± SEM.

In cells transfected with GFP and wild type K_v7.2, application of voltage steps produced slowly activating outward K⁺ currents at a threshold voltage around −40 mV (Fig. 4A–E Fig. S3). Consistently, the resting membrane potential of these cells was hyperpolarized to −37.0 ± 1.1 mV (Table 1), indicating the significant presence of K⁺ currents. In contrast, K_v7.2-A343D channels failed to display voltage-dependent outward K⁺ current and did not alter resting membrane potentials (Fig. 4A–E, Supplemental Fig. 3, Tables 1, 2), consistent with their retention in the ER (Alaimo et al., 2009; Cavaretta et al., 2014; Etxeberria et al., 2008). The K_v7.2-M518V channels produced very small outward currents that lacked voltage-dependence (Fig. 4A–D, Fig. S3) but were sufficient to hyperpolarize membrane potential to −28.5 ± 1.6 mV (Table 1). In sharp contrast to HEK293T cells (Fig. 2), the total protein expression of K_v7.2-A343D and K_v7.2-M518V was comparable to the wild-type K_v7.2 in CHO hm1 cells (Fig. 4E).

The cells expressing K_v7.2-R333W or K_v7.2-K526N generated slowly activating voltage-dependent outward K⁺ currents that were slightly smaller or larger than the cells expressing wild type K_v7.2 (Fig. 4A–D, Fig. S3, Table 1–2), although this difference did not reach statistical significance. Consistent with this trend, decreased K_v7.2-R333W expression and increased K_v7.2-K526N expression was observed in CHO hm1 cells (Fig. 4E). K_v7.2-K526N channels also displayed a small left shift in voltage-dependence at −50 mV to −30 mV (Fig. 4D). Although total expression of K_v7.2-R532W was comparable to wild-type K_v7.2 (Fig. 4E), the cells expressing K_v7.2-R532W generated 40% less outward K⁺ currents (Fig. 4A–C, Fig. S3) with a right shift in voltage-dependence at −10 mV to +10 mV (Fig. 4D). In addition, the R532W mutation significantly slowed the channel activation (Table 2). These findings indicate that the M518V and R532W mutations significantly impaired current expression of K_v7.2 channels.

PIP₂ stimulation of K_v7.2 current is decreased by R333W and R532W mutations and increased by K526N mutation.

Phosphatidylinositol-4,5-bisphosphate (PIP₂) is critical for proper function of all K_v7 channels (Gamper and Shapiro, 2007; Logothetis et al., 2015). Previous studies have suggested that PIP₂-dependent potentiation of K_v7 currents may be mediated by the regions close to helices A and B and regulated by Ca²⁺-bound CaM (Hernandez et al., 2008; Tobelaim et al., 2017b). In K_v7.1, PIP₂ interacts with lysine residues in helix B, where Ca²⁺-bound CaM competes to stabilize the channel open state (Tobelaim et al., 2017b). The analogous basic amino acid residues in helix B of K_v7.2 are structurally close to the beginning of helix A, which contains another cluster of positively charged residues K331, R332, and R333 (Fig. 1B), suggesting that EE mutations located peripheral to CaM contact site in helix B (R333W, K526N, R532W) may alter PIP₂ sensitivity of K_v7.2 channels. To test this possibility, we repeated whole-cell patch clamp recording in CHO hm1 cells in the presence of short-chained diC8-PIP₂ (100 μM) in the intracellular solution.

Addition of diC8-PIP₂ induced a 50% increase in voltage-activated outward K⁺ currents through wild type K_v7.2 channels compared to those measured in the absence of diC8-PIP₂ (Fig. 5A–D, Fig. S4). Addition of diC8-PIP₂ also caused a left shift in voltage-dependence at −60 mV to −0 mV (Fig. 5C). In contrast, R333W mutation abolished the diC8-PIP₂-mediated increase in current density (Fig. 5A–D, Fig. S4). In the absence of diC8-PIP₂, K_v7.2-R532W channels displayed a >50% reduction in voltage-dependent outward K⁺ currents compared to wild-type channels (Fig. 4B, 5D). No further increase in outward K⁺ currents through K_v7.2-R532W channels was observed in the presence of diC8-PIP₂ (Fig. 5A–D). Introduction of diC8-PIP₂ produced larger K⁺ currents through K_v7.2-K526N channels compared to those measured without diC8-PIP₂ (Fig. 5A–B), caused a left shift in voltage-dependence at −30 mV to 10 mV (Fig. 5C), and further hyperpolarized the resting membrane potential to −53.1 ± 2.2 mV from −43.1 ± 0.8 mV (Table 1). Surprisingly, the currents through K_v7.2-K526N channels in the presence of diC8-PIP₂ were significantly larger than those through wild type channels (Fig. 5A–D, Fig. S4). These results together indicate that R333W and R532W mutations reduced whereas K526N mutation enhanced sensitivity of K_v7.2 channels to gating modulation by PIP_2.

Hippocampal neuronal excitability is reduced upon expression of wild type K_v7.2 but not K_v7.2 containing R333W, K526N, or R532W mutation.

Since axonal K_v7 channels inhibit excitability of hippocampal neurons (Shah et al., 2011; Shah et al., 2008), multiple defects in K_v7 axonal surface density and current expression induced by epileptic encephalopathy mutations (Fig. 2–5) could impair their function in suppressing neuronal excitability. To test this, we performed whole cell patch clamp recording of action potentials in cultured hippocampal neurons transfected with wild-type or mutant K_v7.2. Since rat hippocampal neurons in dissociated culture display mRNA and protein expression of K_v7.2 and K_v7.3 as well as K_v7 current (Lee and Chung, 2014; Lee et al., 2015)(Fig.S5–6), we hypothesized that transiently expressed wild-type K_v7.2 would form functional homomeric channels with endogenous K_v7.2 and heteromeric channels with endogenous K_v7.3 at the axonal plasma membrane and reduce excitability (Fig. 6A). Consistent with this hypothesis and our previous reports (Cavaretta et al., 2014), transfection of wild type K_v7.2 increased expression of K_v7.2 about 4-fold compared to untransfected neurons and decreased action potential firing rates compared to untransfected neurons and GFP-expressing neurons for all current injections from 40 pA (Fig. 6B–E, Fig. S6). Although transfection of wild type K_v7.2 did not affect passive intrinsic properties of hippocampal neurons (Table 3), K_v7.2 expression increased AP half-width and AP decay time compared to untransfected neurons and GFP-expressing neurons (Table 4) most likely due to increased K⁺ efflux through transfected K_v7.2-containing channels. There was no change in AP height and AP rise time (Table 4), which are largely regulated by Na⁺ influx through endogenous voltage-gated sodium channels.

Figure 6. Hippocampal neuronal excitability is reduced upon expression of wild-type K_v7.2 but not K_v7.2 containing R333W, K526N, and R532W mutations.

Whole-cell current clamp recording in hippocampal neurons (DIV 6–8) transfected with GFP and K_v7.2 WT or K_v7.2 containing epileptic encephalopathy mutations. Spike trains were evoked by delivering constant somatic current pulses for 500 ms duration at a resting potential of –60 mV. (A) Hypothesis by which exogenously expressed K_v7.2 WT would decrease neuronal excitability by increasing expression of functional homomeric and heteromeric channels formed with endogenous K_v7.2 and K_v7.3 subunits. (B) Representative traces of action potentials induced by 100 pA injection. (C) Average instantaneous AP firing rates (Hz) induced by 0–200 pA injection. (D) Average firing rates at 100 pA from untransfected neurons (None: n = 9), or neurons transfected with GFP (n = 10), GFP and WT (n = 11), R333W (n = 9), K526N (n = 9), or R532W (n = 9). Data shown represent the Ave ± SEM (** p < 0.01 compared to untransfected, ^#p < 0.05 compared to GFP).

Table 3.

Passive properties of hippocampal pyramidal neurons transfected with GFP and K_v7.2 containing epileptic encephalopathy mutations.

Transfection	n	Vm (mV)	Cm (pF)	Rin (MΩ)
Untransfected	9	−54.5 ± 1.6	46.7 ± 2.5	873 ± 85
GFP	10	−52.4 ± 1.9	46.3 ± 3.3	819 ± 42
GFP + Kv7.2 WT	11	−52.4 ± 1.8	45.3 ± 1.3	858 ± 65
GFP + Kv7.2 R333W	9	−53.6 ± 2.0	46.3 ± 1.7	844 ± 44
GFP + Kv7.2 K526N	9	−52.8 ± 0.9	47.6 ± 2.7	881 ± 98
GFP + Kv7.2 R532W	9	−52.5 ± 2.1	48.6 ± 3.5	880 ± 34

n, number; V_m, resting membrane potential; C_m, whole cell membrane capacitance; R_in, input resistance. Each value represents the Ave ± SEM.

Table 4.

Action potential properties of hippocampal pyramidal neurons transfected GFP and K_v7.2 containing epileptic encephalopathy mutations.

Transfection	VT (mV)	AP height (mV)	AP rise time (ms)	AP decay time (ms)	AP HW (ms)	fAHP (mV)
Untransfected	−31.0 ± 1.4	47.7 ± 3.9	1.04 ± 0.08	2.18 ± 0.27	2.50 ± 0.15	−17.4 ± 1.9
GFP	−33.9 ± 1.1	53.3 ± 3.8	1.10 ± 0.08	2.62 ± 0.27	2.78 ± 0.20	−15.2 ± 1.3
GFP + Kv7.2 WT	−32.8 ± 1.0	56.5 ± 2.5	1.14 ± 0.11	3.91 ± 0.40*, #	3.54 ± 0.26*, #	−12.1 ± 2.1*
GFP + Kv7.2 R333W	−32.8 ± 1.1	48.7 ± 2.2	1.05 ± 0.06	2.92 ± 0.43$	3.03 ± 0.23	−12.8 ± 1.7
GFP + Kv7.2 K526N	−35.0 ± 0.8*	54.3 ± 3.1	1.20 ± 0.08	2.93 ± 0.22$	2.98 ± 0.19	−10.5 ± 1.5*
GFP + Kv7.2 R532W	−34.2 ± 1.0	53.5 ± 3.3	1.21 ± 0.10	2.57 ± 0.23$	2.83 ± 0.15$	−12.8 ± 0.8

V_T, voltage threshold for action potential; AP, Action potential; rise, 10–90% rise time of AP; decay, 10–90% decay time of AP; HW, half-width; fAHP, fast after-hyperpolarization. AP properties were measured from the first action potential evoked by a current step to 100 pA at a holding potential of −60mV. Each value represents the Ave ± SEM.

^*p<0.05 compared to untransfected;

^#p<0.05 compared to GFP;

^$p<0.05 compared to WT.

In contrast, transfection of K_v7.2-R333W, K_v7.2-K526N, or K_v7.2-R532W did not reduce action potential firing frequency (Fig. 6B–D), although their expression at the AIS is comparable to wild-type K_v7.2 (Fig. 6E). Interestingly, action potential firing rates induced by 30–120 pA injections were significantly enhanced in neurons expressing K_v7.2-R532W compared to neurons expressing GFP alone (Fig. 6B–D), suggesting that hippocampal neurons expressing K_v7.2-R532W are hyperexcitable. Expression of mutant K_v7.2 did not affect membrane capacitance, input resistance, and resting membrane potential compared to untransfected or GFP-transfected neurons (Table 3). Furthermore, none of these mutations affected AP height, half width, and rise and decay times compared to untransfected or GFP-transfected neurons (Table 4). Together, these data indicate that K_v7.2 proteins containing R333W, K526N, and R532W mutations are defective in reducing excitability of hippocampal neurons.

K_v7.2-M518V proteins undergo ubiquitination and proteasome-dependent degradation whereas K_v7.3 coexpression prevents this degradation.

When investigating the effect of M518V mutation in hippocampal neuronal excitability, we faced difficulty in patching K_v7.2-M518V transfected hippocampal neurons due to frequent plasma membrane rupture. We had similar difficulties with CHO hm1 cells expressing K_v7.2-M518V in Figure 4. These results suggest that M518V mutation may cause other dysfunction that could compromise neuronal health, in addition to severely reducing current and axonal surface expression of K_v7 channels (Fig. 3–4).

Indeed, we observed accelerated degradation of K_v7.2 proteins containing M518V or A343D mutation in HEK293T cells by 24 h post transfection (Fig. 7A, B), although both wild type and mutant K_v7.2 proteins were equally expressed at 10 h post transfection (Fig. 7A). Interestingly, only K_v7.2-M518V proteins were heavily ubiquitinated (Fig. 7C, D). Our immunostaining revealed that M518V mutation decreased K_v7.2 expression by half whereas treatment with the proteasome inhibitor MG132 blocked this degradation (Fig. 7E, F), suggesting that ubiquitinated K_v7.2-M518V proteins underwent proteasome-dependent degradation. MG132 treatment also induced aggregation of K_v7.2-M518V proteins (Fig. 7E), consistent with previous reports that inhibition of proteasome-dependent degradation drives polyubiquitinated proteins in the ER to form large aggregates (Jaeger and Wyss-Coray, 2009).

Figure 7. M518V mutation causes ubiquitination and proteasome-dependent degradation of K_v7.2.

(A-B) Immunoblot analyses of K_v7.2 WT, or mutant K_v7.2 (A343D and M518V) in HEK293T cells at 10 h and 24 h post transfection. Representative immunoblots (A). Quantification at 24 h post transfection for n=4 independent experiments. (B). (C-D) Immunoprecipitation of K_v7.2 WT or mutant K_v7.2 (A343D and M518V) from HEK293T cells cotransfected with HA-ubiquitin. Representative immunoblots (C). Quantification of coIP and lysate (D) are shown for n=4 independent experiments. (E-F) Immunocytochemistry of HEK293T cells showing total K_v7.2 WT or K_v7.2-M518V proteins after 24 h treatment with vehicle control DMSO (1% v/v) or MG132 (1 μM). Dapi staining shows nuclei. (E) Representative images with scale bars: 10 μm. (F) Background subtracted mean intensity of total K_v7.2 fluorescence per transfected cell for DMSO+WT (n=20), DMSO+M518V (n=21), MG132+WT (n=22), MG132+M518V (n=22). Data shown represent the Ave ± SEM (*p<0.05, **p<0.01, ***p<0.005).

Surprisingly, cotransfection of K_v7.3 subunits blocked degradation (Fig. 2C, D) but not ubiquitination of K_v7.2-M518V proteins (Fig. S7A), resulting in a 2-fold increase in their expression compared to expression of K_v7.2-M518V alone (Fig. 8A–B, S7B). In the presence of K_v7.3, abnormal accumulation of K_v7.2-M518V proteins was also evident compared to wild-type K_v7.2 expression (Fig. S7B). These findings indicate that K_v7.3 prevents degradation of ubiquitinated K_v7.2-M518V proteins and stabilizes their expression.

Expression of K_v7.2-M518V induces cell and nuclei shrinkage and hippocampal neuronal injury.

Intracellular accumulation of misfolded proteins has been shown to induce ER stress and cause cell death (Su et al., 2011; Tyedmers et al., 2010). Indeed, we observed that expression of K_v7.2-M518V increased the number of transfected HEK293T cells that underwent cell shrinkage, a hallmark of the early events during programmed cell death by apoptosis (Bortner and Cidlowski, 2002; Maeno et al., 2000), compared to wild type K_v7.2 expression (Fig. 8A, C). Transfection of K_v7.2-M518V and K_v7.3 in HEK293T cells also caused cell shrinkage compared to coexpression of wild type K_v7.2 and K_v7.3 (Fig. 8A, C). Nuclei condensation, a late-stage event in apoptosis (Bortner and Cidlowski, 2002; Maeno et al., 2000), was observed in cells expressing K_v7.2-M518V alone compared to those expressing wild type Kv7.2 and was further enhanced by cotransfection of K_v7.3 (Fig. 8A, D). These results indicate that expression of K_v7.2-M518V induces cell death and K_v7.3 co-expression aggravates this effect in HEK293T cells.

To test if expression of K_v7.2-M518V may compromise neuronal health, we examined the morphology of hippocampal neurons transfected with wild type K_v7.2 or K_v7.2-M518V. At 48 h post transfection, fluorescence imaging of co-transfected GFP and differential interference contrast (DIC) imaging was employed to identify the dead neurons, which displayed damaged soma and beaded or broken neurites. Recent studies have shown that disruption of the AIS by rapid, irreversible proteolysis of cytoskeleton proteins including ankyrin-G and βIV spectrin occurs upon neuronal injury independently of cell death or axon degeneration (Schafer et al., 2009). Therefore, we also examined the presence of ankyrin-G as an indication of health (Fig. 8E). Despite similar expression of wild type K_v7.2 and K_v7.2-M518V proteins in hippocampal neurons (Fig. 6E, S6), we observed a significant decrease in the number of GFP-positive neurons expressing K_v7.2-M518V that showed prominent ankyrin-G immunostaining at the AIS compared to neurons expressing GFP alone or together with wild type K_v7.2 (Fig. 8E, F). Concurrently, expression of GFP and K_v7.2-M518V increased the number of dead transfected neurons compared to expression of GFP with or without wild type K_v7.2 (Fig. 8F). In the presence of HA-K_v7.3, expression of K_v7.2-M518V also increased the number of dead transfected neurons and decreased the number of transfected neurons with ankyrin-G immunostaining compared to expression of wild type K_v7.2 (Fig. 8G). These results suggest that expression of K_v7.2-M518V increases neuronal injury and death.

DISCUSSION

In this study, we investigated the molecular pathogenetic effects of de novo epileptic encephalopathy mutations concentrated at and near helices A and B of K_v7.2 (Fig. 1). In vitro studies with K_v7.2 helix A or helix B have shown that apoCaM binds equally to helices A and B (Xu et al., 2013) whereas Ca²⁺-bound CaM favors interaction with helix B (Alaimo et al., 2012; Gamper and Shapiro, 2003; Xu et al., 2013; Yus-Najera et al., 2002). We have previously shown that CaM binding to the IQ motif of K_v7.2 helix A is critical for the ER export and axonal surface expression of K_v7 channels (Cavaretta et al., 2014), which are responsible for potently suppressing spontaneous and bursting firing of APs in hippocampal neurons (Shah et al., 2011; Shah et al., 2008). Our characterization of selected epileptic encephalopathy mutations of K_v7 helix A (R333W), helix B (M518V, K526N) and helix B-C linker (R532W) revealed the unexpected findings that each mutation displayed a unique combination of K_v7 channel defects including disruption of CaM binding to K_v7.2 (Fig. 2), reduction in axonal surface expression of K_v7 channels (Fig. 3), changes to voltage-dependent activation and PIP₂ sensitivity (Fig. 4, 5), and reduced inhibition of hippocampal neuronal excitability (Fig. 6). The M518V mutation also accelerated degradation by ubiquitination as well as neuronal injury and death (Fig. 7–8).

Pathogenetic effects of M518V mutation at CaM contact site of K_v7.2 helix B

The de novo M518V mutation was found in a boy who displayed drug-resistant neonatal tonic-clonic seizures, and later developed profound mental retardation, speech impairment, spasticity, and autism (Weckhuysen et al., 2012). In this study, we demonstrate that M518V mutation induced multiple defects in K_v7 channels. The M518V mutation located at the CaM N-lobe contact site in helix B severely reduced CaM binding to K_v7.2 (Fig. 1, 2) and surface expression of heteromeric K_v7.2/ K_v7.3 channels in distal axons (Fig. 3). M518 and R525 residues are critical residues that mediate electrostatic interaction with CaM N-lobe (Xu et al., 2013). Considering that BFNE mutation R525Q disrupts CaM interaction with K_v7.2 (Ambrosino et al., 2015), M518V mutation may primarily impair the electrostatic association between K_v7.2 and CaM N-lobe (Strulovich et al., 2016; Xu et al., 2013) by disrupting hydrophobic packing at the interface between K_v7.2 and CaM N-lobe (Fig. 1D). Consistent with previous studies in Xenopus oocyte by Maljevic and colleagues (Orhan et al., 2013), K_v7.2-M518V mutant channels displayed low level of K⁺ currents in CHOhm1 cells, which were still larger than in GFP transfected control cells, and enough to modestly hyperpolarize resting membrane potentials (Fig. 4, Table 1). Considering the crucial roles of CaM binding to K_v7.2 for polarized axonal surface expression, K_v7 channel conductance, and inhibition of hippocampal neuronal excitability (Cavaretta et al., 2014; Shahidullah et al., 2005), a severe reduction of CaM binding serves as one major pathogenic mechanism underlying the M518V mutation.

One unexpected effect of the M518V mutation is ubiquitination and accelerated proteasome-dependent degradation of K_v7.2 when expressed alone (Fig. 7), indicative of ER-associated degradation (Ellgaard and Helenius, 2003). Although CaM binding was severely reduced by A343D mutation in the CaM contact site of helix A and M518V mutation in the CaM contact site of helix B (Fig. 2), no significant ubiquitination was found in K_v7.2-A343D proteins. we speculate that disruption of helix B structure and subsequent loss of CaM binding to helix B by M518V mutation (Fig. 1–2) may reveal a unique misfolded conformation that can be recognized by ubiquitin ligases. Previous studies have shown that a BFNE frameshift mutation creates a novel proteasomal degradation signal RCXRG in K_v7.2 (Soldovieri et al., 2006; Su et al., 2011). Furthermore, ubiquitination of K_v7.3 on XPXPPY motif by the E3 ubiquitin ligase Nedd4–2 has been reported to decrease surface and current expression of K_v7.3-containing channels (Ekberg et al., 2007). Although the amino acid sequence surrounding M518V mutation does not conform to either RCXRG or XPXPPY motif, it is possible that M518V mutation itself may provide a novel ubiquitination signal. Since coexpression of K_v7.3 increased expression of K_v7.2-M518V proteins (Fig. 8, S7), K_v7.3 binding to K_v7.2-M518V may most likely hinder the recognition of ubiquitinated K_v7.2-M518V by the proteasome. Understanding the molecular mechanisms underlying ubiquitination and degradation of K_v7.2-M518V warrants future studies.

Misfolded proteins can perturb normal cellular functions and lead to cytotoxicity when the cells fail to perform chaperone-assisted refolding, ubiquitin-proteasome mediated degradation, or aggresome-autophagy dependent clearance of these proteins (Su et al., 2011; Tyedmers et al., 2010). Indeed, we were surprised to find that a large fraction of HEK293T cells transfected with K_v7.2-M518V alone displayed cell shrinkage and nuclear condensation, indicative of apoptosis (Fig. 8). Similarly, expression of K_v7.2-M518V alone or together with K_v7.3 in hippocampal neurons increased neurite fragmentation and the loss of ankyrin-G at the AIS (Fig. 8). Although intracellular deposits of ubiquitinated proteins have been found in neurons affected by multiple neurodegenerative diseases (Dantuma and Bott, 2014) and Dravet Syndrome (Kang et al., 2015), this study is the first to introduce neuronal injury as a possible pathogenic mechanism for K_v7.2 epileptic encephalopathy (Fig. 9).

Figure 9. A working model for pathogenic mechanisms of epileptic encephalopathy mutations in helices A and B of K_v7.2.

Epileptic encephalopathy mutations (R333W, K526N, and R532W) localized peripheral to CaM contact sites in helices A and B reduced axonal surface expression of K_v7 channels, altered their sensitivity to PIP₂, and disrupted their ability to inhibit excitability in hippocampal neurons. In contrast, M518V mutation at the CaM contact site in helix B severely decreased CaM binding, axonal surface expression, and voltage-dependent activation. Intracellular accumulation of K_v7.2-M518V also caused cell death. We propose that a combination of these multiple yet diverse defects in K_v7 channels could exert more severe impacts on neuronal excitability and health, and thus serve as pathogenic mechanisms underlying Kcnq2 epileptic encephalopathy.

It is unclear how expression of K_v7.2-M518V causes cell death. Transfection of K_v7.2-M518V alone increased cell death in HEK293T cells and hippocampal neurons (Fig. 8), suggesting that ubiquitination and accelerated degradation of K_v7.2-M518V may contribute to cell death. Previous studies reported that disruption of the ubiquitin-proteasome system leads to cell cycle disruption and programmed cell death (Bence et al., 2001; Ross and Poirier, 2005). Although misfolded K_v7.2-M518V proteins are initially degraded by the ubiquitin-proteasome system (Fig. 7), it is possible that failure of chaperone-assisted refolding and continuous production of misfolded K_v7.2-M518V proteins could saturate the capacity of the ER and proteasome, leading to ER stress, proteasome inhibition, and activation of apoptosis (Su et al., 2011; Tyedmers et al., 2010). Prevention of degradation by K_v7.3 (Fig. 8, S7) may further aggravate this situation by accumulating ubiquitinated K_v7.2-M518V proteins. Importantly, increased ER stress response has also been detected in the brains of human patients with temporal lobe epilepsy (TLE) and acute induction of seizures in rodent models of TLE (Engel et al., 2013; Kitao et al., 2001; Yamamoto et al., 2006), which are associated with sclerosis and apoptosis at or near the seizure foci (Coan et al., 2014). Severe reduction in axonal surface expression and current expression of K_v7.2-M518V-containing channels (Fig. 3–4) is expected to increase excitability of neurons compared to those expressing wild-type K_v7.2. Though highly speculative, such neuronal hyperexcitability induced by K_v7.2-M518V may further enhance the ER stress and apoptosis in addition to activating excitotoxicity.

The effects of K_v7.2-M518V expression on neuronal death and injury is particularly interesting since the brain magnetic resonance imaging (MRI) study of a patient with the M518V mutation shows substantial brain lesions and neuronal loss indicated by small frontal lobes, thinned splenium of corpus callosum, ventriculomegaly, and increased cerebral spinal fluid space (Weckhuysen et al., 2012). Neuronal death induced by K_v7.2-M518V accumulation may subsequently facilitate reorganization of circuits that results in hyperexcitability and seizures. In support of this notion, genetic inactivation of autophagy prevents the clearance of aggregated misfolded ubiquitinated proteins and causes neurodegeneration and spontaneous seizures in mice (Komatsu et al., 2006; McMahon et al., 2012), suggesting that epileptic seizures may arise as a consequence of abnormal accumulation of misfolded ubiquitinated proteins. An important task for future studies will be to dissect the molecular mechanisms underlying K_v7.2-M518V-mediated toxicity and its role in the etiology of epileptic encephalopathy.

Pathogenetic effects of R333W, K526N, and R532W mutations near CaM contact sites of K_v7.2 helices A and B

The R333W mutation was found in a boy who suffered from drug resistant seizures, Rolandic epilepsy, and psychomotor delay (Schmitt et al., 2005). The de novo R333W mutation was recently associated with intellectual disability and autism with high levels of recurrence (Geisheker et al., 2017). Our study is the first to characterize the effect of this mutation on CaM binding, expression, and function of K_v7.2. We found that the R333W mutation located proximal to the CaM C-lobe contact site in helix A reduced axonal surface expression of heteromeric channels by 30% (Fig. 3), consistent with its modest effect on reducing apoCaM binding to K_v7.2 (Fig. 2). Furthermore, K_v7.2-R333W channels displayed insensitivity to PIP₂ despite normal voltage-dependent activation (Fig. 4–5). Importantly, the R333W mutation blocked the ability of K_v7 channels to inhibit hippocampal neuronal excitability (Fig. 6) without affecting channel assembly (Fig. 2). Methylation of R333 has been recently reported to enhance K_v7.2 binding to PIP₂ (Kim et al., 2016), further implicating R333 as a critical residue for gating modulation by PIP₂. Considering the crucial roles of CaM binding to K_v7.2 in M-current expression and inhibition of hippocampal neuronal excitability (Shahidullah et al., 2005), a combination of reduced binding of CaM and PIP₂ to K_v7.2 as well as decreased axonal expression of heteromeric channels may underlie the inability of K_v7.2-R333W containing channels to suppress excitability in hippocampal neurons (Fig. 6).

The K526N mutation was found in all affected members in a BFNE family, who exhibited BFNE alone or together with severe epileptic encephalopathy that displayed drug-resistant seizures, mental retardation, white matter reduction, and enlarged lateral ventricles (Borgatti et al., 2004). Here, we show that K526N mutation located distal to the CaM N-lobe contact site in helix B caused a 30% reduction in axonal surface expression of heteromeric channels but did not significantly decrease K_v7.2 binding to apoCaM and K_v7.3 (Fig. 1–3). The K526N mutation induced a modest right shift in voltage-dependence of K_v7.2 channels without altering peak current densities in CHO hm1 cells (Fig. 4–5). A previous study by Borgatti et al reported a right shift in voltage-dependence of homomeric K_v7.2-K526N channels in CHO cells (Borgatti et al., 2004). The difference in voltage-dependence is unclear, but could be attributed to the use of different CHO cell lines. While Borgatti et al used CHO cells, we used CHO hm1 cells which stably express human muscarinic receptors (Kosenko et al., 2012). Another unexpected result was that the K526N mutation enhanced PIP₂ sensitivity in CHO hm1 cells (Fig. 5). The PIP₂ concentration is estimated to be 10 μM and in the range of 1–3% of total lipid (McLaughlin et al., 2002; van den Bogaart et al., 2011), which is much lower than the saturating concentration of diC8-PIP₂ used in this study. Although soluble diC8-PIP₂ will partition between the aqueous and membrane lipid phases, diC8-PIP₂ has a lower effective concentration in the membrane than in the pipette solution (Collins and Gordon, 2013). Furthermore, mice lacking a single copy of the Kcnq2 gene display increased susceptibility to convulsant agents (Watanabe et al., 2000). Therefore, the 30% reduction of axonal K_v7 channels by the K526N mutation, rather than increased PIP₂ sensitivity, may be sufficient to block their function in dampening hippocampal neuronal excitability (Fig. 6). It is also possible that this mutation may impair critical roles of K_v7 in mediating medium afterhyperpolarization (Tzingounis and Nicoll, 2008) or disrupt regulation of K_v7 channels by other signaling proteins including AKAP79/150 and syntaxin1A (Delmas and Brown, 2005; Etzioni et al., 2011; Hernandez et al., 2008; Hoshi et al., 2005; Hoshi et al., 2003; Regev et al., 2009; Suh and Hille, 2002; Suh and Hille, 2007).

Although the R532W mutation is located in the linker between helix B that mediates CaM binding (Alaimo et al., 2012; Gamper and Shapiro, 2003; Xu et al., 2013; Yus-Najera et al., 2002) and helix C that mediates subunit interaction (Schwake et al., 2006), this mutation did not alter K_v7.2 binding to apoCaM or K_v7.3 (Fig. 2). Instead, this mutation decreased axonal surface expression by 30% (Fig. 3). Furthermore, R532W mutation decreased K⁺ current density of K_v7.2 channels by half, right-shifted their voltage-dependence, slowed their activation kinetics, and impaired their gating modulation by PIP₂ (Fig. 4–5). Remarkably, expression of K_v7.2-R532W resulted in hyperexcitability in hippocampal neurons (Fig. 6), suggesting that this mutant subunit may dominantly suppress channel function upon forming heteromeric channels with endogenous K_v7.3. These multiple defects in K_v7 channels and subsequent neuronal hyperexcitability may underlie the etiology of de novo R532W mutation associated with drug-resistant neonatal tonic-clonic seizures, profound mental retardation, speech impairment, severe spastic quadriplegia, and mild microcephaly (Weckhuysen et al., 2012).

Implication of PIP₂ modulation residues near CaM contact sites in K_v7.2 helices A and B.

In the C-terminal tail of K_v7.1, PIP₂ has been reported to bind basic residues K354, K358, R360, and K362 in prehelix A (Thomas et al., 2011) and K526 and K527 in helix B (Tobelaim et al., 2017b). Consistently, introduction of mutations in these residues decreases PIP₂ interaction with the K_v7.1 C-terminal tail in vitro and gating modulation by PIP₂ (Thomas et al., 2011; Tobelaim et al., 2017b). In the cryo-EM structure of homomeric Xenopus K_v7.1, residues K526, K527, and K528 in K_v7.1 are well positioned near the membrane to interact with phospholipid head groups (Sun and MacKinnon, 2017). In K_v7.2 helix B distal to the CaM contact site, a cluster of basic residues (K524, R525, and K526 in protein accession CAA 75348.1) are structurally located close to the beginning of helix A where another cluster of positively charged basic residues (R325, K331, R332, and R333) are located (Fig. 1C). These positively charged residues in proximal helix A and distal helix B may constitute the site for PIP₂ modulation. This possibility is strongly supported by a previous report demonstrating reduced PIP₂ modulation of K_v7.2-R325G channels (Soldovieri et al., 2016), and by our findings that K526N enhanced whereas R333W reduced PIP₂ sensitivity of K_v7.2 channels (Fig. 5). Considering that K_v7.1-K528N proteins retain PIP₂ interaction to a similar extent as wild type K_v7.1 (Tobelaim et al., 2017b), increased PIP₂ sensitivity of K_v7.2-K526N channels may be caused by local conformational change. Our demonstration of PIP₂-insensitive K_v7.2-R532W channels (Fig. 5) not only reveal R532 as a novel PIP₂ interacting residue, but also suggests the helix B-C linker is another possible region that mediates PIP₂ sensitivity.

The proximity of PIP₂ and CaM binding sites raises an interesting possibility that coordinated actions of PIP₂ and CaM may regulate function and axonal surface expression of K_v7 channels. Competitive binding of PIP₂ and Ca²⁺-bound CaM to helix B has been recently demonstrated in K_v7.1 and K_v7.2 (Tobelaim et al., 2017a; Tobelaim et al., 2017b). However, current density of wild type K_v7.2 channels but not PIP₂-insensitive K_v7.2-R325G channels can be enhanced by coexpression of CaM or CaM1234 that cannot bind to Ca²⁺ (Ambrosino et al., 2015; Soldovieri et al., 2016), suggesting that both apoCaM and Ca²⁺-bound CaM may enhance current by increasing PIP₂ binding to K_v7.2 (Alberdi et al., 2015; Kosenko and Hoshi, 2013). Though highly speculative, we propose that PIP₂ binds a composite site formed by both CaM lobes and residues in K_v7.2 helices A and B.

Revisiting the role of CaM binding in polarized axonal surface expression of K_v7 channels.

While the A343D mutation in the IQ motif of helix A abolished apoCaM binding to K_v7.2 and axonal surface expression of heteromeric K_v7 channels (Fig. 2) (Cavaretta et al., 2014), the R333W mutation located proximal to the IQ motif in helix A modestly decreased apoCaM binding to K_v7.2 by 20% (Fig. 2) and reduced axonal surface expression of heteromeric channels by 30% (Fig. 3). In contrast to A343D mutation, the M518V mutation located at the CaM contact site in K_v7.2 helix B severely reduced apoCaM binding to K_v7.2 by 85% (Fig. 2) and surface expression of heteromeric channels in distal axons (Fig. 3). CaM binds to the antiparallel coiled-coil architecture of helices A and B of K_v7.2 with helix A binding to CaM C-lobe, and helix B interacting with the CaM N-lobe (Strulovich et al., 2016; Xu et al., 2013; Yus-Najera et al., 2002). Therefore, disruption of helix B interaction with the CaM N-lobe severely compromises helix A binding to CaM C-lobe. The residual interaction between apoCaM and K_v7.2-M518V through helix A (Fig. 2) could account for the low surface expression of heteromeric K_v7.3/K_v7.2-M518V channels at the AIS (Fig. 3). Since none of these mutations affected subunit interaction (Fig.3), these results together suggest that helix A provides a dominant interaction site for apoCaM that promotes trafficking of heteromeric channels from the ER to the axonal surface and that the degree of apoCaM binding to K_v7.2 helix A governs the amount of axonal K_v7 channels on the neuronal plasma membrane.

Surprisingly, the K526N mutation located distal to the CaM N-lobe contact site in helix B decreased axonal surface expression of heteromeric channels by 30%, although the observed 35% reduction in bound apoCaM did not reach statistical significance (Fig. 1–2). This result was unexpected because CaM binding to the K_v7.2 C-terminal tail is significantly reduced by BFNE mutation R525Q (Ambrosino et al., 2015), which immediately precedes K526. In addition, the R532W mutation in the helix B-C linker also decreased axonal surface expression by 30% (Fig. 3) without altering K_v7.2 binding to apoCaM or K_v7.3 (Fig. 1–2). It is unclear how K526N and R532W mutations modestly decreased axonal surface expression of K_v7 channels without significantly reducing K_v7.2 interaction with CaM or K_v7.3. We have previously shown that axonal surface expression of K_v7 channels can be reduced by 30% upon coexpression of CaM1234, which cannot bind to Ca²⁺ (Cavaretta et al., 2014), suggesting that Ca²⁺-bound CaM can further enhance channel targeting to the axonal surface. Considering that Ca²⁺-bound CaM favors interaction with helix B (Alaimo et al., 2012; Gamper and Shapiro, 2003; Xu et al., 2013; Yus-Najera et al., 2002), we speculate that K526N and R532W mutations may modestly decrease axonal surface expression of K_v7 channels by reducing K_v7.2 affinity to Ca²⁺-bound CaM.

CONCLUSIONS

Here, we discovered molecular pathogenetic mechanisms for epileptic encephalopathy mutations in helices A and B and helix B-C linker of K_v7.2. These mutations cause a reduction in K_v7 polarized axonal surface expression and K_v7 function to inhibit neuronal excitability by variably decreasing CaM binding, voltage-dependent activation, and PIP₂ sensitivity (Fig. 9). One particular mutation M518V induced abnormal accumulation of ubiquitinated K_v7.2 in the presence of K_v7.3 and caused neuronal injury (Fig. 9), opening a new avenue for exploring therapeutic drugs that can alleviate accumulation of misfolded proteins. Since inhibition of K_v7 currents during early neonatal development induces hippocampal sclerosis in addition to increased seizure susceptibility (Peters et al., 2005), our discovery supports the critical importance in early identification of a mutation and its pathogenic mechanism in order to develop effective mutation-specific therapy to combat neuronal hyperexcitability and remediate neuronal injury associated with drug-refractory K_v7.2 epileptic encephalopathy.

ABBREVIATIONS

CaM

calmodulin

K⁺

potassium

AP

action potential

AIS

axonal initial segments

BFNE

benign familial neonatal epilepsy

EEG

electroencephalogram

PIP₂

phosphatidylinositol 4,5-bisphosphate

AKAP

A-kinase-anchoring protein

ER

endoplasmic reticulum

YFP

yellow fluorescent protein

TLE

temporal lobe epilepsy