Early-onset epileptic encephalopathy caused by a reduced sensitivity of Kv7.2 potassium channels to phosphatidylinositol 4,5-bisphosphate

Kv7.2 and Kv7.3 subunits underlie the M-current, a neuronal K+ current characterized by an absolute functional requirement for phosphatidylinositol 4,5-bisphosphate (PIP2). Kv7.2 gene mutations cause early-onset neonatal seizures with heterogeneous clinical outcomes, ranging from self-limiting benign familial neonatal seizures to severe early-onset epileptic encephalopathy (Kv7.2-EE). In this study, the biochemical and functional consequences prompted by a recurrent variant (R325G) found independently in four individuals with severe forms of neonatal-onset EE have been investigated. Upon heterologous expression, homomeric Kv7.2 R325G channels were non-functional, despite biotin-capture in Western blots revealed normal plasma membrane subunit expression. Mutant subunits exerted dominant-negative effects when incorporated into heteromeric channels with Kv7.2 and/or Kv7.3 subunits. Increasing cellular PIP2 levels by co-expression of type 1γ PI(4)P5-kinase (PIP5K) partially recovered homomeric Kv7.2 R325G channel function. Currents carried by heteromeric channels incorporating Kv7.2 R325G subunits were more readily inhibited than wild-type channels upon activation of a voltage-sensitive phosphatase (VSP), and recovered more slowly upon VSP switch-off. These results reveal for the first time that a mutation-induced decrease in current sensitivity to PIP2 is the primary molecular defect responsible for Kv7.2-EE in individuals carrying the R325G variant, further expanding the range of pathogenetic mechanisms exploitable for personalized treatment of Kv7.2-related epilepsies.

The Kv7.2 R325G mutation affects PIP 2 -dependent regulation. The R325 residue, highly conserved among Kv7 subunits, is located in the short linker connecting the S 6 segment with the calmodulin (CaM)-binding A-helix of Kv7.2 C-terminus (Fig. 4a) 23 . In Kv7.1 channels, PIP 2 binding to this proximal cytosolic linker is a critical determinant of open pore stability 24,25 ; therefore, the potential contribution of the R325 residue to PIP 2 binding was investigated by molecular modeling studies. In particular, we grafted a short-chain derivative of PIP 2 (dioctanoyl-PIP 2 ; diC8-PIP 2 ), whose crystal coordinates were taken from the diC8-PIP 2 -bound configuration of the Kir 2.2 channel 26 onto a Kv7.2 homology model built on a Kv7.1 homology model integrating the crystal structure of the Kv7.1 proximal C-terminus including the A and B helices 27 . Starting from this model, docking experiments were performed to find the best-scoring Kv7.2/PIP 2 configuration. The data obtained revealed that, within this region, the negatively-charged PIP 2 molecule is involved in an intricate network of electrostatic interactions with the side chains of residues in the S 2 -S 3 linker (F163, R165), in the S 4 -S 5 linker (S223), and in the pre-helix A region (K319, E322, R325, and Q326) (Fig. 4b). In particular, the R325 side chain interacts with both the 3′ OH and the 4′ PO 4 2− of the PIP 2 molecule; the stability of the interaction between the ζ -carbon of the Kv7.2 R325 residue and the phosphorus atom at C4′ of PIP 2 was confirmed by molecular dynamics experiments over a 10 ns time range (Suppl. Fig. 1).
The R325G mutation does not interfere with the subcellular localization of Kv7.2 subunits in rat hippocampal neurons in vitro. In neurons, Kv7.2 and Kv7.3 subunits are primarily expressed at the axon initial segment (AIS) 37,38 . Epilepsy-causing mutations in Kv7.2 may interfere with such AIS targeting 39,40 . To assess whether the Kv7.2 R325G mutation also prompted similar effects, wild-type or mutant Kv7.2 subunits were expressed (together with Kv7.3 subunits) by transient transfection in embryonic rat hippocampal neurons. Confocal immunofluorescence experiments in non-permeabilized neurons revealed that plasma membrane expression of Kv7.2 and Kv7.2 R325G subunits could be almost exclusively detected at the Ankyrin-G-positive AIS ( Fig. 7a and b); both wild-type and mutant Kv7 subunits displayed an identical AIS expression pattern (Fig. 7c), with immunoreactivity steeply increasing within the more distal regions of the AIS (Fig. 7b), as reported for native channels in rat neocortical neurons 38 . By contrast, Kv7.2 subunits carrying a different EE-associated variant affecting a pore residue (A294V) failed to localize at the AIS, as previously described 39 (Fig. 7).

Discussion
The primary defect of Kv7.2 R325G channels is a decreased PIP 2 sensitivity. Currents carried by all five members of the Kv7 family are characterized by their absolute functional requirement for PIP 2 3,41,42 . In the present experiments, homomeric channels formed by Kv7.2 subunits carrying a mutation (R325G) found independently in four individuals affected with Kv7.2-EE 15-17 , were non functional, despite being expressed at the plasma membrane; Kv7.2 R325G subunits strongly suppressed channel function when incorporated into heteromers with Kv7.2, Kv7.3 or Kv7.3* subunits. The lack of Kv7.2 R325G homomeric channel function was partially rescued by co-expression with PIP5K, a lipid kinase which elevates PIP 2 concentration in the plasma membrane to millimolar levels 28,30 and reduces the ability of G q -coupled receptors to suppress Kv7 currents 36 , strongly suggesting that the primary dysfunction triggered by the R325G mutation is a drastic reduction in Kv7.2 sensitivity to endogenous levels of PIP 2 . Consistent with this view is also the decreased potency shown by the water-soluble PIP 2 analogue diC8-PIP 2 in activating Kv7.2 channel carrying the R325A mutation 43 . Noteworthy, at variance with Kv7.2 R325G channels, homomeric Kv7.2 R325A channels, similarly to Kv7.1 channels carrying the equivalent R360A mutation both in the absence 44 or in the presence 45 of KCNE1 subunits, were functional, although they displayed a reduced current amplitude 43,46 ; the α -helix-stabilizing propensity of alanine relative to glycine 47 provides a plausible explanation for such relevant functional difference. Noteworthy, all Kv7.2 missense variants causing EE (including the c.973A > G leading to the R325G mutation herein investigated) are de novo substitutions of a single nucleotide 7,48 . Instead, a substitution of at least two nucleotides would be needed to generate the arginine-to-alanine mutation studied by Telezhkin 43 ; thus, the R325A mutation is less likely to occur in vivo.
Kv7.2 R325G mutant subunits, when incorporated in heteromeric channels with Kv7.2 and Kv7.3 subunits at a ratio reproducing the genetic balance of epilepsy-affected patients, exerted dominant-negative effects, enhanced current suppression by PIP 2 depletion with VSP, and slowed current recovery kinetics after VSP turnoff; all these results, beside confirming the molecular mechanism of the primary dysfunction, also provide strong support for a significant role of the decreased PIP 2 sensitivity in severe epilepsy pathogenesis in patients carrying the R325G variant.

Mechanistic and structural consequences of the reduced PIP 2 sensitivity of Kv7.2 R325G subunits.
In Kv7.1 channels, PIP 2 is not required for voltage-sensing domain (VSD) activation, being rather indispensable for coupling VSD activation to pore opening 25 . In particular, molecular dynamics experiments revealed that PIP 2 stabilizes the open pore configuration by reducing the electrostatic repulsion among positively-charged residues in the S 4 -S 5 linker, in S 6 , and in the proximal C-terminus immediately past S 6 24 . Unfortunately, a truncated version of the Kv7.1 channel (residues 122 to 358) was used in these experiments 24 ; thus, the potential role of the Kv7.2 R325 residue (corresponding to R360 in Kv7.1) could not be evaluated.
In Kv7.2 channels, as herein confirmed, PIP 2 up-regulates current density and facilitates voltage-dependent opening 6,49 ; dynamic repositioning of PIP 2 from the VSD to the open pore gate occurs during activation 49 , whereas the reverse movement correlates with channel deactivation 46 . Thus, a critical PIP 2 binding site in Kv7.1 and Kv7.2 channels is located at the VSD-pore domain interface, where PIP 2 headgroups are engaged in electrostatic interactions with basic residues in the VSD and the proximal C-terminus (including the S 6 gate) 3,41 , thereby bridging these two domains and providing structural stabilization 1 . The present modeling studies confirm the critical contribution of the R325 residue to PIP 2 binding in the proximal C-terminal pocket in Kv7.2 subunits; substitution of the positively-charged arginine with a smaller, non-polar glycine may significantly weaken such structural stabilization, impeding the translation of the electromechanical forces triggered by VSD displacement into pore opening. Direct crystallographic evidence for a contribution to PIP 2 binding of positively-charged residues located at the interface between the transmembrane and the cytoplasmic domains has been achieved in Kir2.2 26 , a PIP 2 -gated voltage-independent channel 50 . Finally, PIP 2 binding to a similar region has been shown to stabilize the voltage sensor of Kv1 channels in a state of decreased voltage sensitivity, thus promoting functional changes opposite to those described for Kv7 channels 51 .
Coordinated regulation of Kv7.2 channels by PIP 2 and calmodulin. The ubiquitous Ca 2+ -binding protein CaM exerts a critical control over Kv7.2 channel function 23,52 . Changes in CaM binding and functional regulation have been described for several disease-causing Kv7.2 mutations 34 , and co-expression with CaM has been shown to rescue Kv7.2 channels rendered non-functional by mutations in helices A or B of the CaM-binding domain 33 . However, CHO cells transfection with either CaM or with a mutant CaM unable to bind Ca 2+ (CaM 1234 ) 32 failed to recover functional homomeric Kv7.2 R325G channels; moreover, CaM or CaM 1234 did not further potentiate PIP5K-induced current enhancement in both Kv7.2 and Kv7.2 R325G channels. These results are consistent with the hypothesis that CaM-induced enhancement of Kv7.2 macroscopic current density is mostly mediated by changes in PIP 2 affinity 34,53-55 . CaM also regulates polarized axonal surface expression of Kv7.2 subunits 56,57 ; normal AIS expression of heteromeric channels carrying Kv7.2 R325G subunits was observed in cultured hippocampal neurons, suggesting that no significant mutation-induced changes in CaM binding occurs, and that changes in PIP 2 -dependent regulation do not impede AIS trafficking of mutant Kv7.2 subunits. Pharmacological implications and conclusions. Retigabine is the prototype anticonvulsant acting as an activator of neuronal Kv7 channels (Kv7.2-5); retigabine causes a variable degree of hyperpolarization shift of the voltage dependence of channel activation, together with an increase in channel maximal opening probability 22,58 . The present observation that increasing cellular PIP 2 levels with PIP5K negatively shifted the V ½ of Kv7.2 channels (as previously described for Kv7.1 59 , Kv7.2 49 , and other heteromeric Kv7 channels 31 ) and decreased retigabine-induced responses in both in Kv7.2 and Kv7.2 R325G channels, indicates that retigabine and PIP 2 act via at least partially overlapping mechanisms to stabilize voltage-dependent pore opening; consistent with this view is the fact that PIP 2 -depleted Kv7.3 channels are insensitive to retigabine 60 .
In conclusion, the present results add severe forms of Kv7.2-related epilepsy to the growing list of channelopathies caused by changes in PIP 2 -dependent regulation 1,59 ; the recent observation that Kv7 channels are critical determinants of the cortical excitability changes occurring upon dynamic regulation of PIP 2 levels 31 , lends further support to the pathogenetic role of the proposed mechanism in individuals carrying the Kv7.2 R325G variant. Future studies will further explore the structural implications of the functional results herein presented, and define whether such specific molecular defect is associated with distinct clinical features.

Mutagenesis and heterologous expression of channel subunits. Mutations were engineered
by Quick-change mutagenesis (Agilent Technologies) in a pcDNA3.1-Kv7.2 plasmid (for electrophysiological and western-blot experiments), or in a dual-tagged Enhanced Green Fluorescent Protein-Kv7.2-hemagglutinin (EGFP-Kv7.2-HA) plasmid (for immunocytochemistry experiments), as described 21,34,61 . In the EGFP-Kv7.2-HA chimeric constructs, in addition to an EGFP at the cytoplasmic N-terminus, an HA epitope was inserted in the extracellular loop that connects transmembrane domains S 1 and S 2 of Kv7.2 subunits, as previously described 61,62 . Protocols for wild-type and mutant cDNAs expression by transient transfection, as well as methods for CHO cell growth, have been previously described 34 . Total cDNA in the transfection mixture was kept constant at 4 μ g, except for Western-blotting experiments (6 μ g).

Patch-clamp recordings. Macroscopic current recordings from transiently-transfected CHO cells, as
well as data processing and analysis, were performed as reported 11 . In the experiments with TEA or retigabine (obtained from Valeant Pharmaceuticals, Aliso Viejo, CA), currents were activated by 3-s voltage ramps from − 80 mV to 0/+ 40 mV at 0.1-Hz frequency. TEA blockade was expressed as the percentage of peak current inhibition produced by a 2-min drug application. Molecular modeling. Homology modeling. A homology model of a Kv7.2 subunit was generated starting from a Kv7.1 homology model which integrates the crystal structure of the Kv7.1 proximal C-terminus 27 , using the SWISS-MODEL software 64 . This model was then aligned with the crystal structure of Kir2.2 channels bound to diC8-PIP 2 26 , using the Chimera matchmaker tool (https://www.cgl.ucsf.edu/chimera/docs/ContributedSoftware/ matchmaker/matchmaker.html); the PIP 2 crystal coordinates were then superimposed onto the Kv7.2 model. Protein preparation. To obtain a satisfactory starting structure for following studies, the Kv7.2 protein was prepared using the Schrodinger Protein Preparation Wizard 65 . The orientation of hydroxyl groups on S, T and Y, the side chains of N and Q residues, and the protonation state of H residues were optimized. N-and C-terminal residues were capped with acetyl and N-methyl-amide residues, respectively. The ionization and tautomeric states of H, D, E, R and K residues were adjusted to match a pH of 7.4. The structure was finally submitted to a restrained minimization (OPLS2005 force field) 66 that was stopped when the root-mean-square deviation (RMSD) of heavy atoms reached 0.30 Å. Docking studies. The Schrodinger Induced Fit Docking Extended Sampling protocol 67,68 was used for docking studies of PIP 2 on the optimized Kv7.2 configuration. The docking space, centered on the PIP 2 molecule, was defined as a 32 Å 3 cubic box, while the diameter midpoint of docked ligands was restrained within a smaller, nested 22 Å 3 cubic box. Residues within 10 Å of ligand poses were refined by the Prime Software (https://www.schrodinger.com/prime). Molecular Dynamics simulations. The stability of the best scoring PIP 2 / Kv7.2 complex was further investigated by Molecular Dynamic (MD) simulations using Desmond MD system (https://www.schrodinger.com/desmond). The simulated environment was built using the system builder utility, with the structures being neutralized by Na + and Cl − ions, which were added to a final concentration of 0.15 M. Simulations were run in explicit solvent, using the TIP4P water model 69 in a Periodic Boundary Conditions orthorhombic box. A series of minimizations and short MD simulations were carried out to relax the model system, by means of a relaxation protocol consisting of six stages: (i) minimization with the solute restrained; (ii) minimization without restraints; (iii) simulation (12 ps) in the NVT ensemble using a Berendsen thermostat (10 °K) with non-hydrogen solute atoms restrained; (iv) simulation (12 ps) in the NPT ensemble using a Berendsen thermostat (10 °K) and a Berendsen barostat (1 atm) with non-hydrogen solute atoms restrained; (v) simulation (24 ps) in the NPT ensemble using a Berendsen thermostat (300 °K) and a Berendsen barostat (1 atm) with non-hydrogen solute atoms restrained; (vi) unrestrained simulation (24 ps) in the NPT ensemble using a Berendsen thermostat (300 °K) and a Berendsen barostat (1 atm). At this point, a 10 ns long MD simulation was carried out at a temperature of 300 °K in the NPT ensemble using a Nose-Hoover chain thermostat and a Martyna-Tobias-Klein barostat (1.01325 bar). Backbone atoms were constrained during the simulation (10 kcal/mol). Trajectory analyses were performed using the Desmond simulation event analysis tool.
Neuronal cell transfection and immunocytochemistry. Hippocampal cultures were prepared from 18-day embryonic rats, as described 70 . At 8 days in vitro (8 DIV), neurons were co-transfected with EGFP-Kv7.2-HA (wild-type or mutant) and Kv7.3 cDNAs (ratio 1:1, total 2 μ g) using Lipofectamine 2000 and immunostaining was performed at room temperature 72 hr after transfection. For surface immunostaining, neurons were fixed in 4% paraformaldehyde/4% sucrose for 10′ at 37 °C, blocked with 10% normal goat serum in PBS and incubated with rabbit anti-HA antibodies (1:60; 745500; Invitrogen) diluted in 10% normal goat serum in PBS for 1 hr. For permeabilized immunostaining, neurons were incubated with mouse monoclonal anti-ankyrin-G antibodies (1:200; clone 106/36; Millipore) diluted in permeabilizing buffer (15 mM phosphate buffer pH 7.4 containing 0.1% gelatin, 0.3% Triton X-100 and 0.4 M NaCl) for 2 hr. After PBS wash, neurons were incubated simultaneously with rabbit AlexaFluor555-and mouse AlexaFluor649-conjugated secondary antibodies (1:400 and 1:300, respectively; LifeTechnologies) for 1 hr in permeabilizing buffer. Coverslips were then mounted with moviol. Confocal image acquisition was performed on a Zeiss LSM510 Meta laser scanning microscope equipped with a 63x oil immersion lens. The ImageJ software was used for image analysis. Axonal Ank-G and HA signals were measured every 0.14 μ m along a 40 μ m-long region starting from the soma; values (expressed as fluorescence arbitrary units of intensity) in each neuron were normalized, and averaged every 5 points to decrease signal noise. AIS/Soma and AIS/Dendrites ratios were calculated by expressing the HA fluorescence (measured in a 20-30 μ m AnkG-positive area) versus the EGFP fluorescence of a 50 μ m 2 rectangle in the soma (AIS/Soma) or versus a 25 μ m-long region of the main dendrite (AIS/Dendrite) 39 .

Animals. Experimental procedures were performed in accordance with the European Communities Council
Directive (86/809/EEC) on the care and use of animals and the UK Animals (Scientific Procedures) Act 1986, and were approved by the Animal Care and Use Committee of the CNR Institute of Neuroscience.
Statistics. Data are expressed as mean ± SEM. Each data point shown in figures or in the text is the Mean ± SEM of at least 4 determinations, each performed in a single cell or in a separate experiment. Statistically significant differences were evaluated with the Student's t-test or with the ANOVA followed by the Student-Newman-Keuls test, with the threshold set at p < 0.05.