Uncoupling sodium channel dimers rescues phenotype of pain-linked Nav1.7 mutation

The voltage-gated sodium channel Nav1.7 is essential for an adequate perception of painful stimuli. Its mutations cause various pain syndromes in human patients. The hNav1.7/A1632E mutation induces symptoms of erythromelalgia and paroxysmal extreme pain disorder (PEPD), and its main gating change is a strongly enhanced persistent current. Using molecular simulations, we demonstrate that the disease causing persistent current of hNav1.7/A1632E is due to impaired binding of the IFM motif, thus affecting proper function of the recently proposed allosteric fast inactivation mechanism. By using native polyacrylamide gel electrophoresis (PAGE) gels, we show that hNav1.7 dimerizes. The disease-linked persistent current depends on the channel’s functional dimerization status: Using difopein, a 14-3-3 inhibitor known to uncouple dimerization of hNav1.5, we detect a significant decrease in hNav1.7/A1632E induced persistent currents. Our work identifies that functional uncoupling of hNav1.7/A1632E dimers rescues the pain-causing molecular phenotype by interferes with an allosteric fast inactivation mechanism, which we link for the first time to channel dimerization. Our work supports the concept of sodium channel dimerization and reveals its relevance to human pain syndromes.

Voltage-gated sodium channels (VGSC/Nav) play a crucial role in the perception and transduction of 2 painful signals (Ahern et al, 2016) and mutations leading to chronic pain syndromes severely affect the 3 patients' quality of life (Lampert et al, 2010). Human VGSC consist of four domains (DI-DIV), each 4 containing six transmembrane segments (S1-S6) ( segments of each domain, and it is opened by tethered voltage-sensing parts, formed by S1-S4, which 7 move outward upon depolarization (Catterall, 2014;Guy & Seetharamulu, 1986;Sula et al, 2017;8 Payandeh et al, 2011). Within milliseconds after pore opening the VGSCs inactivate: Three amino acids 9 located in the DIII-DIV linker were identified to form the so-called IFM motif or inactivation particle. 10 Until very recently, it was thought that the IFM motif blocks ion permeation by binding to the 11 cytoplasmic side of the pore, thus causing fast inactivation (Armstrong et al, 1973;West et al, 1992). 12 This process has been described as "hinged lid" mechanism (Eaholtz et al, 1994; West et al, 1992). 13 However, with the recently published high-resolution structure of VGSC subtypes NavPas, Nav1.4 and 14 Nav1.7, a new fast inactivation mechanism has been proposed: Instead of a direct occlusion of the 15 pore by the IFM motif, the inactivation particle may bind to a hydrophobic binding pocket in the 16 periphery of the S6 helical bundle and thus cause fast inactivation due to an allosteric effect. This 17 induces movement of the DIVS6 helix towards the ion permeation pathway. At the same time DIIIS6 is 18 pulled towards the ion pore, possibly causing the DIS6 and DIIS6 helices to move in the same direction 19 to close the inner gate of the permeation pathway ( protein difopein (dimeric-fourteen-three-three-peptide inhibitor) (Masters & Fu, 2001). Difopein 30 blocks 14-3-3 and is therefore suggested to be able to inhibit functional coupling of VGSC dimers 31 (Clatot et al, 2017; Masters & Fu, 2001). 32 In the present study, we focus on the pain-related VGSC subtype Nav1.7, which is expressed in role of Nav1.7 in the generation of human pain. 5 We study the previously reported gain-of-function mutation hNav1.7/A1632E mutation, which induces 6 a combination of IEM and PEPD in the heterozygous carrier and is characterized by an incomplete fast 7 inactivation leading to a prominent persistent current (Estacion et  dimerizes and we find evidence that this channel-dimerization affects the allosteric fast inactivation 17 mechanism also in unmutated WT channels. In addition, we present data that reveal how functionally 18 uncoupling dimerization significantly reduces the size of the disease-relevant persistent current of 19 hNav1.7/A1632E. 20 Results 1 hNav1.7/A1632E impairs fast inactivation -evidence for an allosteric mechanism 2 Using patch-clamp of HEK cells overexpressing the pain-linked mutation hNav1.7/A1632E, we 3 confirmed its reported prominent persistent current ( Fig. 1A-C

5
A1632 of DIVS5 is shown as spheres, the IFM motif is shown as orange sticks. 6 B) Close-up on the IFM-binding pocket, a hydrophobic cavity formed by residues from the DIIIS4-S5 linker, 7 DIIIS5, DIIIS6, DIVS5, and DIVS6 in WT and A1632E Nav1.7 (thin sticks, pocket-forming residues; thick sticks, 8 IFM-motif residues; spheres, residue 1632). When the same samples were resolved by high resolution Clear Native-PAGE (hrCN-PAGE), both 1 hNav1.5 GFP and hNav1.7 GFP migrated under non-denaturing conditions at masses of about 560 kDa (Fig.  2 3B, lanes 1 and 3). We compared our results with the migration of the co-resolved hTrpV1 GFP , a 492 3 kDa homotetramer (Julius, 2013) consisting of four protomers with a calculated mass of 123 kDa each 4 ( Fig. 3B). Treatment with a low concentration of 0.1 % of the denaturing detergent lithium dodecyl 5 sulfate (LiDS) resulted in a dissociation of the hTrpV1 GFP homotetramer into the homodimer and the 6 protomer with calculated masses of 246 and 123 kDa, respectively (Fig. 3B, lanes 5 and 6). Also, the 7 VGSC proteins dissociated into faster migrating species. Based on the migration of the mass marker 8 hTrpV1 GFP , hNav1.7 GFP migrated at 570 kDa and 290 kDa in the absence and presence of LiDS, 9 respectively ( Fig. 3B, lanes 3 and 4). We conclude from these results that both hNav1.5 and hNav1.7 10 have a strong propensity to assemble in X. laevis oocytes as homodimers. 11 To address a contribution of the nature of the solubilization detergent on the oligomeric state, we 12 solubilized hNav1.7 GFP -expressing oocytes in three additional non-ionic detergents, NG310, GDN and 13 digitonin besides DDM. hrCN-PAGE resolved identical migration patterns, consistent with 14 homodimeric and monomeric states in the absence or presence of the denaturing detergent LiDS, 15 respectively (Fig. 3C). Thus, our data clearly indicate that apart from hNav1.5, also hNav1.7 is able to 16 form homodimers. 17 1 Figure 3: Nav1.5 and Nav1.7 migrate as non-covalent homodimers in native PAGE gels.    cell line led to a significantly reduced current density (Fig. 4B). Adding difopein, which is supposed to 7 interfere with functional channel coupling, partially restored the current density, suggesting that a 8 dimerization between hNav1.7/R896Q and WT can be responsible for the reduction in current density. 9 Difopein has no impact on the current density of hNav1.7 WT (Fig. 4B). 10 The truncation mutation hNav1.7/G375Afs abolishes sodium current (

12
Data are shown as mean ± SEM.

13
Now that we have established evidence for functional dimerization of hNav1.7 we set out to 14 investigate how this may affect the allosteric inactivation mechanism: We co-expressed difopein with 15 hNav1.7/A1632E in order to suppress functional dimerization. The presence of difopein cut the 16 persistent current of hNav1.7/A1632E in half, reducing it from 14.9 ± 1.4% without to 7.0% ± 0.6% with 17 difopein (Fig. 5 A+B). Thus, functional dimers of Nav1.7 seem to affect the disease causing persistent 18 current of hNav1.7/A1632E, adding additional pathophysiological evidence to channel dimerization.   In the present study we show that the pain-linked mutation hNav1.7/A1632E impairs the allosteric fast 2 inactivation mechanism of the channel, we demonstrate that Nav1.7 forms functional dimers and that 3 uncoupling dimers reduces disease-relevant persistent current of the hNav1.7/A1632E mutation. 4 Allosteric fast inactivation is impaired by hNav1.7/A1632E leading to persistent currents 5 The prominent persistent current of hNav1.7/A1632E is likely to underlie the symptoms associated Xu et al, 2019). According to this allosteric mechanism, the IFM motif binds to a hydrophobic pocket 12 which is built by residues including A1632 on the S4-S5 linker ( Fig. 2A). Substitution of A1632 with 13 glutamate adds additional size, which may sterically impair IFM binding, but it also increases hydration 14 of the pocket. This seems to severely interfere with IFM binding and thus likely causes the prominent 15 persistent current of hNav1.7/A1632E. 16 Interestingly, replacing A1632 with aspartate, which is smaller than glutamate but also carries a 17 Nav1.7 forms functional dimers 26 Our data show that next to hNav1.5 and hNav1.2, hNav1.7 also forms functional dimers. Two binding 27 sites for 14-3-3 were reported: One between amino acids 416 -467, the second one between amino 28 acids 517 -555, both on the DI-DII linker (Supplementary Figure S2). In addition, a 14-3-3 independent 29 channel interaction site, the so-called α-α subunit interaction site, was reported to reside between 30 these two 14-3-3 binding sites (Clatot et al, 2017). In fact, we suggest that channel-dimerization may 31 be considered as a general feature of VGSC, since alignment of the suggested dimerization sites 32 revealed conserved amino acids for almost all subtypes in comparison to hNav1.5 except for hNav1.4. 1 (Supplementary Figure S2). 2 Fast inactivation of hNav1.7 WT is modified by difopein, suggesting that regulation by interaction with 3 14-3-3 may function as a modulator of channel gating under physiological conditions. Depolarizing 4 voltage-depencence of fast inacitvation to a similar extent as that observed by difopein expression 5 with Nav1.7 WT were reported before to be induced by mutations identified in pain patients (e.g. Wu 6 et al, 2013), revealing that very small changes can have a significant impact on cellular excitability. The The absence of dimerization could allow the WT channel to travel to the cell membrane, causing more 32 sodium current to occur. We showed that although the current density was significantly reduced in 33 the presence of WT and hNav1.7/R896Q, a small sodium current was still present. Our results go along 1 with a recent study on hNav1.7 in iPSC-derived nociceptors of CIP patients: The authors showed that 2 in neurons with a bi-allelic expression of a CIP mutation, restoring one deficient hNav1.7 allele was 3 enough to regain some but not all of the electrophysiological functions of these neurons (McDermott 4 et al, 2019). Channel dimerization affects the functioning channels in a dominant-negative way but the 5 remaining channel function seems still sufficient for pain perception. Thus, it appears that one Nav1. 7 6 allele produces enough functioning channels to generate sodium current and support pain perception, 7 and therefore CIP mutations occure in patients always as homozygous or compound heterozygous 8 mutations. 9 The truncation mutation hNav1.7/G375Afs, on the other hand, does not seem to dimerize with the WT 10 likely because the mutant channel is truncated before of the assumed dimerization site (Clatot et al, 11 2017). It is also possible that the mutant channel did not express adequately, and that dimerization 12 was thus absent. However, since our other mutants expressed sufficiently and 14-3-3 is present in our 13 expression system (Supplementary Figure S1), we argue that the conditions for dimerization were 14 given and that the abolishment of the putative dimerization site is likely to be the reason for missing 15 effects of difopein. Here, we show that hNav1.7 forms functionally relevant dimers, that this dimerization modifies the 24 mutation-induced phenotype of the pain-linked hNav1.7/A1632E substitution, and that this 25 modification most likely appears via the allosteric inactivation mechanism suggested by recently 26 published VGSC structures. Our work supports the concept of sodium channel dimerization, its 27 physiological function and reveals its relevance to human pain syndromes. 28 Only green fluorescent cells were used for whole-cell patch clamp recordings 27-60 hours after 6 transfection using a HEKA EPC 10 USB patch-clamp amplifier (HEKA Electronics, Lambrecht, Germany). 7

Materials and Methods
Sampling rate was set to 10kHz. Patch Pipettes were pulled with a DMZ pipette puller (ZEITZ 8 Instrumente Vertrieb GmbH, Martinsried, Germany) and pipettes with a tip resistance in-between 0.8 9 and 3.0 MΩ were used. All recordings were performed at room temperature (21°C±2C) and the liquid 10 junction potential was not corrected. 11 For patch clamp recordings, the following bath solution was used: 140mM NaCl, 3mM KCl, 1mM MgCl2, 12 1mM CaCl2, 10mM HEPES, 20mM Glucose. PH was adjusted to 7.4 using NaOH, the osmolarity of the 13 solution was 310±10 mOsm. Internal pipette solution contained the following: 10mM NaCl,140mM 14 CsF, EGTA 1mM, 10mM HEPES, 18mM Sucrose. The pH was adjusted to 7.33 and the osmolarity was 15 310±10 mOsm. 16 For HEK cell recordings, series resistance was below 7MΩ at any time for all of the cells and 17 compensated to ≥65%. Leak current subtraction was performed online via a P/4 procedure. After 18 reaching whole cell configuration, cells were held at a holding potential of -120mV for 3min and pulsed 19 with 0.1Hz to allow stabilization of the inward current. Immediately afterwards, the current-voltage-20 relationship was measured by stepwise 40ms depolarizations from -90mV to +40mV in 10mV 21

increments. 22
A Boltzmann equation was used for fitting: G/Gmax=(Gmax-Gmin)/(1+exp[(V1/2−Vm)/k]). (Gmax= maximum 23 sodium conductance, V1/2= membrane potential at half maximal activation, Vm= Membrane potential, 24 k= slope factor). Gmax was set to 1 and Gmin to 0. 25 Fast inactivation was measured 10 s after the end of the activation protocol by applying a test-pulse 26 to 0mV for 40ms after pre-pulses of 500ms. Pre-pulses were increased in 10mV increments from -27 130mV to -10mV. Cells were held at a holding potential of -120mV. The same equation as shown above 28 was used for fitting. Imax was set to 1, Imin was not defined. 29 The persistent current measured between 34ms and 39.6ms of each 40ms test-pulse was normalized 30 to the transient peak inward current of the same cell. The maximum persistent current of each cell 31 was then used for comparison . 32 Polymerase Chain Reaction (PCR) 1 PCR was performed with RNA extracted from untransfected HEK293T cells and the WT cell line. For 2 extraction, a Nucleospin RNA kit from Macherey-Nagel (Düren, Germany) was used and the 3 instructions were followed as provided by the company. cDNA was synthesized with a sensifast cDNA 4 synthesis kit by bioline (London, UK). To test for different 14-3-3 isoforms, seven different human-5 primers fabricated by eurofins genomics (Ebersberg, Germany) were used (Table1). Taq-Polymerase, 6 Thermopol Buffer and dNTP from New England BioLabs inc. (Frankfurt am Main, Germany) were used. 7 The protocol included 35 cycles and an annealing temperature at 52°C.  Biochemical analysis of hNav1.5 GFP and hNav1.7 GFP expressed in X. laevis oocytes Pearson test. In case of non-parametric testing, a Mann-Whitney test was applied; in case of 5 parametric testing an unpaired t-test was used. When comparing more than two groups, an ordinary 6 one-way ANOVA with Bonferroni post-hoc correction was used. To quantify variation, the standard 7 error of the mean (SEM) is displayed. Concerning the persistent current and current density, outliers 8 were checked for significance using Grubb's test (GraphPad QuickCals, San Diego CA, USA) and 9 excluded from analysis. Data are shown as mean ± SEM for all figures. 10

11
We