Introduction

Voltage-gated sodium channels (Na_V) are critical for all aspects of neuronal excitability, from action potential (AP) initiation to axonal propagation, transmitter release, and dendritic excitability^1-6. In mature neocortical pyramidal cells, electrogenesis and subsequent propagation of APs is systematically regulated with Na_V recruitment first initiated in the axon initial segment (AIS), with forward propagation along the axon and backpropagation into soma and dendrites^2,3,7. This coordinated process is supported by membrane expression of the Na_V isoforms Na_V1.6 and Na_V1.2. Current models, based on immunostaining of channels and empirical measurements of excitability, suggest that Na_V1.2 and Na_V1.6 are differentially expressed across neurites. In the AIS, Na_V1.6 predominates, with highest membrane density in the AP initiation region of the AIS most distal to the soma. Na_V1.2, by contrast, has higher relative membrane density in the proximal AIS^2,8-12. Along the axon, Na_V1.6 appears enriched at nodes of Ranvier and terminals^13,14. Somatodendritic densities are far lower, and current models suggest that the somatic and perisomatic membrane expresses equal levels of Na_V1.6 and Na_V1.2, whereas dendritic regions more distal to the soma are enriched exclusively with Na_V1.2^4-6,15. Thus, each channel isoform likely has a unique role in AP initiation and propagation simply based on their differential distribution in neuronal compartments.

To date, efforts to understand the unique contributions of different Na_V isoforms has relied largely on experiments in which channel genetic expression has been manipulated, either through constitutive or conditional knockout approaches. While these approaches have strong merit, interpretation is complicated by compensatory changes in other Na_V isoforms or other ion channel classes^4,5,16,17. Pharmacological approaches, which are both acute and reversible, would be preferred, but identifying highly selective compounds that target particular Na_V isoforms is difficult, as ScnXa gene family isoforms have high degrees of amino acid sequence homology. Some compounds exhibit high potency, but differences in the half-maximal inhibitory concentration (IC₅₀) for individual isoforms can be limited^18-21. This is especially true for separation of Na_V1.2 from Na_V1.6, as these two channels have especially high sequence similarity^22,23.

Aryl sulfonamide compounds (ASCs) constitute a unique class of Na_V inhibitors that potently bind and stabilize Na_V isoforms in the inactivated state^24-28. The binding pocket is shielded from ASCs when channels are closed due to the positioning of the S4 voltage sensing domain. In spiking neurons, this essentially imparts use-dependence to ASC block²⁴, as the vast majority of channels enter the inactivated state after a brief period of activation. ASCs exhibit high potency for a subset of Na_V isoforms: Na_V1.2, Na_V1.6 and Na_V1.7^24-27. These isoforms share sequence homology at the binding pocket, each containing a tyrosine-tryptophan (YW) motif that helps stabilize ASC binding. By contrast, Na_V1.1 and Na_V1.3 harbor a serine-arginine (SR) sequence at the same site. Mutagenesis of Na_V1.7 from YW to SR results in a 145-fold decrease in ASC binding without affecting channel biophysical properties²⁴. Here, we engaged a similar strategy for Na_V1.2 and Na_V1.6, using knock-in mice with YW->SR substitutions in either or both channels. This allows for selective, potent ASC-mediated block of YW-containing channels while preserving the function of SR-containing channels. With these tools, we dissected the individual roles of Na_V1.2 and Na_V1.6 in neocortical pyramidal cell excitability. We found that Na_V1.2 and Na_V1.6 have unique—and at times conflicting—effects on overall AP excitability. Moreover, results suggest that ASCs can function as ‘on-demand’ pharmacological inhibitors that help normalize cellular excitability in seizure-like conditions. Together, this work highlights the importance of distinct Na_V isoforms localized to specific cellular domains and their respective contribution to AP properties to help facilitate activity across pyramidal neurons.

Results

Aryl sulfonamides selectively bind and inhibit Na_V isoforms containing the YW motif

ASCs exhibit high affinity for an extracellular region within voltage sensing domain IV (VSD IV) of select Na_V channels with a conserved tyrosine-tryptophan (YW) motif²⁴ (Fig. 1A, B). This YW motif found on Na_V1.2, Na_V1.6 and Na_V1.7 allows for ASC stabilization following channel inactivation (Fig. 1B). By contrast, an SR motif present in Na_V1.1 and Na_V1.3 limits binding appreciably²⁴. Thus, we hypothesized that converting either Na_V1.2 or Na_V1.6 channels from those that contain the YW motif to those that contain an SR motif would alter ASC binding to those channels significantly.

Figure 1:

To test for the effects of this motif substitution, we first examined currents generated by wild type and YW->SR mutated channels expressed in immortalized cell lines (Fig. 1C). HEK cells were primarily used for most experiments. But due to known low transfection efficiency of Na_V1.6 in HEK cells, a subset of experiments were performed using an ND7/23 cell line engineered to lack most native Na_V conductance (ND7/LoNa_V)³⁵. Both Na_V1.2 and Na_V1.6 wildtype channels expressed in cell lines were inhibited markedly by 1 μM GNE-4076 (Fig. 1C). By contrast, SR knock-in mutants continued to flux sodium in the presence of 1 μM GNE-4076. Dose response curves revealed an IC₅₀ of 5.1 nM for wildtype Na_V1.2 that increases 365-fold to 1861 nM for mutant SR channels (Fig. 1D). For Na_V1.6 channels, GNE-4076 IC₅₀ increased 475-fold from 184 nM to 87 μM with the SR mutant (Fig. 1D). To assess whether SR mutations affect channel gating properties in the absence of GNE-4076 binding, steady state activation and inactivation curves were assessed for wildtype or mutant channels (Fig. 1E). Consistent with prior work studying a similar mutation in Na_V1.7²⁴, the YW to SR mutation had no effect on voltage-dependent activation or inactivation for either Na_V1.2 or Na_V1.6 (Table 2).

While GNE-4076 binds potently to either Na_V1.2 (Fig. 1D, F) or Na_V1.7 expressed in HEK cells²⁸, its affinity was lower for Na_V1.6 expressed in ND7/LoNa_V cells. To test if this was due to reductions in affinity imposed by the ND7/23 cell line, we examined current in the few HEK cells in which Na_V1.6 current could be measured (Fig. 1F). While peak currents were small (400 pA vs 3.2 nA for Na_V1.2 transfected using identical protocols), we found that 30 nM GNE-4076 exhibited strong block of Na_V1.6 (Fig. 1F), but due to limitations in expression efficiency in HEK cells, we were unable to collect full dose-response curves to determine IC₅₀.

Results in heterologous expression systems indicate that YW->SR substitution reduces GNE-4076 binding affinity markedly. Given these results, we constructed transgenic mice containing the YW->SR mutation. Two distinct lines, termed Scn8a^SR/SR and Scn2a^SR/SR, were generated by mutating both alleles of each sodium channel gene (Fig. 1G). SR/SR mutations were validated by PCR (see Methods) and mice were bred to be homozygous for the mutations. Comparisons were then made between wild type mice, mice with YW->SR mutations into Na_V1.2 or Na_V1.6 alone (2a^SR/SR or 8a^SR/SR, respectively), or YW->SR mutations in both Na_V1.6 and Na_V1.2 (dual 8a/2a^SR/SR). In principle, one could determine a concentration of GNE-4076 in which native channels are blocked potently, sparing SR variant channels (Fig. 1H). Thus, these mice should enable selective, acute manipulations of Na_V1.2 and Na_V1.6-dependent aspects of neuronal excitability using a chemical genetics approach.

Given the disparate binding results in heterologous cell lines, we first tested the effects of a dose that should potently block WT channels but not affect SR variant channels to verify that native channels behave similarly to those expressed in HEK cells. Neocortical pyramidal cell somata are thought to express similar levels of Na_V1.2 and Na_V1.6 on their membranes^4,5,8. Thus, excised patches from this region can provide insight into block of both isoforms. Acute coronal slices containing mPFC were prepared from WT or 8a/2a^SR/SR mice and nucleated patches were excised from layer 5 pyramidal cells. After establishing a stable baseline of Na_V-mediated current evoked from a holding voltage of -80 mV, a value comparable to the resting membrane potential studied in later current-clamp experiments, cells were pulsed to -12 mV for 10 sec, then returned to -80 mV (Fig. S1A). In untreated WT conditions, currents recovered to near-baseline levels within 5 sec (Fig. S1B). A similar quick recovery was observed in cells from 8a/2a_SR/SR mice in the presence of 200 nM GNE-4076. By contrast, WT neurons exposed to the same dosing were inhibited markedly, with no appreciable recovery within 25 sec (Fig. S1A, B). Taken together, these results show that the YW to SR mutation reduces ASC binding and inhibition of Na_V chimeras both in immortalized cell lines and pyramidal neurons.

Activity-dependent effects of ASCs on Na_V channels and subsequent inhibition alters AP properties and firing rate

Non-selective pharmacological inhibition of Na_V channels in neurons greatly reduces cellular excitability and AP firing^44,45. ASCs are unique in that they stabilize inactivated states, and this stabilization is voltage dependent^24-27. In spiking neurons, ASCs would therefore be predicted to exhibit use-dependence, progressively blocking channels in proportion to a neuron’s activity rate.

To first test this concept, we generated prolonged periods of activity in current-clamp by injecting 300 pA into the somatic pipette for 10 sec (Fig. 2A). GNE-4076–induced changes to AP waveform were visualized with phase plane plots, which plot membrane voltage vs. membrane voltage velocity (Fig. 2B and S3A). Within phase plots, spike threshold appears as sudden deviation from rest and is defined as the voltage at which voltage velocity (dV/dt) first exceeds 15 V/s (Fig. S3A). Following spike threshold, a neuron’s voltage transits through two components of Na_V recruitment, first in the AIS and then in the soma^7,46. These result in characteristic “humps” in the depolarizing aspect of the phase plot^2,11. Thus, these different components of the phase plot can aid one’s understanding of effects on specific components of Na_V-mediated excitability.

Figure 2:

Drug-naïve WT neurons responded to somatic current injection with repetitive spiking at a rate of 14.67 ± 0.47 Hz (n = 12). Neurons fired at a steady state after the first second, with stable AP threshold and instantaneous AP frequency (Fig. 2C). Spiking characteristics were identical in 8a/2a^SR/SR cells treated with GNE-4076, indicating that, at 200 nM, neurons are insensitive to GNE-4076 when the YW->SR motif substitution is present (Fig. 2C, D). However, in WT neurons treated with GNE-4076, threshold continued to depolarize after the first second, with corresponding decrements in instantaneous frequency and peak AP dV/ dt throughout the stimulus (Fig. 2C, D). This demonstrates that GNE-4076 suppresses firing of neurons with a baseline firing rate of ∼15 Hz (Table 3).

Differential roles for Na_V1.6 and Na_V1.2 in AP initiation and somatic excitability

Na_V1.2 and Na_V1.6 are differentially distributed in neuronal arbors and are thought to contribute to different aspects of AP initiation and propagation. In mature neocortical pyramidal cells, Na_V1.6 is enriched in the distal AIS, the site where APs initiate^4,8,16,47. Following initiation, APs forward propagate along the axon via Na_V1.6-enriched nodes of Ranvier, but also backpropagate into the soma through a region enriched with a mix of Na_V1.2 and Na_V1.6⁸. Several studies have sought to identify specific roles for each isoform by conditional deletion of either isoform^4,16. Unfortunately, in such conditions, the residual isoform compensates for loss to some degree, making interpretation of individual isoform roles difficult. We therefore leveraged ASC-based block to study acute, differential inhibition of Na_V isoforms to better understand the individual roles of Na_V1.2 and Na_V1.6 in AP excitability.

To provide a priori predictions of potential ASC-based effects on excitability, we constructed a compartmental model in which each channel’s density could be modulated. Channels were distributed based on predictions from empirical anatomical and physiological studies^4,6,8,16,48. Na_V1.6 was enriched in the distal AIS, Na_V1.2 was enriched in the proximal AIS, both channels were expressed at equal levels in the soma and the first 20 microns of dendrite closest to the soma, and Na_V1.2 was expressed exclusively in all other dendrites (Fig. S2A). Within this model, each channel’s density was modulated in 10% increments, from 100 to 0% (Fig. S2B).

In WT conditions (100% density of both channels), models generated APs with a threshold and AP kinetics comparable to empirical observations (Fig. S2B, C). Within this model, reducing Na_V1.2 or Na_V1.6 density had clear, dissociable effects. Progressive reduction of Na_V1.6 produced a progressive depolarization of AP threshold and a corresponding decrement in dV/dt throughout the entire rising phase of the AP. Na_V1.2 reduction, by contrast, had no effect on AP threshold or the initial speed of the AP initiated in the distal AIS. Instead, components of the AP related to backpropagation and recruitment of somatic Na_V channels were impaired only, with a decrement in peak dV/dt and peak AP membrane potential.

With these predictions in hand, we then repeated empirical experiments described above—where APs were generated with somatic depolarization over 10 sec—in mice where either Na_V1.6 or Na_V1.2 was mutated to be insensitive to 200 nM GNE-4076 binding (Fig. 3A, B). For clarity and simplicity, we will describe studies based on which channel is inhibited rather than which channel was rendered insensitive to GNE-4076. For example, a study in the Scn2a^SR/SR/Scn8a^+/+ animal is a case where Na_V1.6 can be blocked.

Figure 3:

In cases where Na_V1.6 alone could be inhibited, AP threshold depolarized in response to GNE-4076 as much as in WT cases (Fig. 3C, D). Furthermore, AP frequency and peak somatic AP dV/dt was reduced to levels observed in WT cells (Fig. 3C, D). In cases where Na_V1.2 alone could be inhibited, changes in AP threshold and instantaneous frequency were no different than 8a/2a^SR/SR conditions. Indeed, the only change in AP properties was a decrease in peak AP dV/dt (Fig. 3D). This suggests that AP threshold and instantaneous frequency are more sensitive to Na_V1.6 antagonism, whereas peak somatic dV/dt can be affected by antagonism of either isoform.

While these experiments provide clues to the differential roles of Na_V1.2 and Na_V1.6, it does not appear that one can block individual channels fully with APs alone at these drug concentrations. Therefore, we developed a hybrid current- and voltage-clamp experiment to study effects of more complete isoform blockade (Fig. 4A). In this protocol, neurons were held to -80 mV in current clamp with constant bias current (if necessary) and baseline APs were elicited with brief somatic current injection (300 ms, amplitude adjusted to evoke ∼4-5 APs). We then promoted Na_V activation and inactivation by voltage-clamping neurons to -12 mV for 30 sec (Fig. 4A). Following voltage-clamp, neurons were returned to current-clamp with the same bias current as used in baseline conditions. We then evoked APs, first at an interstimulus interval of 2 sec, then at successively increasing intervals of 5 to 60 sec, allowing for somewhat consistent sampling on a log-base timescale that aligns well with both channel recovery from inactivation and recovery from GNE-4076 block (Fig. 4D, F).

Figure 4:

In untreated WT cells, AP threshold following voltage-clamp recovered to baseline values immediately (Fig. 4C, D). By contrast, peak somatic dV/ dt recovered more slowly (Fig. 4C, F), likely reflecting recovery from slow inactivation of channels in the perisomatic region^49-51. Peak dV/dt recovered to within 87.12 ± 0.86% by 12-20 sec of voltage-clamp offset (Fig. 4F, G). Subsequent recovery of the residual peak dV/dt was slower, taking another 15-30 sec to recover the next 5% of peak dV/dt (Fig. 4F). Identical results were observed in 8a/2a^SR/SR cells treated with 200 nM GNE-4076, suggesting that any effect observed in single knock-in recordings will be due to block of GNE-4076 sensitive channel (Fig. 4B-G). We therefore focused on the 12-20 sec after voltage-clamp offset for subsequent analysis, as it is a period in which most channel-intrinsic recovery has occurred, but also a period in which we would still expect significant block from GNE-4076.

When individual channel isoforms were blocked more completely with voltage steps to -12 mV, dramatic changes in AP threshold and peak dV/dt were observed. When Na_V1.6 was blocked, threshold depolarized by 5.815 ± 0.7346 mV (Fig. 4E). When Na_V1.2 was blocked instead, threshold was unaffected (Δ Vm: -0.065 ± 0.155). Thus, AP threshold and AP initiation is mediated exclusively by Na_V1.6. Peak somatic dV/dt was affected by block of either channel isoform, with Na_V1.2 having a larger relative contribution (ΔdV/dt: -200.7 ± 6.861 and -139.2 ± 8.075 V/s, respectively; Fig. 4G). This seems unlikely to be due to differential isoform affinity for 200 nM GNE-4076, as we observed complete block of Na_V1.6 with much lower concentrations in HEK cells (Fig. 1). Instead, it may reflect differential roles of Na_V1.2 and Na_V1.6 in supporting backpropagation of APs into the and recruitment of somatic Na_V channels.

To examine these relative contributions further, we analyzed the rising phase of the AP more closely (Fig. S3A). As described above, the first component of the rising phase of the AP reflects recruitment of Na_V channels localized to the AIS (Fig. S3A, middle). Based on changes in voltage acceleration (second derivative) within this period, the AIS component can be further divided into initiation and AIS backpropagation components^8,37 (Fig. S3A, right). These periods were divided based on a timepoint during the AIS phase where voltage acceleration first peaked (AIS inflection point) or afterwards decreased, creating a trough in an acceleration vs. time graph (AIS max; Fig. S3B). The dV/dt at the AIS inflection point was affected only by Na_V1.6 block (Fig. S3C), whereas the dV/dt at the trough was affected by block of either isoform (Fig. S3E). This reinforces the model where APs are initiated via the Na_V1.6-enriched distal AIS, and where the depolarization produced by these distally localized channels recruits Na_V1.2 in the more proximal AIS.

Acute block of Na_V1.2 increases pyramidal cell AP output

Previously, we showed that conditional knockout of Scn2a increased AP output⁴. We hypothesized that this was due to the lack of Na_V1.2 in dendrites. When they are absent from this compartment, the dendrite does not depolarize as effectively, leading to a corresponding decrease in the recruitment of dendrite-localized potassium channels (K_V). Consequently, neurons repolarize less between APs, making it easier to evoke the next AP in the Na_V1.6-enriched AIS.

Though this hypothesis was supported by compartmental modeling demonstrating that acute block of Na_V1.2 mirrored empirical observations, we could not eliminate the possibility that some form of cellular compensation occurred in the weeks between conditional knockout induction and acute slice experiments⁵². Therefore, we leveraged GNE-4076 to test the effects of acute Na_V1.2 block (Fig. 5A), eliminating the possibility of genetic or post-translational compensation. We further compared this manipulation to block of Na_V1.6 or both isoforms by examining overall AP output from protocols described in Fig. 4A.

Figure 5:

Remarkably, acute Na_V1.2 block mirrored conditional knockout, and was the only condition in which AP output increased (Fig. 5B) and was associated with a depolarization in afterhyperpolarization voltage (Fig. 5C). In contrast, Na_V1.6 block decreased AP output, as did block of both channels (Fig. 5B). This indicates that blocking Na_V1.2 alone can increase AP output, independent of compensatory changes to other channels that may occur with genetic manipulations.

Leveraging use-dependent, isoform-selective Na_V pharmacology as anticonvulsants

Epilepsy is often associated with aberrant, excessive excitability in neocortical networks, whether from direct hyperexcitability in pyramidal cells^53,54 or disinhibition of pyramidal cells via alterations in inhibitory networks^37,55. Several use-dependent Na antagonists with structures similar to GNE-4076 are being developed, exhibiting differential selectivity for Na_V1.2 and Na_V1.6²⁵. Use-dependence may have advantages as antiepileptics. In theory, they would have minimal effect on neurotypical network activity unless such networks were hyperactive enough to promote drug binding. This may occur preferentially in seizure states. But given results above demonstrating that block of Na_V1.2 and Na_V1.6 can have different effects on overall AP output, it is critical to determine how block of either channel affects overall activity in seizure-like conditions.

To test this, we mimicked synaptic input by repeatedly injecting a 60 sec-long somatic current composed of Poisson-distributed EPSC and IPSC-like waveforms designed to elicit ∼10 Hz spiking in baseline conditions (Fig. 6A). Following this baseline, the PSC-like train was repeated, imposed upon a 400 pA standing current injection to increase spiking to ∼30 Hz (Fig. 3B), mimicking spike rates commonly observed during cortical seizure^56,57. After this seizure-like event, cells were returned to baseline voltages and recovery was assessed with 4 more repetitions of the 60 sec PSC stimulus (Fig. 6A-C).

Figure 6:

In the presence of 200 nM GNE-4076, baseline firing rates were stable in all conditions, independent of whether Na_V1.2, Na_V1.6, or both isoforms were blocked (Fig. 6C, E). By contrast, a slight depolarization in AP threshold was observed at baseline when both channels were sensitive to GNE-4076 (e.g., WT plus drug; Fig. 6C, D). Given that this dose can completely block both Na_V1.2 and Na_V1.6 in HEK cells, a small amount of drug block was expected at baseline. Seizure-like activity induction depolarized AP threshold markedly in all genotypes, and threshold depolarization persisted into the recovery phase for all neurons experiencing Na_V1.6 or dual inhibition (Fig. 6C, D).

Seizure-like activity increased spike rate across all genotypes at the onset of seizure-like activity induction (Fig. 6C, E). All cells exhibited some degree of AP accommodation during this stimulus, including drug-naïve controls and cases where both Na_V1.2 and Na_V1.6 were insensitive to GNE-4076 block. WT cells exposed to GNE-4076 accommodated markedly, returning to baseline firing rates at the end of the seizure-like stimulus (Fig. 6C, E). A less dramatic effect was noted when Na_V1.6 was blocked alone (Fig. 6E). By contrast, cells where Na_V1.2 could be blocked alone were not appreciably different than untreated controls (Fig. 6E). Together, these data suggest that use-dependent pharmacology that targets Na_V1.2 and Na_V1.6 may be most beneficial if designed for higher potency at Na_V1.6.

Discussion

Here, we combined genetic and pharmacological strategies to transiently, selectively and reversibly inhibit either Na_V1.6 or Na_V1.2 function using activity-dependent ASCs, which bind channels in the inactivated state. We show that acute blockade of either isoform has opposing effects on neuronal output: inhibition of Na_V1.6 decreases AP output, whereas Na_V1.2 increases AP output. Given this, we found that block of Na_V1.6 rather than Na_V1.2 was more effective at tempering spiking activity in cells driven to seizure-like levels of AP output. This suggests that pharmacology tuned to preferentially block Na_V1.6 over Na_V1.2 in a use-dependent manner may be useful as an anti-epileptic.

Functional implications of differentia compartmental Na_V isoform expression

Electrophysiological recordings of AP propagation delays between somatic and axonal recordings demonstrate that APs initiate ∼35 to 50 microns from the soma, in the distal AIS^2,3,7,8. This region is enriched with Na_V1.6^8,9,47,58,59. Consistent with this, we find that inhibition of Na_V1.6 alone alters AP initiation, with an increase in AP threshold and decrease in total AP number from block of this isoform. Following distal AIS AP initiation, APs sequentially propagate across different neuronal compartments to influence local excitability^2,3,7. Forward propagation was not examined here, but prior reports demonstrate that it is supported almost exclusively by Na_V1.6 in nodes of Ranvier and boutons^9,13,14. By contrast, backpropagation through the proximal AIS and soma recruits a mix of Na_V1.6 and Na_V1.2.

Prior work has suggested that Na_V1.2 channels localized to the proximal AIS are critical for backpropagation, with modeling suggesting that APs would fail to effectively backpropagate into the soma in the absence of proximal AIS Na_V1.2 channels⁸. While some efforts have been made to test this empirically with conditional Scn2a knockout, such experiments are imperfect, as Na_V1.6 redistributes in the AIS of cells that lack Na_V1.2⁴. Here, we were able to test the role of Na_V1.2 in backpropagation without compensatory effects, using acute pharmacological inhibition. Backpropagation failures were not observed, consistent with results from conditional Scn2a knockout⁴. However, we did observe a reduction in the speed at which the AIS component of the AP depolarized when Na_V1.2 was blocked selectively (Fig. S3B-C). This suggests that proximal AIS-localized Na_V1.2 do aid in boosting backpropagating APs as they transit from the distal AIS to the soma. Nevertheless, Na_V1.2 channels are still dispensable for recruitment of somatic Na_V channels, at least in mouse mPFC layer 5b pyramidal cells. Whether similar effects are observed in other cells, including those with larger somata that may be more difficult to depolarize in the absence of Na_V1.2, remains to be tested.

Na_V1.2 expression in pyramidal cell dendrites appears to have two roles in neuronal excitability. Intuitively, these channels provide local inward current that boosts dendritic excitability, leading to a recruitment of voltage-gated calcium channels^5,6. But this depolarization also appears to be critical for recruitment of dendritically localized voltage-gated potassium channels that contribute to AP repolarization and net membrane potential between spikes⁴. Indeed, conditional knockout of Na_V1.2 paradoxically increases AP output, and we suggested previously that this was due in large part to loss of interactions between dendritic Na_V channels and K_V channels. But given observed changes in Na_V1.6 function as well as potential for other compensatory changes in non Na_V ion channels, it was difficult to ascribe hyperexcitability purely to the loss of Na_V1.2 alone ⁵². Here, we showed that identical increases in excitability—associated with depolarization of membrane potential between APs—could be observed with acute Na_V1.2 block. This indicates that such effects can be due purely to the interplay between dendritic Na_V1.2 and K_V channels, and that modifications to potassium channel distribution or function are not necessary for such effects.

Similar to Na_V1.2 conditional knockout, where Na_V1.6 is upregulated, knockout of Na_V1.6 results in an increase in Na_V1.2 expression, at least in the AIS. In pyramidal cells from mice constitutively lacking Na_V1.6, Na_V1.2 occupies the entirety of the AIS rather than just the region proximal to the soma^10,16,47. Thus, it has not been possible to evaluate the role of Na_V1.6 in its normal distribution using genetic manipulations. Here, we find that acute Na_V1.6 block reduces AP output with a concomitant increase in AP threshold. This contrasted markedly with block of Na_V1.2, which had no effect on threshold. Of note, this distinction can be leveraged to assess the specificity of other Na_V-targeting pharmacology that has been suggested to have specificity at select isoforms, as any change in threshold indicates that drugs are interacting with Na_V1.6^10,21.

ASC pharmacology for suppression of neuronal hyperexcitability

Epilepsy can arise from genetic and non-genetic factors, often with unexplained etiology^60,61. Seizure onset is broadly classified as an electrical imbalance of cellular and network activity that favors hyperexcitability⁶², whether it be cell intrinsic, synaptic, or due to complex network effects^63,64. Thus, seizure suppression can be targeted at multiple levels. Nevertheless, proper dosing can be difficult, as one aims to provide drug concentrations that temper excess activity but limit side effects like sedation associated with elevated drug concentrations.

For genetically defined sodium channelopathies, attention has been directed to neuronal cell-types that express each isoform (e.g., SCN1A: Na_V1.1, SCN2A: Na_V1.2 and SCN8A: Na_V1.6)^65,66. Seizures resulting from SCN1A loss of function limit inhibitory neuron excitability, thereby disinhibiting excitatory pyramidal cells^37,55,67. SCN2A gain-of-function results in hyperexcitability in pyramidal cells, especially in early development when Na_V1.2 channels are the sole isoform expressed in the AIS^12,68. Later in development, SCN8A-encoded Na_V1.6 channels are expressed more ubiquitously in the AIS of most cell classes, supporting AP initiation in both excitatory and inhibitory cells^37,53,59,69.

With these cellular distributions and mechanisms of action, guidelines have emerged for treatment. For SCN1A loss-of-function, sodium channel blocking anti-epileptics are typically counter-indicated^70,71. This is because currently prescribed sodium channel blockers are nonselective, with little preference for Na_V1.1, 1.2 or 1.6⁷². Thus, further block of the remaining Na_V1.1 leads to even more disinhibition of excitatory cells before any beneficial effects of blocking Na_V channels in excitatory cells are realized. Similarly, nonspecific sodium channel blockers are counter-indicated for SCN2A loss-of-function seizures, as they tend to increase seizure severity. Instead, non-selective Na_V inhibitors are useful when channel gain-of-function affects excitatory cells, as is the case for both SCN2A and SCN8A gain-of-function cases.

ASCs may be useful within each of these domains, as their chemistry can be adjusted to bias binding to specific isoforms in an activity-dependent manner^24-27. Indeed, compounds similar to those used here that inhibit both Na_V1.2 and Na_V1.6 (but not other Na_V channels) are effective at suppressing chemoconvulsant-induced seizures in ex vivo models²⁵. This parallels our observations, where excess activity was dampened most effectively by dual block of Na_V1.2 and Na_V1.6 (Fig. 6C). This effect appears due in large part to block of Na_V1.6, since, ultimately, this is the isoform responsible for AP initiation. Consistent with this, Na_V1.6-preferring ASCs show promise in protecting from chemoconvulsant-induced seizures in vivo^26,27. Thus, it may be that preferential block of Na_V1.6 would be beneficial in multiple sodium channelopathy conditions, including those associated with channel loss, since network hyperexcitability is ultimately dictated by Na_V1.6 function in the AIS of excitatory neurons after the first months of life.

Beyond selectivity, there may also be advantages to ASC activity-dependent properties. The ASC binding pocket is hidden from drug in a channel’s closed state. This feature, combined with on- and off-rate kinetics, results in an accumulation of block primarily for highly-active neurons (Figs. 4, 6). Similar to on-demand, closed-loop electrical or optogenetic approaches for seizure intervention^73-75, ASCs are essentially biased to suppress prolonged, high-frequency activity, including the type of activity commonly observed during seizures.

Methods

Resources availability

Lead contact

Any additional information or enquires related to reagents or resources should be directed to the lead contact, Kevin J. Bender (kevin.bender@ucsf.edu).

Materials availability

The transfer of unique reagents generated for this study will be made available upon request. A Materials Transfer Agreement may be required.

Data and code availability

This study did not generate any unique datasets or code. Data reported here will be made available by lead contact upon reasonable request.

Experimental models and subjects details

Mouse strains

All animal procedures are in accordance with the Institutional Animal Care and Use Committee (IACUC) guidelines in accordance with the University of California, San Francisco (UCSF). The following mouse strains were used in this study: C57BL/6J, YW–>SR Na_V1.2 KI (Scn2a^SR/SR), YW®SR Na_V1.6 KI (Scn8a^SR/SR) and YW–>SR dual KI (8a/2a^SR/SR). All experimental procedures were performed on mice maintained in-house on a 12:12 hour light-dark cycle under standard conditions with ad libitum access to food and water. For genotyping, genomic DNA was isolated from tail clip biopsies for PCR. Both male and female mice aged postnatal day (P)18-59 were used across all genotypes. C57BL/6J mice were obtained from Jackson Laboratories and YW–>SR KI mice were developed by the Hackos Lab (Genentech).

Generation of SCN8A or SCN2A YW->SR KI mouse

As described previously²⁹, CRISPR/Cas9 technology^30,31 was used to generate a genetically modified mouse strain with either an Scn8a or Scn2a YW–>SR knock-in mutation. A single guide RNA (sgRNA) target and protospacer adjacent motifs (PAM) were identified for Scn8a ENSMUSG00000023033 or Scn2a ENSMUSG00000075318 genomic regions of interest using the CRISPR design tool (Benchling) that uses the algorithm described by Hsu et al.³² to provide ‘MIT’ specificity scores for each sgRNA, as well as the top 15 predicted off-target loci and corresponding MIT off-target scores. The same guide target and PAM were used for both genes. Guide target: 5’ CATTCTCTACTGGATTAATC 3’; PAM: TGG with an algorithm score of 42.3.

For the YW–>SR mutation on the Scn8a gene, predicted cut sites are between 100,933,463-100,933,464 genome coordinates. The following oligonucleotide donor sequence was used: 5’ ATGCTTATCTGCCTTAACATGGTGACCATGATGGTGGAGACAGACACA CAGAGCAAGCAGATGGAGAACATTCTCTCTCGGATTAATCTGGTCTTC GTCATCTTCTTCACCTGCGAGTGTGTGCTCAAAATGTTTGCCTTGAGA CACTACTATTTC 3’. The first point mutation of Y1553S (TAC®TCT) is located at 100,933,454-100,933,456 genome coordinates, and a second point mutation of W1554R (TGG®CGG) is located at 100,933,457-100,933,459 genome coordinates.

FortheYW–>SRmutationontheScn2agene,predictedcutsitesarebetween 166,900-166,901 genome coordinates. The following oligonucleotide donor sequence was used: 5’ GAAATAGTAGTGTCTCAAGGCAAACATTTTGAG CACACACTCGCAGGTGAAGAAGATGACGAAGACCAGGTTGATCCGA GAGAGAATGTTCTCCATCTGCTTGCTCTGTGTGTCTGTCTCCACCATC ATGGTCACCATGTTAAGGCAGATAAGCAT 3’. The first point mutation of Y1553S (TAC®TCT) is located at 101035573-101035575 genome coordinates, and a second point mutation of W1554R (TGG®CGG) is located at 101035576-101035578 genome coordinates. Additionally, two silent mutations were created in the beginning of the gRNA to prevent Cas9 from cutting the donor oligo. The first silent mutation (ATT®ATC) is located at 101035579-101035581 genome coordinates and the second silent mutation (AAT®AAC) 101035582-101035584 genome coordinates.

After homology-directed repair of Cas9-induced chromosome breaks with the oligonucleotide donor, the YW–>SR protein will be expressed. Once a sgRNA decision was finalized, the off-target list was used to identify the top 15, and next-generation sequencing (NGS) amplicon primers were designed for the on-target locus, and each of the off-targets synthetic guide RNA was obtained from Synthego. CAS9 protein was obtained from PROTEIN SOURCE and complexed with sgRNA before microinjection. Reagent concentrations for microinjection were as follows: 25 ng/µl Cas9 mRNA (Thermo Fisher; A29378) + 13 ng/µl sgRNA (Synthego), Oligonucleotide donor (50 ng/µl) (IDT).

After zygote microinjection and embryo transfer, genomic DNA was prepared from tail tip biopsies of potential G0 founders, and G0 animals were first analyzed by droplet digital PCR³³ (Bio-Rad). Primers were used to amplify the HDR event (ON Target) and the 15 most likely off-target sites. Only G0 mosaic founders positive for the intended mutation were screened by targeted amplicon NGS. Amplicons were submitted for NGS analysis.

Founders were selected for mating with wild-type C57BL/6N mice for germline transmission of the gene edited chromosome. Subsequent analysis of genomic DNA from G1 pups was used to confirm germline transmission of the targeted gene and the absence of off-target hits elsewhere in the genome.

To generate the dual 8a/2a^SR/SR line, Scn2a^SR/SR and Scn8a^SR/SR mice were crossed and analysis of genomic DNA was used to confirm germline transmission of the targeted gene.

Synthesis of GNE4076

GNE-4076 was synthesized as previously described in Roecker et al.²⁸, and is is Compound 5 in original manuscript. We use GNE-4076 throughout this manuscript.

Generation of SCN2A YW->SR and SCN8A YW->SR constructs

The double mutation Y1564S/W1565R was introduced into the adult splice isoform of recombinant human Na_V1.2 (NCBI accession number NM_021007; AddGene #162279) using site directed mutagenesis as previously described³⁴. The corresponding mutations (Y1555S/W15556R) were engineered in recombinant human Na_V1.6 (adult isoform) as previously described³⁵. Mutagenic primer sequences are presented in Table 1. All plasmids were nanopore sequenced (Primoridium Labs, Arcadia, CA) to confirm the variants and exclude unwanted mutations.

Electrophysiology in immortalized cell lines

HEK293 cells were transfected with WT or mutant Na_V1.2 using the Invitrogen Lipofectamine LTX kit, whereas Na_V1.6 plasmids were electroporated into the LoNav derivative of ND7/23 cells using Maxcyte technology as previously described³⁵.

Na_V channel currents were recorded from HEK293 or ND7/23 cells using whole-cell patch clamp using a Molecular Devices Axopatch 200B amplifier. The recording pipet intracellular solution contained (in mM): 120 CsF, 10 NaCl, 2 MgCl₂, 10 HEPES, adjusted to pH 7.2 with CsOH. The extracellular recording solution contained (in mM): 155 NaCl, 3 KCl, 1 MgCl₂, 1.5 CaCl₂, 10 HEPES, adjusted to pH 7.4 with NaOH. Currents were recorded at 20 kHz sampling frequency and filtered at 5 kHz. Series resistance compensation was applied at 80%. Solutions containing GNE-4076 were applied using a Fluicell Dynaflow perfusion system.

We characterized the activation and steady-state inactivation properties of both WT and mutant Na_V1.2 and Na_V1.6 channels. To measure activation, we used a holding voltage of -80 mV, a short 20 ms pre-pulse to -120 mV, and a 30 ms pulse to voltages ranging from -100 mV to -20 mV in steps of 5mV at a rate of once per 3 sec. P/4 leak subtraction was used to reduce leak currents. The peaks of the resulting Na_V currents were measured. Conductance was calculated using the equation G = I / (V – V_Na), normalized, and plotted as a function of voltage. To measure inactivation, we started at a holding voltage of -120 mV to bring the channels fully into the closed state (about 1 min). We then pulsed to 0 mV while reducing the holding voltage from -120 mV to +30 mV in steps of 5 mV at a rate of once per 5 sec. The peaks of the resulting Na_V currents were measured, normalized and plotted as a function of voltage. To quantify these biophysical properties, we fit the activation and inactivation curves to the following Boltzmann equations:

V_1/2 is the voltage where half-maximal peak conductance (or current) was observed, and sf is the slope factor.

Ex vivo electrophysiology

Mice aged P18-P59 were anesthetized with isoflurane prior to harvesting the brain. Dissected brains were immediately placed in cutting solution (4°C) containing 87 mM NaCl, 25 mM NaHCO₃, 25 mM glucose, 75 mM sucrose, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 0.5 mM CaCl₂ and 7 mM MgCl₂ that is bubbled with 5% CO₂ / 95% O₂. Coronal slices were prepared from the medial prefrontal cortex (PFC) at a thickness of 250 microns and placed in a holding chamber warmed to 33C for 30 minutes. Slices were then allowed to recover at room temperature until recording. Recordings were performed at 31-33°C in a solution containing 125 mM NaCl, 2.5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 25 mM NaHCO₃, 1.25 mM NaH₂PO₄ and 25 mM glucose that is bubbled with 5% CO₂ / 95% O₂. Osmolarity of the recording solution was adjusted to approximately 309 mOsm.

Neurons were identified using differential interference contrast (DIC) optics for conventional visually guided whole-cell recordings. For current clamp experiments, pipettes were pulled from Schott 8250 glass, with a tip resistance of 3-4 M, and filled with a K-gluconate based internal solution containing 113 mM K-Gluconate, 9 mM HEPES, 4.5 mM MgCl₂, 0.1 mM EGTA, 14 mM Tris₂-phosphocreatine, 4 mM Na₂-ATP, 0.3 mM Tris-GTP; 290 mOsm; pH: 7.2-7.25. All data were corrected for measured junction potentials of 12 mV.

All electrophysiology data were acquired through custom protocols generated in IgorPro (Wavemetrics) via Multiclamp 700A or 700B amplifiers (Molecular Devices). Action potential (AP) waveform measurements were acquired at 50 kHz and low-pass Bessel filtered at 20 kHz for all experiments except for data presented in Fig. 6 (acquired at 10 kHz and filtered at 3 kHz). Pipette capacitance was compensated in all current-clamp recordings to 50% of the fast capacitance measured after membrane seals were established in voltage-clamp. The bridge was balanced, and series resistance was kept < 18 MΩ in all recording. Any cells with input resistance changes exceeding ± 15% were omitted from dataset. A quartz electrode holder (Sutter Instrument) was used to collect all recording to minimize physical drift of recording electrodes.

All recordings were made from medial prefrontal cortex (PFC) layer 5b, thick-tufted pyramidal tract (PT) neurons. Pyramidal neuron identity was confirmed by assessing membrane responses to hyperpolarizing current (−400 pA, 120 ms), with PT neurons defined as those that exhibit membrane depolarization overshoot that peaks withing 90 ms of current step offset³⁶. AP threshold, AIS dV/dt and peak dV/dt measurements were determined from either the first AP or all APs within a spike train evoked by a stimulus. AP threshold was defined as the membrane potential (V_m) when the AP speed (dV/dt) exceeds 15 V/s. Somatic peak (V/s) was defined as the max value in the first derivative (dV/dt) of an AP. AIS max (V/s) was defined as the trough or saddle point in the second derivative (d²V/dt²) that occurs during the depolarizing phase of an AP before the somatic peak³⁷. AIS inflection point was defined as the peak of the second derivative (d²V/ dt²) that occurs prior to the AIS max or somatic peak. Frequency was defined as the spike number in a train elicited per second (spikes/sec or Hz). Afterhyperpolarization (AHP) was defined as the minimum voltage between 2 spikes in a train.

Spike trains were evoked by injecting current (250-350 pA, 10 sec or 300 ms duration) in the presence or absence of 200 nM GNE-4076. In recovery experiments, neurons were held at -12 mV for 30 sec (voltage-clamp) to depolarize cells, activate Na_V channels and expose the ASC binding pocket to maximize GNE-4076 onboarding and Na_V inhibition. The interstimulus interval given during the recovery period was as follows: 2 sec, ∼20-30 sec duration; 5 sec, ∼30-45 sec duration; 15 sec, ∼60-90 sec duration; 30 sec, ∼120-180 sec duration; and 60 sec, ∼60-240 sec duration. In experiments mimicking cellular activity, a post-synaptic potential train was randomly generated using a Poisson probability distribution function for 60 sec with a frequency of 50 Hz and amplitude standard deviation of 200 pA.

For nucleated patch voltage-clamp experiments, recordings were made at 31-33°C in a solution containing 125 mM NaCl, 2.5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 25 mM NaHCO₃, 1.25 mM NaH₂PO₄, 15 mM glucose, 4 mM TEA, 1 mM 4-AP, and 10 µM nifedipine, with or without 200 nM GNE-4076, (bubbled with 5% CO₂ / 95% O₂). A Cs-methanesulfonate based recording solution was used: 110 mM CsMeSO₃, 40 mM HEPES, 1 mM KCl, 4 mM NaCl, 4 mM Mg-ATP, 10 mM Na-phosphocreatine, 0.4 mM Na₂-GTP, and 0.1 mM EGTA; 290 mOsm; pH: 7.2-7.25. Whole-cell recordings were established before withdrawing pipettes from the slice, pulling a region of the somatic membrane with the pipette. Neurons were held to -80 mV and stepped to -12 mV every 2 sec, 5-10 times, then neurons were held to -12 mV for 10 sec to onboard GNE-4076, then returned to -80 mV and probed for Na_V recovery with steps to -12 mV every 2 sec. Leak currents were subtracted with a P/8 protocol using steps from -80 to -90 mV.

Compartmental modeling

A compartmental model was constructed within the NEURON environment to simulate a layer 5 pyramidal neuron as described before^4,38,39. A multi-compartmental model, originally developed by the Blue Brain Project, was implemented to reflect the morphology detailed by Ramaswamy and Markram⁴⁰ and electrophysiological features were adjusted to reflect empirically obtained data^4,38. We modified the model by replacing the aggregated sodium conductances (NaT and NaP) with distinct Na_V1.2 and Na_V1.6 channels to match empirically observed distributions⁸. Na_V1.2 and Na_V1.6 channels were distributed throughout the cell with equal levels in the soma and 20 µm of the proximal dendrites. Na_V1.2 was solely expressed in dendrites more distal to the soma, inferred based on AP-evoked sodium imaging observations in Scn2a^+/-conditions⁶.

The two axonal compartments were subdivided into an axon initial segment and distal axon. Within the AIS, Na_V1.2 and Na_V1.6 were distributed with increased Na_V1.2 in the proximal AIS and increased Na_V1.6 in the distal AIS to recapitulate the channel distribution as previously observed empirically⁸. _V1.2 was not included in the distal AIS or axon where only Na_V1.6 is present, including an enriched region to model a node of Ranvier. To simulate blocking of Na_V1.2 and Na_V1.6, each channel’s density was globally reduced in 10% increments from 100% to 0%. Both Na_V1.2 and Na_V1.6 channels were represented using the Hodgkin-Huxley formalism⁴¹. Parameter optimization for both channels was conducted using an evolutionary algorithm from BluePyOpt⁴² and adapted for use with the computational resources at the National Energy Research Computing Center similar to Ladd et al.⁴³.

Quantification and statistical anaylysis

Data are reported as absolute values or the absolute difference from baseline (delta, Δ). For Δ values, data were normalized either to the initial 500 ms of the stimulus (Fig. 2-3) or baseline spiking (Fig. 4-5). Time course graphs are represented as a mean ± standard error. Summary graphs are represented with boxplots showing the median, quartiles and 90% tails or with violin plots. All summary graphs overlay individual datapoints and represent recordings from single cells (reported n) for all electrophysiology experiments. Data were obtained from 5-9 animals (both sexes) per condition, which are standard group sample sizes used in the field. Analysis was performed blind to genotype ± drug. Statistical analysis was performed using Prism 10 (Graphpad Software). Quantified mean ± standard error and statistical test used is noted in figure legends. Significance was set at an alpha value of 0.05 and ‘ns’ indicates no significance.

Acknowledgements

We thank Drs. JP Johnson, Natali Minassian, Fiona Scott, and members of the Bender Lab for extensive discussions related this work. This work was supported by NIH grants K00 MH134674 (JDG), MH125978 and MH126960 (KJB), by the Hartwell foundation through an Individual Biomedical Research Award (RBS) and by FamiliesSCN2A through an Action Potential award (RBS).

Additional information

Author Contributions

DHH and KJB conceived the project. JDG, CW, EB, TF, JMD, TVA, ALG, RBS, DDH and KJB designed and performed experiments. JDG, CW, EB, TF, JMD, TVA, ALG, RBS, DDH and KJB analyzed the data. JDG and KJB wrote the manuscript. EB, TF, ALG, RBS, and DDH reviewed and edited the manuscript.