Initiation of migraine-related cortical spreading depolarization by hyperactivity of GABAergic neurons and NaV1.1 channels

Spreading depolarizations (SDs) are involved in migraine, epilepsy, stroke, traumatic brain injury, and subarachnoid hemorrhage. However, the cellular origin and specific differential mechanisms are not clear. Increased glutamatergic activity is thought to be the key factor for generating cortical spreading depression (CSD), a pathological mechanism of migraine. Here, we show that acute pharmacological activation of NaV1.1 (the main Na+ channel of interneurons) or optogenetic-induced hyperactivity of GABAergic interneurons is sufficient to ignite CSD in the neocortex by spiking-generated extracellular K+ build-up. Neither GABAergic nor glutamatergic synaptic transmission were required for CSD initiation. CSD was not generated in other brain areas, suggesting that this is a neocortex-specific mechanism of CSD initiation. Gain-of-function mutations of NaV1.1 (SCN1A) cause familial hemiplegic migraine type-3 (FHM3), a subtype of migraine with aura, of which CSD is the neurophysiological correlate. Our results provide the mechanism linking NaV1.1 gain of function to CSD generation in FHM3. Thus, we reveal the key role of hyperactivity of GABAergic interneurons in a mechanism of CSD initiation, which is relevant as a pathological mechanism of Nav1.1 FHM3 mutations, and possibly also for other types of migraine and diseases in which SDs are involved.


Graphical abstract-Legend
Focal GABAergic neurons' hyperactivity (long lasting firing at moderate frequency) (1) caused by pathologic dysfunctions (e.g. NaV1.1 gain-of-function) can lead to spiking-induced accumulation of extracellular K + (2), which drives the network to hyperexcitability (3) and eventually induces depolarization block and CSD initiation, independently from synaptic transmission. This mechanism can be associated not only to gain-of-function NaV1.1 mutations, but possibly also to other dysfunctions that induce GABAergic neurons' hyperactivity. As in all episodic disorders, homeostatic mechanisms can control dysfunctions in the period between the attacks, but triggering factors (e.g. hormonal/neuromodulatory changes or increase of incoming neuronal signals from the periphery) may focally affect neuronal excitability and activities of cortical networks, leading to long lasting GABAergic neurons' hyperexcitability and CSD induction.

Supplementary methods
Brain slices: preparation, electrophysiological recordings and imaging.
Brain slices were prepared as previously described (1, 2). Briefly, mice were killed by decapitation under isoflurane anesthesia, the brain was quickly removed and placed in ice-cold artificial cerebrospinal fluid (ACSF), which contained (in mM): 125 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2, 1.25 NaH2PO4, 25 NaHCO3 and 25 glucose, saturated with 95 % O2 -5 % CO2. Acute coronal slices 400μm thick (300μm for spatial illumination and patch-clamp experiments) were prepared with a vibratome (HM650V, MicroM, Germany) in ice-cold ACSF. Slices were then stored in a submerged chamber with ACSF at 34°C at least one hour before the beginning of the recordings. Selective brain expression of the ChR2(H134R)/tdTomato transgene and thus absence of germline recombination of the floxed allele was systematically confirmed by visual inspection of slices and of other tissues of the mouse using the NightSea Flashlight (USA). Inductions of CSD started > 10min after placing an individual slice in the recording chamber (RC-26GLP; Warner Instruments, USA), which was perfused with "recording ACSF" (rACSF) at 34°C, containing (in mM): 125 NaCl, 3.5 KCl, 1 CaCl2, 0.5 MgCl2, 1.25 NaH2PO4, 25 NaHCO3 and 25 glucose, saturated with 95% O2 -5% CO2. This rACSF has been already used for studying CSD (3), and our control experiments did not show statistically significant differences in CSD properties compared to standard ACSF ( Supplementary Fig.7) Slices were visualized using infra-red differential interference contrast (DIC) microscopy (Nikon Eclipse FN1, Japan) equipped with a CCD camera (CoolSnap ES2, Photometrics, USA), filter cubes for visualization of fluorescent proteins (Semrock, USA) and a filter for optogenetic 475nm blue light illumination (FF02-475/50; Semrock, USA). Acquisitions of large fields of view were obtained with a 0.35X camera adapter lens. Electrophysiological signals were recorded with a Multiclamp 700B amplifier (CV-7B headstage), a Digidata 1440A acquisition board and pClamp 10.3 software (Molecular Devices, USA). DC extracellular field potential recordings were performed using borosilicate glass micropipettes (∼0.5MΩ) filled with rACSF. Whole-cell patchclamp recordings were performed using borosilicate glass pipettes of 3-5MΩ resistance containing (in mM): 120 K-gluconate, 15 KCl, 2 MgCl2, 0.2 EGTA, 10 HEPES, 20 P-Creatine, Na2-0.2 GTP, 2 Na2-ATP, 0.1 leupeptin, adjusted to pH 7.25 with KOH. Whole-cell access resistance (10-25MΩ) was monitored and cells showing instable access resistance (> 20%) were discarded. Recordings were started 5min after obtaining the whole-cell configuration. Juxtacellular-loose patch recordings were performed with the same pipettes used for whole cell experiments, but filled with rACSF, and action potentials were recorded in the I=0 mode.
For pyramidal neurons, we have recorded randomly from these neurons in neocortical layer II-III.
NaV1.1 is expressed also in pyramidal neurons, even if at much lower level than in most GABAergic neurons, but there is not yet consensus about which type of excitatory neuron expresses it. In fact, although some works have reported expression of NaV1.1 in discrete subpopulations of neocortical pyramidal neurons and no expression in hippocampal pyramidal neurons (4,5), numerous others have instead found low level of NaV1.1 expression in most pyramidal neurons of both hippocampus and neocortex, although loss of function of NaV1.1 does not in general modify firing properties of these neurons (6)(7)(8)(9)(10)(11). Moreover, in a BAC transgenic mouse expressing exogenous NaV1.1 (12), which is a model similar to that used by (5), Na + currents generated by the exogenous NaV1.1 were observed in all the neurons (both GABAergic and pyramidal) randomly selected upon dissociation from the neocortex (12), although the amount of current recorded in the different neurons was dependent on the specific transgenic line. Thus the lack of effect of Hm1a in pyramidal neurons in our whole-cell recordings of Figure 2 shouldn't be caused by the random selection of pyramidal neurons not expressing NaV1.1. Additionally, the neocortical specificity of CSD induction that we have observed cannot anyway be explained by the selective expression of NaV1.1 in cortical excitatory neurons, because we obtained neocortical-specific CSD induction also with optogenetic stimulations, which are selective for GABAergic neurons.
Ion sensitive electrodes were built with glass pipettes (tip diameter ∼ 2μm) as previously described (13). Briefly, they were first cleaned with absolute ethanol, then pre-treated with dimethylchlorosilane vapors (Sigma-Aldrich, USA), dried at 100°C for 2h, and finally the tip was backfilled with the K + ionophore I-cocktail B (Fluka, Sigma-Aldrich, USA) using a 28G microfil (WPI, USA). The calibration was done for each microelectrode using solutions of different K + concentrations (2.5, 3.5, 35, and 90mM: KCl added to the ACSF), and electrodes were selected according to the slope of their response (a minimum of 50mV/decade increase in K + concentration). To minimize the leak current of the headstage in ion sensitive recordings, we used Rf = 5GΩ and compensated the residual leak current: we zeroed the pipette offset observed with a 10MΩ load, we then switched to a 10GΩ load and we re-zeroed the offset observed in this condition (which is mainly due to the residual leak current) with the current injection circuitry of the amplifier (holding current); the estimated leak current (equal and opposite to the holding current applied) was 0.73pA at 25°C and stable over time. For accurate evaluation of [K + ]out dynamics, the field potential signal was subtracted from the K + -sensitive recording (the ion sensitive and the extracellular potential pipettes were placed at <100µm of distance from each other).
Electrophysiological signals were filtered at 10kHz and sampled at 25kHz. Multi-unit activities (MUA) were obtained band-pass filtering off-line the LFP trace at 300-500Hz with pClamp (single pole RC filters). Intrinsic optical signal (IOS: near-infrared light transmittance) (14) was monitored acquiring images with the CoolSnap ES2 CCD camera controlled with micromanager (15) at 1 image/s (unless otherwise indicated); note that videos in supplementary material are accelerated (5 to 50 image/s).

Optogenetic illumination of brain slices.
Activation of ChR2 was obtained illuminating brain slices through the 4x objective. A white light source (130W mercury lamp, Intensilight, Nikon, Japan) was connected to the epi-illumination port of the microscope with a light guide containing a 420nm UV blocker filter (series 2000, Lumatec, Germany); the white light was filtered with a 475/50 filter (Semrock) placed in the optical path and delivered to the objective with a FF685-Di02 dichroic beamsplitter (Semrock). The area of illumination was 38.5mm 2 . The blue light power density measured with a power meter (Ophir Photonics, USA) was 2.8mW/mm 2 at the slice surface (photodetector placed at the level of the slice). For spatial illumination experiments, the light guide was connected to a digital micromirror device (DMD)-based patterned photostimulator (Polygon 400, Mightex, Canada), which was connected to the microscope with a custom adapter and a NI-FLT6 Epi-fluorescence Cube Turret equipped with a FF685-Di02 dichroic beamsplitter and a 475/50 filter. The area of the spatial illumination was between 0.34 and 1.1mm 2 .
ON/OFF of the illumination was controlled with the Intensilight shutter or with the Polygon 400 software. We mainly used continuous illumination, but trains of illumination have been also successfully tested (with spatial illumination: 5Hz, duty cycle 50%; see Video 7). The infrared filter of the microscope, besides for patch-clamp experiments, was used in most experiments to eliminate the blue light of the optogenetic illumination from the acquired images and to obtain near-infrared IOS.

Induction of CSD in brain slices by application of KCl.
Cortical spreading depression was induced by brief puffs of KCl (130mM) with a glass micropipette (2-4MΩ) connected to an air pressure injector (holding pressure: < 1PSI, injection pressure: 7 to 10PSI; PV820 Pneumatic Picopump, WPI, USA) (1). The field potential recording pipette was placed at least 500μm away from the CSD induction area. For CSD induction with a solution containing 12mM KCl (12mM KCl, 125mM NaCl, negative control solution was 137mM NaCl saline solution), long local perfusions of KCl were performed with a glass pipette of 0.5MΩ (3PSI) until CSD induction. For both brief puffs and long local perfusions, Fastgreen (0.1%, SIGMA-Aldrich, USA) was added to the pipette solutions to visualize the injection area. Application of the control NaCl solutions with Fastgreen never ignited CSD.

Processing and analyses of IOS images.
Image analysis with ImageJ-Fiji was used for identifying the CSD wave by intrinsic optical imaging, as in (1). Image processing was performed to evaluate latency of CSD initiation and to quantify CSD propagation speed with a custom made macro (available at https://www.ipmc.cnrs.fr/~duprat/scripts/imagej.htm): to eliminate the background and isolate the wave from the raw image, a representative image acquired before the 470nm illumination was subtracted from the others and then contrast was enhanced. To determine the speed of the propagating wave, successive line plots of the wave front from processed images (every 2s) were drawn manually and the spatial distances between them were measured by means of the peak finder ImageJ plugin. For each slice, speed was estimated on a minimum of 4 time points (8s). For few experiments, the quality of the images was not sufficient for a reliable quantification of CSD propagation speed, and they were not included in the analysis for the quantification of the speed.
Latencies of CSD initiation were quantified off-line using processed images. Only CSDs that initiated in the visual field of the camera were used for latency analysis.
For the evaluation of the percentage of CSD induction, individual slices were exposed to 470nm illumination for 90s maximum. When the illumination did not induce CSD within this time limit, we tested that the slice could generate CSD by injecting a puff of 130mM KCl solution as described above.
The percentage of optogenetically induced CSD was quantified considering as non-successful optogenetic inductions only experiments in which the slice could generate CSD with a subsequent application of KCl. The same protocol was used to determine the success rate of spontaneous CSD induction following bath application of Hm1a toxin: individual slices from VGAT-ChR2 mice were perfused with rACSF or with Hm1a (10nM) dissolved in rACSF for 15min. If no CSD occurred in this time window, then optogenetic CSD induction was tested. The percentage of CSD triggered following Hm1a was quantified considering as non-successful inductions the experiments in which the slice could generate CSD with a subsequent 470nm illumination. Threshold of CSD induced by short puffs of 130mM KCl was quantified as in (1), evaluating the area of injection by drawing the limits of the dark zone observed upon injection of the fast-green containing solution and obtaining with ImageJ-Fiji the value of the enclosed area, which was increased by repeating the injections (inter-injection interval 1min) until CSD was ignited, which defined the CSD threshold. "Aborted CSD" refers to CSD that rapidly decelerate after initiation and stop propagating within <800µm from the initiation site.

Patch-clamp recordings in cell-lines.
Plasmids, cell culture and transfections. The cDNAs of the NaV1.1 Na + channel (GenBank sequence NM_006920.4) and of the NaV1.2 Na + channel (GenBank sequence NM_021007) were obtained from Jeff Clare (Glaxo-SmithKline, UK) and subcloned into the pCDM8 vector to reduce rearrangements, as already described (17,18

In vivo experiments.
Male or female mice (4-8 weeks of age) were deeply anesthetized with ketamine/xylazine (100mg/kg and 5mg/kg, respectively) and placed in a Faraday cage-shielded stereotaxic frame (Narishige Instruments, Tokyo, Japan). Supplemental doses of anesthesia were applied on appearance of withdrawal reflex in response to limb pinching. The body temperature was maintained at 37-37.5°C (rectal probe) with a heating pad connected to a temperature controller (TCAT2DF, Physitemp Instruments, USA); the stability of the respiratory activity was monitored with a piezoelectric transducer (MLT1010 Pulse Transducer, AD Instruments, UK) and that of the heart rate with ECG recordings.
For optogenetic experiments, a craniotomy was performed (-2mm antero-posterior, 3.5mm lateral with respect to bregma), the dura was carefully removed, and mineral oil (Sigma-Aldrich) was applied on the cortex to prevent cortical surface drying. A glass pipette filled with 0.9% NaCl (1-2µm tip diameter, Ag/AgCl electrode) was lowed into the barrel cortex for recording DC field potentials. The pulses, delivered till a CSD was triggered; after 100s the stimulation was considered unsuccessful. Note that trains of illumination were able to induce CSD also in brain slices (Video 7).
For evaluating the effect of Hm1a on KCl-induced CSD, two burr holes were drilled at the level of the somatosensory cortex (coordinates relative to Bregma: AP= - 1.7 or - 0.4mm, respectively; lateral= ± 3.0mm, according to the mouse brain Atlas of Paxinos and Watson (19) in ACSF) or ACSF were injected (2.5µl) at the same stereotaxic coordinates at a rate of 0.33µl/min and the needle was left in situ for an additional 5min to prevent back flow before removal. Two KCl-induced CSD were generated at 15min intervals before and after injection of Hm1a or ACSF. Latencies between the beginning of cortical KCl injection and the onset of CSD were measured. For each hemisphere, the mean of latencies obtained for the two CSD before Hm1a or ACSF injection was compared to that observed after the injection.
DC field potentials were recorded with a glass micropipette filled with ACSF and a differential extracellular amplifier (EX4-400, Dagan, USA), and acquired with a Digidata 1440A and pClamp software (Molecular Devices, USA). A reference Ag/AgCl electrode was placed on the cortical surface.
Anesthetized mice were sacrificed at the end of the recordings by cervical dislocation.

Computational model
We revised the model developed in (20)  is the membrane capacitance per area unit, the membrane potential, the time, the valence and a conversion factor. We replaced the Wang-Buzsáki model of the interneuron that we used in (20) with a more recent model, which better models features of fast-spiking cortical interneurons (21).
Moreover, we modified the GABAergic neuron leak current implementing Na + (0.012 mS cm −2 ) and K + (0.05 mS cm −2 ) leak conductances, so that the GABAergic neuron does not spike in the absence of external input and its resting potential is in the physiological range.
Compared to (20), we set the ratio of the GABAergic neuron/pyramidal neuron volume to 2/3, and, to include the impact of excitatory synaptic currents on the dynamics of ion concentrations, we separated its Na + , and K + , components, assuming an equal permeability of the glutamatergic receptors to both ions: where is the maximal conductance, the pyramidal neuron synaptic variable, and and the reversal potentials. We modeled external inputs (drives) to the neurons using constant glutamatergic currents (which did not depend on the glutamate released by the synapses of the modeled neurons), whose amplitude was set by varying their conductances: , for the pyramidal neuron and , for the GABAergic neuron. We included the activity of the Na/K ATPase in the dynamics of ion concentrations for both neurons, not only for the pyramidal one as in (20). We replaced the expression describing the dependence of the pump current on the intracellular Na + and extracellular potassium concentrations with a more realistic one developed by (22), which is based on experimental data: We introduced a voltage dependence of the pump as in (23), to prevent the membrane potential from reaching excessively negative values when recovering from a depolarization block: We used the half activation concentration values and from (24) and a maximal pump rate at −70 mV ,−70 = 30 µ −2 . We also modified a number of additional minor aspects. First, we increased the maximal conductance of the Na + leak current of the pyramidal neuron to 0.015 mS cm −2 , so that its resting membrane potential is not overly negative. We also set the maximal conductance of Supplementary Figure S1. Test of the selectivity of our synthetic Hm1a peptide on Na + channel isoforms expressed in the central nervous system: Nav1.1, Nav1.2 and Nav1.6. A. Representative whole-cell patch-clamp recordings of Na + currents evoked with 100ms depolarizing steps to -10mV from a holding potential of -100mV in tsA-201 cells expressing Nav1.1, Nav1.2, or Nav1.6 in control condition or in the presence of 10, 50 or 300nM Hm1a. Apparent EC50 (efficacy concentration at 50% of maximum effect) was 3nM for Nav1.1 (n=5 cells), 53nM for Nav1.2 (n=5 cells) and 41nM for Nav1.6 (n=6 cells). Note the selective enhancing effect of 10nM Hma1 on Nav1.1 compared to Nav1.2 and Nav1.6. At higher concentrations, Hma1 loses its selectivity on Nav1.1. B. Quantification of the effect of 10nM Hm1a displayed as fold increase of INaP, showing that Hm1a is selective towards Nav1.1 over Nav1.2 and Nav1.6 isoforms. Whiskers box-chart plots represent median, minimum and maximum.
Supplementary Figure S3. CSD is readily triggered by focal puffs of KCl 130mM in neocortex, striatum and hippocampus. A. Experimental design of KCl-induced CSD, which was triggered by a focal puff of 130mM KCl applied using a picopump (10PSI, 500ms). Fastgreen (0.1%) was added to the KCl solution to visualize the effective KCl injection and quantify its area (mean=0.030 ± 0.005mm², n=10 slices). B. Representative CSD induced in the neocortex and revealed by both a negative DC shift in the LFP and the propagating wave observed with intrinsic optical signal (IOS) imaging. The four lower panels show representative IOS images of CSD at different time points (image processing of raw images to better highlight the CSD wave, see methods). Scale bars 250µm. See Video 2. C. CSD was induced by 130mMKCl and had similar properties in different mouse lines: the dot plot displays the propagation speed in slices from VGAT-ChR2 mice (line that we used for optogenetic experiments, in which CSD has been induced with 130mM KCl after unsuccessful optogenetic illuminations; n=7 slices), WT littermate mice (n=14), VGAT.cre mice (n=10) and ChR2.lox mice (n=10) (Kruskall-Wallis test, p=0.32); for the last three conditions the slices are those presented in Fig.3D (success rate plot). Bars correspond to medians. Pooling all the data, the CSD ignited by the KCl puff propagated in the cortical tissue at the speed of 3.01 ± 0.08mm/min (mean ± SEM, n=41 slices).    Representative slice from VGAT-ChR2 mice (which express ChR2 selectively in GABAergic neurons) containing neocortex, hippocampus, dorsal striatum and thalamus, in which CSD was induced only in the neocortex by the illumination with blue light of a complete hemisphere. The optogenetic illumination has been filtered out for clarity and was larger than the imaged area; illumination was stopped at CSD initiation. The pipette is the LFP extracellular recording electrode. Time is indicated as min:sec. 5 images/s. See Figure 3 for details.

Video 6.
Optogenetic CSD induction using continuous spatial illumination (illumination area: 1.1mm 2 ), visualized by intrinsic optical imaging in raw images (left panel) and after image processing (right panel). For better illustrating the experiment, in these images the optogenetic illumination has not been filtered out and was stopped at CSD initiation. Time is indicated as min:sec. 5 images/s.

Video 7.
Optogenetic CSD induction using 100ms-5Hz trains of spatial illumination (illumination area: 1.1mm 2 ), visualized by intrinsic optical imaging in raw images; the optogenetic illumination has not been filtered out for better illustrating the experiment, and was stopped at CSD initiation. CSD has been induced with discontinued illumination in 62.5% of the slices (n = 8 slices in total). Time is indicated as min:sec. The video has been recorded at 5 images/s and is shown at 50 images/s.