Cortical Activation in Generalized Seizures

Summary

Objective

Generalized epilepsy patients exhibit different epileptiform events including asymptomatic interictal spikes (IS), absence seizures with spike-wave-discharges (SWDs), and myoclonic seizures (MS). Our objective was to determine the spatiotemporal patterns of cortical activation in SWDs, IS and MS in the Gabra1^+/A322D juvenile myoclonic epilepsy mouse.

Methods

We fabricated affordable, flexible high-density EEG (HdEEG) arrays and recorded spontaneous SWD, IS, and MS with video/HdEEG. We determined differences among the events in amplitude spectral density (ASD) in the δ/θ/α/β/γ-frequency bands at baseline (3.5–4.0s before the first spike time, t₀) and the pre-spike period (0.1–0.5s before t₀) and elucidated the spatiotemporal activation during the t₀ spike.

Results

All three events had an increase in ASD between baseline and pre-spike in at least one frequency band. During pre-spike, MS had the largest δ-band ASD, but SWD had the greatest α/β/γ-band ASD. For all three events, the ASD was largest in the anterior regions. The t₀ spike voltage was also greatest in the anterior regions for all three events and IS and MS had larger voltages than SWD. From 7.5–17.5 ms after t₀, MS had greater voltage than IS and SWD and maximal voltage was in the posterior parietal region.

Significance

Changes in spectral density from baseline to pre-spike indicate that none of these generalized events are instantaneous or entirely unpredictable. Prominent engagement of anterior cortical regions during pre-spike and at t₀ suggest common anterior neural circuits participate in each event. Differences in pre-spike ASD signify that although the events may engage similar brain regions, they may arise from distinct pro-ictal states with different neuronal activity or connectivity. Prolonged activation of the posterior parietal area in MS suggests that posterior circuits contribute to the myoclonic jerk. Together, these findings identify brain regions and processes that could be specifically targeted for further recording and modulation.

Introduction

Generalized seizures (GS) are those that rapidly activate bilaterally distributed brain networks¹ and are the prominent symptom of many epilepsy syndromes including the genetic (aka idiopathic) generalized epilepsy and the developmental and epileptic encephalopathy syndromes². Many patients with generalized epilepsy exhibit multiple types of GS with different EEG patterns and behavioral correlates. Typical absence seizures confer abrupt altered awareness with EEG spike-wave discharges (SWDs) while myoclonic seizures (MS) are lightening-like body jerks with spikes or polyspikes. Patients also have interictal spikes (IS) that resemble MS on EEG but lack any observable behavior change.

Because GS activate bilateral networks so rapidly¹, they appear on clinical EEG recordings as though they engage all brain areas without discrete foci that could be specifically targeted. However, advanced neurophysiology recordings in humans and rodents revealed that typical absence GS do preferentially engage specific brain regions^3–8. High density EEG (HdEEG) recordings in a rat model of typical absence epilepsy identified a prominent role of the somatosensory cortex in SWDs⁴. These HdEEG experiments, in combination with other GS studies^3,9–12, provided the basis for a current model in which typical absence seizures involve reverberation of the somatosensory thalamocortical network and with possible initiation in the cortex. Importantly, selective targeting of nodes of the thalamocortical network has modulated typical absence seizure frequency in some of these models^6,13–16.

Within the last decade, a multitude of human monogenic generalized epilepsy mutations have identified and some of them have been expressed in mouse models¹⁷. We developed the (Gabra1^+/A322D) mouse model of juvenile myoclonic epilepsy (JME) mouse which possesses the human missense mutation (A322D) of the GABA_A receptor α1 sbunit¹⁸. Significantly, the Gabra1^+/A322D mouse exhibits spontaneous IS, MS and typical absence seizures/SWDs^19–21 and thus provides an opportunity to elucidate the brain areas activated during these events within a single model. Here, we developed a method to fabricate affordable, flexible HdEEG arrays for mice and determined cortical regions activated in each of these epileptiform events.

Methods

Animals

All protocols were approved by the Vanderbilt University Institutional Animal Care and Use Committee. We previously described Gabra1^+/A322D mouse creation²⁰ and that, unlike their wild type littermates, Gabra1^+/− and Gabra1^+/A322D mice have frequent ethosuximide-sensitive SWDs with behavior arrest and EMG attenuation and MS with generalized spikes^20,21. We used six adult (21 ± 3 weeks, 19 – 26 weeks) Gabra1^+/A322D mice of both sexes. The mice were housed in a facility with a temperature and humidity-controlled environment and a twelve-hour light-on/light-off cycle and were provided unlimited access to food and water.

HdEEG array fabrication

Using the open-source drawing software (Inkscape), we designed HdEEG arrays with 0.93 mm diameter electrode pads (EPs, circles) that overly supplementary motor (sM), motor (M), somatosensory (S), barrel somatosensory (B), visual (V), and association (A) cortices (Fig 1A). Two additional EPs for the reference (R) and ground (G) overlay cerebellum. Lines (0.20 mm) from the EP to the zero-insertion-force connector pads (ZP) are conducting wires. HdEEG arrays were fabricated in complementary halves to be positioned separately over the left and right hemispheres. HdEEG designed with these dimensions allowed the greatest density of electrodes while still enabling copper etching and component mounting.

Figure 1: HdEEG recording of generalized seizures.

A) Scale diagram of a mouse dorsal brain indicating the location of the HdEEG electrodes. Electrodes with odd and even numbers are over the left and right hemispheres, respectively, and electrode names indicate the region of functional cortex; supplementary motor (sM), primary motor (M), somatosensory (S), barrel somatosensory (B), parietal association (A), and visual (V). Reference (R) and ground (G) electrodes are over cerebellum. B) Template for left and right complementary halves of HdEEG array. The small circles are the electrode pads (EP) where the electrode connectors are soldered. The lines from the electrode pads are the wires that connect to the zero-insertion force pads (ZPs) that connect to the headbox. C) Examples of a spike wave discharge (SWD, blue), interictal spike (IS, green), and myoclonic seizure (MS, red) recorded from the same mouse. Each discharge exhibits frontally-predominant generalized spikes. MS resemble IS on visual inspection but are associated with a myoclonic jerk; time of jerk marked by dotted line (47 ± 31 ms after t₀).

The design was printed onto dextrin-coated toner transfer paper (Pulsar, 50–1101) using a common office laser printer (Ricoh, MPC3004). Copper-clad polyimide film (Cu 610 g/m², polyimide 25 μm thickness; Dupont, Pyralux LF9210R) was pre-treated by lightly abrading it with 320 grit sandpaper and then by washing it, consecutively, with a thiourea-based metal polishing agent (Tarn-X, Jelmar), distilled water, and acetone. The image was transferred from the paper to copper-clad polyimide by passing it six times through a heated laminator (Tamerica, SM330) and the toner on the image was sealed with a plastic toner-reactive foil (Pulsar, 50–1225) according the manufacturer’s instructions. Unprotected copper was dissolved with a FeCl₃/HCl etching solution (MG Chemicals, 415) at 50–60° C for approximately ten minutes.

Electrode connectors were gold-plated electronic pin receptacle connectors (0.52 mm inner diameter, Mill-Max, 4428-0-43-15-04-14-10-0) and were soldered (247 Solder) to the EP in an electric oven (Black & Decker, TO1705SB) and then strengthened with epoxy (J-B Weld, 8276). The arrays were bent at a right angle and the wires were protected with acrylic conformal coating (ACL Staticide, 8690) and a biocompatible²² silicone sealant (Dow Corning, 734).

Electrodes were straight tungsten rods (254 μm diameter, A-M Systems) cut to approximately 3 mm and sharpened to a point (80 ± 20 μm) with a rotary tool (Dremel). A small amount of epoxy was applied to the top of the electrode to facilitate handling and prevent it from being inserted too far into the connector.

The electrode cups from standard, disposable DIN 42802 connector-containing EEG electrodes were removed, and the bare wires were soldered to the terminals of two flat flexible connectors (FFC, 12 contact, 1 mm pitch, Amphenol, SLW12R-1C7LF). These FFC were used to connect the HdEEG ZP to the EEG headbox.

Surgery

Under isoflurane anesthesia, the dorsal scalp was removed, and the skull was cleaned with saline and dried. Bregma was identified and the two halves of the HdEEG array were placed with the B5 and B6 EP positioned at AP −1.2 mm, M/L ±3.5 mm and the reference and ground electrodes over the cerebellum. The array was firmly attached to the skull with cyanoacrylate adhesive and dental cement. Through the lumen of each electrode connector, a 29 Ga needle gently drilled the underlying skull. Electrodes were then inserted through the connector into the subdural space without brain penetration. The mice were observed every day after surgery to monitor for any signs of distress.

Video/HdEEG recording

Video/HdEEG was obtained using a Nicolet V32 amplifier with 2000 Hz EEG sampling and 30 fps video. A short recording was obtained on postoperative day zero or one to verify adequate electrode impedances and to acclimate the mice to the recording chambers. Video/HdEEG used for quantification was performed on postoperative day seven or eight starting at approximately 10:00 am, three hours after the lights-on phase of the mouse facility and continued for 5.4 ± 0.6 hours. After euthanasia, the skulls and brains were examined to ensure correct placement of the electrodes and absence of cortical injury.

Signal analysis

We previously demonstrated that Gabra1^+/− and Gabra1^+/A322D mice had substantially more frequent epileptiform events than their wild type littermates and described the EEG and behavioral properties of SWDs (characteristic 6–8 Hz complexes with behavior arrest) and MS (spikes with head or body jerks)^19,20,23. The HdEEG record was visually inspected to identify these epileptiform discharges (Fig 1C). To ensure signal quality for quantitative analysis, we employed strict inclusion criteria and rejected samples that contained artifact in more than two channels or in two adjacent channels. The video that accompanied EEG epochs that met the inclusion criteria was analyzed, frame-by-frame and SWDs accompanied by movement were rejected. A MS was a generalized spike with an abrupt jerk of the head, body, tail/limb, or multiple body parts within 100 ms of the M4 electrode peak and events with voluntary movement within one second of the peak were excluded. IS were generalized spikes without movement within one second of the peak.

Quantitative HdEEG analyses was performed using the FieldTrip MATLAB toolbox (Donders Centre for Cognitive Neuroimaging, University Nijmegen)²⁴. HdEEG was digitally high-pass filtered at 0.3 Hz and notch filtered at 60, 120, and 180 Hz using zero-phase fourth-order Butterworth filters and downsampled to 400 Hz. Because SWDs, IS, and MS all exhibited prominent spikes in the M3 and M4 electrodes, discharge onset (t₀) was defined as the time of the first M4 peak.

Spectral power was determined from Morlet wavelet time-frequency (2–50 Hz) transformations in 50 ms windows and the amplitude spectral density (ASD) was calculated as the square root of the power. The baseline period was defined as the mean ASD from four seconds to 3.5 seconds before t₀ (−4.0 to −3.5s). The pre-spike period was defined as the closest 50 ms time interval before t₀ in without overlap of the wavelet and the t₀ spike. The pre-spike time would be expected to be most biologically relevant for spike initiation and exhibit the greatest difference from baseline but still be amenable to time-frequency analysis without spectral contamination from the first spike. Because we used a six cycle wavelet width, the half-widths were slightly less than each frequency’s wavelength²⁵. Therefore, the pre-spike periods for δ- (2–3 Hz), θ- (4–7 Hz), α- (8–12 Hz), β- (13–29 Hz), and γ- (30–50 Hz) frequency bands occurred at, respectively, 0.50, 0.25, 0.15, 0.10, and 0.10 s before t₀. Spatiotemporal analysis of SWD, IS, and MS spikes that were all time-locked to the M4 spike, a spike present in all three epileptiform events.

Statistical analyses

All statistical comparisons were made using the intrasubject-averaged epileptiform events of each type. Fieldtrip’s nonparametric cluster-based permutation (CBP) method allows the comparison of numerous multidimensional neurophysiological data (time, channel, frequency). Dependent-samples CBP tests were used to evaluate a) within-subject differences among the three event types (SWD, IS, and MS) and b) changes in ASD from baseline to pre-spike. Fieldtrip’s CBP implementation has been thoroughly described²⁶. Briefly, 1) all data points in the two conditions were initially compared using the dependent-sample student’s T-test and those reaching a significance of p < 0.05 were clustered together based on temporal, spectral, and spatial (electrode center within 2.9 mm) adjacency. The sum of T-values was calculated for each cluster. 2) 500 artificial datasets were created by randomly assigning the original data points to either condition 1 or condition 2. The student’s T-test in step 1 was repeated for each artificial dataset to produce 500 sets of artificial T-value sums for the different clusters. 3) The percentile rank of the T-value sums from the original dataset relative to the 500 artificial datasets is the probability that two conditions differ.

As a second method to compare a) ASD/voltage differences among the three epileptiform events and b) differences in ASD baseline and pre-spike within each event, we constructed linear mixed effect models (LMEM) using the lme4 package for R²⁷ with either a) event type or b) baseline/pre-spike as one fixed factor. In both cases, brain region was a second fixed factor and experimental subject was a random factor. ASD or voltages were averaged from bilateral electrodes corresponding to one of four pre-defined brain regions, motor, (MO, electrodes sM1–2, M1–M4), somatosensory (SS, electrodes S3–4, B5–8), posterior parietal association (PA, electrodes A1–2), and visual (V, electrodes V1–6). ASD was integrated for each frequency within each frequency band (δ-γ). LMEM of the t₀ spike (Fig 4), was performed in three time points relative to t₀, early (−17.5 to −7.5 ms), middle (−5.0 ms to 5.0 ms), and late (7.5 ms to 17.5 ms). The data were sequentially fit to multiple LMEM starting with the least complex (random effect only) and then progressing in complexity with individual addition of each of the fixed factors, additive factor combination, and, finally, interacting factor combination. The models were compared by an analysis of variance test and the least-complex model that significantly better-fit the data was chosen. Post-hoc pairwise contrasts were calculated using estimated marginal means and p-values were adjusted by the Tukey HSD method.

Figure 4: Spatiotemporal analysis of initial spikes.

The initial epileptiform spikes from each event were time-locked. Within-subject time-locked voltages from each event type were averaged from −20 ms to 20 ms with a 2.5 ms time resolution. A) Average inter-event differences (SWD-IS, SWD-MS, and IS-MS) in voltage are plotted topographically from −20 ms to 20 ms relative to t₀ in 5 ms bins. Blue colors indicate that the voltage of the second event type is greater than the first, red colors indicating that the voltage of the first event type is greater than the second, and green signifies no voltage difference (0 ± 4% full scale) between the two event types. White circles mark clusters with CBP test p-values < 0.05. B) Scatter plots depict within-subject averaged voltages averaged into three time points relative to t₀, early (−17.5 to −7.5 ms), middle (−5 ms to 5 ms) and late (7.5 ms to 17.5 ms) and pooled from four brain regions (MO = motor, SS = somatosensory, PA = parietal association, V = visual). Data depicted with the same color and symbol shape were obtained from the same subject and the lines connecting the symbols show the within-subject voltage-differences among event type (upper) or brain region (lower). Asterisks represent statistically-significant post-hoc pairwise contrasts (* p < 0.05, ** p < 0.01, *** p < 0.001, N = 6 subjects).

Diagrams of the mouse dorsal cortex were made using the Brain Explorer 2 software (Allen Institute for Brain Science). Topographic plots of ASD and voltage were plotted over representations of the electrode arrays that were superimposed over diagrams of the dorsal cortex. All reported p-values are two-tailed and Bonnferoni (CBP) or Tukey (LMEM) corrected p-values are reported in the text as appropriate. Because referential surface negative EEG potentials reflect postsynaptic intracellular depolarization, we report surface negative and surface positive EEG voltage as positive and negative voltage, respectively, to simplify presentation and to maintain consistency in sign between voltage and ASD.

Results

HdEEG arrays were easily-prepared from affordable, commercially-available materials. Implantation was well-tolerated; all six mice recovered rapidly from the surgery and did not demonstrate any sign of distress postoperatively. The average electrode impedance was 23 ± 17 kῼ on postoperative day zero/one and 22 ± 14 kῼ on postoperative day seven/eight with an average change of −1 ± 14 kῼ in electrode impedance during the week. Postmortem examination revealed that the electrodes were positioned as expected without damage to the cortex.

HdEEG monitoring captured spontaneous SWD, IS, and MS (Fig 1C). From the six Gabra1^+/A322D mice, we identified 26 SWDs, 93 IS, and 23 MS that met the inclusion criteria for quantitative analysis. SWDs consisted of 6–8 Hz spike-wave complexes similar to those described in rat and mouse models of absence seizures^28,29. Both IS and MS appeared as generalized spikes that appeared similar to visual inspection (Fig 1C). However, unlike IS, MS were associated with jerks of the head (4), body (14), tail/limb (3), and multiple body parts (2). The jerks occurred 47 ± 31 ms after the M4 peak; only one jerk (head) occurred prior to the M4 spike (−19 ms).

Baseline and pre-spike spectral density differs among the epileptiform events.

Previous local field potential recordings from depth electrodes in somatosensory cortex and thalamus in a rat absence epilepsy model found that δ and θ spectral density increased prior to SWD onset³⁰. It was not known if pre-spike changes in spectral density are associated with IS or MS or occur in regions other than somatosensory cortex.

Time-frequency transforms for each event type within a given subject were averaged. CBP testing revealed that, at baseline, (−4.0 to −3.5s), SWDs exhibited significantly reduced δ-, α-, and β-band spectral density than IS and reduced α-band ASD than MS (Fig 2A, p = 0.012). Similar results were obtained fitting the data to the LMEM (Fig 2B, Fig S1); δ-ASD was significantly different among all three epileptiform events (MS > IS > SWD, p < 0.027), IS events had significantly greater α- and β-band ASD than MS and SWD (p < 0.033), and SWD had greater γ-band ASD than MS (p = 0.020). There were no significant differences among the event types in baseline θ power in CBP or LMEM analysis (Fig 2A, Fig S1) and there was no significant interaction between epileptiform event type and brain region. In all frequency bands, ASD was greater in the motor (MO) and somatosensory (SS) regions than the visual (V) areas (p < 0.003), and, in the θ-γ frequency bands, there was also greater ASD in the posterior parietal association area (PA) than the visual region (p < 0.001, Fig 2B, Fig S1).

Figure 2: Baseline spectral density.

Amplitude spectral density (ASD) at baseline (−4.0 to −3.5s relative to t₀) was averaged for all events within each subject. A) average inter-event differences (SWD-IS, SWD-MS, and IS-MS) in baseline density are plotted topographically for each frequency band (δ/θ/α/β/γ). Blue colors indicate that the ASD of the second event type is greater than the first, red colors indicating that the ASD of the first event type is greater than the second, and green signifies no ASD difference (0 ± 4% full scale). White circles mark clusters with CBP test p-values < 0.05. B) Scatter plots depict within-subject averaged baseline δ-, α-, and β-band ASD grouped by event type (upper) or brain region (lower, MO = motor, SS = somatosensory, PA = parietal association, V = visual). Data depicted with the same color and symbol shape were obtained from the same subject and the lines connecting the symbols show the within-subject ASD changes among event type (upper) or brain region (lower). Asterisks represent statistically-significant post-hoc pairwise contrasts between event types (upper) or brain regions (lower) (* p < 0.05, ** p < 0.01, *** p < 0.001, N = 6 subjects).

CBP tests demonstrated that SWD had significantly increased θ-, α-, β-, and γ-band ASD between the baseline and the pre-spike periods (Fig 3A, p = 0.004). Scatter plots that depict the time course of the change in ASD reveal that the spectral density in the motor cortex in the θ- and α-frequency bands began to increase at approximately one second prior to t₀ while the spectral density in the β-, and γ-bands began to increase at approximately 0.25 s prior to t₀ (Fig S2A). LMEM (Fig 3B, Fig S2B) showed similar results as CBP and identified significant baseline to pre-spike increases in the δ-γ bands in SWD, β-band in IS, and θ- and β-bands in MS.

Figure 3: Pre-spike spectral density.

For each frequency band (δ/θ/α/β/γ), amplitude spectral density (ASD) was determined at baseline (−4.0 to −3.5s relative to t₀) and pre-spike time (−0.50, −0.25, −0.15, −0.10, −0.10s relative to t₀ for δ, θ, α, β, γ-bands, respectively). Within-subject averaged differences between pre-spike and baseline ASD were determined for each event type, separately (A-B), as well as within-subject differences in pre-spike ASD among the three type of epileptiform events (C-D). A,C) Averaged differences in pre-spike and baseline ASD for each event type (A) or inter-event differences in pre-spike density (C) are plotted topographically for each frequency band. Blue colors indicate decreased pre-spike ASD relative to baseline (A) or that the second event type has greater pre-spike ASD than the first (C) and red colors signify an increase in pre-spike ASD relative to baseline (A), or that the first event type had greater pre-spike ASD than the second (C). Green signifies no difference (0 ± 4% full scale) between the conditions. White circles mark clusters with CBP test p-values < 0.05. B,D) Scatter plots depict B) within-subject averaged baseline (b) and pre-spike (p) θ- and β-band ASD or D) pre-spike ASD (δ-, θ-, α-, and β-band). LMEM testing demonstrated that SWD, but not IS or MS had significant interactions between brain region (MO = motor, SS = somatosensory, PA = parietal association, V = visual) and baseline/pre-spike grouping (B) and that, in the α-, and β-bands, there were significant interactions between brain region and event type in and thus these data are plotted separately for each brain region. Data depicted with the same color and symbol shape were obtained from the same subject and the lines connecting the symbols show the within-subject ASD changes. Asterisks represent statistically-significant post-hoc pairwise contrasts (* p < 0.05, ** p < 0.01, *** p < 0.001, N = 6 subjects). Positive voltage corresponds to surface negative EEG potentials.

CBP testing of the pre-spike period (−0.50 to −0.10s), revealed significant differences between SWD and both IS and MS in the in the β-, and γ-frequency bands as well as differences between SWD and MS in the θ- and α-frequency bands (Fig 3C). Prominent differences between SWD and both IS and MS were localized to the anterior cortex. Data fit to LMEM revealed that in the α-, β-, and γ-frequency bands, there was a significant interaction between epileptiform event type and brain region. SWDs had increased ASD relative to both IS and MS in the motor cortex (p < 0.006, Fig 3D, Fig S2C). In addition, SWD exhibited increased pre-spike α-band density relative to MS in the somatosensory cortex (p = 0.042, Fig 3D). MS had increased δ-density than IS and MS (p < 0.008) and there were no differences among epileptiform event types in the θ-frequency band (Fig 3D).

Spatiotemporal evolution of initial spikes differs among SWD, IS, and MS

CBP analyses of averaged events from each subject demonstrated that IS and MS had significantly higher voltage than SWD (p = 0.012) with prominent differences between SWD and IS in the anterior region and a prominent difference between SWD and MS more posterior (Fig 4A). The mean voltage was grouped in three time periods (Fig 4B), early (−17.5 to −7.5 ms), middle (−5.0 to 5.0 ms), and late (+7.5 to 17.5 ms). In the early period, there was no difference in mean voltage among the three event types, but voltage in the motor and somatosensory regions was greater than the visual areas (p < 0.040). In the middle time period, both IS and MS had greater mean voltage than SWD (p < 0.001) and, like the first period, voltage in motor and somatosensory areas was greater than visual regions (p ≤ 0.014). In the late period, MS has greater voltage than both IS and SWD (p ≤ 0.007). Unlike the other time periods, voltage in the largest voltage was in the posterior region and was greater than that in the somatosensory area (p = 0.015).

Discussion

HdEEG recordings of mouse models of genetic epilepsy reveal potentially seizurogenic regions of cortex and underlying subcortical areas to allow us to generate initial hypotheses regarding potential GS networks. We developed this simple technique to fabricate affordable HdEEG arrays rather than use commercially-available arrays³¹ because we anticipate that we will apply this approach to multiple genetic mouse models in several different forms of generalized epilepsy and thus will require a method that is cost-feasible. Another affordable method of mouse HdEEG recording is achieved through the use of pre-fabricated electrical pin connectors³². However, we found that our copper-etched arrays were better suited to our needs because 1) the flexible polyimide support allowed the array to bend with the curvature of the skull, 2) copper etching our own design allowed placement of electrodes in areas of interest, 3) surgery times were shortened because the craniotomy holes could be drilled directly through the lumen in the electrode connector after the array was attached to the skull, and 4) the thin (80 μm) tungsten electrodes were better-tolerated, in our experience, than 400 μm electronics pins.

Our HdEEG arrays were well-tolerated by the mice. They recovered rapidly after surgery, showed no signs of distress/pain, and the electrode impedances remained stable during the week after surgery until recording. Therefore, these HdEEG arrays are well-suited for the interrogation of GS networks in mouse models.

All three epileptiform events activate anterior cortex.

The similarities and differences among SWD, IS, and MS are summarized in Fig 5. A principal finding was that all events strongly activated the anterior cortex. For all three epileptiform events, the largest voltage at t₀ was in the motor cortex and there were statistically significant differences between the voltages in both the motor and somatosensory regions and the visual area (Fig 3). In addition, the pre-spike spectral density was also localized to the anterior regions prior to all events with the greatest density in the motor cortex and statistically significant differences between the visual area and the other three brain regions at most frequencies.

Figure 5: Common and differentiating features of epileptiform events.

Following the baseline, all three epileptiform events exhibit the greatest spike voltage in the anterior regions as well as significant increases in β-spectral density from baseline to pre-spike. IS and MS spikes had significantly greater voltage than SWD spikes and MS and SWD events and significantly increased θ and γ spectral density from baseline to pre-spike. SWDs had increased δ- and α-band ASD from baseline to pre-spike, had the largest pre-spike α-, β-, and γ-band ASD, and had the greatest pre-spike α- and β-band ASD located in motor cortex (MO). MS had increased latent spike voltage 7.5–17.5 ms after t₀.

The activation of the anterior cortex in all three events is consistent with the previous studies that demonstrated a leading role of the somatosensory cortices in rodent SWDs^{4–6,21,30,33} and the activation of frontal cortex in human absence seizures⁷. This result suggests that similar brain networks are engaged in each of these generalized epileptiform events. The finding that the same, or nearby brain regions may participate in different types of GS would have important implications for future neuromodulation therapy because it may be possible to stimulate a single brain region and reduce multiple types of GS exhibited by the same patient.

Based on the results of the prior studies of rodent SWDs^{4–6,21,30,33}, it was surprising that the area exhibiting the largest t₀ voltage and pre-spike activity was in the motor, and not somatosensory cortex. It should be emphasized that increased voltage/spectral density does not necessarily imply a leading role of that region. Identification of the somatosensory cortex as a critical component in SWDs was determined by disconnection, pharmacological, and millisecond-resolution measurements from electrodes placed directly on- or within the cortex. Now that this study identified anterior cortex as a region of interest for all three events, future modulation and high-resolution depth recordings can differentiate the roles of somatosensory and motor cortices

SWDs have the greatest increase in pre-spike spectral density.

A previous depth electrode study in a rat model of absence epilepsy found increased δ- and θ-band activity in thalamus and somatosensory cortex seconds before SWD onset^30,34,35, results that suggested that absence seizures are not instantaneous and unpredictable events. The greater δ- and θ-activity associated with increased thalamocortical functional connectivity that forms a temporary “pro-ictal” state favoring SWD formation. Similarly, we found that, compared with baseline, SWDs have increased pre-spike spectral density in all frequency bands (Fig 3A–B, Fig S2). In addition, we found that IS have increased β-band density, and MS have increased θ- and γ-band activity. Therefore, it may be possible to predict occurrence of all these types of epileptiform events and deliver on-demand therapy.

Although spectral density increases from baseline to the pre-spike period in at at least one frequency band in all three events, SWDs have substantially greater increases than IS and MS. Therefore, SWDs have significantly greater α-, β-, and γ-ASD in the pre-spike period, particularly in the motor cortex, than the other two events (Fig 3D). Future experiments with depth electrodes placed in cortex and thalamus can determine if the faster-frequency (α, β, and γ) SWD pre-spike activity and the modestly-raised spectral density in IS and MS also correspond to different neuronal activity or functional connectivity and the formation of the hypothesized pro-ictal state.

Delayed activation of posterior cortical or subcortical regions may contribute to jerks in MS.

For all three epileptiform events, the maximal spike voltage during the early and periods is in the anterior cortex (Fig 4A–B). However, during the late period, the maximal voltage was in the posterior parietal region (Fig 4B) and, during this time, MS had significantly greater voltage than the other two events (Fig 4B). Although there was no statistically-significant interaction between brain region and event type, MS had the greatest voltage over posterior cortex. Possibly, either posterior cortex or subcortical areas underneath posterior cortex contribute to the MS jerks. The posterior association cortex is a complex region involved in many functions including decision-making, sensory processing, and, importantly, motor planning^36,37 and thus could potentially play a role in mediating MS jerks. It is also possible that subcortical regions ventral to the posterior association cortex (e.g. midline thalamus or midbrain) may be activated during MS cause increased voltage on HdEEG by volume conduction. Future HdEEG in combination with depth electrode recordings in the association cortex and underlying regions will further elucidate the origin of these discharges.

In summary, our detection of increased baseline to pre-spike activity indicates that SWD, IS, and MS are potentially predictable events, and the differences in pre-spike spectral density among the events suggest that they arise from distinct “pro-ictal” states that may be important for spike formation. Prominent activation of anterior cortex in all events indicates that similar brain regions are engaged in each of them but increased late period voltage in only MS suggests posterior circuits may also contribute to myoclonic seizures. Future studies with HdEEG/depth electrodes targeting these areas of interest will extend our understanding of their roles in seizure-genesis and whether specifically targeting these regions will reduce seizures.

Supplementary Material

Supp FigS1-2

Supplementary Figure 1: Baseline θ- and γ spectral density: Amplitude spectral density (ASD) was determined at baseline (−4.0 to −3.5s relative to t₀). Scatter plots depict within-subject averaged baseline θ- and γ-band ASD grouped by event type (upper) or brain region (lower, MO = motor, SS = somatosensory, PA = parietal association, V = visual). Data depicted with the same color and symbol shape were obtained from the same subject and the lines connecting the symbols show the within-subject ASD changes in ASD among event type (upper) or brain region (lower). Asterisks represent statistically-significant post-hoc pairwise contrasts between event types (upper) or brain regions (lower) (*** p < 0.001, N = 6 subjects).

Supplementary Figure 2: Pre-spike spectral density. For each frequency band (δ/θ/α/β/γ), amplitude spectral density (ASD) was determined at baseline (−4.0 to −3.5s relative to t₀) and from −2.0 seconds to the pre-spike time (−0.50, −0.25, −0.15, −0.10, −0.10s relative to t₀ for δ, θ, α, β, γ-bands, respectively) in 0.05 second intervals. The ASD was pooled from electrodes in four brain regions (MO = motor, SS = somatosensory, PA = parietal association, V = visual). A) Among the six subjects, the average difference between the ASD at each time point and the baseline ASD (± standard deviation) in the motor area (filled symbols) and visual area (open symbols) was plotted for each event type (blue, SWD, green IS, red MS). B,C) Scatter plots depict B) within-subject averaged baseline (b) and pre-spike (p) δ-, α- and γ-band ASD or C) pre-spike γ-band ASD. LMEM testing demonstrated that SWD, but not IS or MS had significant interactions between brain region and baseline/pre-spike grouping (B) and γ-band pre-spike density (C), and thus these data are plotted separately for each brain region. Data depicted with the same color and symbol shape were obtained from the same subject and the lines connecting the symbols show the within-subject ASD changes. Asterisks represent statistically-significant post-hoc pairwise contrasts (*** p < 0.001, N = 6 subjects).

Key Points.

• High density EEG recording arrays can be fabricated easily and affordably and are well-tolerated in a genetic epilepsy mouse model

• SWD, IS, and MS preferentially engage anterior cortical regions during pre-spike periods and at the time of the first spike.

• The three epileptiform events differ in the prominent frequency bands activated during the baseline and pre-spike periods.

• MS are associated with persistent posterior activation after the first spike