Dynamics of Sensorimotor Cortex Activation During Absence and Myoclonic Seizures in a Mouse Model of Juvenile Myoclonic Epilepsy

Abstract

Objective

Generalized epilepsy syndromes often confer multiple types of seizures, but it is not known if these seizures activate separate or overlapping brain networks. Recently, we reported that mice with a juvenile myoclonic epilepsy mutation [Gabra1(A322D)] exhibited both absence and myoclonic generalized seizures. Here, we determined the time course of sensorimotor cortex activation and the spatial distribution of spike voltage during these two seizures.

Methods

We implanted Gabra1^+/A322D mice with multiple EEG electrodes over bilateral somatosensory cortex barrel fields (S1) and anterior (aM1) and posterior (pM1) motor cortices and recorded absence seizures / spike-wave discharges (SWDs) and myoclonic seizures. We used nonlinear-association analyses and cross-correlation calculations to determine the strength, leading regions, and time delays of cortical coupling from the preictal to ictal states and within the spike and interspike periods. The distribution of spike voltage was also measured in SWDs and myoclonic seizures.

Results

EEG connectivity among all electrode pairs increased at the onset of both SWDs and myoclonic seizures. Surprisingly, during spikes of both seizure types, S1 led M1 with similar delay times. Myoclonic seizure spikes started more focally than SWD spikes with a significant majority appearing first only in S1 electrodes while a substantial fraction of SWD spikes were detected first in S1 and at least one M1 electrode. The absolute voltage of myoclonic seizure spikes was significantly higher than that of SWD spikes and there was a greater relative voltage over M1 during myoclonic seizure spikes than in the first one to two SWD spikes.

Significance

The leading sites in S1 and similar delay times suggest both SWDs and myoclonic seizures activate overlapping networks in sensorimotor cortex and thus, therapeutically targeting this network could potentially treat both seizures. Spike focality, absolute voltage and voltage distribution provide insight into neuronal activation during these two seizure types.

Introduction

Generalized seizures are classified as those that rapidly engage bilaterally distributed cerebral networks.¹ Juvenile myoclonic epilepsy (JME), the most prevalent generalized epilepsy syndrome, is associated with three types of generalized seizures, myoclonic, generalized tonic-clonic, and, sometimes (~30%), absence seizures.² Anticonvulsant drugs control many JME seizures, but drug-resistant seizures are present in a substantial fraction of JME cases. In particular, JME patients who have absence seizures are more likely to have medically-intractable seizures of all types.^3;4

Determining the cerebral networks that participate in different types of generalized seizures is critical to understand their epileptogenic mechanisms⁵ and for the development of potential network-specific therapies (e.g. neurostimulation).^6-9 Different types of generalized seizures associate with very distinct behavioral and electrographic manifestations,^10-12 observations suggesting that the epileptic brain either uses different networks to generate different seizure types or that it uses the same network but activates it differently. For example, absence seizures confer 5-20 seconds of behavior arrest and loss of awareness associated with spike wave discharges (SWDs),¹² whereas myoclonic seizures produce very brief (< 1 s), lightening-like body jerks associated with polyspike and wave discharges.¹² Possibly, distinct networks are activated in these two seizure types. In rat models of SWDs, ictal activation is first found in the somatosensory cortex (S1) barrel fields.^13-15 Although the brain regions involved in myoclonic seizures have not been studied in detail, the prominent jerks suggested that motor circuits are involved, a hypothesis supported by brain imaging studies.^16;17

Recently, we developed a mouse model¹⁸ that expresses a human JME-associated missense mutation (A322D)¹⁹ of the GABA_A receptor α1 subunit. Early in life (postnatal day 35, P35), Gabra1^+/A322D mice have spontaneous absence seizures with SWDs, and, later in life (P120), they have spontaneous SWDs, myoclonic seizures, and a greater sensitivity to pentylenetetrazol- (PTZ-) evoked myoclonic seizures than wild type mice.¹⁸ Therefore, the study of SWDs and myoclonic seizures in Gabra1^+/A322D mice will reveal the cerebral networks engaged in these two seizure types, results applicable to JME and, possibly, to other generalized epilepsy syndromes that confer also absence and myoclonic seizures. Here, we recorded SWDs and myoclonic seizures in Gabra1^+/A322D mice using subdural electrodes placed in bilateral S1 barrel cortex and in the anterior (aM1) and posterior (pM1) M1 regions and determined the dynamics of sensorimotor cortex activation and distribution of spike voltage.

Methods

Animals

The Vanderbilt University Institutional Animal Care and Use Committee approved all protocols. Mice were housed in a controlled facility with a twelve hour light/dark schedule, a temperature and humidity controlled environment, and ad libitum water and food. We previously reported the construction of mice that heterozygously expressed the Gabra1(A322D) mutation in a congenic C57BL/6J background.¹⁸ Although wild type mice have very rare SWDs and polyspike complexes, we have previously shown that Gabra1^+/A322D and Gabra1^+/− mice have very frequent SWDs that are associated with behavior arrest and EMG attenuation and are inhibited by the anti-absence seizure drug, ethosuximide.^18;20 In addition, in mutant, but not wild type mice, the polyspike discharges are often accompanied by myoclonic jerks and the SWDs.¹⁸

We used four female Gabra1^+/A322D mice (P137-P145) since we previously found that female mice had greater numbers of SWDs²⁰ and that mice older than P120 had more frequent myoclonic seizures.¹⁸ In addition, we recorded stereotactic EEG from four age-matched wild type mice; consistent with our previous results,^18;20 the wild type mice did exhibit any spontaneous SWDs and had substantially fewer myoclonic seizures (not shown).

Stereotactic EEG

Commercially-available screw electrodes (Pinnacle Technology) were sanded to a length of 1 mm using a Dremel tool. Under isoflurane anesthesia, we stereotactically implanted electrodes (Fig 1A) in the bilateral aM1 (AP +1.0 mm, ML ± 1.5 mm), pM1 (0 mm AP, ± 1.5 ML) and S1 barrel field (0 AP mm, ± 3.0 ML mm). A reference electrode and ground electrode were implanted over the posterior cortex/cerebellum junction. Mice recovered at least seven days before the EEG.

Figure 1. Corticocortical relationships in spontaneous SWDs.

A) Scale diagram of a Gabra1^+/A322D mouse brain indicates the location of the stereotactic placement of electrodes in the anterior motor cortex (aM1), posterior motor cortex (pM1), and somatosensory cortex barrel field (S1) as well as the location of the reference (Ref) and ground (Grd) electrodes. B) Example of a referential EEG recording of a SWD. The origin (t = 0s) of the time scale at the top of the panel is at the time of the first spike. C-E) Nonlinear association analyses was performed in overlapping windows starting from 1953 ms prior to first spike to 977 ms after the first spike among the electrodes in the left (left) and right (center) hemispheres as well as between the corresponding electrodes in each hemisphere (right). C) Plots of the mean nonlinear association constant (h²) at the start of each time window demonstrates that h² increases in all electrode pairings at the time of the first spike and that the increase persists for at least the first 977 ms of the SWD. Panel D depicts the time delays between the pairs of electrodes at the start of each time window. Positive delays indicate that the EEG signal in the second electrode (to the right of the arrow) precedes the first electrode (to the left of the arrow). E) The median pre-ictal (t = −1953 to −1855 ms) and ictal (t = 244 to 342 ms) times are shown as a horizontal line within the box plots and the 25^th to 75^th percentile delay times are depicted by the box length. During the ictal, but not pre-ictal period, the median S1 to aM1 lag time was 2 ms (0 - 6 ms) on the left and 3 ms (1 – 7 ms) on the right. In addition, in the right hemisphere, S1 preceded pM1 by 2 ms (0 – 5 ms). There was no significant delay in EEG signal between corresponding electrodes in the left and right electrodes. Asterisks = Bonferroni-corrected P values, *** P < 0.001, ** P < 0.01, * P < 0.05, ns = nonsignificant. N = 30 SWDs from 4 mice.

Video/EEG was obtained using a Natus EEG with a 1024 Hz sampling rate. We recorded baseline EEG for three hours to capture spontaneous SWDs. Although P120 Gabra1^+/A322D mice experience spontaneous myoclonic seizures,¹⁸ they only occur approximately once per day. Administration of low-dose PTZ^18;21 evokes myoclonic seizures with a similar electrographic morphology and genotype- and age-dependency as spontaneous myoclonic seizures.¹⁸ Therefore, after the baseline, we gave 25 mg/kg PTZ intraperitoneally (i.p.) and an additional 10 mg/kg i.p. 45 minutes later. The skulls and brains were examined after euthanasia to ensure correct electrode placement and lack of cortical injury.

EEG analysis

Sliding window nonlinear association analysis

The EEG was first high-pass filtered at 1 Hz prior to analysis. SWDs were identified by their characteristic EEG pattern^18;20 and myoclonic seizures were defined as spike and polyspike discharges associated with a definite myoclonic jerk on video.

We performed linear and nonlinear association analyses^13;22-24 to measure the coupling strength, direction, and time lag among pairs (x, y) of EEG electrodes using a MATLAB program that we adapted from the one included in the FieldTrip analysis toolbox (Donders Centre for Cognitive Neuroimaging, University Nijmegen).²⁵ Unlike linear analysis which assumes a constant relationship between the two signals at all voltages within the time window, nonlinear analysis allows for different relationships at different voltages. Nonlinear physiological processes (e.g. rectification, greater association at higher voltages, etc.) could potentially be involved in the coupling of different brain regions that would not be apparent with a linear analysis. The linear analyses were performed by calculating the least squares regression line between the pairs of electrodes (x,y) using all the voltages within the time window. Nonlinear analyses were performed by sorting the voltages of x and y within the window based on ascending values of x and then dividing the sorted data into seven equal bins. A nonlinear curve that relates x to y was estimated by constructing seven straight line segments using the bin centers of x as the independent variable and the average of the y elements within the bin as the dependent variable.

Using the calculated linear regression lines and the nonlinear curves, the theoretical y value [f(x)] for each value of x was calculated. Equation 1 was used to calculate the association constant, h² [N = sample number (150), ‹y› = mean y value in the time window, y_i = observed y values]. For each time window, h² was calculated 101 times, once for each time shift from −50 to 50 samples. The time shift producing the maximal h² was the time delay, τ. The optimal window duration (146 ms) was empirically determined by using different window sizes and finding that the 146 ms duration was the shortest window that that adequately fit our data. If the relationship between the EEG signals from the electrode pairs is nonlinear, h² from the nonlinear analysis would be greater than h² from the linear analysis.

$$h^2=\frac{\sum _{i=1}^N{(y_i-\langle y\rangle )}^2-\sum _{i=1}^N{[y_i-f\left(x_i\right)]}^2}{\sum _{i=1}^N{(y_i-\langle y\rangle )}^2}$$ Equation 1

For each SWD and myoclonic seizure, h² and τ were calculated in 61 overlapping windows beginning 1953 ms before the start of the seizure through 977 ms after the seizure onset. In the figures, we show one nonlinear coupling of electrode signals (x → y). However, similar results were obtained with the opposite coupling (y → x, not shown).

For SWDs, the first spike defined the ictal start time. Because myoclonic seizures could have one-three spikes, we used the time of the last spike as the ictal start time because this spike was always associated with the myoclonic jerk (jerk spike). Separate analysis of myoclonic seizures associated with one, two, or three spikes provided similar results, and thus we grouped these data together.

Spike voltage, duration and inter-channel association during spike and interspike periods

To determine the coupling and time delays among the electrodes specifically within the spike and interspike periods, we manually centered a small time window (73 ms) on the spike peaks and performed cross correlation analysis with possible time shifts of from −49 to 49 ms. The interspike period was defined as the time 73 ms before the beginning of each spike and 73 ms after the end of the last spike. The time shift that maximized the correlation constant, R, was the time lag, τ. Since the most prominent changes in spike amplitude and morphology were within the first three to four spikes of the SWDs, we limited our analyses to the first four SWD spikes.

To determine the fraction of spikes that first originated in S1, M1 (anterior or posterior), or diffusely in S1 and M1 (S1/M1), we performed cross-correlation analysis between the channel with highest voltage and the other five channels. We omitted channels without a spike by excluding channels with a correlation constant (R) less than 0.75. The channel(s) that had the shortest delay time, τ, was/were considered to contain the leading spike. If only S1 electrodes (unilateral or bilateral) contained the leading spike, it was considered of an S1 spike, and if only aM1 and/or pM1 electrodes (unilateral or bilateral) contained the leading spike, it was considered an M1 spike. If a combination of S1 and M1 electrodes contained the leading spike, it was considered diffuse S1/M1 onset.

Spike voltages in each channel were determined with the baseline voltage defined as the minimum voltage 20-29 ms before the peak. Spike duration (σ) was calculated by fitting each spike voltage, V, to a Gaussian function (equation 2), where B is the baseline voltage, A is the peak amplitude, t is time, μ is the spike midpoint, and σ is the Gaussian root mean square width (half width at half maximal voltage).

$$V=B+A{e}^{-[{(t-\mu )}^2∕2{\sigma }^2]}$$ Equation 2

Statistical analyses

Statistical analyses were performed using the R 3.0.1 Statistical Package for Windows (R Foundation for Statistical Computing). Delay times were deemed statistically different from zero using a one-sample Wilcox test. Differences in cross-correlation constant (R) between in the preictal period and the spike and interspike periods were compared using a paired t-test. Distributions among fractions of leading spikes in the S1, M1, and S1/M1 brain regions were compared using a χ² test. Two-factor ANOVA tests, with Tukey post-tests, compared the effects of spike type and brain region on voltage and on rank-transformed spike durations (σ). Dunn’s test compared the effects of spike type on M1 to S1 normalized voltage. Bonferroni corrections for multiple comparisons were applied to the P-values from the one-sample Wilcox, paired t-test, and χ² tests. All tests are two-sided, and only P values less than 0.05 (Bonferroni-corrected, if necessary) were considered statistically significant. In the text and in figures, mean values are given ± 5-95% confidence intervals and median values are given with 1^st-3^rd quartile range in parenthesis.

Results

Nonlinear association analysis in SWDs and myoclonic seizures

We used linear and nonlinear association analyses^13;22-24 to determine changes in sensorimotor cortex connectivity and time lags from 1953 ms prior to a SWD until 977 ms after SWD onset. As was seen in corticocortical associations in rat SWDs,¹³ there was a small, but significant, nonlinear component to the coupling with ictal h² significantly higher (0.05 ± 0.01; P ≤ 0.026) using the nonlinear method in all, but one, electrode pair (right aM1 → pM1, P = 0.059).

An EEG segment from a typical SWD is shown in Figure 1B. Because we used PTZ to evoke myoclonic seizures in subsequent experiments, we also separately performed nonlinear analyses on SWDs recorded in the presence of PTZ (SWD + PTZ Supplementary Figure 1A) to confirm that PTZ did not substantially alter SWD connectivity. Figures 1C and Supplementary 1B show the time course of the changes in nonlinear h². In all intrahemispheric and interhemispheric pairings of electrodes, h² started to increase approximately 250 ms before the first SWD spike, continued to rise until the first spike (time 0), and remained elevated through at least 977 ms after SWD onset.

Figures 1D and Supplementary 1C depict the median lag times among the electrode pairings and the differences in lag times are compared in Figures 1E and Supplementary 1D. There was no significant time lag in the preictal period. During the ictal period, S1 preceded aM1 by a median of 3 ms bilaterally (left: 0 – 6 ms; P = 0.003; right: 1– 7 ms; P = 0.003). In the right hemisphere, the S1 preceded pM1 by a median of 2 ms (0 – 5 ms; P = 0.032). Similar results were obtained for the SWD + PTZ discharges (Supplementary Figure 1D). These data showing S1 leading M1 during SWDs are consistent with the findings during SWDs in Wag/Rij rats.¹³

Next, we used linear and nonlinear association analysis to determine sensorimotor connectivity during myoclonic seizures. As with SWDs, nonlinear association analysis of myoclonic seizures produced ictal h² values that were modestly (0.05 ± 0.01), but significantly (P ≤ 0.033), increased compared with linear analysis in all channel combinations, a result showing that there was a small nonlinear component to corticocortical coupling in these seizures.

Figure 2A shows the three types of EEG patterns of associated with myoclonic seizures; namely, those with one, two, or three spikes; the last spike had the highest amplitude and was also always the one associated with the myoclonic jerk (jerk spike). Figure 2B shows changes in nonlinear h² from the preictal state to the myoclonic seizure. The nonlinear association constants increased 300-400 ms prior to the jerk spike, reached a maximum 100-200 ms prior to the jerk spike, and returned to baseline 100-200 ms after the jerk spike.

Figure 2. Corticocortical relationships in myoclonic seizures.

Stereotactic placement of electrodes is the same as in Figure 1A. A) Examples of referential EEG recordings of a myoclonic seizure with one (left), two (middle), and three (right) spikes. The 2-4 Hz slow and wave-spike activity seen before the single-spike myoclonic seizure and after the three-spike myoclonic seizure was seen frequently, but not universally, in the PTZ-treated mice. B-D) Nonlinear association analyses among the electrodes in the left (left) and right (center) hemispheres as well as between the corresponding electrodes in each hemisphere (right, IH) was performed in overlapping windows from 1953 ms prior to the ultimate spike (jerk spike) to 977 ms after the jerk spike. B) Plots of the mean nonlinear association constant (h²) at the start of each time window demonstrates that h² increases in all electrode pairings starting approximately 500 ms prior to the jerk spike. Panel C depicts time delays between the pairs of electrodes; positive delays indicate that the EEG signal of the second electrode (to the right of the arrow) precedes the first electrode (to the left of the arrow). E) The median pre-ictal (t = −1953 to −1855 ms) and ictal (t = −98 to 0 ms) times are shown as a horizontal line within the box plots. During the ictal, but not pre-ictal period, the EEG signal in S1 significantly preceded that of the aM1 by a median of 2 ms (1 - 4 ms) on the left and 4 ms (4 – 5 ms) on the right. Right pM1 and S1 preceded the corresponding electrodes in the left hemisphere by 2 ms (pM1: 0 – 3 ms; S1: 0 – 5 ms). Asterisks mark Bonferroni-corrected P values, ***P < 0.001, **P < 0.01, *P < 0.05, ns = nonsignificant. N = 20 myoclonic seizures from 4 mice.

Figure 2C shows the time course of changes in the median lag times and Figure 2D compares differences in the lag times between the preictal and ictal periods. There was no significant time lag during the preictal period. During the ictal period, S1 led aM1 with a median lag time of 2 ms on the left (1 – 4 ms; P = 0.009) and 4 ms on the right (4 – 5 ms; P = 0.009). For the interhemispheric electrode pairings, right pM1 and right S1 preceded left pM1 and left S1 by 2 ms (pM1: 0 – 3 ms; P = 0.009; S1: 0 – 5 ms; P = 0.042).

Corticocortical association during spike and interspike periods in SWDs and myoclonic seizures

The sliding window analyses in demonstrated that, similar to corticocortical coupling in rat SWDs,¹³ there was only a small nonlinear component (ictal h² increased 0.05 ± 0.01) to the corticocortical association during SWDs and mycolonic seizures in Gabra1^+/A322D mice (Figs 1-2). Therefore, we next used a linear method, cross-correlation analysis to determine the cortical coupling and delay times specifically in the spike and interspike periods. The advantage of this linear method is that it can be used with the short time windows that can encompass a single spike. We manually centered a small time window over each spike and interspike period and calculated the cross correlation association constant (R) at different time lags. Figure 3A shows a typical SWD that is depicted on an expanded time scale in Figure 3B. Compared with the preictal state, the association constants (R) between S1 and both aM1 and pM1 were significantly increased during all four SWD spikes as well as during several of the interspike periods (Fig 3D).

Figure 3. Corticocortical relationships during the spike and interspike periods of SWDs.

Spontaneous SWDs were recorded with electrodes placed in the same regions as in Figure 1A. Panel A shows the EEG during a SWD. The region highlighted in the green box is depicted on an expanded time scale in panel B to show the first four spikes (Sp1-Sp4, yellow outline) and interspike periods (I1-I5). The vertical dashed lines through the spikes mark the leading spikes in the left (red) and right (blue) hemispheres and the lagging spikes in each hemisphere are marked by arrowheads that are outlined red (right) or blue (left) and filled with green to indicate a delay from S1 to aM1 or red to indicate a delay from S1 to pM1 . C) Cross-correlograms of the Sp4 spike from panel B among electrodes from the left (top) or right (bottom) hemispheres are shown with colored vertical dashed lines placed at the maximum of the corresponding R and projected to the time lag on the x-axis. D) Mean changes in R (ΔR ± 5-95% confidence intervals) between the pre-ictal period and either each spike or interspike period (Sp1-Sp4, I1-I5) among electrodes in the left and right hemisphere and between the corresponding electrodes in each hemisphere (interhemispheric, IH). Statistical significance is indicated by asterisks at the bottom of the plot. Panel E depicts the median lag times (error bars represent 25^th and 75^th percentiles); asterisks at the bottom of the graph indicate statistical significance relative to no lag (0 ms). In both hemispheres, there were significant lags in aM1 relative to S1 (green) in Sp2, Sp3, and Sp4. F) Graphs depict the fraction of spikes originating exclusively in S1, exclusively in M1, or diffusely in S1 and M1 (S1/M1). The portions of the S1 bars colored red, blue and purple represent the fraction of S1 spikes originating unilaterally from the left, right, and bilateral S1 regions, respectively. Bonferroni-corrected P values: ***P < 0.001, **P < 0.01, *P < 0.05, ns = nonsignificant. N = 35 SWDs from 4 mice.

Visual inspection and the cross-correlograms (Fig 3C) reveal that spikes at S1 precede M1. Importantly, significant lag times were found only during spikes, and not the interspike periods (Fig 3E). The median lag times between left S1 and aM1 were 0 ms for spike 2 (0 – 1 ms, P = 0.036), 2 ms for spike 3 (0 – 3 ms, P = 0.001) and 3 ms (0 – 5 ms, P < 0.001) for spike 4. On the right, the S1 to aM1 lag times were 1 ms for spike 2 (0 – 2 ms, P = 0.007), 2 ms for spike 3 (1 – 4 ms, P < 0.001) and 2 ms for spike 4 (1 – 6 ms, P < 0.001). There were no significant delay times interhemispherically. Similar results were found for SWDs + PTZ spikes (Supplementary Figure 2A-B).

The preponderance of SWD spikes originated either exclusively in S1, or diffusely in S1/M1; only a small minority began in M1 (Fig 3F). There was no significant difference between the fraction of S1 and S1/M1 spikes. A similar result was obtained for the SWD + PTZ spikes although, for spike 2, there was a significantly greater fraction detected at S1 than S1/M1 (Supplementary Figure 2D).

We next determined the coupling among the aM1, pM1, and S1 electrodes during the spike and interspike periods of myoclonic seizures. An EEG during a typical myoclonic seizure is shown in Figure 4A and on an expanded time scale in Figure 4B. Relative to the preictal period, R was significantly increased in both spikes 1 and 2 bilaterally as well as the second interspike period on the left (Fig 4D).

Figure 4. Corticocortical relationships during the spike and interspike periods of myoclonic seizures.

PTZ-evoked myoclonic seizures were recorded with electrodes placed in the same locations as in Figure 1A. A) EEG during a typical myoclonic seizure that is depicted on an expanded time scale in (B) showing the two spikes (Sp1, Sp2, yellow outline) and interspike periods (I1-I3) with dashed vertical lines (red = left; blue = right) placed through the leading spikes and arrowheads marking the trailing spikes outlined in red (right) or blue (left) and filled with green to indicate a delay from S1 to aM1 or red to indicate a delay from S1 to pM1. C) Cross-correlograms of the Sp1 and Sp2 spikes among electrodes from the left (top) or right (bottom) hemispheres are shown with colored vertical dashed lines placed at the maximum of the corresponding R and projected down to time lag producing the maximal R. D) Mean changes in R (ΔR ± 5-95% confidence intervals) between each spike/interspike period (I1-I3, Sp1,Sp2) and the pre-ictal period among electrodes in the left and right hemisphere and between the corresponding electrodes in each hemisphere (interhemispheric, IH). Statistical significance is indicated by asterisks at the bottom of the plot. Panel E depicts the median lag times (error bars depicting 25^th and 75^th percentiles) from the cross correlations shown in panel D; asterisks at the bottom of the graph indicate statistical significance relative to no lag (0 ms). F) Graphs depict the fraction of spikes originating in S1 or M1 brain regions or diffusely throughout the sensorimotor cortex (S1/M1). The portions of the S1 bars colored red, blue, and purple represent the fraction of S1 spikes originating from the left, right, and bilateral S1 regions, respectively. G) Comparison of the fraction of spikes initiating exclusively in S1 between SWDs (from Fig 3) and myoclonic seizure spikes. Bonferroni-corrected P values: ***P < 0.001, **P < 0.01, *P < 0.05, ns = nonsignificant. N = 33 myoclonic seizures form 4 mice.

Visual inspection and cross-correlation (Fig 4C) revealed delays in the peaks of spike 1 and spike 2 in aM1 and pM1 relative to S1. In spike 1, right S1 preceded ipsilateral aM1 and pM1 by 2 ms (Fig 4E, 1 – 3 ms, P < 0.001). In spike 2, S1 preceded aM1 and pM1 by 2 ms on the left (aM1: 0 – 3 ms, P = 0.002; pM1: 1 – 5 ms, P < 0.001) and 5 ms on the right (aM1: 4 – 6 ms, pM1: 2 – 5 ms, P < 0.001). In contrast to SWDs, there was a significant delay between the corresponding interhemispheric pairs of S1 and pM1 electrodes.

Figures 4F-G demonstrate that, in contrast to SWDs, significantly larger fractions of the myoclonic seizure spikes were first detected exclusively in S1 (spike 1: 82%; spike 2: 87%) compared with M1 (spike 1: 0%, P < 0.001; spike 2: 3%, P < 0.001) and S1/M1 (spike 1: 18%, P = 0.009; spike 2: 10%, P < 0.001). Also, unlike SWDs, there was a significantly higher fraction of S1 spikes originating in the right rather than the left hemisphere (spike 1 P = 0.023; spike 2 P = 0.002).

Spike duration and voltage in SWDs and myoclonic seizures

Overlaid EEG traces (Fig 5A) demonstrate the spike durations and voltages in a typical SWD and myoclonic seizure. There was no significant effect of electrode location (aM1, pM1, S1) on spike duration (σ, P = 0.459). The median duration of the first SWD spike was 12 ms (8 -17 ms), which was slightly longer than the median duration of SWD spikes 2-3 and myoclonic seizure spikes 1-2 (range 8 ms – 9 ms, P ≤ 0.015 ). The lack of a substantial effect of spike type on duration indicates that the differences between SWD and myoclonic seizure spike focality (Fig 4G) do not result from SWDs having broader spikes obscuring leading and trailing waveforms.

Figure 5. Voltage changes in SWDs and myoclonic seizures.

Panel A shows EEG samples from spontaneous SWDs (spikes 1-4, N = 33) and PTZ-evoked myoclonic seizures (spikes 1 and 2, N = 23). The EEG channels from the left anterior motor cortex (aM1, blue), posterior motor cortex (pM1, green), and somatosensory cortex (S1, red) are overlaid to demonstrate the similarities in spike duration and the differences in voltage. The plots in panel B depict the mean voltages (± 5-95% CI) from SWD spikes 1-4 and from myoclonic seizure spikes 1-2 in the left and right hemispheres. Graph bars are colored in sections corresponding to the voltages in aM1 (blue), pM1 (green), and S1 (red). A two-factor ANOVA compared the effects of spike (SWD spike 1-4 and myoclonic spikes 1-2) and brain region (aM1, pM1, S1) on voltage. Although there was no significant interaction between spike type with brain region on voltage (P ≥ 0.177), there were large independent effects (P < 0.001). In both hemispheres, myoclonic seizure spike 2 had higher voltage than SWD spikes 1-4 and myoclonic seizure spike 1 (P < 0.001). Myoclonic spike 1 also had greater voltage than the first two SWD spikes (P < 0.004) and, on the right, had greater voltage than the third and fourth SWD spike (P ≤ 0.011). SWD spike 1 had significantly lower voltage than spike 3 or 4 (P ≤ 0.007). For both SWD and myoclonic seizure spikes, the voltage in aM1 was significantly greater than that in pM1 or S1 (P ≤ 0.005). Panel C depicts the median voltage ratio between aM1 and S1 (top) and between pM1 and S1 (bottom). The ratio between pM1 and S1 was greater in myoclonic spike 2 than the first two SWD spikes in both hemispheres and the ratio between aM1 and S1 was greater in myoclonic spikes 1 and 2 than the first two SWD spikes on the left. Differences in the aM1 to S1 voltage ratios between the myoclonic and SWD seizure spikes on the right were not statistically significant (# P = 0.081).

Spike voltage depended on both electrode location and spike type (Fig 5B). Importantly, myoclonic seizure spike 2 (jerk spike) had higher voltage than all other spikes. For both SWDs and myoclonic seizures, voltage in aM1 was significantly greater than that of pM1 (P <0.005) and S1 (P < 0.001) and S1 had greater voltage than pM1 (P = 0.002). Similar results were obtained when the myoclonic seizure spikes were compared with the SWDs + PTZ spikes (Supplementary Figure 3A).

To control for effects spike amplitude variability on potentially significant interactions between spike type and electrode location on voltage, we also compared normalized spike voltages (aM1/S1 and pM1/S1, Fig 5C). Myoclonic seizure spike 2 (the jerk spike) had a greater pM1/S1 voltage ratio than SWD spikes 1 and 2 in both hemispheres (P ≤ 0.020), and a greater aM1/S1 voltage ratio in the left hemisphere (P = 0.005). A similar trend was seen for the SWD + PTZ spikes, although the results are only statistically significant when comparing pM1/S1 in myoclonic seizure spike 2 and SWD spike 1 on the right (Supplementary Figure 3B).

Discussion

SWDs and myoclonic seizures activate overlapping networks in sensorimotor cortex

Identifying the brain networks that participate in different seizure types in generalized epilepsy syndromes has implications for the development of network-specific therapies (e.g. neurostimulation).^6-9 Here, we determined the direction and time course of activation of the sensorimotor cortex during SWDs and myoclonic seizures. During both seizure types, S1 activates two to five milliseconds prior to M1, a similar time lag as that seen in a rat model of SWDs.¹³ To our knowledge, this is the first report quantifying the spread of cortical activation during myoclonic seizures, and the first that measured coupling specifically during spike and interspike periods in either seizure type.

Because absence and myoclonic seizures are associated with different behaviors, we anticipated that the time course and direction of cortical activation in these two seizure types would be different – possibly with a leading role of M1 in myoclonic seizures. However, our observation that both seizure types activate sensorimotor cortex with a similar dynamics suggest that they both engage an overlapping seizure network within the cortex.

With the limited number of electrodes used here, we cannot determine whether SWDs or myoclonic seizures initiate in S1 or in another brain region that then activates S1 before M1. Time-lagged correlation between two electrodes simply indicates that measurement of the waveforms at the first electrode predicts the waveforms at the second electrode and does not necessarily imply that neuronal activity at the leading site drives activity at the lagging site. However, our measurements of S1 to M1 dynamics during SWDs and myoclonic seizures closely resemble those seen in SWDs in absence epilepsy rats¹³ which were subsequently found to initiate in S1.^13-15 Moreover, because EEG spikes associate with the firing of cortical neurons,²⁶ our finding that the S1 to M1 time lag is present during spikes and not the interspike period suggests that M1 activation is dependent on S1 neuronal firing. Therefore, we think it is likely that both SWDs and myoclonic seizures will be found to initiate in S1 barrel fields.

In addition to precluding an unambiguous determination of the SWD and myoclonic seizure initiating site, the limited spatial resolution also prevents us from identifying brain regions other than M1 that are secondarily activated and shape the behavioral phenotype. Possibly, the secondary activation in a subcortical region in myoclonic seizures, but not SWDs could contribute to the motor behavior. Future experiments with electrodes simultaneously placed in S1 barrel cortex and subcortical brain regions such as thalamic relay nuclei and basal ganglia will determine whether S1 activation leads these subcortical sites and whether differential activation of subcortical regions distinguishes SWDs from myoclonic seizures.

Differential activation of the thalamocortical network in SWDs and myoclonic seizures

Although our study demonstrates that SWDs and myoclonic seizures engage the sensorimotor cortex with similar dynamics, we also found that they differ in hemispheric lateralization, spike focality and voltage. These differences in sensorimotor cortex activation may contribute to the behavioral differences exhibited in these two types of seizures.

Our study showed that although SWDs were not lateralized to the left or right hemisphere, myoclonic seizures exhibited a greater fraction of spikes starting in the right S1 than the left. Our finding of myoclonic seizure lateralization onset to right S1 in Gabra1^+/A322D mice is consistent with previous reports showing EEG lateralization in human JME,²⁷ rodent audiogenic seizures,^28;29 and rodent sensory processing.^30-32

In addition, we found that myoclonic seizures initiate more focally within the sensorimotor cortex than SWDs. The majority of spikes in myoclonic seizures were initially detected exclusively in S1 electrodes whereas a substantial fraction of SWD spikes were detected in S1 electrodes and at least one M1 electrode. This observation suggests that although both SWD and myoclonic seizure spikes spread from S1 to M1 with a similar timecourse, the initial activation in S1 is more diffuse in SWD spikes than myoclonic seizure spikes.

Finally, we determined that myoclonic seizure spikes have a higher voltage than SWD spikes and a greater relative distribution of voltage in M1 than S1 compared with the first one to two SWD spikes. Importantly, we found that the second myoclonic seizure spike, the jerk spike, has the highest absolute voltage and the greatest relative voltage in M1 relative to S1. Because voltage differences reflect different densities of synchronously firing neurons,³³ this latter result suggests the myoclonic jerk is associated with a large number of synchronously firing neurons in M1. Therefore, in addition to the possible activation of different subcortical sites as described above, the distinct behaviors seen in SWDs and myoclonic seizures may also result from differences in the recruitment of synchronously firing neurons within S1 and M1.

In Figure 6, we present a model depicting different numbers of synchronously firing cortical neurons that is consistent with our findings. Based on findings in rat SWDs,^14;23;34 we placed the synchronously firing neurons in the lower cortical layers. At onset, almost all myoclonic seizure spikes were detected first in S1, a finding consistent with synchronously firing neurons occurring focally in S1 at the start of the spike. In contrast, in SWDs, there was no significant difference between the fraction of spikes appearing first in S1 and those that were detected in S1/M1. This suggests that some spikes originate with synchronous firing in S1, but others occur with synchronous neurons in S1 and in some regions of M1. Abundant cortico-cortial connections between S1 and M1, important for coordinating sensory input with motor function,^35-37 may serve to synchronize S1 and M1 neurons. In both SWD and myoclonic seizure spikes, there is widespread involvement of synchronously-firing neurons in S1 an M1 two to five milliseconds after onset.

Figure 6. Neuronal activation model in SWDs and myoclonic seizures.

The gray crescents represent coronal sections through the S1 and M1 cortices. Blue arrows represent abundant S1 to M1 corticocortical connections. Triangles within the cortices depict populations of pyramidal neurons; red triangles indicate synchronously firing neurons and green triangles indicate relatively inactive neurons. Based on findings in rat SWDs, we placed the synchronously firing neurons in the lower cortical layers. The top row depicts seizure onset and the bottom row the spread of the seizure 2-5 ms after onset. Some SWDs (left column) and almost all myoclonic seizures (right column) start exclusively in S1 with populations of synchronously firing neurons only within this cortical region. After spread of the spike (2-5 ms later) synchronously firing neurons are present in M1. Other SWDs (middle column) have synchronously firing neurons in S1 and M1 at onset. The greater absolute voltage of myoclonic seizure spikes is represented by the increased number of synchronously firing neurons in the myoclonic seizure and the increased M1 to S1 ratio in myoclonic seizures is depicted by a higher ratio of synchronously firing neurons in M1 relative to S1.

In conclusion, we demonstrated that both SWDs and myoclonic seizures are associated with a similar pattern of activation of sensorimotor cortex in which spikes in S1 lead those in M1 by 2-5 ms. This similar pattern of activation suggests that overlapping brain networks in sensorimotor cortex are involved in both seizures types. In addition, we identified differences in spike focality and voltage that may contribute to the distinct behaviors associated with absence and myoclonic seizures.

Supplementary Figure 1: Corticocortical relationships in SWDs in the presence of PTZ: A) Example of a referential EEG of a SWD recorded in the presence of PTZ. The origin (t = 0s) of the time scale at the top of the panel is at the time of the first spike. D) Nonlinear association analyses was performed in overlapping windows starting from 1953 ms prior to first spike to 977 ms after the first spike among the electrodes in the left (left) and right (center) hemispheres as well as between the corresponding electrodes in each hemisphere (right). C) Plots of the mean nonlinear association constant (h²) at the start of each time window demonstrates an increase in association among all electrodes at the time of the first spike that persists for at least the first 977 ms of the SWD. Panel D depicts the time delays between the pairs of electrodes at the start of each time window. Positive delays indicate that the EEG signal in the second electrode (to the right of the arrow) precedes the first electrode (to the left of the arrow). E) The median pre-ictal (t = −1953 to −1855ms) and ictal (t = 244 to 342 ms) times are shown as a horizontal line within the box plots and the 25th to 75th percentile delay times are depicted by the rectangle length. During the ictal, but not pre-ictal period, the EEG signal in S1 significantly preceded that of aM1 both in the left (median 4 ms; 1st-3rd quartile 2 - 5 ms) and right (median 3 ms; 1st-3rd quartile 2 – 4 ms) hemispheres. In addition, in the right hemisphere, the EEG signal in S1 significantly preceded that of the pM1 (median 2 ms; 1st-3rd quartile 1 – 3 ms). There was no significant delay in EEG signal between corresponding electrodes in the left and right electrodes. Asterisks = Bonferroni-corrected P values, *** P < 0.001, ** P < 0.01, * P < 0.05, ns = nonsignificant. N = 14 SWD + PTZ from 4 mice.

Supplementary Figure 2: Corticocortical relationships during the spike and interspike periods of SWDs in the presence of PTZ. SWDs were recorded in the presence of PTZ with electrodes placed in the left and right anterior motor cortex (aM1), posterior motor cortex (pM1), and somatosensory cortex (S1), A) Mean changes in R (ΔR ± 5-95% confidence intervals) between the pre-ictal period and each spike (Sp1-Sp4, yellow) or interspike period (I1-I5) among electrodes in the left and right hemisphere and between the corresponding electrodes in each hemisphere (interhemispheric, IH). Statistical significance is indicated by asterisks at the bottom of the plot that are colored the same as the pairs of electrodes. Relative to the pre-ictal period, R was significantly increased between S1 and both aM1 (green) and pM1 (red) in spikes Sp3 and Sp4 as well as interspike period, I4 in both hemispheres and between S1 and aM1 during spikes Sp1 and Sp2 and interspike, I2, in the left hemisphere. In both the left and right hemispheres, R was significantly increased between S1 and aM1 (green) and pM1 (red) in all spikes (Sp1-Sp4) as well as the I3 interspike period. Interhemispherically, association constants were significantly increased between aM1 electrodes in Sp1 and Sp4, between pM1 electrodes in Sp3 and Sp4, and between the S1 electrodes in Sp2 and Sp4. Panel B depicts the median lag times (error bars represent 25^th and 75^th percentiles) from the cross correlations shown in panel A; asterisks at the bottom of the graph indicate statistical significance relative to no lag (0 ms). In both hemispheres, there were significant lags in aM1 relative to S1 (green) in Sp3, and Sp4. There was no significant lag in the interspike periods or between the any of the three pairings of interhemispheric electrodes. C) Graphs depict the fraction of spikes originating in exclusively S1, exclusively M1, or diffusely in S1 and M1 (S1/M1). The portions of the S1 bars colored red, blue, and purple represent the fraction of spikes originating from the left, right and bilateral S1 regions. Bonferroni-corrected P values: ***P < 0.001, **P < 0.01, *P < 0.05, ns = nonsignificant. N = 16 SWDs from 4 mice.

Supplementary Figure 3: Spike voltages of SWDs in the presence of PTZ . The plots in panel A depict the mean voltages (± 5-95% CI) from SWD + PTZ spikes 1-4 (N = 16) and from myoclonic seizure spikes 1-2 (N = 23) in the left and right hemispheres. Both the SWDs and myoclonic seizures were recorded in the presence of PTZ and the plots of the myoclonic seizures are the same as in Figure 5. Graph bars are colored in sections corresponding to the voltages in aM1 (blue), pM1 (green), and S1 (red). A two-factor ANOVA compared the effects of spike (SWD spike 1-4 and myoclonic spikes 1-2) and brain region (aM1, pM1, S1) on voltage. In both hemispheres, myoclonic seizure spike 2 had higher voltage than SWD spikes 1-4 and myoclonic seizure spike 1 (P < 0.001). In the left hemisphere, myoclonic seizure spike 1 had greater voltage than SWD spike 1 and, in the right hemisphere, myoclonic seizures spike 1 had greater voltage than SWD spikes 1 and 2. The voltage in aM1 was significantly greater than that in pM1 and it was greater than S1 in the left hemisphere. Panel B depicts the median voltage ratios between aM1 and S1 (top) and between pM1 and S1 (bottom). The ratio between aM1 and S1 was significantly greater in myoclonic spike 2 than SWD spike 1 (P = .047) on the right and there was a trend toward increased aM1 to S1 ratios between myoclonic seizure spike 1 and SWD spike 1 on the left (#, P = 0.082) and on the right (#, P = 0.053).

Key Points.

• In sensorimotor cortex, both SWDs and myoclonic seizures are first detected in S1 and propagate to M1 with similar lag times of 2-5 ms.

• In SWDs and myoclonic seizures, lag times between S1 and M1 are only present during spikes and not in the interspike periods.

• A substantial majority of myoclonic seizure spikes are first detected only in S1 electrodes and thus are more focal than SWD spikes.

• Myoclonic seizure spikes have a greater voltage than SWD spikes and a larger relative voltage over M1 than the first one to two SWD spikes.