2.1.

Animal Preparation

All mice preparation and experimental procedures followed the institutional guidelines of the Korea Institute of Science and Technology (KIST). Twelve adult female ICR mice [ $8–10\text{weeks}$ , $31.2\pm 3.53\mathrm{g}$ $(\text{mean}\pm \mathrm{SD})$ , Orientbio Inc., Seongnam, South Korea] were anesthetized with Avertin (2%, $20\mu \mathrm{l}∕\mathrm{g}$ body weight) and mounted on a stereotaxic apparatus (David Kopf Instruments, Model 902, Tujunga, California). Following a $1\text{-}\mathrm{in.}$ midline incision, the scalp was opened and held with microclamps. For the EEG recording, two screw electrodes with an impedance of $20\mathrm{k}\Omega$ ( $1∕32\mathrm{in.}$ diam) were implanted on the skull (anterior-posterior: $+2.0\mathrm{mm}$ , lateral: $\pm 2.5\mathrm{mm}$ with respect to the bregma point). A ground electrode was implanted on the cerebellum $1\mathrm{mm}$ dorsal to the $\lambda$ point. A small amount of dental cement (Ketac Cem Easymix, 3M ESPE Worldwide) was applied to secure the electrodes on the skull.

In order to drive seizures, we administered intraperiotoneally $\gamma$ -butyrolactone [(GBL), $500\mathrm{mg}∕\mathrm{kg}$ body weight, Sigma, St. Louis, Missouri] and 4-aminopyridine [(4-AP), $10\mathrm{mg}∕\mathrm{kg}$ body weight, Sigma] to six mice in each group. The GBL and 4-AP are two of the frequently used seizure inducing drugs for absence seizure and tonic-clonic seizure, respectively. The applied dosages of GBL and 4-AP were selected based on similar protocols used to generate seizures in other in vivo mouse studies.⁹ In our study, we increased administration of GBL to enhance the electrical and hemodynamic responses, and the dosage of the 4-AP was reduced because unexplained sudden deaths were observed in mice pretreated with Avertin in our preliminary tests.

2.2.

NIRS Setup

A two-wavelength eight-channel frequency-domain NIRS (Imagent, ISS, Champaign, Illinois) was applied to monitor the cerebral hemodynamic responses. The light sources of Imagent are modulated at high frequency of $110\mathrm{MHz}$ , and the modulation amplitude (AC) is provided in addition to the intensity. Each channel contains two wavelengths (690 and $830\mathrm{nm}$ ), and each laser diode was turned on and off sequentially under time-based multiplexing. Two optical fibers with a $400\text{-}\mu \mathrm{m}$ core diameter and 0.39 numerical aperture (NA) (FT-400EMT, Thorlabs, Newton, New Jersey) carrying different wavelengths were carefully colocated on each source point by using a homemade manipulating arm. To collect diffuse photons, two detector fibers were tied and placed at the midpoint between the bregma and $\lambda$ points, each of which were relayed to two photomultiplier tubes. We used two detectors to enhance the signal-to-noise ratio (SNR). Eight source optical fibers were arranged in a circle around the mouse skull, as described in Fig. 1 , and the source-detector distances were maintained at $\sim 0.5\mathrm{cm}$ from the detector. Modulation of light, which is the ac signal in Imagent system, was collected at a sampling rate of $31.25\mathrm{Hz}$ . Phase shifts between each source-detector pair were monitored continuously during the measurement, and the channel-to-channel variance was not significant. All experiments were performed in a dark and quiet room.

Fig. 1

(a) Arrangement of optical sources (S1–S8, two circles indicate separate light sources of 690 and $830\mathrm{nm}$ ), two detectors ( ${\mathrm{D}}_1$ and ${\mathrm{D}}_2$ ) and EEG electrodes (V, R, and G, which stand for active, reference, and ground electrode, respectively) and (b) photo of the mouse ready for data acquisition.

2.3.

EEG Setup

A bipolar, one-channel EEG was used in this study. The raw signals from the skull EEG were amplified 10,000 times using an ac preamplifier (Model QP511, Grass Technologies, West Warwick, Rhode Island) and collected with a data-acquisition device (Digidata 1440A, Axon Instruments Inc., Union City, California) at a sampling rate of $1\mathrm{kHz}$ . Total recording time was $\sim 1\mathrm{h}$ for all cases. Additionally, to reduce ambient noise, we connected the stereotaxic frame and EEG system to a common ground. We recorded the NIRS and the EEG signals simultaneously and coregistered a marker for each signal so that direct temporal comparison would be possible.

2.4.

Signal Processing of EEG Data

EEG signals were bandpass filtered with a 10th-order zero-phase Butterworth filter with a cutoff frequency of 0.1 and $50\mathrm{Hz}$ . Epileptic episodes were identified by detecting epileptiform discharges.¹⁰ The most common epileptiform discharges induced by GBL and 4-AP are known to be spike-and-wave discharge complexes and spike train, respectively.¹¹ In our study, other epileptiform discharges, such as polyspike or polyspike-and-slow-wave complexes, were also observed. The EEG signals with a repetitive high-amplitude activity were judged as epileptic events. To avoid the false positive, we did not count any patterns that were found in the predrug state.

2.5.

Signal Processing of NIRS Data

Two detector signals were averaged to increase the SNR, and the changes of ac signal amplitudes were used in our analysis. The temporal patterns of ac and dc were almost identical, but ac has much less irregular movement compared to dc. The concentration of hemoglobin was calculated based on the modified Beer–Lambert law¹²

The rhythmical activities of $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ and $\Delta \left[\mathrm{HBr}\right]$ were analyzed using various spectral analysis methods, such as fast Fourier transform. The phase relationship between oxy- and deoxyhemoglobin was further analyzed by evaluating instantaneous phase using Hilbert transform,¹³ which is defined by

Spatial correlation between hemodynamic signals from eight NIRS channels was evaluated using a cross-correlation function called “xcorr” in Matlab, which is defined by

3.1.

Drug-Induced Seizures in Mouse EEG

As the first step in isolating the epileptic moments in vivo in mice, we performed cortical EEG recordings of mice, simultaneously with NIRS recordings. In general, the epileptic activities in GBL-treated mice were episodic lasting a few seconds, whereas the ictal activities of 4-AP-treated mice were successive, and even after the cessation of ictal activities, intermittent pulses were observed in the EEG. In terms of spectral properties, it was found that the EEGs of GBL-treated mice had a reduced power in all frequency-bands-compared to the anesthetic baseline state [Fig. 2 ]. On the other hand, the EEGs of 4-AP-treated mice were slightly enhanced after drug administration for all frequency bands [Fig. 2]. Therefore, it is clear that the electrical activities in brain caused by those two drugs differed; from this we infer that the seizure-inducing mechanism was also different. The epileptiform episodes were successfully identified in five out of six GBL-treated mice and in four out of six 4-AP-treated mice.

Fig. 2

(a) Histogram of the mean relative power spectrum of EEGs before (black bar) and after administration of GBL (gray bar). Error bars represent the standard deviation. Note the significant decrease in all the EEG bands ( ${}^{**}$ , $p<0.01$ ; ${}^{*}$ , $p<0.05$ , Kruskal-Wallis nonparametric one-way ANOVA test). The EEG data were normalized in all mice. The total power between 0.1 and $250\mathrm{Hz}$ during baseline measurements was used as the normalization factor. (b) The same histogram for the mice administered with 4-AP. Note a significant increase in the $\gamma$ band. The power spectrum tends to increase in $\delta$ , $\theta$ , and $\alpha$ bands also, but not significantly.

Various EEG patterns of GBL-treated mice infrequently appeared during seizure events. Figure 3 show the frequently appearing epileptiform discharge in GBL-treated mice. The onset and termination of ictal periods were recognized as abrupt changes in the amplitude and frequency, mostly with the appearance of spikes or sharp waves or a combination of the two. The epileptiform discharge started to appear about $2–4\min$ after the administration of GBL and appeared irregularly, as summarized in Table 1 . The morphological feature of epileptiform discharge, and frequency, were different between each mouse. The duration between onset and termination of the ictal discharges was $2.5\pm 0.9\mathrm{s}$ and the peak-to-peak amplitude during ictal discharge was two to five times higher than those during the interictal period.

Fig. 3

Typical patterns of epileptiform discharges observed in (a) GBL-treated mice. Various signals such as sharp wave, spike-and-wave complexes, polyspike complex, and polyspike-and-wave complex were observed. (b) Baseline (left) and ictal tonic activity (right) in 4-AP-treated mice.

Table 1

Summary of epileptic responses recorded by EEGs for GBL-treated mice.

The EEG activities after the 4-AP injection were characterized by a long period of repetitive spikes, initially low in amplitude but increasing as time passed [Fig. 3] three out of six mice showed convulsive EEG within $3\min$ after the dose. One mouse experienced epileptiform discharges $30\min$ after the dose. The repetition rate of spikes was approximately 0.5 or $1\mathrm{Hz}$ . The interictal periods could not be clearly defined on the EEGs of 4-AP-treated mice because of the intermittent appearance of paroxysmal discharge. In the case of 4-AP, the peak-to-peak amplitudes during a seizure ranged from 13 to 25 times higher than the standard baseline signals, as summarized in Table 2 . Sometimes, strongly synchronized spikes with amplitudes of nearly 50 times higher than other spikes were observed.

Table 2

Summary of epileptic responses in the EEGs of 4-AP-treated mice.

3.2.

Systemic Hemodynamic Responses to Seizure Measured by NIRS

Figure 4 illustrates the systemic changes in $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ and $\Delta \left[\mathrm{HBr}\right]$ after the administration of seizure-inducing drugs. The long-term responses in hemodynamic concentrations were notably different for each drug applied: in general, mice treated with GBL had a slow but large increase in $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ ; on the other hand, mice treated with 4-AP had multiple sharp inflows without long-term drifts.

Fig. 4

Systematic hemodynamic responses after administration of (a) GBL and (b) 4-AP. The vertical line represents the time of i.p. injection. $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ and $\Delta \left[\mathrm{Hbr}\right]$ are given in solid and dotted lines, respectively. The scale bar in $y$ -axis is $100\mu \mathrm{M}$ . All data were obtained by averaging hemodynamic responses from eight regions of the brain. Note the apparently different hemodynamic patterns with time between these two drug groups.

The systemic hemodynamic response records show diverse patterns between each mouse even when treated with the same drug. For example, four out of six GBL-treated mice experienced global increases in $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ (mouse 1, 2, 4, and 5): Mouse 3 died $20\min$ after dose and the $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ increase was observed only during the first several minutes. Mouse 6 did not show any significant systemic responses. In 4-AP-treated mice, fast and sporadic blood inflow was observed in four out of six mice (mice 7–10).

3.3.

Spatial Correlation of Hemodynamic Signals during Seizure

The hemodynamic data obtained in different regions of the mouse brain were analyzed to extract information on the dynamical correlation between two regions. The degree of spatial correlation during epileptic events was compared to the values during baseline. The level of spatial correlation was obtained by calculating the cross-correlation coefficient of $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ at different positions. For eight NIRS channels, the cross-correlation coefficients for 28 pairs $\left({}_{8}C_2\right)$ were calculated every minute. The mean and standard deviation of the spatial correlation are plotted with respect to time for each mouse along with the seizure occurrence caused by GBL and 4-AP in Figs. 5 and 5 , respectively. Neither significant change of spatial correlation nor correlation to the epileptic events was observed in GBL-injected mice [Fig. 5]. The spatial correlation of hemodynamics increased significantly after injection of drug compared to the baseline in the 4-AP group, except mouse 7 (paired $t$ -test, $p\text{-value}<0.05$ ). In addition, it was observed that the synchronized behavior of hemodynamic flow precedes the cease of epileptic episodes [mouse 8 and 9 in Fig. 5].

Fig. 5

Mean (black, thick solid lines) and standard deviation (black, thin solid lines) of the spatial correlation $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ for each mouse plotted with seizure activity (red, dotted line) in (a) GBL and (b) 4-AP groups. The ictal ratio is the period of ictal activity per minute. The spatial correlation was obtained by calculating the cross-correlation coefficient of $1\min$ $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ time series at different two positions. Prior to cross-correlation calculation, each $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ was detrended by removing the systemic signals. No significant responses observed in scalp EEG in mouse 1, 7, and 12. The spatial correlation coefficients significantly increased in 4-AP group $(p<0.05)$ ; however, no significant change was observed in the GBL group. (Color online only.)

3.4.

Hemodynamic Responses to GBL-Induced Seizure

The durations of GBL-induced seizure were $<2\mathrm{s}$ in most of the cases. We analyzed each hemodynamic change for each epileptic event. In regard to hemodynamic response to seizure, five distinct patterns can be classified: “no response,” “wash-out” ( $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]\uparrow$ and $\Delta \left[\mathrm{HBr}\right]\downarrow$ ), “joint-decrease” ( $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]\downarrow$ and $\Delta \left[\mathrm{HBr}\right]\downarrow$ ), “joint-increase” ( $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]\uparrow$ and $\Delta \left[\mathrm{HBr}\right]\uparrow$ ), and deactivation ( $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]\downarrow$ and $\Delta \left[\mathrm{HBr}\right]\uparrow$ ), as schematically shown in the inner box in Fig. 6 . Sample traces of each case were shown in Supplements 1–4. Prior to categorizing hemodynamic responses, a zero-phase Butterworth high-pass filter with a cutoff frequency of $0.01–\mathrm{Hz}$ was applied to eliminate any drift. 40–70% of epileptic responses were not accompanied by any hemodynamic responses. After each epileptic event, the subsequent hemodynamic signals were averaged. Mouse 1 was excluded in this process because the epileptic episodes were hardly to be distinguishable in the EEG recording. It was observed that the joint decrease was the dominant hemodynamic response (37–52%), and the wash-out responses were observed (0–17%) third. The opposite pattern of wash-out, i.e., the de-activation response was observed only in mouse 6 with a significant ratio (37.5%). We have not noted any single case of joint increase.

Fig. 6

(a) Number of various hemodynamic responses counted for $1\min$ only when epileptic episodes appear in the GBL group. The small box inside of the graph illustrates how we defined the type of hemodynamic responses. In the early period after the dose, no response or joint decrease was dominant. Wash-out response was observed additionally in $8–15\min$ after the dose. A simultaneous decrease in oxyhemoglobin and increase in deoxyhemoglobin responses appeared at $12\min$ after the dose and became a dominant response in the late period after $20\min$ after the dose. (b) The averaged temporal phase relationship between $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ and $\Delta \left[\mathrm{Hbr}\right]$ of the GBL-treated (solid line) and 4-AP-treated (dashed line) mice groups. Their standard deviation for each data point is represented by a dotted line. The raw data were filtered by the zero-phase Butterworth $0.05–0.2\mathrm{Hz}$ bandpass filter to evaluate the LFO range, and the Hilbert transform was applied. Then, the phase difference between $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ and $\Delta \left[\mathrm{Hbr}\right]$ was obtained by subtracting ${\theta }_{\mathrm{Hbr}}$ from ${\theta }_{\mathrm{Hb}{\mathrm{O}}_2}$ as described in Eq. 4. The ninth polynomial fitting was performed for all data. Note that the in-phase relationship is observed in GBL-treated mice while an out-of-phase relationship is maintained in 4-AP-treated mice.

Figure 6 shows a summary of hemodynamic responses of all GBL-treated mice with respect to time. For the first $5\min$ after the dose, neither epileptiform discharge nor hemodynamic response was observed for all cases. Afterward, various hemodynamic responses were recorded, such as “no response,” “joint-decrease,” “wash-out,” and “de-activation.”

3.5.

Phase Relationship between $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ and $\Delta \left[\mathrm{HBr}\right]$

The relative phase changes between oxy- and deoxyhemoglobin concentrations in the frequency band of low-frequency oscillation (LFO) were investigated. $180\mathrm{deg}$ out-of-phase oscillations in $\Delta \left[\mathrm{Hb}{\mathrm{O}}_2\right]$ and $\Delta \left[\mathrm{HBr}\right]$ are typically recorded⁸ and in-phase changes are associated with modifications to the total hemoglobin concentration or blood partial volume.¹⁴ The instantaneous phase was obtained by calculating the Hilbert phase after filtering by a zero-phase bandpass filter (Butterworth, $0.05–0.2\mathrm{Hz}$ ). The temporal development of the phase shift in both drug groups is shown in Fig. 6. Because we are only interested in the relative phase relationship between oxy- and deoxyhemoglobin, phases between 180 and $360\mathrm{deg}$ were converted to those between 180 and $0\mathrm{deg}$ . We observed that the phase relationship between oxy- and deoxyhemoglobin during baseline recording in the mouse was are almost out of phase, and its mean value $(\Delta \theta ={\theta }_{\mathrm{Hb}{\mathrm{O}}_2}-{\theta }_{\mathrm{Hbr}})$ was $\sim 145\mathrm{deg}$ . In the 4-AP group, no statistically different phase shifts between oxy- and deoxyhemoglobin were observed after drug administration. In the GBL group, the out-of-phase relationship moved to the in-phase state approximately $20\min$ after the dose and then returned to the out-of-phase state $40\min$ after the dose.