We used 2-to 3-month-old (n = 16) male Bama pigs (Taihe Biotechnology Co., Ltd.^®) with weights ranging from 25 to 35 kg. The pigs were screened to ensure that they passed animal quarantine requirements and they were then housed in temperature-and humidity-controlled rooms during the study. SEEG electrodes were implanted under animal anesthesia followed by a confirmation of a structure MRI scan for the electrode location. The baseline recordings were conducted 2 days after implantation. KA doses were injected after the completion of baseline recording. Video monitoring was conducted 7/24 until the study was completed. EEG recordings were performed 2 h per day for 9–28 days, depending on the animal health and electrode cap endurance. After the experiments were completed, animals were euthanized. All of the experimental procedures followed the regulations of the Institutional Animal Care and Use Committee (IACUC) of Shenzhen Children’s Hospital.

Of the three pigs that underwent dosage testing (Table 1), all of the quantities of KA triggered interictal spikes after 30 min of injection. For the 5 μl dose, the acute epileptic events only occurred in the first 2 h but there were no seizure activities for the first 24 h. The 10 μl KA dose produced approximately 6 h of interictal spikes and 11 acute seizures within the first 24 h. The 20 μl dose showed similar numbers of interictal spikes and triggered more than 50 acute seizures within 24 h. However, the pig exposed to a 20 μl dose died on the second day. Based on this, we chose 10 μl for the KA injection experiment in consideration of the survival rate and the effectiveness of the mTLE model.

For the 16 animals exposed to the 10 μl KA injection, 12 animals successfully completed the 4-week experiments, two animals died in the first week following injection and two animals lost their cap and were excluded from the dataset. Chronic-tonic seizures were captured in four pigs in weeks 2–4. Besides the four pigs with seizure occurrence, interictal spikes or short bursts of seizure-like events were observed in another four pigs. These eight animals were classified as the epileptic group (E +, n = 8) in this study. The remaining four animals showed no epileptic-like activities during the experimental period and were classified as the animals do not have epilepsy group (E–, n = 4).

Electrical activity was recorded under mild xylazine hydrochloride anesthesia. Power spectrum density (PSD) analysis suggested a peak frequency at 4 Hz and 10 Hz in baseline recording (Figure 2A), which represents the delta wave and alpha wave in EEG. They are related to the non-rapid eye movement (NREM) sleeping stage (49, 50). Compared to the baseline, PSD analysis of 2 h LFPs from the post-KA period indicated a decrease of delta and alpha power and an appearance of the 18–20 Hz peak in the hippocampus (Figure 2C). The multi-taper analysis also revealed differences in time-frequency patterns of the two experimental conditions (Figures 2B,D). The alpha wave disappeared in the data after KA injection.

A total of 46 tonic-chronic electrophysiological seizures were identified in the 4-week post-KA period from four pigs (Figure 3). The earliest seizure onset was at day 9 (in subject 102) and the latest was recorded at day 28 (the last day of measurement). The average seizure onset day was (18 ± 5.96), and the number of onsets was 4, 13, 4, and 20 in each animal. The seizure duration ranged from 12 to 45 s (mean = 27.21 ± 5.20).

We quantified 812 HFOs from the sensorimotor and hippocampal channels. All HFO events were manually verified in the data with a 500 ms LFP and time-frequency plot centered with HFO peaks after computational detection (see Figures 4A,B and Supplementary Figure 1 for examples). HFO rates were calculated on each channel, with the highest rate displayed in ipsilateral hippocampus (14.36 ± 6.76/min) and contralateral hippocampus (11.47 ± 7.06/min; Figures 4C,D). Though the HFO rates were much larger in these two channels (F = 2.49, p = 0.002) they were statistically similar to each other (p = 0.29). Peak frequencies ranged from 132 to 157 Hz, with an average of 147.3 ± 20.45 Hz. We found no significant difference of peak frequencies (F = 1.30, p = 0.20) between each channel (Figure 4E).

We studied the within-and between-group differences of HFO rates corresponding to different experimental periods (baseline vs. post-KA). In the within-group comparison, there were significant increases of HFO rates in the E + group (F = 10, p < 0.001) in the post-KA period compared to the baseline, specifically observed in week1 (p = 0.47) and week2 (p < 0.001). However, there were no significant differences in HFO rates in the E-group (F = 2.8, p = 0.065), though there was a slight decrease in the HFO rate in week 1 (p = 0.43). For the between-group comparison, we found E+ > E– in week1 (p = 0.005), week2 (p < 0.001), and week4 (p = 0.48). The HFO rates of E + and E – in the baseline were similar (p = 0.93; Figure 5A).

The stability of the HFO measurements was assessed daily from the 2 h recordings. HFOs from both hippocampus and sensorimotor cortex were compared between baseline and during epileptogenesis. In total, the HFO rates were significantly larger in the post-KA period compared to baseline (p = 0.39). Among the time epochs, most of the epochs showed significantly larger HFO rates in epileptogenesis than baseline (Figure 5A). There was no difference in the HFO rates in total and in each epoch in the sensorimotor cortex (except for some channels where no HFOs were detected). Analysis of peak frequency revealed no differences between baseline and in epileptogenesis except for channel 6 (p = 0.011).

The ICC (1, k) = 0.81 [LB = 0.54, UB = 0.94, F = 5.14, p < 0.001] was computed from the weekly HFO rate data, suggesting excellent (ICC > 0.8) test–retest reliability of results. These data were obtained by grouping all of the subjects, including both E + and E – (Supplementary Table S1 and Figure 5B). We also studied the data by separating the E + and E – groups, and found fair reliability [ICC (1, k) = 0.49] in the E + group and fair reliability [ICC (1, k) = 0.59] in the E– group.

A limitation in the experimental design of this study was the lack of prior knowledge, especially for the swine model of mTLE. Most prior studies on pigs have used subcortical penicillin (PCN) injections, which acted as a γ -aminobutyric acid (GABA)_A antagonist. The results demonstrated effective physiological changes in vital signs that occurred during convulsive activities in the treated pigs (13, 51–54). Various dosages of penicillin were used ranging from low (10 μl; (53)) to high (100 μl; (52)). However, even with the relatively larger size of the pig brain, a lower dose of PCN (10 μl) is sufficient to produce focal epileptiform activities and solid clinical seizure activities from grade II to IV (15).

Compared with the subcortical epilepsy model using PCN, the KA-model in the pig hippocampus produces an efficient local lesion with the chronic manifestation of spontaneous seizures that histopathological results showed as mTLE (3). KA-injection is a mature model in rodents. KA can preferentially destroy hippocampal pyramidal cells and trigger neurodegeneration in the hippocampus, but the affected region varies with dosage and the hippocampus is very sensitive to the KA level. Intraventricular injection of KA (at 0.5 nmol) in Sprague–Dawley rats caused pyramidal cell degeneration in CA3 at the rostral pole of the hippocampus whereas higher doses (0.8 μg) caused neuronal loss in more caudal regions of the hippocampus (19). Doses higher than 0.8 μg induced neurodegeneration in CA1 and CA2 (19). In rats, intrahippocampal administration of KA at a dosage between 0.4 and 2.0 μg is usually effective in inducing a convulsive status epilepticus after 5–60 min following the injection. Considering the PCN-swine model and the KA-model of rodents, we initially used a relatively narrow dosage range of 5, 10, and 20 μl. Initial testing with three doses indicated that 10 μl is close to the optimal concentration for the KA-model in swine (Table 1). All three initial cases showed epileptic activities but only case2 (10 μl) and case3 (20 μl) evoked tonic–chronic seizures. Case3 exposed to a 20 μl KA injection had a severe neuromuscular blockade and vigorous convulsions lasting more than 5 h and died within a short time. The remaining animals tested with 10 μl had a high (87.5%, 14/16) survival rate. These data were similar to results reported using the rodent model (with a mortality rate of 12%) (55–57).

Interictal patterns during slow wave sleep have been widely reported in animal epilepsy models. Similarly, an obvious change in frequency components was observed pre- and post-KA injection. There were 4 and 10 Hz frequency peaks on the power spectrum in baseline recordings. However, the 10 Hz frequency peak vanished after KA injection and a frequency peak of 20 Hz appeared within a short period (Figure 2). While the power spectrum only displayed the overall compositions of frequency, time-frequency analysis with the multi-taper method were further involved to display the frequency distribution along 2-h interval recordings. A 10 Hz frequency component, also known as sleep spindles, was usually seen in stage 2 of NREM sleep and plays an important role in memory consolidation (58). The disappearance of sleep spindles after status epilepticus may indicate the occurrence of pathological changes in cerebral areas, and the post-epilepsy memory injury related to human patients. Sleep disturbance is also a common sign in epilepsy patients (59, 60).

Despite the KA-evoked acute epileptic responses, we captured a total of 46 electrographic seizures from four animals from 9 to 28 days. As the first onset was more than 7 days after the KA injection, and each animal had more than one onset, we considered these events to be tonic-chronic seizures. Compared to the rodent intrahippocampal KA model, we found the first onset of the chronic seizures was a 9 days, which is later than rats (5 days) (19). The majority (66%, 8/12) of the tested pigs showed clear signs of epileptogenesis, and this rate was higher than a previous study of the intrahippocampal KA rat model (54%) (41).

The HFOs occur in various brain areas, in both normal and pathological conditions (28, 61). In previous studies using a rodent KA-model of mTLE, there was an increase of ripple, and an appearance of fast ripple activities (41, 62, 63) during the latent period of epileptogenesis. With two 16-channel SEEG electrodes, each with a contact area of 2 mm, we successfully detected HFO activities in the swine hippocampus in the baseline and after KA injection. A noticeable difference was the significant increase of the HFO rate after KA injection, compared to the baseline period. Consistent with the rodent model, the increase of HFO rates indicated changes in the local circuits based on the formation in pathologically interconnected neuron clusters (PIN-cluster) (64). We did not find any fast ripple activities; all of the HFO events were in the 100–200 Hz frequency range of ripple. This may be because the 2 mm contact area in the SEEG channels is too large to collect the local activates yield with fast ripples.

The HFO events in baseline displayed an evenly distributed pattern, while after KA injection, HFO distributions were mostly constrained within interictal spikes. Previous studies found that the pathological high frequency oscillations (pHFOs) often co-occur with interictal spikes (32, 65, 66). Interictal spikes (and co-occurring pHFOs) may disrupt memory processing (67). Most studies have demonstrated that interictal events disrupt memory consolidation processes, as interictal spikes have negative impacts when they occur during retrieval versus encoding phases of memory tasks (68). During sleep, interictal spikes inappropriately initiate spindle activity in the cortex and disrupt memory consolidation processes that engage cortical areas. The highest HFO rate was detected from the hippocampus area, which is similar to reports on mTLE patients (69). A high test–retest reliability of HFO rates was observed in the post KA period. The weekly stability suggested a reliable measurement of HFOs as critical biomarkers of epileptogenesis (25, 27, 31). There was no change or decrease of the HFO rates in the pigs do not have epilepsy (the E– animals). This is different from findings using the rodent model (41), where both E + and E – rats showed an increase of HFOs in the post-KA phase during the entire 5-week experimental period.

The animal phenotyping for the E + and E – groups was based on the occurrence of recurrent seizures observed in the post-KA period as well as the interictal spikes. Although interictal spikes normally indicate epileptic events, they are not a consistent biomarker for the development of chronic epilepsy. Other severe neurological conditions, such as stroke also shared the similar patterns of interictal spikes (70, 71). Our study duration was 4 weeks and omitted animals that had only been observed with IEDs but not seizures. However, the four animals classified as the E– group may have a chance to develop spontaneous epilepsy. This pitfall could be addressed by increasing the experimental period but doing so would increase the maintenance issue as the pigs gained weight and increased in size. The daily movement behavior of animals is also a challenge for the endurance of the EEG caps. For simplicity, this study only used two SEEG electrodes. The use of this model could be further investigated on a large scale by planning the subdural EEG and more intracranial EEGs. Overall, this study suggests that a KA-induced swine model is useful for studying the mechanisms and translational applications of epilepsy.