Laminar distribution of electrically evoked hippocampal short latency ripple activity highlights the importance of the subiculum in vivo in human epilepsy, an intraoperative study.

Abstract

Objective

The goal of this study was to define the pathology and anesthesia dependency of single pulse electrical stimulation (SPES) dependent high-frequency oscillations (HFOs, ripples, fast ripples) in the hippocampal formation.

Methods

Laminar profile of electrically evoked short latency (<100ms) high-frequency oscillations (80–500 Hz) was examined in the hippocampus of therapy-resistant epileptic patients (6 female, 2 male) in vivo, under general anesthesia.

Results

Parahippocampal SPES evoked HFOs in all recorded hippocampal subregions (Cornu Ammonis 2–3, dentate gyrus, and subiculum) were not uniform, rather the combination of ripples, fast ripples, sharp transients, and multiple unit activities. Mild and severe hippocampal sclerosis (HS) differed in the probability to evoke fast ripples: it decreased with the severity of sclerosis in CA2–3 but increased in the subiculum. Modulation in the ripple spectrum was observed only in the subiculum with increased fast HFO rate and frequency in severe HS. Inhalational anesthetics (isoflurane) suppressed the chance to evoke HFOs compared to propofol.

Conclusion

The presence of early HFOs in the dentate gyrus and early fast HFOs (>250 Hz) in the other subregions indicate the pathological nature of these evoked oscillations. Subiculum was found to be active producing HFOs in parallel with the cell loss in the hippocampus proper, which emphasize the role of this region in the generation of epileptic activity.

1. Introduction

Temporal lobe epilepsy (TLE) is the most common form of adult localization-related epilepsy, which is caused by the malfunction of the hippocampus due to neuronal hyperexcitability(Tatum, 2012).

Hippocampus (Hc) plays essential roles in learning, spatial navigation(Moser and Moser, 1998) and episodic memory processes(Vargha-Khadem et al., 1997) that are distinguished by characteristic oscillations. In awake exploring condition when encoding and immediate retrieval of the newly acquired memory traces take place, a slow oscillation emerges in the hippocampus in theta (5 – 10 Hz) range in rodents and delta (2 – 4 Hz) in humans(Buzsaki and Moser, 2013; Jacobs, 2014). Memory consolidation under resting awake or slow-wave sleep condition is followed by bouts of fast oscillations forming sharp wave-ripple complexes(Behrens et al., 2005; Bragin et al., 1999b; Jirsch et al., 2006) (SWR) consist of transient network oscillations around 200 Hz emerging in the CA3 region and spreading toward CA1 and subiculum(Buzsaki et al., 1992; Girardeau and Zugaro, 2011).

High frequency oscillations have been strongly linked to epileptogenesis in humans(Bragin et al., 1999b; Jirsch et al., 2006), in slices of the human epileptic neocortex(Roopun et al., 2010) and animal models of epilepsy(Foffani et al., 2007). Several studies have shown that HFOs specify the epileptogenic zone better than interictal discharges (IID)(Fabo et al., 2008; Jacobs et al., 2008; Ulbert et al., 2001; Ulbert et al., 2004; Valentin et al., 2005; van ‘t Klooster et al., 2017). Slow and fast ripples (80–200 Hz for ripples(Bragin et al., 1999a) and 250–500 Hz for fast ripples(Bragin et al., 1999b; Köhling and Staley, 2011)) presumably have distinctive roles; slow ripples may have a physiological role in memory consolidation while fast ripples appear mostly upon epileptic transformation, however, this correlation needs to be clarified(Menendez de la Prida et al., 2015). The appearance of fast ripples correlates to the seizure onset zone (Bragin et al., 2002; Jacobs et al., 2008), to disease severity(Jacobs et al., 2009), and the frequency of seizures(Steward, 1976). Furthermore, the removal of the ripple generating area is correlated with good surgical outcome(Akiyama et al., 2011; Jacobs et al., 2010).

Single pulse electrical stimulation (SPES) is a promising method to assist controlled delineation of the epileptic focus by evoking pathological events. In epileptic patients, cortical SPES within the seizure onset zone can evoke spikes and HFOs(van ‘t Klooster et al., 2017), as well as pathological delayed responses(Valentin et al., 2005). However, evidence for the presence of evoked HFOs in the highly epileptogenic human hippocampus is lacking.

The goal of this present study was to investigate SPES evoked HFOs and their laminar distribution within the anatomically identified human hippocampal formation in vivo, under general anesthesia, using multiple channel microelectrodes in patients with drug-resistant temporal lobe epilepsy. High spatial resolution laminar field potential gradient (FPG), and multiple unit activity (MUA) from the hippocampal formation were analyzed to describe the features of evoked oscillations.

2. Materials and Methods

2.1. Patients

Fourteen patients with medically intractable mesial temporal lobe epilepsy with unilateral seizure onset were involved in this study (nine females, five males, mean age 42 years, range 25–57 years). From the available recordings, three cases were excluded because of missing or low-quality histological reconstruction, two cases where the stimulation protocol was applied after partial resection of the temporal lobe, and another recording due to noise. Information about the remaining 8 patients is given in Table 1. Presurgical evaluation of all the cases was performed in the Epilepsy Centre of the National Institute of Clinical Neurosciences, Budapest (from 2007) (NICN) and the former National Institute of Psychiatry and Neurology, Budapest (2004–2007). Results from high-resolution MRI, interictal, ictal EEG with scalp and/or invasive subdural electrodes and the clinical history data were consistent of unilateral mesial temporal lobe epilepsy in all cases. The surgery and electrode implantations were done in the NICN. The study was approved by the Hungarian Medical Research Council (ETT TUKEB 20680–4/2012/EKU) and performed following the Declaration of Helsinki.

Table 1.

Summary of patient characteristics. Abbreviations: Impl. side: implantation and resection side, MRI: Magnetic Resonance Imaging, HS: hippocampal sclerosis, TU: tumor, FCD: focal cortical dysplasia Hc.: hippocampal, sHS: severe cell loss and reorganization of the hippocampus (severe hippocampal sclerosis), mHS: mild cell loss and reorganization of the hippocampus (mild hippocampal sclerosis). Outcome: (Engel classification) 1A: Aura and seizure free; 2B: Rare seizures; 3A: Often seizures with a worthwhile reduction.

Patient code	Age at operation (years)	Age at epilepsy onset	Impl. side	Sex	Duration of epilepsy (years)	MRI finding	Hc damage	Analysed electrode localization	Anesthesia	Outcome	Followup time (years)
P17	31	9	Right	male	22	Right HS	sHS	body	Isoflurane	1B	3
P22	46	44	Left	female	10	Bilateral HS	sHS	digitation s	Propofol	3A	9
P25	36	30	Left	female	6	Left HS	sHS	head	Propofol	1A	3
P26	40	17	Right	female	23	Bilateral HS	sHS	body	Isoflurane	2B	3
P33	51	19	Left	female	32	TU (Left amygdala)	mHS	head	Propofol	3A	2
P36	39	1.5	Left	female	37.5	Left HS	sHS	body	Propofol	1A	9
P38	57	34	Left	male	23	Left HS	mHS	body	Isoflurane	1A	5
P47	38	36	Left	female	3	Left FCD + HS	mHS	head	Propofol	1A	6

2.2. Surgery and recording

The intrahippocampal recordings were performed during electrocorticography (ECoG) assisted tailored anterior temporal lobectomy. The type of surgery, the application of ECoG and the type of anesthesia were selected solely based on clinical grounds and had no interaction with the experimental procedure. Premedication included 5–10 mg benzodiazepine and 50–100 mg pethidine. Two types of anesthesia were used, inhalation (gas) with isoflurane (average MAC value was 1.1–1.5) or intravenous with propofol (1,5–2,5 mg/bwkg/min). Continuous nitrogen-monoxide inhalation was present beside both types of anesthesia.

The recording technique was previously described(Fabo et al., 2008; Ulbert et al., 2001). Briefly, a long shafted deep structure laminar multi-microelectrode (dME) was used to record the activity from the hippocampus. As part of the ECoG procedure, an eight contact clinical strip electrode (Ad-Tech Medical Instrument Corporation, Racine, USA) was placed around the temporal pole. Spontaneous interictal discharges and 50Hz stimulation evoked epileptic afterdischarges(Fabo et al., 2008) were on-line analyzed to provide intraoperative evidence for the epileptogenic nature of the resected area.

The clinical procedure remained unchanged during the whole surgery with the only exception of the placement of the electrode (a 10 centimeters long, 350 μm diameter, stainless steel needle with 24 linearly arranged contacts with Ø: 25μm, Platinum-Iridium wires, 200 μm center to center distance, first contact is 5 mm far from the tip) and performance of the SPES stimulation(Jiruska et al., 2017; Magloczky and Freund, 2005; Schomburg et al., 2014).

The tip of the microelectrode was positioned to the surface of the hippocampus under visual control. The gradual tissue penetration was provided by the μm-precise microdrive system at approximately 200 μm/sec speed in 2–4 mm increments. The slow insertion minimalized the traumatic effect of electrode penetration. The field potential gradient recording was continuous during the insertion and extraction. The recordings were taken with custom made equipment with two settings, in the low-frequency band (0.2–500 Hz, sampled at 2 kHz/channel with 16-bit precision), and in the high-frequency band (150–5000 Hz, 20 kHz/channel on 12 bit), described previously (Fabo et al., 2008; Ulbert et al., 2001).

As part of the intraoperative ECoG procedure, an eight contact clinical strip electrode (Ad-Tech Medical Instrument Corporation, Racine, USA) was placed around the temporal pole (Figure 2A). The strip electrode was used to deliver single pulse electrical stimulation (SPES) near the parahippocampal region in order to stimulate the input pathways of the hippocampus. Low frequency (0.5 Hz, 5–10-15 mA, 0.2 ms pulse duration, 25 or 50, bipolar, biphasic square wave) SPES stimuli were applied through the adjacent electrode contact sites, to elicit potentials from the hippocampus. The stimulation site was estimated relative to the temporal pole.

Figure 2.

A) Schematic illustration of the position of the stimulating strip and the microelectrodes. Evokability of the HFA and the HFO depends on stimulus location (1–5 cm from temporal pole). B) Evoking capability according to the subregions and the severity of the HS. Notice the significant difference between mild and severe HS in CA2-3 and subiculum. C) Frequency distribution of the sevHFOs in beta-gamma (12–80 Hz), slow ripple (80–250 Hz) and fast ripple (250–500 Hz) band according to subregions of hippocampus and severity of sclerosis (mild and severe/orange and black). In the slow ripple band, two or three Gaussians were fitted to gain the best fit. D) The ratio of fast ripple power according to the hippocampal subregions and the severity of the HS. Slow (80–250 Hz) and fast (250–500 Hz) ripple power ratio is significantly different in SUB and CA2-3 subregions of hippocampal formation. Dark grey cases originate from mild HS, light grey ones illustrate severe HS cases. In severe CA2-3 the slower components became overweight, and fast reduced compared to mild CA2-3. This ratio is just the opposite in subiculum, the slower reduced and the fast components came in greater numbers while DG did not change. E) HFO evokability depending on anaesthesia (propofol vs isoflurane) according to the subregions. F) SevHFO and evMUA overlap across subicular layers (P22, proximal and distal subiculum). Line with square represents the average HFO amplitude derived from different stimulations. Maximal MUA activity (4600–4800 μm) appears below the sevHFO peak (4200 μm). HFOs mainly appeared in the dendritic layers of the subiculum, as higher MUA delineates the cell layer. The x-axis shows mm distance from the surface (0 mm) of the hippocampus, where the electrode was inserted.

2.3. Histology

After the SPES protocol, the hippocampal formation and the entorhinal cortex were removed en-block for histological processing according to the original surgical plan. The electrode trajectory was reconstructed during the histological post-processing(Ulbert et al., 2004). The penetration of the electrode was identified by light microscopic examination; the trajectory was reconstructed from multiple stained sections(Fabo et al., 2008) and was analyzed in digital picture editing software Adobe Photoshop CS5. Based on the histology, all electrode contact points at all penetration positions during the recording were specified. All recording sites were grouped and allocated to the different subregions of the hippocampal formation: Cornu Ammonis 2 −3 (CA), dentate gyrus (DG) and subiculum (SUB). The severity of hippocampal cell loss was categorized into mild (minimal to moderate cell loss in the CA1 region) and severe hippocampal sclerosis (HS) (more than 90% cell loss in the CA1 region). These categories are defined as mild HS (mHS) and severe HS (sHS)(Wittner et al., 2002).

2.4. Data analysis

Data processing was performed with Neuroscan (Neuroscan Inc.) and Matlab (Mathworks Inc.). Free accessible, open source Matlab packages such as EEGLAB(Delorme and Makeig, 2004), Fieldtrip (Donders Institute for Brain, Cognition and Behavior and the Max Planck Institute in Nijmegen), as well as previously described(Fabo et al., 2008) script package developed by our laboratory and self-written Matlab scripts were used.

To remove the effect of the sharp transient of the stimulation artifact, a 4 ms segment around the stimulation was spline interpolated before analysis.

Recording channels were grouped into 3 hippocampal regions 1: Cornu Ammonis 2 or 3 (CA 2–3), 2: dentate gyrus (DG), and 3: subiculum (SUB) based on the result of the histological reconstruction of the electrode trajectory. CA regions were grouped due to the low hit ratio in these areas. CA1 region was missed by all the trajectories in the examined patients.

The analysis of evoked high-frequency activities was built up by a series of steps. First, the evoked high-frequency activities were characterized by the power increment in the 80–500 Hz range. To determine the intensity of the evoked response, the filtered (2^nd order Butterworth, zero phase shift), rectified data was smoothed with root mean square using a sliding 11 ms window. The maximal amplitude between 10–100 ms after stimulation was converted to statistical z-score value using the mean and standard deviation of the baseline (−450 ms to −50 ms before the stimulation). Evoked potentials higher than five z-score were considered as statistically significant(Bragin et al., 2010) and these events were referred as significant evoked high-frequency activities (sevHFA). Significant evoked high-frequency oscillations (sevHFOs) were defined with at least 4 cycles by visual selection and were considered identical to evoked ripples.

In this study, evokability, the probability to produce HFO (evoking capability -evokability) is considered as a measure of the ability of the hippocampal regions to produce evoked HFO as a response to the stimulation from the temporo-basal structures. Evokability was defined by the ratio (expressed in %) of the sevHFOs events to all the SPES stimuli within a delivered train.

The frequency components of the evoked HFOs were measured with event-related spectral perturbation (ERSP from EEGLAB) in a −450 ms to +150 ms window around the stimulation. The maximal ERSP coefficients were collected into three frequency distributions (12–80 Hz, 80–250 Hz and 250–500 Hz). The mean frequency and standard deviation were calculated with Gaussian distribution fitting (Curve Fitting Toolbox, Matlab). Two-sided Wilcoxon’s rank sum test was applied with 95% or 99% confidence interval (CI).

To address the question how likely the multiple unit activity (MUA) is responsible for the HFO generation, MUA amplitude was calculated from the 20 kHz sampled recording with bandpass filtering (2^nd order Butterworth) between 500 – 5000 Hz, using the same method as for sevHFA.

The laminar distribution of different oscillatory activities (sevHFA, MUA) was calculated within each channel correlating the amplitudes of the evoked sevHFA and MUA during each stimulation trial. Overlap percentages along the trajectory were defined based on the ratio of channels showing significant (CI 95%) correlation.

Laminar reconstruction of the LFP data was performed using real size photo of the histological sections. All the figures were color edited in CorelDraw (Corel Corporation, Ottawa) to gain the highest image quality.

3. Results

Multiple channel microelectrodes recordings were obtained from the hippocampus of therapy-resistant epileptic patients. The reconstructed electrode trajectories of the eight included cases are shown in Figure 1/A. Supporting Table 1 shows the sampled hippocampal regions according to the hippocampal sclerosis severity and anesthesia. Severe hippocampal sclerosis (sHS) was found in five patients, mild (mHS) in three. Inhalation type anesthesia (isoflurane) was conducted in three cases, and intravenous (propofol) in five cases. Based on histological categorization, Cornu Ammonis (CA) regions were sampled in five cases (P17, P25, P26, P38, and P47), dentate gyrus (DG) in five cases (P17, P25, P33, P38, P47) and subiculum (SUB) in six cases (P17, P22, P25, P26, P33, P36).

Figure 1.

A) Electrode trajectory reconstructions on histological slices from all patients. B) Schematic illustration of the hippocampal formation. Examples of typical evoked HFOs waveforms in the subregions (A-D). Black triangles indicate the stimulus. Note that SPES can evoke HFOs with different latencies in the DG (A) and the subiculum (D), whereas HFOs were evoked in the CA2-3 region with similar latency. CA: Cornu Ammonis, DG: dentate gyrus, SUB: subiculum

3.1. Analysis of evoked high-frequency activity (sevHFA) and oscillations (sevHFO)

All together 130 cortico-hippocampal evoked potential (CHEP) stimulation trains containing 2275 stimuli (25 or 50 stimuli per train) were included. After automatic detection for short latency (‹100ms) sevHFA, the best channels containing sevHFO were selected visually in each train and region (DG, CA2–3, SUB) (see Methods 2.4).

Figure 1/B shows typical waveforms in different regions. Overall 862 sevHFO were found, 104 in DG (51 in mild, 53 in severe HS), 241 in CA2–3 (81 in mild, 160 in severe HS), and 517 in SUB (13 in mild, 504 in severe HS). Amplitude, latency, duration, frequency and the evoking probability were calculated for all events.

Latency and duration

The latency of the sevHFOs showed a bimodal distribution in the DG (22.8±2.5 ms; 37.1±1.7 ms) and the SUB (15.2±6.7 ms; 39.6±4.7 ms), whereas it was unimodal (20.3±5.9 ms) in CA2–3. The average duration of all hippocampal sevHFOs was 18.3 ±4.5 ms and 16.7 ±1.4 ms in the CA2–3 region. Comparison of the duration of the earlier and later peaks showed that subicular sevHFOs had a similar duration (18.7 ± 2.1 ms), while in the DG, the earlier responses were slightly longer (29.4 ±4.9 ms) than the later responses (20.2 ±1.4 ms).

3.2. The probability to produce sevHFO

Figure 2/A illustrates the average evokability. The most effective stimulation (74.7%) was applied 2–4 cm from the temporal pole on the medial surface of the temporal lobe. Among the recorded hippocampal regions, in mild HS cases the CA2–3 was the most active region (mean 88%), followed by DG (mean 32%) and SUB (mean 19%), while in severe HS, the highest ratio was found in the SUB (mean 70%), followed by CA2–3 (mean 29%), and DG (mean 14%). The difference between mild and severe HS was statistically significant in SUB (p < 0.05) and in CA2–3 (p < 0.01). (Figure 2/B)

3.3. Frequency of sevHFOs

The frequency content of every sevHFO was characterized with the distribution of frequencies corresponding to the maximum power within three frequency bands: 1) beta-gamma (12–80 Hz), 2) slow ripple (80–250Hz) and 3) fast ripple (250–500 Hz). Multiple Gaussians were fitted to the histograms in each band to show relevant frequency peaks. Figure 2/C and Table 2 describe the results.

Table 2.

Mean frequencies of the evoked HFOs according to the hippocampal subregions and the severity of hippocampal sclerosis. DG-Dentate Gyrus, CA2–3- Cornu Ammonis 2 or 3 areas, SUB-subiculum.

Mean frequency according to HS severity
	DG		CA2–3		SUB
	Mild	Severe	Mild	Severe	Mild	Severe
Beta/Gamma (12–80 Hz)	28 ±1.23	24 ±1.15	28 ± 1.08	26 ±1.34	30 ±4.89	30 ±3.54
Ripple (80–250 Hz)	100 ±3.34 195 ±3.93	80 ±0.35 170 ±5.42	78 ±0.47 115 ±1.45 180 ±0.38	80 ±1.44 125 ±0.7 180 ±4.38	78 ±1.84 180 ±1.48	78 ±0.92 145 ±4.24 235 ±5.18
Fast Ripple (250–500 Hz)	no FR component	no FR component	305 ±2.82	280 ±1.22	no FR component	400 ±7.85

Low and high-frequency peaks were found in all regions and HS conditions. Most events showed beta frequency components, varied between 24 and 30 Hz in every region in both pathological conditions. In the slow ripple range, the frequency peaks were centralized around three sub-bands: 78–100 Hz (low), 115–145 Hz (middle) or 170–235 Hz (high). There were only slight differences between the regions: in mild HS DG showed higher frequencies compared to CA2–3 and SUB which relation was reversed in severe HS. The middle frequencies (115–145 Hz) were seen mostly in the CA2–3 region.

Fast ripples were found in CA2–3 (both mild and severe HS) and in severe HS SUB cases reaching exceptionally high-frequency values (around 400Hz) in the latter case.

3.4. Slow – fast ripple ratio

Fast ripple ratio was plotted by the fast ripple (250–500) to total frequency (80–500 Hz) band power ratio for all the sevHFOs (Figure 2/D). In accordance with the transition from mild HS to severe HS, the sevHFOs were shifted significantly (t-test, p<0.01) toward the slow ones in CA2–3, remained in the slow range in the DG (no significant difference) and shifted significantly (t-test, p<0.01) toward the fast ripples in the SUB. As a result, there were two conditions with relatively high fast ripple ratio (around 50% median): CA2–3 in mild HS and SUB in severe HS.

3.5. Anesthesia considerations

Two types of anesthetics were compared: inhalation type isoflurane and intravenous propofol. sevHFO evokability in the SUB was significantly (t-test, p<0.05) lower with isoflurane than with propofol (isoflurane 32%, propofol 70%) (Figure 2/E). There was no statistically significant difference between DG and CA2–3.

The type of the anesthesia had a small effect on the frequency of HFOs (Table 3). Unfortunately, only CA2–3 and SUB had enough sample to analyze. In CA2–3, mean gamma and fast ripple frequency increased, ripple frequency decreased under isoflurane compared to propofol. The same effect in subiculum led to gamma and fast ripple frequency decrease.

Table 3.

Mean frequencies of sevHFOs, grouped by anaesthesia. DG-Dentate Gyrus, CA2–3- Cornu Ammonis 2 or 3 areas, SUB-subiculum, no data – no data in this combination.

Mean frequency according to anaesthesia
	DG		CA2–3		SUB
	Propofol	Isoflurane	Propofol	Isoflurane	Propofol	Isoflurane
Beta/Gamma (12–80 Hz)	28 ±1.51	no data	26 ± 1.14	28 ±1.06	32 ±2.80	26 ±2.53
Ripple (80–250 Hz)	100 ±3.59 195 ±4.18	no data	78 ±0.33 125 ±0.75 220 ±1.39	76 ±0.42 120 ±1.97 180 ±1.20	78 ±1.51 150 ±4.24 240 ±4.04	160 ±6.25
Fast Ripple (250–500 Hz)	no FR components	no data	295±3.39	310 ±1.10	400 ±7.61	365 ±9.32

3.6. The spatial extent of the HFO generating area

The extent of the sevHFOs including slow and fast ripples was measured by the extent of sevHFO detections throughout the recording channels (evokability higher than 0%). SevHFO events were assigned into slow or fast ripple categories based on their highest power frequency with cutoff frequencies at 150, 200 or 250Hz, to locate very high-frequency oscillations separately.

Broadband sevHFOs (80–500 Hz) were generated within 1–1.6 mm wide area, fast ripples emerged within 0.2–0.8 mm; 0.2–0.6 mm; or 0.2–0.4 mm area according to the increasing frequency cutoff value, respectively.

3.7. Evoked multiple unit and high-frequency activity overlap

To define the overlap, a ratio of interference between the summed evoked action-potential firing of the local neurons (evMUA) and the high-frequency oscillation generators (sevHFA) was described. All stimulation trains on all channels were included, not exclusive for detected sevHFO oscillations. Latency was defined as the time for the maximum amplitude between 10 to 100 ms.

The peaks of sevHFAs preceded evMUA peak in 50% of all the trains with an average of 31±23 ms, and followed it in 31% with 26±21 ms, and coincided in 18.8% (within 5 ms).

The highest amplitudes over the channels were identified for both sevHFA and evMUA. Figure 2/F illustrates the difference in one particular case (P22). High sevHFA amplitudes were shifted toward the apical dendrites of pyramidal neurons by several hundred μm compared to high MUA activity (3400–4200 μm versus 4600–4800 μm). The overlap from all patients is shown in Figure 3. The low-frequency sevHFA (80–250 Hz) and MUA showed a statistically significant correlation (Pearson, p < 0.05) in average 38.74 % of the recording contacts (range among patients varied between 0 – 92%) and high frequency (250–500Hz) sevHFA and MUA showed a significant correlation in average 45.71% of the recording contacts (range: 3 – 88 %).

Figure 3.

SevHFO and evMUA overlap of all patients plotted on mm scale distance from the surface (0 mm) of the hippocampus, where the electrode was inserted. Dark blue circles indicate the evokability of the sevHFO in analyzed electrode trajectories. The diameter of the circles proportionate to the degree of the evokability. The rectangles indicate 100%. Cyan blue stars show the subselected best channels with significant HFOs.

4. Discussion

This study aimed to describe the laminar profile and spectral properties of the single pulse electrical stimulation evoked early (<100 ms) cortico-hippocampal potentials (CHEP) in the human epileptic hippocampus in vivo with microelectrodes during anesthesia for the first time. A major finding is that these CHEPs contained abundant amount of evoked HFOs in each sub-region and the probability to produce HFO in these regions depended differently according to the tissue damage and the type of anesthesia.

The hippocampal sub-regions are innervated by the entorhinal cortex (EC) that is an area integrating multimodal imputs from different brain regions including various aspects of the temporal lobe (Naber et al., 2001; Steward, 1976; Wilson et al., 1991). In our study, the electrical stimulation of the temporal pole can induce short-latency high-frequency activities in each recorded sub-region of the hippocampal formation and the high evokability ratio (74.7% at 2–4 cm from temporal pole) confirms the functional connections between the temporal cortex and the hippocampus, whose strength could be altered by the underlying epileptogenic reorganization (Coste et al., 2002). Based on electrophysiological studies, the initial discharges simultaneously occurring within each hippocampal sub-regions after entorhinal cortical stimulation are followed by weaker excitatory responses transmitted through the trisynaptic pathway (Yeckel and Berger, 1990). Based on the variable short latencies of evoked HFOs (earliest 7 ms, average: 15–22 ms), one can assume the presence of oligo-synaptic transmission in humans.

Analyzing the action potential firing (MUA), a higher (46%) correlation was found with the fast frequency range (250–500 Hz) than with the slow (80–250 Hz) evoked HFAs (39%). Based on the spatial distribution patterns, the generation of HFO relies on postsynaptic potentials of the dendrites (mostly apical) rather than neuronal firing, which can be observed in higher frequency (>250 Hz) cases. On the other hand, HFOs may encompass a range of cellular and synaptic processes in overlapping frequencies from 80 Hz to 800 Hz (Menendez de la Prida et al., 2015). In humans, the power spectrum of HFO is not clearly defined, since there is evidence that the HFO associated spike activity shifts the spectrum to the higher (>300 Hz) and the slower (<80 Hz) frequency bands(Bragin et al., 1999a; Le Van Quyen et al., 2010; Staba et al., 2002). In our study, the broadband frequency range (12–500 Hz) during the evoked HFOs showed beta (24 −30 Hz) with co-occurring slower (78–250 Hz) and higher (250–400 Hz) ripple frequency components as described in animal(Skaggs et al., 2007), and human (Alvarado-Rojas et al., 2015) studies. The dominant frequencies (around 25, 80, and 120 Hz) in CA2–3 found by us were similar to the previously studied rat CA1 region (Schomburg et al., 2014). We found that the 80–250 Hz slow ripple band is separated into three subcategories: 78–100 Hz (slow), 115–145 Hz (middle), and 170–235 Hz (high). It is still a question whether these sub-bands represent functionally different activities or not. The high percentage of MUA activity in the very fast (>300 Hz) oscillations and the wide frequency range supports the theory that fast HFOs are the summation of discharges made from in- or out-of-phase action potentials(Jiruska et al., 2017).

Hippocampal sclerosis features reorganization of both excitatory and inhibitory circuits of the CA regions and the dentate gyrus within the hippocampus, manifesting in cell loss, axonal sprouting and gliosis(Magloczky and Freund, 2005). In contrast to the damaged regions of the hippocampus proper (CA3 and CA1 regions), the structure of the subiculum is remained fairly preserved(Cavazos et al., 2004). The most evident differences between mild and severe sclerosis in this study were the increase of the HFO evokability, the increase in frequency of these evoked events and the higher fast ripple ratio in subiculum, compared to CA2–3. Interestingly, DG produced only slow ripples with both pathologies Previous animal data showed that ripple activity in DG was a pure consequence of the epileptogenic process within the hippocampus both in spontaneous and evoked conditions(Bragin et al., 1999a; Bragin et al., 1999b). The finding that DG showed evoked slow ripples unrelated to the tissue damage implies that the evoked HFOs under anesthesia, either slow or fast, are rather linked to pathological process than the activation of physiological ripple generators. The transition of the high frequency activity from CA to subiculum in parallel with the cell loss in the former areas and the almost complete loss of CA1 region can be explained by an activity emerging in the subiculum after the reorganization of hippocampal connections linked to sclerosis(Chung et al., 2015). Subiculum has been demonstrated to be autonomously active in isolated human epileptic hippocampal slices(Cohen et al., 2002) and to generate rhythmic discharges in animal models(Knopp et al., 2005), even without hippocampal sclerosis(Wozny et al., 2005). Drexel et al. showed the importance of the potent inhibitory action mediated by parvalbumin cells in CA1/subiculum and its potential role in the mechanism of the TLE(Drexel et al., 2017). The increased frequency of the HFOs can be the result of the disruption of neuronal firing causing the ripple spectrum becoming unstable(Foffani et al., 2007; Jefferys et al., 2012) and thus, HFOs are appearing in the fast ripple range(Alvarado-Rojas et al., 2015). This type of modulation was observed only in the subiculum in this study, which further emphasized the role of this region in cases of severe cell loss within the hippocampus. These results could indicate that the subiculum alone might have a significant role in TLE.

Regarding the anesthesia, isoflurane and propofol have a different effect on the excitability level of the hippocampus. The effect of propofol is contradictory since despite high-dose propofol is antiepileptic and reduces the number of epileptic HFOs, it has no effect on spikes(Samra et al., 1995). Both depolarizing and hyperpolarizing effects of isoflurane can be ascribed to a decrease in excitatory and inhibitory synaptic drives(Roth and MacIver, 1986). Takeda et al found that isoflurane decreased the amplitude of HFO waves, but had little effect on their oscillation rate(Takeda and Haji, 1993). We found that isoflurane, compared to propofol, significantly decreased the number of sevHFOs and their amplitude, slowed down the gamma and ripple frequency in the SUB, the ripple frequency in CA2–3, and shifted the gamma and fast ripple frequency higher in CA2–3. Based on the degree of the evokability, inhalational anesthesia is recommended to be able to record SPES evoked HFOs.

5. Conclusion

Early (latency<100 ms) high-frequency oscillations can be evoked by parahippocampal single pulse electrical stimulation in all sampled subfields of the human hippocampus (DG, CA2–3, subiculum). The presence of early HFOs in the dentate gyrus and early fast HFOs (>250 Hz) in the other subregions indicate that SPES evoked rather epilepsy-related pathological oscillations. Fast HFOs emerged in the cell layers, in correlation with increased cell firing, while slow HFOs (80–250 Hz) appeared in the dendritic regions and seemed to be separated into different frequency bands. Subiculum was autonomously active producing HFOs in parallel with the cell loss in the hippocampus proper that further brings attention to the importance of this region in temporal lobe epilepsy.

Highlights.

• HFOs can be evoked by parahippocampal stimulation in all subfields of the human hippocampus.

• Inhalational anesthetics (isoflurane) suppressed HFO generation compared to propofol.

• Subiculum was active in coincidence with the severity of HS, signifying its importance in TLE.