Human entorhinal cortex electrical stimulation evoked short‐latency potentials in the broad neocortical regions: Evidence from cortico‐cortical evoked potential recordings

Abstract Objective We aimed at clarifying the clinical significance of the responses evoked by human entorhinal cortex (EC) electrical stimulation by means of cortico‐cortical evoked potentials (CCEPs). Methods We enrolled nine patients with medically intractable medial temporal lobe epilepsy who underwent invasive presurgical evaluations with subdural or depth electrodes. Single‐pulse electrical stimulation was delivered to the EC and fusiform gyrus (FG), and their evoked potentials were compared. The correlation between the evoked potentials and Wechsler Memory Scale‐Revised (WMS‐R) score was analyzed to investigate whether memory circuit was involved in the generation of the evoked potentials. Results In most electrodes placed on the neocortex, EC stimulation induced unique evoked potentials with positive polarity, termed as “widespread P1” (P1w). Compared with FG stimulation, P1w induced by EC stimulation were distinguished by their high occurrence rate, short peak latency (mean: 20.1 ms), small peak amplitude, and waveform uniformity among different recording sites. A stimulation of more posterior parts of the EC induced P1w with shorter latency and larger amplitude. P1w peak amplitude had a positive correlation (r = .69) with the visual memory score of the WMS‐R. In one patient, with depth electrode implanted into the hippocampus, the giant evoked potentials were recorded in the electrodes of the anterior hippocampus and EC near the stimulus site. Conclusions The human EC electrical stimulation evoked the short‐latency potentials in the broad neocortical regions. The origin of P1w remains unclear, although the limited evidence suggests that P1w is the far‐field potential by the volume conduction of giant evoked potential from the EC itself and hippocampus. The significance of the present study is that those evoked potentials may be a potential biomarker of memory impairment in various neurological diseases, and we provided direct evidence for the functional subdivisions along the anterior–posterior axis in the human EC.


| INTRODUC TI ON
The medial temporal lobe has been regarded as an important structure for episodic memory. The medial temporal lobe system for episodic memory is composed of the hippocampus and the perirhinal, entorhinal, and parahippocampal cortices. The entorhinal cortex (EC), which is located in the anterior parahippocampal gyrus, is considered the gateway between the hippocampus and neocortex (Squire & Wixed, 2011). Thus, it is important to know how the EC interacts with both the hippocampus and neocortex for a more detailed understanding of the episodic memory system. Subdural electrode implantation is commonly performed for epileptic surgery. The two main purposes of electrode implantation are identification of the epileptogenic focus and functional brain mapping. We developed the cortico-cortical evoked potential (CCEP) method as an invasive interventional approach to probe the causal influence, the so-called "effective connectivity" (Matsumoto et al., 2004;Matsumoto & Kunieda, 2018). In this method, we apply single-pulse electrical stimulations to specific cortical areas and record CCEPs time-locked to the stimuli from remote cortices. This method can identify the effective connectivity between the stimulus site and remote neocortical region, which are presumably connected by a long association fiber tract (Yamao et al., 2014).
The CCEP study found in preliminary investigations that uniform sharply contoured positive potentials of short latency can be recorded from various neocortical regions after the stimulation of the medial temporal lobe. We termed such broadly recorded potentials as widespread P1 (P1w). P1w were observed in response to stimulation of the anterior parahippocampal gyrus, especially the EC. Such broad evoked response on various neocortical areas outside the limbic system has not been investigated yet, at least not in electrophysiological studies. Previous anatomical and electrophysiological studies investigating the connectivity between the EC and neocortex showed that connections are mainly found within the temporal lobe and that connections outside the temporal lobe are almost limited to limbic structures such as the cingulate gyrus and orbitofrontal area (Catenoix et al., 2005;Catenoix, Magnin, Mauguiere, & Ryvlin, 2011;Enatsu et al., 2015;Koubeissi, Kahriman, Syed, Miller, & Durand, 2013;Kubota et al., 2013;Lacruz, Seoane, Valentin, Selway, & Alarcon, 2007;Munoz & Insausti, 2005). However, in the recent human CCEP study using an effective connectivity model, it was revealed that the hippocampus could act as a signal amplifier of afferent information flow from the EC (Krieg et al., 2017). Therefore, EC electrical stimulation can have significant effects on broad neocortical regions outside the temporal lobe via such signal amplification of the hippocampus.
Along with CCEP, as part of the presurgical evaluations for epileptic surgery, patients in our institute usually undergo a series of neuropsychological batteries, including Wechsler Memory Scale-Revised (WMS-R). The WMS-R has been used clinically as one of the standard batteries for the assessment of memory function. Medial temporal lobe epilepsy patients frequently experienced memory impairment, as reflected by the decrease in WMS-R scores. The present study aimed to clarify the significance of P1w evoked by the human EC electrical stimulation. For this purpose, we investigated the characteristics of P1w, including its correlation with WMS-R scores to verify whether P1w reflect the medial temporal lobe episodic memory circuit activity.

| Subjects
We recruited nine patients (five men and four women, 21-52 years old) with medically intractable medial temporal lobe epilepsy who underwent subdural electrode implantation into the EC (left: seven; right: two) from April 2011 to January 2015 in our hospital. The demographics of all patients are shown in Table 1. All patients provided written informed consent. The protocol was in accordance with the Declaration of Helsinki and was approved by the ethics committee of our institute (IRB #443).

| Electrode placement
Electrode locations for all patients are shown in Figure 1a. The electrodes were implanted in medial and lateral temporal, orbitofrontal, lateral frontal, and lateral parietal areas. After excluding the electrodes inappropriate for the analysis due to bad recording condition because of the disconnection of the electrode wire or floating of the electrode from the brain surface, the number of total electrodes and electrodes per patient was 792 and in the range of 63-107 [88 ± 14.3, mean ± standard deviation (SD)], respectively.

| Stimulus sites
Stimulus sites for all patients are summarized in Figure 1b, which shows the stimulus sites as the midpoints of electrode pairs in impairment in various neurological diseases, and we provided direct evidence for the functional subdivisions along the anterior-posterior axis in the human EC.

K E Y W O R D S
cortico-cortical evoked potential, electrical stimulation, entorhinal cortex, hippocampus, memory MNI standard space. The total numbers of stimulus sites in the EC and the fusiform gyrus (FG) were 24 and 20, respectively. The number of stimulus sites per patient varied between one and five (2.7 ± 1.4) for the EC and between one and five (2.2 ± 1.2) for the FG. Current intensities of the stimuli ranged from 6 to 12 mA (9.4 ± 1.6 mA).

| Cortico-cortical evoked potentials
The methodology of CCEPs has been previously described in detail (Matsumoto et al., 2004(Matsumoto et al., , 2007(Matsumoto et al., , 2012 The aim of CCEPs was to identify functional and seizure networks by stimulating the majority of implanted electrodes for research and clinical purposes Matsumoto, Kunieda, & Nair, 2017). We used the highest current intensity at which (a) the patient did not notice the stimulation and no apparent symptoms were evoked, (b) adjacent electrodes did not show excessive artifacts interfering with the recording, and (c) no afterdischarges were detected.
Current intensity was adjusted by increments of 1 or 2 mA reaching a maximum current between 6 and 12 mA. In some stimulus pairs, we could not reach a maximum current intensity (12 mA) due to symptoms such as pain, excessive artifacts in adjacent electrodes, or afterdischarges. Patients were awake during stimulation, except during the sleep investigations described below. Electrocorticogram (ECoG) recordings were referenced to a scalp electrode on the skin over the mastoid process contralateral to the side of electrode implantation.

| Definition of and inclusion criteria for P1w potentials
As stated in the Section 1, we considered the short-latency poten- 1. P1w is the earliest deflection of positive polarity.
2. The peak amplitude of P1w is more than three times larger than the SD within the baseline window.
3. The peak latency of P1w is in the range from +5 to +50 ms after the stimulation.
4. P1w is not followed by a typical or late N1 potential (described below).
We defined only potentials satisfying all four criteria as P1w.
Here, N1 is the earliest deflection of negative polarity. In typical CCEP recordings, we usually observe local responses near the stimulus site and discrete remote responses. Both nearby and remote CCEP waveforms typically show N1 and N2 potentials. Remote responses are usually spatiotemporally separated from nearby responses, and those responses form one or more electrical fields (herein defined as "remote isolated field") on the area far from the stimulus site (see also Figure 2a,b). Typically, CCEP studies consider only N1 with a peak latency between 5 and 50 ms after the stimulation ("typical N1"; Matsumoto et al., 2004Matsumoto et al., , 2007Matsumoto et al., , 2017; Figure S1A).
However, in our preliminary analysis, the peak latency of N1 evoked by either the EC or FG stimulation occasionally exceeded 50 ms, especially in electrodes located in lateral temporal and parietal areas ("late N1"; Figure S1B).
Based on these observations, we defined N1 as the earliest deflection of negative polarity with an amplitude exceeding six times the SD within the baseline window (Keller et al., 2014). In addition, we defined that it is between +5 and +50 ms when the upslope of N1 crosses the six SD threshold for the first time ( Figure 2d, Figure S1).  The caveat is that in our previous CCEP study of the dorsal language network, we called the preceding trough of N1 with a positive polarity as P1, which is regarded as the first component of a direct white matter pathway (Yamao et al., 2014). That P1 is different from P1w potentials investigated in this study.

| Representative indices of P1w potentials
We analyzed peak latencies and peak amplitudes along with their presence as representative indices of P1w and compared them for EC and FG stimulation. We also investigated whether the locations of stimulus sites within the EC had an influence on these parameters.
Recording electrodes located within 25 mm from a stimulation electrode were excluded from the analysis to eliminate local effects, presumably through short U-fibers (Keller et al., 2011). Electrodes in bad recording conditions, for example, disconnected or floating electrodes, were also excluded. For the detection of P1w peaks, we used custom MATLAB scripts (developed by M.M.).

| P1w peak amplitude (μV)
In a preliminary analysis of P1w evoked by EC stimulation, stimulus artifacts were almost always followed by a small notch of negative polarity. We defined the P1w peak amplitude as the voltage difference between the peaks of the poststimulus negative notch and the P1w potential ( Figure 2d). The timing of the poststimulus negative notch was determined for each stimulus site by visual inspection.
When P1w had multiple positive peaks, the largest peak was used.

| P1w peak latency (ms)
The peak latency of P1w was defined as the time between stimulation and P1w peak (Figure 2d).

| P1w presence
We also analyzed how many electrodes per stimulus site recorded a P1w potential. We assigned each recording electrode to a nominal "P1w presence" scale, with P1w being either present or absent.

| P1w presence ratio (%)
This parameter was defined as the percentage of electrodes that recorded a P1w potential. Subdural electrodes within 25 mm from the stimulus site were excluded.

| Contour map of P1w peak latency and amplitude
To analyze the effects of stimulus locations within the EC on P1w peak latencies and amplitudes, we generated 3D contour map in MNI standard space for both indices on the cortical surface of the EC, which reflected P1w peak latencies and amplitudes. The midpoint of each electrode pair was employed to represent the coordinates of this stimulus site. P1w indices were averaged among P1w-present electrodes for each stimulus site.
Then, these averaged values were plotted at the midpoint of each stimulation electrode pair. We also correlated P1w peak indices, that is, latency and amplitude, with the locations of EC stimulus sites. The correlation was investigated along the X-and Y-coordinates of the MNI standard space, representing the medial-lateral and anterior-posterior axes, respectively.

| Anatomical definition of the EC and FG stimulus sites
Based on the method used by Franko, Insausti, Perula, Insausti, and Chavoix (2014), the anterior and posterior borders of the EC were defined according to positions of key landmarks (e.g., limen insulae and the end of the gyrus intralimbicus) in the preoperative brain MRI. An MRI taken during implantation was coregistered to the preoperative MRI to determine whether an electrode was located within the border of the EC in the anterior-posterior axis. The lateral border of the EC was defined as the collateral sulcus (ColS).
The ColS was also identified in MRIs taken during implantation. The inclusion criteria for EC stimulation were defined as follows: 1. Anterior-posterior axis: Either electrode of the stimulus pair must be located in the EC.
2. Medial-lateral axis: Both electrodes of the stimulus pair must be located in the parahippocampal gyrus medial to the ColS.
To clarify the features of P1, we used the fusiform gyrus as a control stimulus site. The inclusion criteria for FG stimulation were defined as follows: 1. Anterior-posterior axis: Both electrodes of the stimulus pair must be located in the FG. The ColS and occipito-temporal sulcus represent the medial and lateral boundaries of the FG, respectively.
2.8 | Differences in functional connectivity between the EC and FG stimulation: the number of remote isolated fields the reviewers, the remote isolated field was defined by agreement.

| Anatomical localization of electrodes in individual and standard spaces
The methods of standard electrode placement and coregistration to the MNI standard space are described in detail elsewhere (Matsumoto et al., 2004(Matsumoto et al., , 2011

| Cortico-cortical evoked potentials during sleep
To exclude the possibility that P1w potentials were stimulus artifacts, we also investigated the state-dependent changes of P1w

| Data analysis
CCEP analysis, including auto-detection of P1 peaks, was performed by MATLAB (R2013a) and custom scripts developed by M.M.
Statistical analyses were performed with JMP Pro 11. 3D projection of P1w latency and amplitude values onto the brain surface in the MNI standard space was achieved by custom MATLAB scripts (developed by T.N.).

| Comparison of P1w indices between EC and FG stimulation
To

| Number of remote isolated fields
For conventional remote CCEP responses, the number of remote isolated fields per stimulus site was significantly smaller in EC stimulation compared to FG stimulation (Figure 3c; EC stim., 1.8 ± 1.9 [n = 24]; FG stim., 4.6 ± 2.1 [n = 20]; p = .0002, Mann-Whitney test). In EC stimulation, the local maxima of remote isolated fields were located mainly in the temporal lobe and the orbitofrontal area, which is consistent with previous human CCEP and monkey tracer studies (Munoz & Insausti, 2005). On the other hand, in FG stimulation, the local maxima of remote isolated fields were located not only in these areas but also in the lateral prefrontal and lateral parietal areas (see Figure S2).

| Sleep CCEPs
The P1w presence ratios were 100% at all sleep stages in all patients except for the awake (W) period of patient I with 91.9% (Figure 4).
Thus, we did not analyze P1w presence ratios further in the sleep CCEP study because the statistical difference of P1w presence ratio among sleep stages is supposed to be absent.  probably due to the limited number of subjects (n = 9). In addition, we analyzed the correlation between WMS-R MQ scores and P1w peak amplitude in FG stimulation. However, there were only mild, nonsignificant correlations (P1w peak amplitude-visual MQ, r = .34, p = .37).

| Giant evoked potentials in the hippocampus and EC near the stimulus site
We found that EC stimulation evoked local responses with very high amplitude, up to a few millivolts ( Figure 6). In one patient, patient F, evoked responses with high amplitude up to one millivolt were also recorded at the subdural electrode in the EC near the stimulus site, and at the depth electrode in the anterior hippocampus ( Figure 6). In this patient, the earliest peak latencies of such giant evoked potentials preceded the peak latency of P1w by a few milliseconds.

| Identified features of P1w induced by EC stimulation
EC stimulation induced in the majority of implanted neocortical electrodes unique evoked potentials with positive polarity, termed  as P1w. The prominent features of P1w in EC stimulation were their waveform similarity, high occurrence rate, short peak latency, and small peak amplitude, demonstrated by a comparison of P1w indices during EC and FG stimulation (Figure 3a,b,d).
We consider the two main reasons for the lack of reports about P1w in previous CCEP studies. First, those CCEP studies mainly focused on spatially localized discrete responses with a typical pattern of negative polarities (N1 and N2; Keller et al., 2014;Swann et al., 2012;Yamao et al., 2014); these features are in contrast to those of P1w, which include positive polarity or nonlocalized spatial distribution. Second, the CCEP studies of the limbic system usually focused on the responses recorded in the depth electrodes, which cover the much narrower neocortical regions than the subdural electrodes do.

| Origin of P1w
The duration of P1w was so long that a stimulus artifact seems unlikely as the origin of P1w. Sleep CCEPs demonstrated the modulation of P1w peak amplitudes by sleep stages (Figure 4) similar to conventional N1/N2 amplitudes . The absence of discrete evoked potentials in the reference electrode, which was located on the contralateral mastoid process, excluded the possibility that P1w were merely a consequence of a reference electrode activation ( Figure S3). These observations led us to regard P1w as the real signal derived from the brain.
The waveform similarities of P1w across recording sites led us to consider the possibility that P1w is equivalent to the far-field potential by the volume conduction. In addition, in three patients, in whom we added the scalp electrode for the sleep staging in sleep CCEP, the P1w potentials were recorded in not only the intracranial subdural electrodes but also in the scalp electrodes ( Figure S4), supporting the hypothesis that the origin of P1w is the far-field potential. In the electrophysiological research field, when the similar waveforms are recorded regardless of the locations of the recording electrodes in the monopolar montage, such responses are typically regarded as far-field potentials. The concept of far-field potential was firstly introduced in the auditory brainstem response study (Jewett & Williston, 1971), followed by the somatosensory evoked potential study (Cracco, 1972). Cracco showed that the evoked potentials in the deep subcortical structures could be recorded as far-field potential in the broad areas on the scalp by the volume conduction.
Previous studies of rodent models showed that the volume-conducted currents from the hippocampus and EC contributed to the local field potentials recorded in the neocortex (Bland & Whishaw, 1976;Gerbrandt, Lawrence, Eckardt, & Lloyd, 1978;Sirota et al., 2008). We found that EC stimulation evoked giant local responses in the EC near the stimulus site and the hippocampus and that their earliest peak latencies preceded the peak latency of P1w by a few milliseconds ( Figure 6). In the recent human CCEP study using an effective connectivity model, Krieg et al. (2017) revealed that the hippocampus could act as a signal amplifier of afferent information flow F I G U R E 4 Cortico-cortical evoked potential modulation during sleep. (a) P1w peak amplitude variations at different sleep stages in patients F, H, and I. *p < .0001, repeated-measure ANOVA for each patient; **p < .05, Tukey-Kramer HSD post hoc analysis. (b) P1w peak latency did not change at different sleep stages. L, light sleep; REM, REM sleep; SWS, slow-wave sleep; W, awake. Black lines: mean ± standard error We found that the values of P1w peak latencies in EC stimulation were surprisingly consistent despite variable distances between the EC and recording sites (Figure 2a,c). However, in a typical CCEP, the N1 onset and peak latency vary according to the fiber length between the stimulus and recording site (Matsumoto et al., 2012).
Thus, direct cortico-cortical projections are unlikely to be the source of P1w.
If P1w were derived from the indirect cortico-subcortical-cortical projections, the hippocampus and the thalamus are the two candidates for a subcortical relay point. The mean peak latency of P1w was 20.1 ms. Yeckel and Berger (1990) showed in a rabbit study that the electrical stimulation of EC afferents to the hippocampus caused multisynaptic excitation of CA1 pyramidal neurons via the trisynaptic pathway with latencies of 16-21 ms. Thus, from the point of conduction velocity, the EC-hippocampus-thalamo-cortical projection is unlikely to be the origin of P1w.
In summary, we regard the far-field potential from medial temporal structures as the most likely origin of P1w. However, we provided only indirect supportive evidence for this hypothesis. We hope a further study to clarify whether P1w is a far-field potential or not.

| P1w evoked by FG stimulation
We observed that the stimulation of the FG also elicited P1w. The FG is adjacent to the perirhinal cortex. The perirhinal cortex is located within the collateral sulcus (Figure 1b, right lower panel) and has strong connections with both the hippocampus and EC. According to the electrical density simulation study (Nathan, Sinha, Gordon, Lesser, & Thakor, 1993), the relatively high stimulation intensity we used for the fusiform gyrus might have activated the perirhinal cortex at the superficial part of the collateral sulcus. Then, the stimulation of the perirhinal cortex would have activated both the hippocampus and EC and resulted in the generation of P1w.

| Neuroscientific implications: memory and P1w
We uncovered that the peak amplitudes of P1w were smaller in patients with a lower visual MQ score ( Figure 5). That correlation between P1w peak amplitude and WMS-R score, which may be explained by the axonal loss in the memory circuit from the EC to the hippocampus due to intractable epilepsy, supports the idea that P1w reflects the activity of the medial temporal lobe memory circuit. The changes of P1w peak amplitudes at different sleep stages (Figure 4) may reflect the sleep modulation on the activity of the memory circuit as the underlying mechanism for efficient memory consolidation during sleep.
Typically, verbal memory is considered as predominantly processed in the left hemisphere, which is in most cases dominant for language function, and visual memory in the right hemisphere.
However, as   (Evensmoen et al., 2015). Notably, the anterior-posterior gradient within the EC, as we found, provides direct evidence for the functional subdivisions along the anterior-posterior axis in the human EC.
Finally, the correlation between P1w peak amplitudes and WMS-R score also suggests that P1w potentials may be potential biomarkers of memory impairment. In three patients, in whom we added the scalp electrode for the sleep staging in sleep CCEP, the P1w potentials were recorded in not only the intracranial subdural electrodes but also in the scalp electrodes ( Figure S4)

| Limitations of the present study
There are several limitations in the present study. The number of subjects recruited for the study was relatively small (n = 9).
According to the clinical purpose of the functional brain mapping, such as language function mapping, the majority of the patients (7 out of 9 patients) had the electrode coverage on the left hemisphere.
The evoked responses to the electrical stimulation of the right EC were recorded in only two patients. Thus, it was difficult to confirm the characteristics of the responses to the right EC stimulation, or whether the unique CCEP responses shown in the present study were specific to the left EC stimulation.
As for the supposed generator mechanism of P1w, we provided only limited evidence supporting the hypothesis that P1w reflects a far-field potential from the medial temporal structures. The majority of the implanted electrodes were subdural electrodes. Therefore, the activity in deep structures of the brain, such as the hippocampus or the thalamus, was not recorded in most patients. Thus, the confirmation of the P1w generator mechanism was difficult.
In the present study, we could not standardize the current intensity of the stimulation, which ranged between 6 mA and 12 mA, due to symptoms such as pain, excessive artifacts in adjacent electrodes,   (1, 2, 4, 6, 8, 12 mA) in one patient (the patient G). However, further study with standardized stimulus parameters is needed to further validate the results in the present study.
Since all subjects suffer from medial temporal lobe epilepsy, we could not completely exclude the possibility that P1w is entirely a pathological response through aberrant epileptic networks involving the EC. However, two out of nine patients (patients C and E) did not have hippocampal sclerosis nor was the epileptic focus located in the parahippocampal gyrus (Table 1). This supports the hypothesis that P1w reflects the activity of a normal memory network.

| CON CLUS IONS
We clarified that human EC electrical stimulation evoked short-latency potentials on the broad neocortical regions. The features of those potentials, termed as P1w, were positive polarity, short peak latency (mean 20.1 ms), and waveform uniformity. The origin of P1w remains unclear, although the limited evidence from the present study suggests that P1w is the far-field potential by the volume conduction of giant evoked potential from the medial temporal structures, especially the EC itself and hippocampus. A stimulation of more posterior parts of the EC induced P1w with shorter latency and larger amplitude. Our results support to a certain extent the idea that P1w reflects memory circuit activity, as suggested by the correlation between the WMS-R score and the peak amplitude of P1w, although that correlation did not survive a correction of multiple comparisons. The significance of the present study is that those evoked potentials may be a potential biomarker of memory impairment in various neurological diseases, and we provided direct evidence for the functional subdivisions along the anterior-posterior axis in the human EC.

Department of Epilepsy, Movement Disorders and Physiology
is an endowment department, supported with a grant from GlaxoSmithKline K.K., NIHON KOHDEN CORPORATION, Otsuka Pharmaceutical Co., and UCB Japan Co., Ltd. Department of Respiratory Care and Sleep Control Medicine is the department donated from Teijin Pharma Limited, Philips Respironics GK, Fukuda Denshi CO., LTD., and Fukuda Lifetech Keiji CO., LTD.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.