Effects of Early-Life Seizures on Coordination of Hippocampal-Prefrontal Networks: Influence of Sex and Dynamic brain states

Abstract

Objective:

Early-life seizures (ELS) alter activity-dependent maturation of neuronal circuits underlying learning and memory. The pathophysiological mechanisms underpinning seizure-induced cognitive impairment are not fully understood and critical variables such as sex and dynamic brain states with regard to cognitive outcomes have not been explored. We hypothesized that in comparison to control (CTL) rats, ELS rats would exhibit deficits in spatial cognition correlating with impaired dynamic neural signal coordination between the hippocampus and medial prefrontal cortex (mPFC).

Methods:

Male and female rat pups were given 50 flurothyl-induced seizures over 10 days starting at postnatal day (P) 15. As adults, spatial cognition was tested through active avoidance on a rotating arena. Microwire tetrodes were implanted in the mPFC and CA1 subfield. Single cells and local field were recorded and analyzed in each region during active avoidance and sleep.

Results:

ELS males exhibited avoidance impairments while female rats were unaffected. During avoidance, hippocampus-mPFC coherence was higher in CTL females than CTL males across bandwidths. In comparison to CTL males, ELS male learners exhibit increased coherence within theta bandwidth as well as altered burst-timing in mPFC cell activity. Hippocampus-mPFC coherence levels are predictive of cognitive outcome in the active avoidance spatial task.

Significance:

Spatial cognitive outcome post ELS is sex-dependent as females fare better than males. ELS males that learn the task exhibit increased mPFC coherence levels at low-theta frequency which may compensate for ELS effects on mPFC cell timing. These results suggest that coherence may serve as a biomarker for spatial cognitive outcome post ELS and emphasize the significance of analyzing sex and dynamic cognition as variables in understanding seizure effects on the developing brain.

1. Introduction

While seizures are considered the defining feature of childhood epilepsy, epilepsy is a spectrum disorder with long-term cognitive, behavioral, and psychiatric symptoms. Among these, cognitive abnormalities are the most common and disabling seizure co-morbidities.¹ Of particular concern, in infants and young children pharmacoresistant seizures are associated with high rates of life-long deficits in cognition.²

Spatial memory is the process of continuous encoding and retrieval ³ of allothetic and egocentric information underlying the ability to self-localize in space and time.^{4, 5} The orchestration of stereotyped patterns of activity in the developing entorhinal cortex, hippocampus and mPFC (medial Prefrontal Cortex) circuits is necessary for the ontogeny of memory and spatial information processing by optimizing plasticity at appropriate synapses.^{6, 7} The importance of these developmental processes for higher order cognitive abilities is exemplified by the establishment of functional connectivity between the mPFC and hippocampus. In the adult brain, the well-timed, coherent coactivation of mPFC and hippocampal networks within oscillatory rhythms⁸ underscores memory, decision making, goal encoding and avoidance.^9–12

Discrete oscillatory bandwidths have been proposed to route communication within and between subregions of the hippocampus ^13–15. CA1 activity synchronizes with neocortical input during the acquisition of information via oscillations in the medium gamma range (60-90 Hz), and synchronizes with the CA3 during long-term memory recollection^{15, 16} or the prefrontal cortex during working memory¹⁷ via oscillations in the slow gamma range (30-60 Hz). Oscillatory activity in these bandwidths emerges within the first three postnatal weeks in rodents, driving the development of spatial processing circuits.^18–20 Development of this oscillatory maturation parallels the development of allocentric spatial learning and memory, arising between postnatal (P) day 21 and P25 in the rodent.²¹ The integration of spatial and temporal information via coordinated hippocampus and mPFC oscillatory activity²² has been proposed to begin at the end of the first postnatal week.⁶ Whereas normal activity patterns are required for circuit maturation, less is known about how aberrant neuronal activity during this sensitive developmental period disrupts the ontogeny of spatial cognition.²³ Seizures are particularly disruptive pathological events in the developing brain as they result in massive bursts of synchronized network activity across brain regions²⁴. There are considerable data indicating the disruptive effects of seizures on developing hippocampal oscillations,⁷ and studies have examined the effects of seizure on the balance of signal coordination on the hippocampal-mPFC circuit with regard to working memory²⁵ and sociability²⁶ deficits. However, the effects of early-life seizures (ELS) on the hippocampal-mPFC circuit in relation to spatial memory deficits²⁷ remain poorly understood. Moreover, ELS cognitive outcome variables such as dynamic coordination efficacy in the hippocampal-mPFC circuit and sex ^28–30 have not been studied.

In this study, the effect of seizures during critical developmental periods on spatial cognition in both male and female rats was evaluated. Our hypothesis is that seizures during early development will alter cellular and oscillatory activity within and between the hippocampus and mPFC. We further hypothesize that the direction and degree of this alteration will determine spatial cognitive outcomes. We assessed the effects of critical period ELS on the coordination of neural signals within and between the mPFC and hippocampus during two states: 1) Active avoidance, where dynamic coherence was measured while the animals were actively engaged in avoiding the shock zone on a rotating arena; and 2) Sleep, where coherence was also measured during Slow Wave Sleep (SWS) and Rapid Eye Movement (REM) stages.

2. Methods

2.1. Animals and Seizure Induction

The study design is provided in Fig. 1A. Sprague-Dawley (Charles River, Montreal) rats were subjects in this study. Pups were weaned at P21 and group housed until approximately P35-40, at which time they were individually housed. Rats were maintained on a 12-hour light/dark cycle. Experimental design, performance and analysis aimed to adopt guidelines for rigor and reproducibility in science.³¹ Group sizes were determined a priori, and animals were randomly assigned to experimental groups. Male (M) and female (F) rat pups (M=17; F=16) were split into control (CTL; M=7; F=8) and Early-Life Seizure (ELS; M=10; F=8) groups. As a model of recurrent generalized tonic-clonic seizures^{25, 26, 32} seizures were chemically induced with flurothyl, a GABA_A antagonist³³. ELS animals were given 5 flurothyl induced seizures a day over a 10-day period between P15 to P24 (Total = 50 seizures). Development of oscillatory maturation in key bandwidths^18–20 parallels the development of allocentric spatial learning and memory, arising between postnatal day 21 and P25 in the rodent.²¹ The P15 to P24 time period was therefore selected to test the putative disruptive effects of ELS flurothyl seizures on the spatial processing circuit and corresponding performance of a complex spatial task. All pups were placed in an octagon shaped plastic container (17.8 cm wide, 11.4 cm high, 7.6 cm at a side) set in an airflow hood. Each pup was positioned in their own wedge of the octagon (Supp. Fig. 1) and facing the open central portal of the container. Approximately 0.1 ml of undiluted flurothyl solution (Bis(2,2,2-trifluoroethyl) ether, 98% pure; Sigma-Aldrich) was loaded into a 1 ml syringe and doses of approximately 0.02 ml were injected by hand onto filter paper in the central portal of the container with each dose separated by 2 mins. The flurothyl evaporated, was inhaled by the pups and caused convulsions, typically between the 2^nd and 4^th dose. Pups were removed from the flurothyl upon tonic extension of both forelimbs and hindlimbs and placed in a holding container. After all pups had exhibited tonic-clonic seizure, they were reunited with the dam. Seizures were spaced every hour, starting at the initial flurothyl exposure. CTL animals experienced no seizures but were separated from dams for a similar amount of time as ELS animals, in order to control for maternal deprivation stress.

Figure 1:

Study design and number of shocks received over the course of 16 training sessions in the active avoidance on the rotating arena. A) Study Design. ELS flurothyl-induction was carried out P15 to P25. Male and female rats then underwent an assay of spatial cognition via the active avoidance task on a rotating arena. Rats were then implanted with a 4-tetrode array in the PFC and 8-tetrode array in CA1. Following recovery, rats underwent LFP and unit recordings while sleeping or engaged in active avoidance on a rotating arena. EEG during both the awake and sleep state were evaluated for spectral power and coherence between the hippocampus and PFC. Sleep recordings were also evaluated for phase locking and number of sharp wave ripples (SWR). B) ELS had no effect on spatial learning or memory in female rats. Both ELS and CTL females reached criterion of 5 or fewer shocks in consecutive sessions by the first day of training (Sessions 1-8). CTL males reached criterion by day 2. ELS males received the most shocks on the second day of training and 40% failed to reach learning criterion. Relative to session 16 in CTL females, the number of shocks on day 2 were equivalent for both female groups and CTL males. ELS males received significantly more shocks on day 2, therefore demonstrating a comparative spatial learning and memory deficit (* = p < 0.05; × = GEE comparator, session 16 CTL female group performance). C) Examples of behavioral performance on the rotating arena for a CTL female (Left) and an ELS male (Right). Colored lines indicate animal path. Green ‘+’ indicates LED and animal position. Shaded red sector indicates the shock zone.

All rats were evaluated for spatial memory and learning in the active avoidance rotating arena. At P70 rats were implanted with a 4 tetrode electrode array in the PFC and an 8 tetrode array in the CA1 region of the dorsal hippocampus. Following recovery, rats underwent local field potential (LFP) and unit recording while engaged in active avoidance or during sleep. EEG during both the awake state while engaged in the active avoidance spatial task and sleep state were evaluated for spectral power and coherence between the hippocampus and PFC. Sleep recordings were also evaluated for phase locking and number of sharp wave ripples (SWR examples shown in Supp. Fig. 2). All procedures were approved by the University of Vermont’s Institutional Animal Care and Use Committee and conducted in accordance with guidelines from the National Institutes of Health. Detailed methods and results of statistical analyses can be found in the supplementary materials.

3. Results

3.1. Active avoidance

Behavioral performance during active avoidance was assayed by measuring the number of shocks received in each session and group (CTL males [N = 7], CTL females [N = 8], ELS females [N = 8], ELS males [N = 10]). Several ELS male rats were unable to meet task criterion in the active avoidance task (6 learners, 4 non-learners) while most CTL male rats did (6 learners, 1 non-learner). One CTL male rat was removed from analysis as its shock pin repeatedly came out during training and interrupted its learning curve. Remarkably, all ELS females (8/8) and all CTL females (8/8) met task criterion. Chi-square analysis showed marginal differences between learners and non-learners in the four groups (chi-square statistic = 7.66, p = 0.0534).

General Estimating Equation (GEE) analysis revealed a significant sex effect (Wald value = 7.18, p = 0.007) where females (13.06 ± 3.10) received a significantly lower mean number of shocks than males (4.5 ± 1.39). There was also a significant group × sex × training session interaction for the number of shocks received during two days of active avoidance training (Wald value = 9.35 × 10¹², p < 0.0001) where CTL male rats improved performance by the 2^nd day of training while ELS male rats did not.

Relative to session 16 in CTL female rats (Fig. 1B), the number of shocks was similar to ELS females throughout training. Vaginal lavage for assessment of estrous cycle showed no difference in number of shocks by estrous stage (proestrus, estrus and diestrus; F = 0.3092, p =0.7368). From sessions 12 to 16, CTL males were on par with CTL females while ELS males received a significantly higher number of shocks (p < 0.05; see Supp. Table 1 for Wald and p-values per group, sex and session comparison). Performance between CTL and ELS female rats was equivalent throughout training, suggesting ELS females were unaffected. While performance of CTL male rats was equivalent to ELS females by the second day of training, ELS male rats demonstrated relative learning and memory impairments throughout training.

3.2. Electrophysiology during active avoidance

3.2.1. Signal coherence during the entire avoidance session

Behavioral analysis indicates a significant active avoidance deficit in ELS males in relation to CTL females. However, there was a distinction within this group with regard to the animals that were able to reach learning criterion and those that could not. We now ask if there were distinct physiological differences with regard to the efficacy of signal coordination between the mPFC and hippocampus in relation to cognitive outcome in active avoidance. Specifically, we wished to know if there were compensatory mechanisms at play in the ELS females and ELS male learners that abrogated cognitive deficit. Active avoidance performance in each group during electrophysiological recordings was similar to that seen at the end of training.

Coherences of mPFC and hippocampal LFPs across frequencies in each bandwidth of interest for animals were assessed in each group (CTL male learners [N = 4], CTL Females [N = 7], ELS Females [N = 7], and ELS male learners [N = 6]) during either the entire active avoidance sessions or as a function of dynamic active avoidance epochs averaged across the session. Data from ELS male non-learners [N = 3] were also collected but were excluded from the GEE analysis due to a lack of statistical power in relation to CTL male non-learners. These data corresponding to the analyses below can be found in Supp. Fig. 3.

Illustrations of signal data collected in relation to behavior during active avoidance and the approach toward statistical analysis can be found in Supp. Fig. 4A–E. Examples of raw or filtered CA1 and mPFC LFP signals during 20 seconds of recording in a CTL female during active avoidance are shown in Fig. 2A–D. Accompanying spectrograms for each raw or filtered trace are shown underneath each plot. Results for analyses in delta (Fig. 2E–F), slow gamma (Fig. 2I–J) and medium gamma (Fig. 2K–L) can be found in supplementary materials.

Figure 2:

Signal examples and mPFC-CA1 coherence across the whole active avoidance recording session in each bandwidth. A-D) Raw LFP trace in CA1 (Top) and mPFC (Bottom recorded from a CTL female) over approximately 10 seconds in the delta/theta range with corresponding spectrograms (A). Filtered LFP signals from the same location and time series are also shown for Hz theta (B), slow gamma (C) and medium gamma (D); E-F) Group × frequency plots (E) and group plots across bandwidth (F) showing that delta frequency coherence is highest in CTL females and lowest in ELS male non-learners. Delta coherence in ELS male learners is higher than CTL males; G-H) Group × frequency plots (G) demonstrates that theta coherence is highest at ~ 8 Hz for CTL males, CTL females exhibit multiple peaks throughout the bandwidth and both ELS females and ELS male learners exhibit peaks at low and mid-range theta. Across theta bandwidth (H), theta frequency coherence is highest in CTL and ELS females and lowest in CTL males. Theta coherence is higher in ELS male learners than CTL males, due in part to the additional low-theta coherence peak; I-J) Group × frequency plots (I) and group plots across slow gamma bandwidth (J) where coherence is highest in CTL and ELS females and equal between CTL males and ELS male learners; K-L) Group × frequency plots (K) and group plots across medium gamma bandwidth (L). Medium gamma coherence is highest in CTL and ELS females and equal between CTL males and ELS male learners. Females exhibit higher coherence than males and ELS male learner coherence is higher than CTL males at low theta frequencies and delta. Color of asterisk indicates significant group differences (p < 0.05) in comparison to ELS male learners.

In the 5-12 Hz theta bandwidth, there was a significant interaction between group × sex × frequency (Wald value = 273.75, p < 0.0001). CTL females showed the highest coherence values while CTL males exhibited the lowest values (Fig. 2G). Coherence was significantly higher in CTL females throughout the bandwidth (p < 0.00001) while in ELS male learners and ELS females, coherence increased significantly between 4.9 and 10.3 Hz (p < 0.05; see Supp. Table 3A for detailed results). Though all groups had a peak theta coherence at ~ 7.8 Hz, only the CTL males did not show an additional low-theta peak.

A significant group × sex interaction (Wald value = 100.87, p < 0.0001) showed that coherence in ELS male learners (Fig. 2H) was lower than ELS females (Wald Value = 6.88, p = 0.009) and CTL females (Wald Value = 62.28, p < 0.0001) but higher than CTL males (Wald Value = 13.14, p < 0.0001). As suggested by the above analysis, lower overall theta coherence in CTL males was likely due to the more selective coherence at ~ 7.8 Hz.

Receiver Operator Characteristic (ROC) analyses tested the predictability of task performance as a function of mPFC-hippocampus coherence within each bandwidth (Fig. 3A–D). The area under the curve (AUC) in each analysis indicates that mPFC-hippocampus coherence in all bandwidths but delta are good predictors of cognitive outcome.

Figure 3:

Receiver Operator Characteristic (ROC) curves for analyzing diagnostic ability of coherence in each bandwidth as a predictor of cognitive outcome. Whole session coherence values within each bandwidth were analyzed for delta (A), theta (B), slow gamma (C) and medium gamma (D). Delta coherence was the only poor predictor of cognitive outcome in the active avoidance task. The red line at the diagonal represents chance levels of diagnostic ability.

3.2.2. Dynamic LFP theta coherence during active avoidance

As the most pronounced whole session group × frequency interactions were in the theta bandwidth, we conducted a detailed analysis of dynamic theta coherence relative to avoidance epochs. As illustrated in Supp. Fig. 4, we used changes in acceleration to infer active avoidance arcs used by rats in each group during each session on the rotating arena. These avoidance arcs allow analysis of the effect of both movement and changes in cognitive demand, such as recall versus increased sensorimotor processing, on the coherence of signals between the dorsal hippocampus and mPFC in each bandwidth of interest. The results of statistical analysis for coherence and avoidance epoch interactions (Fig. 4A–E) can be found in supplementary material.

Figure 4:

A-E) Dynamic CA1-mPFC theta coherence relative to peak acceleration epochs −3 to −1 seconds prior to peak acceleration (Left), at peak acceleration (Middle), and +1 to +3 seconds post peak acceleration (Right) for CTL females (A), ELS females (B), CTL males (C), ELS male learners (D). GEE analysis revealed that the most prominent significant coherence changes within the theta bandwidth occurred at −1 second prior to peak acceleration in all but ELS males. Relative to avoidance epochs, ELS male learners showed less coherence dynamism than CTL Males. The dynamic coherence analysis from group × epoch × frequency analyses was collapsed to group × frequency to analyze dynamic signals across epochs for each group (E-H). CTL females and ELS females exhibited similar dynamic coherence patterns at low and mid-range theta. ELS male learners likewise exhibited a peak coherence at low-theta frequency. In contrast to the other groups, CTL males were most coherent at mid to high-theta frequency. K) ROC curves for dynamic theta coherence at 8 Hz across all active avoidance epochs and at −3 to + 3 seconds from peak acceleration. Theta coherence oscillations at 8 Hz have the greatest diagnostic ability for cognitive outcome post ELS at epochs associated with memory recall (−3 seconds) and while actively moving away from the shock zone (+1 second).

As the group × epoch × frequency interactions during avoidance are complex, the dynamic theta analysis was collapsed across epochs in order to compare these results with the averaged coherence across the session. In accordance with the whole session analysis, collapsing dynamic theta coherence across epochs revealed that the key difference across groups was the degree to which peak values fell at approximately 6 Hz or between 7.5 Hz – 9.5 Hz (Fig. 4F–J). Therefore, the low-end of the theta bandwidth in the CTL females was used as a reference for GEE analysis, revealing a significant group × frequency interaction (Wald value = 332.47, p < 0.0001). For CTL females, mean dynamic coherence at 5.4 Hz was equal to coherence at ~ 7.5 Hz but was significantly higher than values at frequencies greater or equal to 8.3 Hz (p < 0.05, Supp. Table 7). In comparison to ELS females, no significant differences were shown at low-theta or across the bandwidth (p > 0.05). Similarly, coherence peaked in ELS male learners and non-learners at 5.4 Hz. For ELS male learners, coherence remained high until 7.4 Hz (p > 0.05, Supp. Table 6). In contrast to all other groups, CTL males tended to exhibit little coherence in the low-theta range. Coherence values were significantly lower at all frequency ranges in CTL males, except at 7.3 Hz and 9.3 Hz (p > 0.05) where coherence tended to peak. In a direct comparison to CTL females at theta frequencies > 7.5 Hz, all groups exhibited similar coherence values between 8.38 and 10.36 Hz, (p > 0.05, Supp. Table 8). Mean dynamic coherence across epochs are in agreement with whole session analysis. Coherence is highest at low-theta frequencies for all groups but CTL males. Increased low-theta coherence in ELS males may therefore abrogate cognitive deficit in the spatial domain as well as in working memory.²⁵

ROC analyses tested the predictability of cognitive outcome as a function of mPFC-hippocampus coherence at 8 Hz, where dynamic theta in ELS male non-learners was the lowest (Fig. 4K). Predictability of outcome for dynamic 8 Hz theta across epochs was similar to predictability for theta bandwidth during the entire session. However, 8 Hz theta coherence at −3 secs and +1 sec were also good predictors of cognitive outcome. These epochs are noteworthy as at −3 seconds rats are more likely to recall the shock zone location, while at +1 sec the rats are actively navigating away from the shock zone.⁷

3.2.3. Cell burst rhythmicity during active avoidance

Burst rhythmicity analysis of cell activity in the mPFC and CA1 compared the timing of cell activity in these regions with the frequency of dynamic CA1-mPFC LFP theta coordination. Autocorrelation examples used in this analysis are shown for a sample CA1 cell (Fig. 5A) and an mPFC cell (Fig. 5B) and illustrate robust mid-theta spike timing modulation. Hippocampal pyramidal cells were isolated in CA1 from rats in each group (CTL-F = 25 cells, CTL-M = 26 cells, ELS-F = 12 cells, ELS-M = 70 cells; Fig. 5C) during active avoidance.

Figure 5:

Example of burst-rhythmicity analysis of a CA1 (A) and mPFC cell (B) via autocorrelation. C-D) Number of cells in each brain region and distribution curve (black line). The majority of hippocampal CA1 cells in all groups burst at ~ 8.5 Hz (C), while the majority of mPFC cells have bimodal distributions and burst at either 6 Hz or ~ 8.5 Hz. However, mPFC cells in ELS males burst at 6 Hz or ~ 10 Hz mirroring the tendency of dynamic CA1-mPFC theta coherence to be high at the extremes of theta but incoherent at ~ 8 Hz. This distribution was significantly different between ELS and CTL males (* = p < 0.05).

In ELS-M, more cells were recorded from non-learners than learners in the hippocampus (N=50 versus N=20) and mPFC (N=74 versus N=36). GLM found no statistical difference in the mean burst rhythmicity of hippocampal cells (ELS-MNL = 8.23 ± 0.323 Hz, ELS-ML = 8.50 ± 0.510 Hz; p = 0.650) or mPFC cells (ELS-MNL = 8.89 ± 0.328 Hz, ELS-ML = 8.63 ± 0.471 Hz; p = 0.655) between each subgroup. A two sample Kolmogorov–Smirnov (KS) test likewise found no significant differences in the distribution of cell rhythmicity in hippocampal (z = 0.567, p = 0.905) or mPFC cells (z = 0.765, p = 0.602). ELS male learners and non-learners were therefore pooled in the following analysis.

The General Linear Model (GLM) analysis found no significant group effect for mean burst rhythmicity of hippocampal cells (CTL-F = 8.17 ± 0.433 Hz, CTL-M = 8.30 ± 0.425 Hz, ELS-F = 8.60 ± 0.626 Hz, ELS-M = 8.31 ± 0.259 Hz; p =0.956), nor was there an effect for the distribution of cell rhythmicity between CTL and ELS males (KS test; z = 0.560, p = 0.913) or between CTL and ELS females (Kolmogorov-Smirnov (KS) test; z = 0.883, p = 0.417).

Cells from the prelimbic region of the mPFC were also isolated from rats in each group (CTL-F = 88 cells, CTL-M = 68 cells, ELS-F = 53 cells, ELS-M = 110 cells; Fig. 5D). Again, GLM found no group effect for mean burst rhythmicity (CTL-F = 7.9 ± 0.275 Hz, CTL-M = 8.23 ± 0.314 Hz, ELS-F = 8.29 ± 0.356 Hz, ELS-M = 8.80 ± 0.247 Hz; p = 0.113) and the KS test found no significant effect of frequency distribution between CTL and ELS females (z = 0.883, p = 0.417). However, the KS test did find a significant difference in frequency distribution between CTL and ELS males (z = 1.39, p = 0.043).

Unlike other groups, mPFC cells in ELS males exhibited a bimodal distribution in their rhythmicity and were either < 8 Hz or > 8 Hz. In ELS male non-learners, this pattern closely corresponded to LFP coherence peaks at theta bandwidth frequency extremes (Supp. Fig. 5A–B). The absence of the 8-9 Hz mPFC frequency range in ELS male non-learners may represent compounded levels of CA1-mPFC neural discoordination; one at the network level and the other at the level of spike-timing.

3.3. Sleep Electrophysiology

In order to compare mPFC-hippocampus signal coordination efficacy between ELS and CTL during other brain states associated with learning and memory, ^{34, 35} without the influence of active movement, we also measured signal properties in each group during SWS and REM sleep. Relative power in delta, theta, slow gamma and medium gamma bandwidths were recorded during SWS and REM sleep in the hippocampus and mPFC. As expected, during SWS delta was the dominant frequency in both the hippocampus and mPFC. In REM theta was dominant in the hippocampus but not the mPFC where relative power of delta and theta were similar (Fig. 6A–D). There were no differences in relative power between the four groups in SWS and REM in either the hippocampus or mPFC (p > 0.05). In SWS sleep there were group differences in total coherence (F(3,121)=8.082, p<0.0001), delta (F(3,121)=5.726, p=0.0011), theta (F(3,121)= 4.261, p=0.0067), slow gamma (F(3,121)=4.395, p=0.0057) and medium gamma (F(3,121)= 4.442, p=0.0053) with Tukey’s multiple comparisons test showing decreased total (p<0.05), delta (p<0.01), slow gamma (p<0.001) and fast gamma (p<0.001) in the ELS males compared to CTL males during SWS (Fig. 6E). In REM sleep there were group differences in total coherence (F(3,121)=12.35, p=0.0063), delta (F(3,121)= 6.714, p= 0.0003) and medium gamma (F(3,121)=4.518, p=0.0053) with Tukey’s multiple comparisons test showing decreased total (p<0.001), delta (p<0.01), and slow gamma (p<0.01) in the ELS males (Fig. 6G). To summarize, during SWS ELS males have the lowest coherence in all bandwidths During REM, the ELS males also have the lowest phase locking in all bandwidths.

Figure 6:

A-D) Relative power in the delta, theta, slow gamma and medium gamma during SWS in the hippocampus (A) and mPFC (C) and during REM in the hippocampus (B) and mPFC (D). During SWS, delta is the dominant frequency in both regions whereas during REM theta was dominant in CA1 but not the mPFC where relative power of delta and theta were similar. There were no group differences in relative power during SWS or REM. Coherences and phase locking during SWS and REM sleep (E-H). During SWS (E) and REM (G) coherence was low in ELS males. There were significant group differences in total coherence across bandwidths and in the delta, slow and medium gamma during SWS (E) and in total, delta and slow gamma during REM (G). Phase locking paralleled the trend found in coherence. Significant differences were found across most bandwidths in both SWS (F) and REM (H). There were no significant coherence differences between the CTL and ELS female rats. (^ = p<0.05; + = p<0.01; # = p<0.001; * = p<0.0001 indicate significant differences between CTL and ELS males).

4. Discussion

The primary finding of this study was that ELS resulted in spatial cognitive deficits in male but not female rats. Cognitive deficits in ELS males were consistent with prior studies demonstrating that seizures during the critical period of physiological connectivity underpinning spatial cognition causes deficits in several tests measuring spatial cognition.^{7, 36, 37} The secondary finding was that CA1-mPFC LFP coherence was a robust predictor of active avoidance performance via ROC curve analysis, revealing coherence as a potential biomarker of cognitive outcome. Within ELS males, those that were able to learn the active avoidance task also demonstrated higher signal coordination at delta and theta bandwidths than CTL males. A previous study hypothesized that increased CA1-mPFC coordination post ELS was a form of network compensation underpinning cognitive rehabilitation during epochs of short-term memory demands.²⁵ This hypothesis is now extended to suggest that increased theta coherence post ELS may in part be due to increased coherence at low-theta frequencies. In addition, increased mPFC-hippocampus theta coordination may compensate for cell timing abnormalities at specific frequencies and abrogate both short-term working memory²⁵ and spatial cognitive deficits. In contrast, CA1-mPFC signal coordination in ELS male non-learners was particularly incoherent at 8-9 Hz in theta bandwidth. Similarly, these animals exhibited an absence of mPFC cellular burst-timing rhythmicity at 8-9 Hz, an active modulation frequency in all other groups. However, without electrophysiological data from CTL animals that were unable to learn the task, it is possible that poor hippocampus-mPFC coordination at 8-9 Hz is related to sex as well as ELS. Since the CTL non-learners make up a small proportion of the male rats, future studies would need to incorporate a substantially larger cohort of rats in order to determine if low coherence and corresponding adverse cognitive outcomes are exclusively caused by ELS or mitigated by an interaction of sex and maternal separation stress.

That female rats fared better than male rats in spatial cognition post ELS was surprising, illustrating the importance of exploring sex as a variable for cognitive outcome post ELS.³⁰ Most prior studies have used only male rats, or found sex differences in cognitive outcomes when seizures were administered in adults.³⁸ Using a comparable approach with early-life seizure induction, Akman et al.³⁹ found that three episodes of kainic acid-induced status epilepticus on P4-6 caused more transient learning delays in the Barnes maze on P16-19 in males but not females. There are developmental physiological reasons that could account for the differences in either vulnerability to seizures or subsequent cognitive outcome in prepubescent pups. While it is unclear if the presence of sex hormones such as estradiol, even at early developmental stages, may offer some protection against early-life neurological insults^40–42, there are established sex differences early in postnatal development relative to depolarizing GABAergic signaling.⁴³ There is a sex-dependent developmental shift of GABA conductance and chloride gradient, leading to differences in the timing of key developmental stages in males and females.⁴⁴ Specifically, hyperpolarizing reversal potentials of GABAergic postsynaptic currents (EGABA) appear earlier in female than male rats because of differential expression of the chloride transporters.⁴⁴ Recurrent induced status epilepticus elicited daily at postnatal days P4-5, reverses the direction of GABAergic responses in both sexes but in opposite directions. In males, recurrent status epilepticus triggered a premature appearance of hyperpolarizing GABAergic signaling at P9, instead of P14, whereas in females EGABA transiently becomes depolarizing at P8–13.⁴⁴ Early cessation of depolarizing GABAA receptor signaling may disrupt neuronal differentiation, as shown for the development of excitatory dendritic spines.^{43, 45}

With regard to CA1-mPFC signal coordination during active avoidance, control females showed the highest coherence levels of all the groups at each bandwidth. ELS female theta coherence was more similar to ELS male learners than control females. Although ELS female coherence is reduced from a higher baseline than males, it is not sufficiently reduced to cause spatial cognitive deficits, as is the case for the ELS male non-learners. Both GEE coherence analysis and ROC analysis suggest there is a critical threshold of hippocampus-mPFC discoordination that is associated with poor cognitive outcomes post ELS. The higher baseline coherence in females may have been a contributing factor to the lack of a cognitive deficit in ELS females, preventing ELS female coherence from reaching this critical threshold.

Lastly, females in both the ELS and CTL groups exhibited an intrinsic flexibility in CA1-mPFC coherence across the theta bandwidth by having multiple peaks in the low and mid-theta ranges. In contrast, CTL males had only one peak at mid-theta. Again, these results indicate the importance of studying baseline physiological differences in cognitive processing by sex. Exploring these underlying sex differences is necessary for understanding the long-term effects of encephalopathic insults as a function of development.

Dynamic 8 Hz coherence was found to be a good predictor of cognitive outcome at −3 seconds and +1 seconds from peak acceleration during active avoidance. These epochs have been shown to be associated with increased memory recollection of the shock zone location and avoidance from the shock zone respectively. ^{7, 46} In line with these findings, CA1-mPFC theta synchrony has been shown to be critical for the transmission of behaviorally relevant information during epochs of anxiety ⁴⁷ or avoidance.^{12, 48} Optogenetic inhibition of direct ventral hippocampal-mPFC projections reduces corresponding theta coordination and avoidance behavior,¹² while sinusoidal optogenetic 8 Hz theta pacing of the mPFC increased avoidance behavior relative to 2, 4 or 20 Hz stimulation frequencies.⁴⁸ The results of the ROC analysis in dynamic theta further suggest that the 8 Hz theta frequency has an important role in facilitating neural transmission between CA1-mPFC and generating avoidance behavior. Future experiments could use optogenetic stimulation to explore the necessity of the 8 Hz bandwidth in controls, as well as the functional role of low-theta frequencies in females and ELS males that learn the active avoidance task.⁴⁹

Although the data indicate the largest and most reliable coherence increases occurred 1 second prior to avoidance, it is not known whether this is solely due to anticipation of avoidance or the activation of aversive neural representations.⁴⁸ As the avoidance measure was relative to peak-acceleration during avoidance arcs, representing the sharpest increases in movement speed, increased coherence 1 second prior to avoidance may also be indicative of the emergence of wide-scale cortical organization in the execution of motor movements.⁵⁰ Yet, the results strengthen the argument that neural coordination across disparate brain regions is necessary for good cognitive outcomes post ELS.

Apart from coherence during the dynamic movement and cognitive demands during active avoidance, coherence and phase locking measurements were conducted during SWS and REM. These measurements are important as they allow assessment of neural coordination during different brain states relevant to cognition and consolidation, without the influence of movement. In SWS, phase locking results paralleled the coherence measures and showed that pooled ELS males exhibited lower levels of CA1-mPFC signal coordination. Notably, no group differences in SWRs were found in the study.

Extrapolating these findings to the clinic is difficult. There is little evidence from the clinical literature that sex has an impact on cognitive outcome in childhood epilepsy, although this information is sometimes unreported. It is known that sex plays a role in developmental disabilities as boys are twice as likely as girls to have attention deficit hyperactivity disorder (ADHD), autism, cerebral palsy, and learning and intellectual impairments.^{51, 52} In a primate study, Goldman et al.⁵³ found that bilateral orbital PFC lesions in Rhesus monkeys at age 1 year or younger resulted in impairments on object reversal or delayed response in males but not females. Yet, after age 15 months, lesions induced deficits in both sexes to the same degree. Going forward, it will be important to consider sex as a biological variable in both clinical and experimental studies assessing seizures during critical developmental periods and their effect on communication between brain regions underlying cognition.

Supplementary Material

s1

Supplementary Figure Legends

Supplemental Figure 1: Flurothyl seizure induction chamber.

s3

Supplemental Fig. 3: As in Fig. 2, Coherence results for ELS male non-learners in each bandwidth (A-H).

s2

Supplemental Figure 2: SWRs in the hippocampus (Hipp) and mPFC; A) Examples of SWRs in the hippocampus; B) Example of synchronous SWRs in the mPFC hippocampus; C) Example of SWRs in four channels in the hippocampus. The initial SWR is an average of approximately 40 SWRs (blue dashed box). The brown dotted lines designate the SWRs while the green line is 1 sec marker.

s5

Supplemental Figure 5: The dynamic theta coherence (A) and burst-rhythmicity patterns (B) of ELS male non-learners are similar in that they represent an absence of neural coordination within and between the mPFC and CA1 regions at ~ 8 Hz, an active frequency in all other groups during active avoidance at the scale of both the LFP and single cells.

s4

Supplemental Fig. 4: Signal processing with regard to behavior and spatial cognitive demands during active avoidance on the rotating arena. A) Illustration of path taken by a CTL female (gray line) during one session of active avoidance. The rat tended to avoid the shock zone in arcs and entered the shock zone twice (red dots); B) These arcs can be divided phases of rest transitioning to recall, increased sensorimotor demands during the avoidance run, and returning to rest; C-D) Spectrograms associated with avoidance behavior showing signal power and frequency in the delta and theta ranges relative to speed (Top C) sorted speed (Bottom C), or acceleration (D). Theta power tends to peak during the middle of the avoidance arc when rats are at their highest speeds or peak acceleration. Speed is transformed to acceleration in order to separate the beginning and end of the avoidance arc (D); E) Peri-stimulus theta spectral properties for a single channel in the CA1 (Left) and a single channel in mPFC (Right) relative to peak acceleration during active avoidance (dashed black line at 0 seconds). Theta amplitude in both channels tends to peak ~ 1 second prior to peak acceleration.

Key Findings.

• Spatial cognitive outcome post ELS is sex-dependent where males fare worse than females.

• Increased theta mPFC-hippocampus coherence in ELS male learners is due to greater coherence distribution across the bandwidth in comparison to CTL males.

• Increased low-theta frequency mPFC-hippocampus coherence in ELS male learners may represent a physiological compensation mechanism that offsets the effects of ELS on mPFC cell timing during active avoidance.

• mPFC-hippocampus coherence levels in theta and gamma bandwidths are predictive of spatial cognitive outcome

Abbreviations:

ELS

Early-Life Seizure

CTL

Control

mPFC

medial Prefrontal Cortex

(P)

Postnatal

SWS

Slow Wave Sleep

REM

Rapid Eye Movement

GEE

General Estimating Equations

LFP

Local Field Potential

ROC

Receiver Operator Characteristic Analysis

AUC

Area Under Curve

KS

Kolmogorov–Smirnov test

LED

Light Emitting Diode

FFT

Fast Fourier transform

EEG

Electroencephalography

PLV

Phase Locking Value

SE

Standard Error