|View Full Text||PubReader|
|Abstract||Print this Article|
|PMC||Export to Citation|
|Article as PDF||Open Access|
Exp Neurobiol 2021; 30(6): 430-440
Published online December 31, 2021
© The Korean Society for Brain and Neural Sciences
Hyunwoo Yang and Yong Jeong*
Department of Bio and Brain Engineering, KI for Health Science and Technology, KAIST, Daejeon 34141, Korea
Correspondence to: *To whom correspondence should be addressed.
TEL: 82-42-350-4324, FAX: 82-42-864-5318
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Alzheimer’s disease (AD) is the most common cause of dementia, characterized by prominent episodic memory dysfunction. Recent studies have suggested that there is a sequential mechanism in the memory deficit, with long-term ones preceding short-term ones. However, there is lack of explanation for these symptoms. Interaction between the hippocampus and retrosplenial cortex (RSC) during slow-wave sleep (SWS) is a crucial step for successful long-term memory formation. In particular, sharp-wave ripple (SWR) is a principal hippocampus oscillation that coordinates with RSC activity. To determine the relationship between memory dysfunction and SWR-related oscillation changes in AD, we implanted local field potential electrodes in the hippocampus and RSC of AD model mice (APP/PS1). We found that the SWR-coupled ripple wave increased in the RSC, while the amplitude of the SWR was preserved. In addition, the corresponding delta power in hippocampus and RSC was elevated, together with altered delta synchrony in AD mice. All these findings showed a significant correlation with long-term memory deficits measured in contextual fear conditions. Our study suggests that altered SWR-coupled oscillations are a possible underlying mechanism of episodic memory dysfunction in AD mice.
Keywords: Alzheimer disease, Episodic memory, Slow-wave sleep, Brain waves
Alzheimer’s disease (AD) is the most common cause of dementia and is characterized by a progressive loss of cognitive function and additional neuropsychiatric symptoms. One of the earliest and most prominent symptoms of AD is episodic memory dysfunction [1-4]. Although the initial episodic memory deficit is well known, the details of amnesia progression with respect to memory storage process remains uncertain because of the variation in disease status of patients and memory tests [5-7]. Several human studies have suggested that episodic memory disruption in AD patients is due to ineffective consolidation or storage of new information [8-11]. These studies showed that AD patients rapidly forget encoded information as the disease progresses. In addition to investigations in humans, studies using AD mice models showed that long-term memory deficits precede short-term memory deficits [12, 13]. These results indicate that there might be a sequential mechanism in the episodic memory dysfunction of AD; however, there is still a lack of explanation for these symptoms.
Appropriate communication between the hippocampus and cortices is critical for proper future use of encoded information. According to previous studies, this interplay is extensively executed during slow wave sleep (SWS), which is characterized by large-amplitude delta waves [14-17]. Neural components related to previous events are reactivated, corresponding synaptic connections are strengthened, and irrelevant ones are globally scaled down during the unique oscillations of SWS [18-20]. Sharp-wave ripple (SWR) is a principal SWS oscillation generated in the hippocampus. It is a high-frequency (100~250 Hz) oscillation that occurs with large-amplitude sharp waves . Sharp-wave ripple leads to hippocampal synaptic plasticity and spreads out and influences interconnected cortical areas .
Recent studies showed that the SWR corresponding oscillation is observed in various cortices including prefrontal cortex, anterior cingulate cortex, posterior cingulate cortex together with retrosplenial cortex (RSC) [23-26]. The common feature of these areas is that they are association cortices having reciprocal connectivity with medial temporal lobe. Therefore, SWR correlated oscillation is considered as the critical factor of information transfer from the hippocampus to cortical regions. Among the areas, the RSC has received a lot of attentions recently because of its direct anatomical projection from the dorsal hippocampus CA1 [23, 24]. A previous report also revealed that the SWR coordinated activity in the RSC was larger than that in other brain areas . Modulation of RSC activity or projection between the hippocampus and RSC during SWS is relevant to memory performance, which implies that valid communication between the regions is essential [27, 28]. Together with SWR, slow oscillation (SO) and spindle events detected in the cortex play an important role in memory consolidation. These oscillations tend to build temporal relationships with each other, indicating mutual information transfer between the hippocampus and cortex [14-17, 21].
Although previous findings have proven that cortical interactions with SWR play a crucial role in long-term memory performance, it is still unclear whether or which interaction is responsible for the preceding long-term memory dysfunction in AD. In this study, to determine the possible relationship, we identified the electrophysiological features between the hippocampus and RSC local field potentials (LFPs) during SWS and investigated their relationship with AD mouse memory performance.
In this study, we used 12-month-old male APPswe/PSEN1dE9 (AD, APP/PS1) mice with a C57BL/6 and C3H hybrid background. We used 25 wild-type (WT ) and 21 transgenic (TG) mice. We housed all the mice under a 12-h light/dark cycle (light phase 09:30~21:30/dark phase 21:30~09:30) and provided them
We performed surgery under tiletamine-zolazepam/xylazine anesthesia (40 mg/10 mg per kg, i.p.). After the administration of anesthesia, we placed the mice on a stereotaxic frame (KOPF instruments) and implanted tungsten electrodes (#797550, A-M systems) on the following sites of the right hemisphere: dorsal hippocampus CA1 (AP-1.82 ML+1.25 DV-1.32) and RSC (AP-2.40 ML+0.30 DV-0.80). We implanted a reference electrode (stainless steel screw) within the skull over the cerebellum. A stainless-steel electromyography electrode (#790700, A-M systems) was located between the neck muscles. Each mouse was single-housed after the surgery and had a recovery time of at least 1 week. We used 18 WT and 14 TG mice (4-day interval group in the contextual fear conditioning test, described below) in the surgery, and excluded five WT and two TG mice because of incorrect recording sites.
Before the behavior experiments, each mouse had an LFP recording session at the home cage. The LFP signals were ×1,200 amplified, recorded, and digitized at a 1,600-Hz sampling rate using a digital system (Twin system, GRASS technologies). We simultaneously video-recorded mouse behavior.
We extracted 3 h of LFP signals during the light phase. We classified sleep state as described in a previous report . First, we manually identified the awake state by the recorded video and filtered electromyography signals (250~500 Hz). Rapid eye movement sleep (REMS) and SWS were distinguished by theta (5~10 Hz)/delta (1~4 Hz) power ratio analyzed using the hippocampus LFP channel. We used the top 25% ratio as the threshold for dividing the sleep state. We excluded SWS epochs of less than 1 min from the analysis. We identified SWR, cortical ripple, SO, and spindle within the classified SWS segments. To detect the SWR, the LFP signal from the hippocampus channel was band-pass filtered (ripple band, 100~250 Hz), and the envelope of the filtered signal was calculated using the Hilbert transform. Events above 3 standard deviations of ripple-filtered signal and longer than 15 ms were defined as SWR. We analyzed power spectrograms of LFPs with wideband LFP around the SWR from the hippocampus and RSC. The center of the spectrogram was set to the minimum trough of the ripple oscillation. Each amplitude change in the frequency blocks was z-scored and averaged within the designated frequency band (delta, 1~4 Hz; theta, 5~10 Hz; beta, 12~30 Hz; gamma, 31~100 Hz; ripple, 100~250 Hz). We adapted the same process to the RSC channel to identify the RSC cortical ripple. Coupling of SWR and RSC ripple was investigated only when RSC ripple occurred within ±50 ms of the SWR trough. We detected SO and spindle as described previously . To detect SO, the LFP signal from the RSC channel was band-pass filtered (SO, 0.5~4 Hz) and zero-crossing points were determined. Within intervals between zero-crossing points, we found the interval lasting for 150~500 ms and satisfying the amplitude criterion: maximum peak is higher than 2 SD or higher than 1 SD with a minimum peak lower than 1.5 SD. We discovered that upward zero-crossing point~maximum peak point~minimum peak sequences lasting for 0.15~0.5 s and included them in further analysis. To detect spindle, the LFP signal from the RSC channel was band-pass filtered (SPD band, 7~15 Hz), and the envelope of the filtered signal was smoothed with a 100-ms Gaussian window. Events higher than 1.5 SD of filtered signal (and peak value is higher than 2 SD) for 0.2 s were identified first. We merged events with intervals less than 0.4 s. Only cases with a final length between 0.4 and 4 s were included for further analysis. Coupling between the hippocampus SWR and RSC SO was defined as occurrence of SO (maximum peak) within 0.25 s from the center of SWR. Coupling between the RSC SO and spindle was defined as occurrence of spindle (maximum peak) within 0.1~1 s from the end of SO. The two couplings above appeared simultaneously (in terms of SO) and we defined these events as triple case [29-31]. All analyses were based on MATLAB (Mathworks) and filtering process was performed using ‘eegfilt’ function from the EEGLAB toolbox.
To identify episodic memory function in mice, we performed a contextual fear conditioning (CFC) test. During the conditioning session, we placed the mouse in a 20×16×20 cm chamber with a metal grid floor (Med Associates). After 3 min of habituation in the chamber, the mouse received two foot-shocks (0.7 mA, 2s) at an interval of 1 min and stayed in the chamber for 1 min after the final foot shock. To distinguish the short-term and long-term aspects of memory function, we divided the mice into two groups: 2-h interval group (2h ITI, WT=7, TG=7) and 4-day interval group (4d ITI, WT=18, TG=14). We returned both groups to their home cage immediately after the conditioning session and placed them back into the same experimental chamber at different time intervals. During the context recall session, the mice stayed in the chamber for 3 min. We video-recorded all sessions, quantified the freezing behavior manually, and excluded freezing behavior lasting less than 1s from the analysis.
To measure locomotion and anxiety levels in mice, we conducted an open field test. The test chamber was 40×40×40 cm and was made of white acryl. The illumination intensity of the chamber was 200 lx. The mouse was placed at the edge of the box and the movement was video-recorded. The total test time was 30 min. We analyzed the distance travelled and time in the center zone (EthoVision XT, Noldus). Only the 4d ITI group was used in this test.
All statistical analyses were performed using the GraphPad Prism version 9. We presented data as mean±standard error of the mean (SEM). We used the Kolmogorov-Smirnov test as a normality test and analyzed the pass data using Student’s t-test. The Mann-Whitney test was used when normality was not satisfied. We applied Pearson’s correlation analysis to determine the significance of linear regression between two parameters.
To assess the short-term and long-term memory performance in AD mice, we performed CFC at distinct time intervals (Fig. 1A). In the 2h ITI experiment, WT and TG mice showed similar level of freezing time in after-shock period (Fig. 1B left; WT, 22.24±3.79, n=7; TG, 16.44±5.39, n=7; p=0.396, Student t-test) and context recall session (Fig. 1B right; WT, 41.21±5.61, n=7; TG, 45.12±6.33, n=7; p=0.652, Student t-test). On the other hand, in the 4d ITI experiment, TG mice showed decreased context recall freezing time (Fig. 1C right; WT, 47.09±3.76, n=18; TG, 28.88±5.01, n=14; **p<0.01, Student’s t-test) without unchanging after-shock freezing (Fig. 1C left; WT, 30.78±5.52, n=18; TG, 20.82±6.40, n=14; p=0.247, Student’s t-test). This result indicates that 12-month-old APP/PS1 mice have intact function in memory encoding and short-term processes, but deficits in relatively longer interval memory. Additionally, we performed an open field test on the 4d ITI group mice to verify whether the reduced freezing behavior was due to altered motor function or anxiety in TG mice. There were no differences in the total distance traveled (Fig. 1D left; WT, 8421.76±575.14, n=18; TG, 9651.04±906.88, n=14; p=0.242, Student ’s t-test) and the time spent in the center zone of the open field chamber (Fig. 1D right; WT, 126.37±19.99, n=18; TG, 142.09±22.80, n=14; p=0.608, Student’s t-test). This result indicates that general motor function and anxiety levels in TG mice are not different from those in WT mice; therefore, the decreased context recall freezing is mainly due to the loss of memory ability.
We implanted the LFP electrodes on the hippocampus and RSC to determine the relationship between electrophysiological interplay and long-term memory dysfunction in AD mice. Slow wave sleep epochs, which are crucial for memory storage, were extracted from the LFP signals recorded in the home cage. First, we identified SWR using hippocampus LFP and gathered LFP traces around individual SWR events (Fig. 2A). We performed spectrogram analysis of the surrounding LFPs (aligned with the SWR trough), individually z-scored, and averaged it. The spectrogram results showed a clear increase in the ripple (100~250 Hz) band and delta (1~4 Hz) band power in both hippocampus (Fig. 2B) and RSC (Fig. 2C). The peak amplitude was much greater in TG mice. Delta oscillations around SWR were significantly higher in both hippocampus (Fig. 2D; WT, 0.228±0.009, n=13; TG, 0.285±0.010, n=12; ***p<0.001, Student’s t-test) and RSC (Fig. 2E; WT, 0.190±0.007, n=13; TG, 0.233±0.011, n=12; **p<0.01, Student’s t-test) in TG mice, but ripple power was increased in RSC only (Fig. 2E; WT, 0.750±0.056, n=13; TG, 0.950±0.081, n=12; *p<0.05, Student’s t-test). In addition, the beta power was decreased in TG RSC during the SWR event (Fig. 2E; WT, 0.436±0.021, n=13; TG, 0.371±0.023, n=12; *p<0.05, Student’s t-test).
Because of the importance of ripple band oscillation in the SWR, we further analyzed the features of the ripple wave. In TG mice, the amount of ripple power during SWR was negatively correlated with context freezing level (Fig. 2F; WT,
Unlike transient ripple events, delta waves have a much longer temporal structure. Therefore, we extracted a wider (0.5 s from the SWR trough) time window for the exact analysis of SWR-coordinated delta activity. Delta-filtered LFP signals from each brain region were averaged and z-scored. Group averaged data showed that delta waves tend to show negative peaks prior to and positive peaks after the SWR center in both areas (Fig. 3A), similar to a previous report . The general shape of delta oscillation did not change in TG mice, but the amplitude (defined as amplitude from negative to positive peak) increased (Fig. 3B HPC; WT, 0.739±0.066, n=13; TG, 1.109±0.083, n=12; ***p<0.001, Mann-Whitney test) (Fig. 3B RSC; WT, 0.883±0.073, n=13; TG, 1.174±0.051, n=12; **p<0.01, Student’s t-test). The duration between peak points was also expanded in RSC in TG mice (Fig. 3C; WT, 0.181±0.002, n=13; TG, 0.191±0.002, n=12; *p<0.05, Student’s t-test). The higher delta amplitudes in TG mice were negatively correlated with the recall freezing time, reminiscent of previous ripple wave results (Fig. 3D left; WT,
High-frequency oscillations are relatively temporary events compared to low-frequency oscillations. Thus, fast waves can be located within a particular phase of slower waves, and this process is known to lead to the integration of neuronal population [15, 16]. To investigate the phasic relationship between the SWR and coordinated delta oscillation, we analyzed the hippocampus and RSC delta phases corresponding to individual SWR outbreak points (Fig. 4A). Most SWRs were nested within a similar phase of HPC and RSC, described as high probability along the diagonal line (Fig. 4B). There was no difference in the average phase of each brain area between the WT and TG mice. However, the paired relationship of the delta phase between hippocampus and RSC was different. (Fig. 4C; WT, n=13, ####p<0.0001; TG, n=12, p=0.109, Paired t-test). This result can be represented as a decreased phase difference between the two regions in the TG (Fig. 4D; WT, 26.901±4.014, n=13; TG, 9.051±5.187, n=12; *p<0.05, Student’s t-test). The amount of the delta phase difference was positively correlated with the context recall freezing (Fig. 4E; WT,
Along with SWR in HPC, SO and SPD are critical SWS oscillations detected in the cortices. The temporal coupling between these events has been regarded as a proxy for memory consolidation. The general features of SO and SPD were similar despite the reduced SWR occurrence in TG mice (Table 1). Temporal coupling can be divided into serial events implying the sequential occurrence of SWR-SO-SPD and events with overlap, indicating an outbreak of SWR within the SO or SPD period  (Fig. 5A; events in serial: green box; event with overlap: magenta box). In the case of coupled SWR and SO, the occurrence ratio did not change, but serial SO and SPD events were decreased in the TG mice (Fig. 5B SWR→SO; WT, 0.208±0.013, n=13; TG, 0.204±0.014, n=12; p=0.837, Student’s t-test) (Fig. 5B SO→SPD; WT, 0.194±0.004, n=13; TG, 0.178±0.005, n=12; *p<0.05, Student’s t-test). The triple events ratio was conserved, which seems to be due to the unchanged SWR→SO ratio (Fig. 5B SWR→SO→SPD; WT, 0.046±0.003, n=13; TG, 0.039±0.002, n=12; p=0.081, Student’s t-test). We then analyzed the event overlapping ratio by comparing the total number of SWRs in each mouse. Trends were significant but in different directions (Fig. 5C SWR within SO; WT, 0.207±0.010, n=13; TG, 0.274±0.009, n=12; ****p<0.0001, Student’s t-test) (Fig. 5C SWR within SPD; WT, 0.214±0.008, n=13; TG, 0.183±0.007, n=12; **p<0.01, Student’s t-test). None of the above features correlated with freezing level in context recall, although they were different from those of the WT mice (Fig. 5D).
We investigated the relationship between memory performance and SWR corresponding to RSC activity in an AD mouse model during sleep. First, AD mice had a long-term memory deficit, as expected. In contrast to previous studies [12, 13], the exact timing of the symptoms were distinct. This difference can be explained by the fact that the onset age of memory deficit varies with previous reports [12, 13, 32-34]. The details of the CFC protocol and the conditions of the behavior setup could also lead to this disparity. Despite this difference, it is quite reproducible, at least in the AD mouse model.
AD mice showed increased amplitude in the SWR-coordinated RSC oscillation, ripple, and delta band. The amplitude of ripple and delta oscillations were negatively correlated with memory performance, specifically in the TG group. Due to the unaltered SWR amplitude in TG mice, elevated ripple amplitude might be a local issue within the RSC. The co-occurrence ratio and coherence between SWR and RSC ripple were intact, implying that the problem might be in the reception unit rather than ripple communication between the hippocampus and RSC. Increased ripple power might indicate increased excitability in the RSC. Increased excitability in an AD mouse model is a well-known characteristic of seizures in various models . In particular, unexpected neural excitation during SWR could hinder the actual component of cognitive information, called a memory engram. Unnecessary action potentials act as noise that obstructs selective synaptic potentiation . This excitatory-inhibitory imbalance could be the result of selective loss of inhibitory neuronal function [37-39]. Therefore, the increased ripple amplitude in the RSC might be a reflection of these underlying mechanisms.
Recent study measured the directional coupling between hippocampus and several cortical areas in other type of APP/PS1 mouse model . The study identified the switched direction of information flow from RSC (including posterior cingulate cortex) to frontal cortex in SWR dependent ripple oscillation. Also, they found the reduced gamma band (20~60 Hz) coupling between hippocampus CA1 and RSC in AD mice. This study implies the disrupted interplay between hippocampus and RSC could be substituted by other cortical areas. Although we did not measure the rearrangement of communication between hippocampus and other cortical areas, RSC seems to be the core area showing the altered relationship with hippocampus in AD mouse model.
Unlike ripple waves, increased nearby delta oscillations were also detected in the hippocampus. Amplitudes of delta waves were closely correlated with each brain region regardless of genotype. The SWR-coordinated delta oscillation also showed a lead and lag relationship between the hippocampus and RSC; however, TG mice showed a high synchrony between them, as inferred by the decreased phase difference between regions. The results indicate the importance of delta amplitude as well as the delta phasic interplay in context memory recall. A previous report found that the RSC negative and positive delta peaks around SWR represent up and down states, respectively . Although the exact role of RSC population activity occurring around SWR remains unknown, it is plausible that the above results connote altered neural activation around SWR.
Previous study identified that the changed SWR-SO and spindle-SO coupling between hippocampus and parietal cortex in AD mouse model of tauopathy and amyloidosis . This paper suggests the impairment in hippocampal-cortical coordination of memory replay events, however it is not sure whether the altered relationship is the general characteristic across the cortical area and regardless of AD mouse model of amyloidosis. Considering the difference in AD model, stage of disease and LFP extraction time point, it is difficult to direct comparison. However, still our results are in the same line in terms of the alteration in spindle-SO serial coupling.
This study covered only the characteristics of the field potentials in the hippocampus and RSC during SWS. Therefore, we could not suggest the source of increased amplitude or determine the correlation between increased amplitude and actual changes in neuronal population activity, in terms of memory engram and replay. Utilizing the simultaneous recording of individual neuronal activity, classifying between type (excitatory or inhibitory) and function (encoded memory engram or unrelated population), and identifying alterations in the neuronal group would be necessary for a better understanding of the current results. Although the role of REMS has been relatively overlooked, recent study identified that Granger causality between the hippocampus and RSC is important factor in successful memory consolidation during REMS . This study raises the possibility of disrupted hippocampus and RSC coupling during REMS, which is not analyzed in our study. Further investigation is needed to expand our knowledge about fundamental sleep oscillation communication between two regions and episodic memory dysfunction under AD pathology.
This research was supported by the Basic Research Lab (BRL) Program (NRF-2020R1A4A1018714) funded by the Korean Government (MSIP) through the National Research Foundation (NRF) and by the Technology Innovation Program (20012464) funded by the Ministry of Trade, Industry & Energy (MOTIE, Korea).
Characteristics of SWS sleep oscillation events
Amplitude (Z-score), occurrence (Events/s), duration (ms : SWR and RSC ripple, s : SPD and SO), frequency (Hz).
p-value indicates the significance of WT and TG comparisons. Significant results are represented in bold type (***p<0.001, Student’s t-test).