High-frequency oscillations (250-500 Hz) in animal models of Alzheimer's disease and two animal models of epilepsy

Abstract

Objective: To test the hypothesis that high-frequency oscillations (HFOs) between 250 and 500 Hz occur in mouse models of Alzheimer's disease (AD) and thus are not unique to epilepsy.

Methods: Experiments were conducted in three mouse models of AD: Tg2576 mice that simulate a form of familial AD, presenilin 2 knock-out (PS2KO) mice, and the Ts65Dn model of Down's syndrome. We recorded HFOs using wideband (0.1-500 Hz, 2 kHz) intra-hippocampal and cortical surface electroencephalography (EEG) at 1 month until 24 months of age during wakefulness, slow wave sleep (SWS), and rapid eye movement (REM) sleep. In addition, interictal spikes (IISs) and seizures were analyzed for the possible presence of HFOs. Comparisons were made to the intra-hippocampal kainic acid and pilocarpine models of epilepsy.

Results: We describe for the first time that hippocampal and cortical HFOs are a new EEG abnormality in AD mouse models. HFOs occurred in all transgenic mice but no controls. They were also detectable as early as 1 month of age and prior to amyloid beta plaque neuropathology. HFOs were most frequent during SWS (vs REM sleep or wakefulness). Notably, HFOs in the AD and epilepsy models were indistinguishable in both spectral frequency and duration. HFOs also occurred during IISs and seizures in the AD models, although with altered spectral properties compared to isolated HFOs.

Significance: Our data demonstrate that HFOs, an epilepsy biomarker with high translational value, are not unique to epilepsy and thus not disease specific. Our findings also strengthen the idea of hyperexcitability in AD and its significant overlap with epilepsy. HFOs in AD mouse models may serve as an EEG biomarker, which is detectable from the scalp and thus amenable to noninvasive detection in people at risk for AD.

Keywords: Alzheimer's disease; epilepsy; high-frequency oscillations; interictal spikes; seizures.

Figure 1:. HFOs in the dentate gyrus of Tg2576 mice

(A) A representative Nissl-stained section confirming electrode placement in the dentate gyrus (DG; red arrow). The asterisk marks a cavity after electrode removal. Electrode targeting for both the control and transgenic mice is provided in Supplemental figure 2. (B) A representative EEG recording from the DG of a control mouse during SWS. Top trace is bandpass-filtered for 250–500Hz. Bottom trace is the raw (unfiltered) trace. (C) Same as in B but for a Tg2576 transgenic mouse. Note the presence of HFOs in the filtered trace (red arrows) and absence of HFOs in controls. HFOs were also visible in the raw trace at the troughs of slow wave activity (asterisks). (D) Expanded traces from the inset in C are shown. Note that HFOs (red arrows) stand out from baseline in the filtered trace. (E) Total number of HFOs (all behavioral states pooled) in control and Tg2576 transgenic mice. The total number is based on three 10min recordings one from each behavioral state (wakefulness, SWS, REM), and units are /min. HFOs were significantly more frequent (Mann-Whitney U-test, U=0, p=0.001) in Tg2576 transgenic mice (n=8) compared to control mice (n=5), where they were absent.

Figure 2:. HFOs in the dentate gyrus of PS2KO mice

(A) A representative Nissl-stained section confirming electrode placement in the DG (red arrow). Electrode targeting for both the control and transgenic mice is provided in Supplemental figure 3. (B) A representative EEG recording from the DG of a control mouse during SWS. Note the absence of HFOs in filtered and raw traces. (C) Same as in B but the mouse was a PS2KO transgenic. Note the presence of HFOs in the filtered (red arrows) and raw traces (asterisks). Please note that the difference in amplitude between PS2KO transgenic and control is because mice were recorded using different recording systems. As we explain in Methods, comparisons were possible across recording systems since normalized data were not significantly different across systems. Please see Methods for more details. (D) Expanded traces from the inset in C are shown. Note that HFOs (red arrows) stand out from baseline in the filtered trace. (E) Total number of HFOs recorded from control (n=4) and PS2KO transgenic (n=4) mice. Data from all behavioral states were pooled. The total number of HFOs was significantly higher in PS2KO vs. control mice (Mann-Whitney U-test, U=0, p=0.02).

Figure 3:. HFOs in the dentate gyrus of a mouse model of Down’s syndrome

(A) A representative Nissl-stained section confirming electrode placement in the DG (red arrow). Electrode targeting for both the control and transgenic mice is provided in Supplemental figure 4. (B) A control EEG recording from the DG during SWS is shown. No HFOs were detected. (C) Same as in B but the mouse was a Ts65Dn transgenic. Note HFOs in filtered (red arrows) and raw (asterisks) traces. (D) Expanded traces from the inset in C are shown. Note an HFO in the filtered trace (red arrow) which is also visible in the raw trace (asterisk). (E) Total number of HFOs recorded from control (n=5) and Ts65Dn transgenic (n=4) mice. Data from all behavioral states were pooled. The total number of HFOs was significantly greater in Ts65Dn vs. control mice (Mann-Whitney U-test, U=0, p=0.01).

Figure 4:. The number of HFOs in AD mouse models is elevated during slow wave sleep

(A) Top trace is filtered in the 250–500Hz frequency, rectified to show HFOs during a representative sleep period that includes both SWS (blue) and REM (green). Detection threshold for HFOs is denoted by a red dotted line and theta/delta ratio by a green line. Bottom trace shows raw EEG during SWS (blue) and REM (green). Note frequent HFOs during SWS and reduced number of HFOs during REM, which corresponds to an increased theta/delta ratio. (B) Number of HFOs during wakefulness, SWS and REM sleep in Tg2576 transgenic mice (n=8). One-way ANOVA revealed a significant difference (Kruskal-Wallis test, H(3)=16.56, p=0.0003) and post-hoc tests showed a significantly greater number of HFOs during SWS vs. REM (p=0.01) or wakefulness (p=0.0003). There was no significant difference between the mean number of HFOs in wakefulness vs. REM (p=0.90). (C) Same as in B but for PS2KO transgenic mice (n=4). One-way ANOVA revealed a significant difference (F(2, 9)=30.39, p<0.0001). Post-hoc tests showed a significantly greater number of HFOs during SWS vs. REM (p=0.0002) or wakefulness (p=0.0002), but not between wakefulness and REM (p=0.81). (D) Same as in C but for Ts65Dn transgenic mice (n=4). One-way ANOVA revealed a significant difference (F(2, 9)=9.90, p<0.01). Post-hoc tests showed a significantly greater number of HFOs during SWS vs. REM (p=0.02) or wakefulness (p=0.006), but not wakefulness and REM (p=0.80).

Figure 5:. Comparison of spectral frequency, number and duration of HFOs between different AD mouse models

(A) Representative example of an HFO (red arrow) recorded in the DG of a Tg2576 transgenic mouse. Top trace is wideband recording (0.1–500Hz). The center is a filtered trace (250–500Hz), and at the bottom are spectral properties of the HFO in the 250–600Hz time-frequency domain. (B) Same as in A but for an HFO recorded from a PS2KO transgenic mouse. (C) Same as in B but for an HFO recorded from a Ts65Dn transgenic mouse. (D) Number of HFOs during wakefulness, SWS, REM and all states combined (“All”) between different models. A two-way ANOVA with model and behavioral state as factors showed a significant effect for model (F(2, 39)=8.45, p=0.0009) and behavioral state (F(2, 39)=49.86, p<0.0001) with no significant interaction (F(4, 39)=1.21, p=0.32). Post-hoc comparisons showed that HFOs during wakefulness or REM were not significantly different between models (all comparisons; p>0.05). However, SWS HFOs were significantly different between Tg2576 vs. PS2KO (p=0.04) and Ts65Dn (p=0.001) mice. The total number of HFOs in all behavioral states (“All”), pooled, was significantly different between models (F(2, 13)=4.97, p=0.02) but post-hoc tests did not confirm statistically significant differences in the number of HFOs in Tg2576 vs. PS2KO (p=0.06), and Ts65Dn mice (p=0.07). (E) HFO duration during SWS between different mouse models. No significant differences were found between models (One-way ANOVA, F(2, 13)=0.54, p=0.59). (F) Spectral frequency of HFOs during SWS in different mouse models. No significant differences were found (Kruskal-Wallis test, H(3)=0.90, p=0.66).

Figure 6:. HFOs in the AD mouse models closely resemble HFOs in 2 animal models of epilepsy

(A) Representative example of an HFO (red arrow) recorded from the hippocampus of a mouse where kainic acid was injected. Top trace is wideband recording (0.1–500Hz). The center is filtered trace (250–500Hz), and bottom shows spectral properties of the HFO in the 250–600Hz time-frequency domain. (B) Same as in A but for an HFO recorded from the hippocampus of a PILO-treated mouse. (C) Comparison of HFO duration between AD, IHKA and PILO models. One-way ANOVA found no significant differences in HFO duration between models (F(2, 23)=1.84, p=0.18). (D) Same as in C but for spectral frequency. As in the case of HFO duration, one-way ANOVA found no significant differences between models (F(2, 23)=1.66, p=0.21).

Figure 7:. HFOs associated with IIS and detection of HFOs from the cortical surface

(A) Detection of IIS during REM sleep in a Tg2576 transgenic mouse. Threshold used for IIS detection is denoted by a red dotted line. (B) Expanded trace of a representative IIS from the inset in A. Note the presence of an HFO (red arrow). (C) Example of an HFO occurring before the peak amplitude of an IIS. (D) Example of an HFO occurring during an IIS. (E) Example of an HFO occurring after an IIS peak. (F) HFOs associated with IIS were significantly shorter in duration compared to HFOs occurring independent of IIS (paired Student’s t-test, t=3.33, df=4, p=0.02). (G) Same as in F but for spectral frequency. HFOs associated with IIS showed significantly higher spectral frequency vs. HFOs independent of IIS (paired Student’s t-test, t=3.01, df=4, p=0.03). (H) HFOs during an IIS recorded simultaneously from left (blue traces) and right DG (red traces) and overlying right frontal cortex (black traces) using a subdural screw electrode. (I) Expanded traces from the inset in H showing filtered traces in the 250–500Hz frequency. Note that HFOs do not start or peak at the same time in all recording channels. Vertical green line denotes onset of the HFO in the left DG. (J) Representative example of an HFO (red arrow) recorded with a subdural screw electrode from the right frontal cortex during SWS. Note presence of the HFO in the unfiltered trace (asterisk).

Figure 8:. Spectral properties of HFOs during a representative seizure in a Tg2576 mouse

(A) Example of a convulsive seizure recorded from the DG of an 18-month-old Tg2576 transgenic mouse. Top trace is recording from the left DG and bottom is from the right DG. Insets 1, 2 and 3 are expanded in B. (B) Spectral frequency of HFOs during a convulsive seizure. Top trace is wideband recording. Bottom panel shows spectral properties of the HFOs with frequency >200Hz in the time-frequency domain. B1: An HFO occurring just after an IIS in left and right DG. B2: An HFO during an IIS in the left and the right DG. B3: An HFO at the time between 2 IIS in the left DG and during an IIS in the right DG.