Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disease with a preclinical/prodromal stage characterized by extracellular accumulation of amyloid-β (Aβ) into senile Aβ plaques and later by intracellular accumulation of hyperphosphorylated tau (hTau) protein into neurofibrillary tangles (NFTs), both of which contribute to neuronal and synaptic loss and clinical manifestations of cognitive dysfunction and memory impairment (Bateman et al., 2012; Jack et al., 2013; Vos et al., 2013). Aβ and hTau are byproducts of normal cellular processes and are cleaved from the Amyloid Precursor Protein (App) and the Microtubule Associated Protein (MAPT), respectively. While MAPT mutations are more common in Frontotemporal Dementia, several missense App mutations have been associated with elevated risk for familial or Early-Onset Alzheimer’s Disease (EOAD) (Carter et al., 1992; Chartier-Harlin et al., 1991; Clark et al., 1998; Goate et al., 1991; Hutton et al., 1998).

AD and sleep exhibit a bidirectional relationship whereby disordered sleep is both a symptom of AD pathologies and has a causative role in their development (Lucey et al., 2019; Musiek & Ju, 2022; Wang & Holtzman, 2020). Although disordered sleep occurs across all disease phases in AD, due to the causal role that disordered sleep may play in disease progression at the prodromal stage, considerable translational interest exists in understanding the mechanisms by which disordered sleep contributes to disease progression and the potential beneficial impacts of sleep-related interventions during the preclinical phase (Kang et al., 2009; Musiek & Ju, 2022; Roh et al., 2012). Sex differences have been documented in AD-related sleep dysfunction (C. E. Johnson et al., 2024) but the role that biological sex plays in determining AD risk is difficult to disentangle from impacts of sex on other biological processes and socioeconomic risk factors (Gong et al., 2023).

Sleep/wake has been characterized in several of the mouse models of AD that have been developed in the last two decades (Katsuki et al., 2022). Several amyloid and tau mutant mouse strains have been shown to exhibit sleep disruption and increased wakefulness consistent with human AD (Holth et al., 2017; Huitron-Resendiz et al., 2002; Jyoti et al., 2015; Kang et al., 2009; Platt et al., 2011; Roh et al., 2012; Schneider et al., 2014; Wisor et al., 2005; Bin Zhang et al., 2005; Zhu et al., 2018). Furthermore, chronic sleep restriction compounds pathology in amyloid and tau mutant mice compared to free-sleeping mutants (Kang et al., 2009; Zhu et al., 2018), whereas immunizing an AD model against Aβ appears to normalize disturbed sleep and daily Aβ cycling (Roh et al., 2012). Together, these studies support the concept of a bidirectional relationship between sleep/wake and the neuropathological components of AD. In the current study, we characterized sleep/wake states in aged App^NL-G-F mice, a murine model that exhibits pronounced amyloid pathology and which was engineered to exhibit three missense mutations in the App gene associated with an elevated risk for EOAD (Manabe & Saito, 2025; Saito et al., 2014). Although previous studies have characterized sleep/wake in 6-12 month old App^NL-G-F mice (Calafate et al., 2023; Maezono et al., 2020; Yao et al., 2023), we report results here from (1) a longitudinal comparison of App^NL-G-F and WT mice at 14-18 vs. 18-22 months of age, and (2) continuous 14-day EEG/EMG recordings of male vs. female App^NL-G-F and WT mice at 18-24 months of age when the pathology has become more severe. We find that App^NL-G-F mice exhibit an “insomniac” long wake/short sleep phenotype that is particularly pronounced in females and suggestive of a defect in mechanisms underlying the transitions between arousal states.

Results

App^NL-G-F mice exhibit region-specific amyloid pathology and microglial activation

As described previously (Miller et al., 2024; Saito et al., 2014; Sasaguri et al., 2017), App^NL-G-F mice are characterized by progressive amyloidosis and gliosis and may thus model preclinical stages of AD. Amyloid deposition in representative 22 month old WT and App^NL-G-F mice is presented in Figures 2A and 2A’. Two-way ANOVA revealed a significant main effect of genotype on Aβ expression (p < 0.0005). Figure 2A” presents, for each of 5 brain areas, the proportion of each brain area in which Aβ expression was observed. Post hoc analysis indicated that Aβ expression was significantly higher in the RSC of App^NL-G-F mice compared to WT. Similarly, Figure 2B and 2B’ compares Iba1 staining, an indicator of gliosis, between the two strains and Figure 2B” presents, for each of 5 brain areas, the proportion of that area in which Iba1 is expressed. Two-way ANOVA also demonstrated a significant main effect of genotype for Iba1 expression (p < 0.0001); post hoc analysis revealed significantly higher Iba1 levels in App^NL-G-F mice relative to WT in all 5 brain regions examined.

Figure 1.

Figure 2.

EEG Spectra in App^NL-G-F and WT mice

Figure 3 presents a heat map of the EEG spectra (0-25Hz) across a 24-h period for a representative individual of each sex and genotype. In comparison to age-matched WT littermates, the EEG spectra of the female App^NL-G-F mouse exhibited less spectral power in the 0-5 Hz range during the dark phase, suggesting less sleep. During the light phase, the sleep bouts of this female also appeared to be shorter in duration than in any of the other three mice in Figure 3.

Figure 3.

Distribution of Arousal States, Activity and Body Temperature at 14-18 vs. 18-22 months of age

Figure 4 presents the time in each state, activity and subcutaneous body temperature (T_sc) based on 24-h recordings collected from WT and App^NL-G-F mice at 14-18 months of age and again at 18-22 months. A mixed-effects model ANOVA with Age, Genotype, and Time of Day as factors indicated significant main effects of Genotype and Time of Day for all five of the parameters measured (Table 1). There was a main effect of Age on activity and T_sc but not for the amounts of Wake, NREM or REM sleep (Table 1). At 14-18 months, App^NL-G-F mice spent significantly more time awake (16.1%; Figure 4A’) and less time in both NREM (-18.2%; Figure 4B’) and REM (-14.0%; Figure 4C’) sleep across the 24-h period than WT mice. By 18-22 months, App^NL-G-F mice spent 18.8% more time awake, 19.9% less time in NREM and 18.2% less time in REM sleep. The mixed model ANOVA also revealed significant Age x Genotype interactions for both activity and T_sc (Table 1). App^NL-G-F mice were hyperactive relative to WT mice at 18-22 months but not at 14-18 months; Figures 4D and 4D’ document this notable age-related effect on activity which likely affected the distribution of arousal states at 18-22 months. Figure 4E presents the diurnal rhythm of subcutaneous T_sc in both genotypes at both ages and Figure 4E’ shows that the mean T_sc across the 24-h period declines with age in both genotypes, as indicated by the main effect of Age in the mixed model ANOVA.

Figure 4.

Table 1.

Figure 5 presents the time spent by WT and App^NL-G-F mice in each state, activity and T_sc during the light and dark phases at 14-18 months and 18-22 months of age. As expected, the distribution of arousal states differed between the light and dark phases in both genotypes with more Wake during the dark phase and more NREM and REM sleep in the light phase in this nocturnal species. At 14-18 months, the 16.1% greater levels of Wake and lower levels of NREM and REM sleep in the App^NL-G-F mice evident in Figure 4 were exclusively due to differences during the dark phase (Figure 5A’, 5B’ and 5C’). In contrast, the changes in arousal states at 18-22 months of age in App^NL-G-F mice relative to WT mice evident in Figure 4 were due to increased amounts of Wake and reduced NREM sleep that occurred in both the light (Figures 5A, 5B) and dark phases (Figures 5A’ and 5B’). Undoubtedly, a major factor that underlies the increased wake and reduced sleep in App^NL-G-F mice at 18-22 months of age was increased activity during the dark phase as the mice aged (Figures 4D, 5D’). Mixed-effects model ANOVA with Age and Genotype as factors confirmed a main of effect of Age on activity during the light phase (p = 0.0359; F_(1,13) = 5.478) and a stronger effect during the dark phase (p = 0.0007; F_(1,13) = 19.40) as well as both a Genotype and Age x Genotype interaction during the dark phase. Mixed model ANOVA also revealed a significant effect of Age on T_sc during both the light (p = 0.0003; F_(1,13) = 24.41) and dark (p < 0.0001; F_(1,13) = 37.95) phases; mean T_sc declined with age in both genotypes during both the light (Figure 5E) and dark (Figure 5E’) phases.

Figure 5.

Sleep Architecture in Cohort 1 mice at 14-18 vs. 18-22 months of age

The EEG/EMG data summarized in Figure 4 were analyzed for conventional measures of sleep architecture, i.e., the number of bouts/state and mean bout duration/state. A mixed-effects model ANOVA with Age, Genotype and Time of Day as factors indicated significant main effects of Genotype and Time of Day on the number of Wake, NREM and REM sleep bouts (p < 0.0001 for all states; Table 2) but, as in Table 1, there was no significant effect of Age on the number of bouts for any of these states.

Table 2.

On the other hand, mixed-effects model ANOVA on the duration of Wake, NREM and REM sleep bouts with Age, Genotype and Time of Day yielded more complex results. Although there was a main effect of Time of Day for all 3 states (p < 0.0001; Table 3), Genotype was significant only for Wake bout duration (p < 0.0001; Figure 6A’) but there were Genotype x Time of Day interactions for all 3 states (Wake: p < 0.0001; NREM: p = 0.0015; and REM: p = 0.0054). There were also main effects of Age for both NREM (p = 0.0294) and REM (p = 0.0337) bout duration (Table 3).

Figure 6.

Table 3.

Figure 6 presents the mean Bout Duration for each state during the light and dark phases. Although the mean Bout Duration did not differ between genotypes for any state during the light phase (Figures 6A, B, and C), the mean Wake Bout Duration during the dark phase was more than twice as long for the App^NL-G-F mice at both ages (14-18 months: p = 0.0099; 18-22 months: p = 0.0064; Figure 6A’) and the mean REM Bout Duration for App^NL-G-F mice was 46.3% shorter at 14-18 months of age (p = 0.0195; Figure 6C’). With one exception, the distribution of mean Wake Bout durations of the App^NL-G-F mice during the dark phase at 18-22 months of age did not overlap with that of WT mice of the same age (Figure 6A’), which led us to investigate the distribution of Wake Bout Durations in more detail. The entire 14-day recording of 18-22 month old WT (n=7) and App^NL-G-F (n=8) mice in Cohort 1 was scored using Somnivore and the number Wake Bouts were binned in durations of <60 sec, 60-100 sec, 100-140 sec, 140-180 sec, 180-220 sec, 220-260 sec and >260 sec. As indicated in Figure 7A, App^NL-G-F mice had significantly fewer short (<60 sec) Wake Bouts than WT mice (p = 2.68 x 10^-6) but more longer (>260 sec) Wake Bouts (p = 1.31 x 10^-6). These longer Wake Bouts are likely due to the increased activity of App^NL-G-F mice during the dark phase at 18-22 months (Figures 4D and 5D’). Figures 7B and 7C show that, although activity increases for individuals of both genotypes with age, the age-dependent activity increase was greater in App^NL-G-F mice during both the light and dark phases.

Figure 7.

Multiple Sleep Latency Test and Response to 6-h Sleep Deprivation

The results presented above indicated that App^NL-G-F mice had less sleep than WT littermates, particularly in the dark phase when Wake bouts were of longer duration, and that activity increased with age. To determine whether the level of sleepiness differed in the two strains, Cohort 1 mice underwent a murine MSLT at 18-22 months of age (Figure 1). Figure S1 shows that there were no significant differences between the two strains in the latency to or accumulation of NREM or REM sleep during the five 20 min nap opportunities. Consequently, App^NL-G-F mice at 18-22 months of age appear to be no sleepier than age-matched WT littermates despite having less sleep and greater activity per day, particularly during the dark phase.

To assess the integrity of the sleep homeostasis, mice of both strains also underwent a 6-h sleep deprivation (SD) followed by a recovery sleep (RS) opportunity at 18-22 months of age. Figure 8 shows that, after cessation of the 6-h SD, NREM time increased in App^NL-G-F mice for several hours during the recovery phase (Figure 8B). Although NREM time did not increase during any particular RS hour in WT mice (Figure 8A), the cumulative amount of NREM sleep over the first 5-h of the RS phase was significantly increased in both genotypes relative to the baseline recorded 24-h earlier (Figure S2B and S2B’). REM sleep amounts did not change during the 5-h RS period during the light phase in either genotype (Figures 8C, 8D and Figure S2C). During the subsequent dark phase, however, REM sleep amounts were greater in both App^NL-G-F and WT mice compared to baseline (Figure S2C’), indicating a delayed recovery of REM sleep. Both strains showed the expected increase in NREM delta power during the first hour after SD cessation that gradually declined over the subsequent 4-h (Figures 8E, 8F). Thus, the homeostatic response to SD appears to be intact in App^NL-G-F mice even at 18-22 months of age.

Figure 8.

Comparison of state-specific EEG spectra in Cohort 1 mice at 14-18 vs. 18-22 months of age

Figure 9 compares the EEG Spectral Power Density during Wake, NREM and REM sleep for WT vs. App^NL-G-F mice at 14-18 vs.18-22 months of age. Mixed-model ANOVA revealed a main effect of Age for all bandwidths during Wake except high gamma. Spectral power density during Wake declined with age in App^NL-G-F mice for 4 of the 5 bandwidths and for WT mice in 3 of the 5 bandwidths (Figures 9D-G), the exception being the delta range which increased with age for both genotypes (Figure 9C). Spectral power density in the alpha range during Wake was significantly lower in the App^NL-G-F mice at 18-22 months compared to WT (Figure 9E). The gamma/delta ratio decreased with age in both genotypes (Figure 9H).

Figure 9.

The spectral power density during NREM sleep was unchanged with age in both genotypes (Figure 9C’-H’).

During REM sleep, however, there were significant effects of both Age and Genotype for the delta (Figure 9C’’) and theta (Figure 9D’’) bands as well as a Genotype effect for the alpha band in which App^NL-G-F mice exhibited lower power density than WT mice at 18-22 months (Figure 9E’’). Spectral power density in the delta range during REM sleep was greater in App^NL-G-F compared to WT mice at 18-22 months (Figure 9C’’) but reduced in the theta band for App^NL-G-F mice at both ages (Figure 9D’’) as also occurs in younger App^NL-G-F mice (Calafate et al., 2023; Maezono et al., 2020). Moreover, the theta/delta ratio declined with age in App^NL-G-F mice and was reduced in 18-22 month old App^NL-G-F vs. WT mice (Figure 9I”).

Sex Differences in State Amounts in App^NL-G-F vs. WT mice at 18-24 months of age

To undertake a rigorous statistical comparison of App^NL-G-F and WT mice at an advanced age, the EEG/EMG data collected from the recording of Cohort 1 mice at 18-22 months of age was combined with the recordings collected from Cohort 2 mice. This combination resulted in 31 mice across 4 groups: a female App^NL-G-F group (N=8) ranging from 20.6-23.8 mos (22.0±0.4 mos), a female WT group (N=6) ranging from 19.8-24.6 mos (22.1±0.6 mos), a male App^NL-G-F group (N=9) ranging from 18.3-24.8 mos (21.1±0.8 mos), and a male WT group (N=8) ranging from 18.3-24.8 mos (21.4±1.0 mos).

Figure 10 shows the average amount of Wake (A), NREM (B), and REM (C) sleep on 14 consecutive days for the 31 mice recorded when they ranged from 18.3-24.8 months of age. App^NL-G-F mice spent more time awake and less time in both NREM and REM sleep compared to the WT mice. Although a short sleep/long wake phenotype was consistent between App^NL-G-F males and females, it was clearly stronger in females.

Figure 10.

Figure 11 presents the amount of each state for both strains separated by sex across the entire 24-h period as well as for the light and dark phases. Two-way ANOVA revealed significant effects of Genotype and Sex for the amounts of Wake, NREM, and REM sleep across the entire 24-h period (Table 4); however, a significant Genotype x Sex interaction was only indicated for the amount of REM sleep (F_{(1, 27)} = 10.95; p = 0.0027). Post hoc tests revealed that both male and female App^NL-G-F mice have more Wake (Fig. 11A) and less NREM sleep (Fig. 11A’) than WT littermates; however, only female App^NL-G-F mice have less REM sleep (Fig. 11A’’). These genotype and sex effects were also evident during the dark phase (Figs. 11C, C’ and C”) but, during the light phase, were only present in the App^NL-G-F females who continued to exhibit an insomniac phenotype with more Wake (Fig. 11B) and less NREM (Fig. 11B’) and REM (Fig. 11B’’) sleep. Overall, female App^NL-G-F mice were more severely affected than males.

Figure 11.

Table 4.

Sex Differences in Bout Durations in App^NL-G-F mice at 18-24 months of age

Across the entire 24-h period, there were significant effects of Genotype on the durations of both Wake (F_{(1, 27)} = 34.34; p < 0.0001) and REM sleep (F_{(1, 27)} = 21.89; p < 0.0001) bouts, for Sex on Wake Bout Duration (F_{(1, 27)} = 10.13; p = 0.0037), and a significant Genotype x Sex interaction was only indicated for NREM sleep Bout Duration (F_{(1, 27)} = 7.772; p = 0.0096; Table Wake (Fig. 12A) and REM sleep (Fig. 12A’’) bouts were longer in both sexes of App^NL-G-F mice than in their WT littermates; NREM bouts were longer only in App^NL-G-F males (Fig. 12A’).

Figure 12.

These same patterns were evident during the light phase (Table 5): there were significant Genotype effects for Wake (F_{(1, 27)} = 29.46; p < 0.0001) and REM sleep (F_{(1, 27)} = 27.80; p < 0.0001) bout durations and a Genotype x Sex interaction for NREM Bout Duration (F_{(1, 27)} = 5.635; p < 0.0250). Wake (Fig. 12B) and REM sleep (Fig. 12B’’) bouts were longer during the light phase in App^NL-G-F mice in both sexes than in their WT littermates but NREM bouts were longer only in App^NL-G-F males (Fig. 12B’).

Table 5.

During the dark phase, there were again significant Genotype effects for Wake (F_{(1, 27)} = 19.35; p < 0.0001) and REM sleep (F_{(1, 27)} = 5.666; p < 0.0246) Bout Durations, as well as Sex effects on Wake (F_{(1, 27)} = 11.86; p = 0.0019) and REM sleep (F_{(1, 27)} = 7.281; p < 0.0119) Bout Durations, and a Genotype x Sex interaction for NREM Bout Duration (F_{(1, 27)} = 6.71; p < 0.0153; Table 5). As in the light phase, Wake Bout Durations during the dark phase were longer in both sexes of App^NL-G-F mice than in their WT littermates (p < 0.01; Fig. 12C); this was also the case for REM Bout Durations in female App^NL-G-F mice (p < 0.05; Fig. 12C”) but not for males. Consistent with an insomniac phenotype, NREM Bout Durations were shorter in female App^NL-G-F mice than in their female WT littermates (p < 0.01; Fig. 12C’).

Sex Differences in the Number of Sleep/Wake Bouts in App^NL-G-F mice at 18-24 months of age

Two-way ANOVA revealed significant effects of Genotype for the number of Wake (F_{(1, 27)} = 18.34; p < 0.0001), NREM (F_{(1, 27)} = 21.83; p < 0.0001), and REM sleep (F_{(1, 27)} = 46.10; p < 0.0001) bouts across the entire 24-h period, as well as during the light and dark phases (Table There was also a significant main effect of Sex for the number of bouts of Wake (F_{(1, 27)} = 7.237; p = 0.0121) and NREM sleep (F_{(1, 27)} = 6.338; p = 0.0181) across the 24-h period and during the dark phase (Wake: p < 0.0001; NREM: p < 0.0001; Table 6). Genotype x Sex interaction effects were significant for number of bouts of Wake (F_{(1, 27)} = 6.97; p = 0.0136), NREM (F_{(1, 27)} = 5.00; p = 0.0338), and REM (F_{(1, 27)} = 6.79; p = 0.0148) sleep across the 24-h period, for Wake (F_{(1, 27)} = 6.862; p = 0.0143) and NREM sleep (F_{(1, 27)} = 4.876; p = 0.0359) during the light phase, and for REM (F_{(1, 27)} = 5.47; p = 0.0269) sleep during the dark phase (Table 6).

Table 6.

Post hoc tests revealed that App^NL-G-F males have fewer Wake, NREM and REM sleep bouts than WT males across the 24-h period (Fig. 13A-A’’), during the light phase (Fig. 13B-B’’), and for Wake (Fig. 13C) and NREM (Fig. 13C’) during the dark phase. Female App^NL-G-F mice have fewer REM periods than WT littermate females across all 3 time periods (Fig. 13A”, B” and C”) and fewer NREM bouts during the dark phase (Fig. 13C’).

Figure 13.

Sex differences in state-specific EEG spectral power in App^NL-G-F mice at 18-24 months of age

Spectral analysis of the EEG (0.5-15 Hz) was conducted for all epochs of Wake, NREM and REM sleep. As expected, EEG quality during Wake epochs varied depending upon the magnitude of locomotor activity during any given epoch. Because of the size of the dataset (31 mice x 14 days/mouse x 24 h/day x 60 min/h x 6 epochs/min = 3.75 x 10⁶ 10-s epochs), it was not practical to inspect all Wake epochs to eliminate artifacts due to activity that might distort the spectral analysis during Wake. Accordingly, Figure 14 presents the EEG spectra only for NREM and REM sleep. Two-way ANOVA revealed significant variation in EEG power across bandwidths during NREM sleep (F_{(59, 1652 )} = 1.590; p = 0.0033). Post hoc analysis indicated reduced spectral power in the EEG of female App^NL-G-F mice relative to female WT mice between 4.88 -10.50 Hz during NREM sleep. A similar analysis of the EEG during REM sleep documented reduced spectral power in the EEG between 6.35 – 11.72 Hz for female App^NL-G-F mice and between 7.08-11.48 Hz for male App^NL-G-F mice relative to sex-matched WT mice.

Figure 14.

Discussion

App^NL-G-F mice at younger ages (6-12 months) have previously been the subject of sleep/wake studies (Calafate et al., 2023; Maezono et al., 2020; Yao et al., 2023). In the present study, we compared the diurnal rhythms of sleep/wake, activity and T_sc of App^NL-G-F mice to WT littermates at 14-18 and 18-22 months of age and evaluated biological sex as a factor in a larger cohort at 18-24 months. In contrast to a single 24-h recording that is typically used to phenotype sleep/wake in most studies including those of App^NL-G-F mice at younger ages, the sleep/wake data in the older cohort of App^NL-G-F and App^WT mice studied here were derived from a continuous 14-day recording. We find that App^NL-G-F mice have a long wake/short sleep phenotype characterized by extended Wake bouts and reduced REM sleep amounts and, as in human AD patients, that these characteristics of partial insomnia are exacerbated in female App^NL-G-F mice.

Sleep/wake and activity phenotypes of 14-18 month vs. 18-22 month old App^NL-G-F mice

App^NL-G-F mice exhibited more wake and less NREM and REM sleep than App^WT mice at both ages (Figure 4). These observations are consistent with those of Maezono et al. (Maezono et al., 2020) who found reduced REM sleep in App^NL-G-F mice at both 6 and 12 months of age and increased Wake and reduced NREM sleep at 12 months of age. A prospective study in humans found that lower REM sleep percentage and longer REM sleep latency were both associated with a higher risk of incident dementia (Pase et al., 2017). More recently, a meta-analysis of 24 EEG studies of AD patients found that the most consistent characteristic observed across studies was increased Wake and reduced NREM and REM sleep relative to controls (Y. Zhang et al., 2022), indicating that the partial insomnia phenotype of App^NL-G-F mice strongly resembles that of AD patients.

Although both App^NL-G-F and WT mice exhibited more activity at 18-22 months than at 14-18 months, increased activity with age was more evident in App^NL-G-F mice, particularly during the dark phase (Figure 7C), which undoubtedly contributed to the increased amount of wake and reduced amounts of NREM and REM sleep. These changes in sleep/wake and activity were restricted to the dark phase in 14-18 month old mice but occurred during both the light and dark phases at 18-22 months (Figure 5D, 5D’). Indeed, App^NL-G-F mice of both ages showed longer Wake Bout Durations than WT mice during the dark phase; very long Wake Bouts (>260 sec) were significantly longer in App^NL-G-F than in WT mice (Figure 7A). Overall activity was greater in the older App^NL-G-F mice during the dark phase (Figure 7C). Whereas previous studies reported that reduced REM sleep occurred during the light phase in 6-12 month old App^NL-G-F mice (Calafate et al., 2023; Maezono et al., 2020; Yao et al., 2023), we found reduced REM sleep only during the dark phase in both 14-18 month and 18-22 month old App^NL-G-F mice as described in 9-month old App^NL-G-F mice (Yao et al., 2023). Irrespective of this difference regarding the timing of reduced REM sleep levels, the results across studies indicate a progressive disruption of the diurnal distribution of sleep/wake that is likely a consequence of the hyperactivity evident in older App^NL-G-F mice.

Impact of biological sex on the phenotype of App^NL-G-F mice at 18-24 months of age

Robust differences in sleep/wake patterns were present within, as well as between, both WT and App^NL-G-F strains at 18-24 months of age. Within the wildtype strain, female WT mice had more REM sleep than WT males (Figure 11A”) due to more REM sleep bouts during the light phase (Figure 13B”). Within the KI strain, female App^NL-G-F mice had more Wake than males (Figure 11A) due to longer Wake bout durations in the dark phase (Figure 12C). Female App^NL-G-F mice also had less NREM sleep than males (Figure 11A’) due to shorter NREM bouts in the light phase (Figure 12B’). Sleep/wake patterns in mice are largely determined by strain genetics (Jan et al., 2020; Tafti & Franken, 2002) with minor effects of the estrus cycle (Koehl et al., 2003; Paul et al., 2006; Swift et al., 2024). Our 14-day recordings conducted in 18-24 month old mice (Figure 10) would minimize any effects of the estrus cycle which is typically 4-5 days in mice (Byers et al., 2012).

The within sex/between strain comparisons of sleep/wake characteristics are likely of greater biological interest, particularly whether both sexes of App^NL-G-F mice exhibit the long wake/short sleep partial insomnia phenotype described above. Figure 11A demonstrates that both male and female App^NL-G-F mice have significantly more Wake than WT mice; this difference was particularly evident in the dark phase (Figure 11C) and was due to longer Wake bouts in both sexes at all times of day (Figure 12A-C). Similarly, both sexes of App^NL-G-F mice exhibited less NREM sleep (Figure 11A’) than WT mice; in males, this was due to fewer bouts of NREM sleep (Figure 13A’) as described at younger ages (Maezono et al., 2020). Although older male App^NL-G-F mice show the same tendency toward fewer (Figure 13A’’), longer (Figure 12A’’) REM bouts that females do, these differences are less extreme than in females and, consequently, male App^NL-G-F mice do not exhibit the significant deficit in REM sleep that females do (Figure 11A’’).

Until recently (Jin et al., 2025; C. E. Johnson et al., 2024; Y. Zhang et al., 2022), there have been relatively few studies of sleep in AD patients and, despite the fact that AD is more common in women, there is limited information about sex differences in sleep of AD patients. A recent meta-analysis simply stated that “being female and of advanced age were associated with increased SL [sleep latency] in AD patients compared with controls” (Y. Zhang et al., 2022). Women experience sleep disorders at a higher frequency than men including higher rates of insomnia (Zeng et al., 2020; B. Zhang & Wing, 2006) and restless legs syndrome (Berger et al., 2004; Hogl et al., 2005). Sleep apnea rates rise dramatically in women after menopause and both sleep apnea and menopause increase AD risk (Hall et al., 2015; Jehan et al., 2015). A recent study suggests that obstructive sleep apnea presents differently and thus may be underdiagnosed in women (Bonsignore et al., 2019). CSF biomarkers and PET imaging studies suggest that women with mild cognitive impairment (MCI) and AD exhibit faster brain atrophy rates in both disorders (Barnes et al., 2005; Hua et al., 2010; Jack et al., 2017; Jansen et al., 2015; K. A. Johnson et al., 2016; Mattsson et al., 2017; Shinohara et al., 2016; Skup et al., 2011). Furthermore, women with MCI with elevated amyloid plaque or NFT burden or CSF biomarkers for AD were more likely to develop dementia or neurodegeneration than males, suggesting an elevated susceptibility to AD-related neuropathology in women (Barnes et al., 2005; Koran et al., 2017). Thus, the more severe partial insomnia phenotype of female App^NL-G-F mice resembles some of the symptomatology observed in women with AD.

Impacts of pathology on sleep/wake and activity

Although both sexes of App^NL-G-F mice exhibited partial insomnia with reduced amounts of both NREM and REM sleep and increased amounts of wakefulness in comparison to WT mice, insomnia is typically characterized by decreased sleep quality as measured by sleep fragmentation in addition to reduced sleep amounts. However, sleep fragmentation was not observed in male App^NL-G-F mice; paradoxically, these mice exhibited greater sleep consolidation with fewer, longer bouts of both NREM (Figure 12A’, 13A’) and REM (Figure 12A”, 13A”) sleep, particularly during the light phase (Figure 12B’, 12B”, 13B’, 13B”). The reduced sleep amounts in male App^NL-G-F mice was thus due to fewer sleep bouts across the 24-h period. A hyperactive hypocretin/orexin or monoaminergic arousal system or a dysfunctional GABAergic sleep onset system could underlie the longer bouts of Wake in App^NL-G-F mice (Figure 12A). However, since NREM and REM sleep bouts are also longer in male App^NL-G-F mice (Figure 12A’ and 12A”), a more parsimonious explanation would be that Aβ deposition impairs the mechanism(s) that underlies the transition between states (i.e., arousal state switching) such that individual bouts of any given state become prolonged. This concept is the converse of that which has been proposed to account for the unstable states that are characteristic of the sleep disorder narcolepsy in which loss of the hypocretin/orexin neurons results in instability of arousal states (Saper et al., 2001; Saper et al., 2010). Impairment of state transitions could either be manifest as a problem to initiate a new state (sleep or wake onset failure) or to terminate the current state.

Female App^NL-G-F mice also showed fewer REM bouts across the 24-h period (Figure 13A”) and longer Wake (Figure 12A) and REM (Figure 12A”) bouts but, in contrast to males, NREM bout duration was not significantly different (Figure 12A’). The longer Wake and REM bouts occurred in females during both the light and dark phases as well as across the 24-h period (Figure 12). Thus, female App^NL-G-F mice may share the same impairment of transitions from Wake and REM sleep that male App^NL-G-F mice exhibit.

Perhaps the most consistently observed characteristic of App^NL-G-F mice across studies is the deficit of REM sleep (Calafate et al., 2023; Maezono et al., 2020; Yao et al., 2023). As stated above, REM sleep bouts are longer (Figure 12A”) and the deficit in REM time is due to many fewer REM bouts in both sexes (Figure 13A”), particularly during the light phase (Figure 13B”). This phenotype could result from an impairment in the REM onset circuitry which involves hypocretin and melanin-concentrating hormone (MCH) neurons in the lateral hypothalamus (LH), GABAergic neurons in the ventrolateral periaqueductal grey (vlPAG), glutamatergic neurons in the sublateral dorsal tegmental (SLD) nucleus of the pons, and glycinergic/GABAergic neurons of the ventromedial medulla (Luppi et al., 2025). Overexpression of the hypocretin/orexin system is often associated with AD and can result in arousals from sleep and impaired REM sleep (Liguori et al., 2014; Musiek & Ju, 2022). A smaller proportion of MCH neurons have been shown to be active during REM sleep in App^NL-G-F mice and MCH axons in both App^NL-G-F mice and AD patients exhibit large “spheroid” swellings (Calafate et al., 2023) that disrupt electrical conduction along axons (Yuan et al., 2022). Moreover, reduced REM sleep in App^NL-G-F mice has previously been correlated with increased Aβ in the pontine and medullary areas responsible for REM sleep (Maezono et al., 2020). This presumptive impairment of the REM onset mechanism in App^NL-G-F mice may correspond to the increased REM latency in AD patients (Pase et al., 2017) that has recently associated with higher Aβ burden, phosphorylated tau-181 (p-tau181), and lower brain-derived neurotrophic factor (BDNF) levels (Jin et al., 2025).

Sleep homeostasis in App^NL-G-F mice

Since App^NL-G-F mice exhibited a partial insomnia relative to WT mice, we conducted a murine MSLT at 18-22 months of age to determine whether the level of sleepiness differed between the two strains. As indicated in Figure S1, there were no significant differences between the two strains in the accumulation of NREM or REM sleep during the five 20 min nap opportunities. We also probed the sleep homeostatic system in these two strains by conducting a 6-h SD beginning at light onset. Both strains increased a conventional measure of sleep homeostasis, EEG Slow Wave Activity, in a comparable manner during the subsequent recovery period (Figure 8E vs. 8F). These results differ from those reported in a previous study in which the ratio of EEG delta power recorded during a 4-h rebound after a 4-h SD period relative to the delta power recorded during baseline was reduced in 6 month old App^NL-G-F compared to WT mice (Calafate et al., 2023). In our study, App^NL-G-F mice increased the amount of NREM sleep more robustly during the first 2-h of the RS period (Figure 8B) than App^WT mice (Figure 8A), suggestive of a stronger homeostatic response. When compared across the entire 5-h RS period during the light phase, however, the amount of NREM sleep did not differ between the strains (Figure S2B). Together, these results suggest that the amyloid pathology evident in the brains of App^NL-G-F mice (Figure 2) has minimal impact on the sleep homeostatic system.

Spectral analysis of the EEG in App^NL-G-F mice

During wakefulness, EEG spectral power density declined with age in App^NL-G-F mice in 4 of the 5 bandwidths and for WT mice in 3 of the 5 bandwidths (Figures 9D-G), indicating an age-related reduction in EEG spectral power in both strains. On the other hand, EEG spectral density in the delta range increased with age for both genotypes, reflecting an overall slowing of the EEG with age (Figure 9C). In contrast to Wake, there were no age-related differences in EEG spectral density during NREM sleep in either strain (Figure 9C’-H’). These results differ from the profound reductions in EEG Slow Wave Activity found in NREM of patients with advanced AD (Prinz et al., 1982) and thus suggest that App^NL-G-F mice may not be a good model of advanced AD. However, EEG spectral power did differ between female App^NL-G-F and WT mice in the 4.88 -10.50 Hz range during NREM sleep at 18-24 months of age (Figure 14A).

During REM sleep, spectral power density in the delta range was greater in App^NL-G-F compared to WT mice at 18-22 months (Figure 9C’’), further indicating a slowing of the EEG with age in App^NL-G-F mice. Spectral power density was also reduced in the theta band for App^NL-G-F mice at both ages (Figure 9D’’) as occurs in younger App^NL-G-F mice (Calafate et al., 2023; Maezono et al., 2020). Consistent with previous observations (Maezono et al., 2020), the theta/delta ratio declined in older App^NL-G-F mice (Figure 9I”). Sex-specific analyses revealed reduced EEG spectral power between 6.35 – 11.72 Hz during REM sleep in female App^NL-G-F mice compared to female WT mice and between 7.08-11.48 Hz for male App^NL-G-F mice compared to male WT mice (Figure 14B). These results are consistent with studies in patients with mild cognitive impairment or early stage AD that report reduced fast oscillatory activity and increased slow oscillatory activity during REM sleep (Brayet et al., 2016; Prinz et al., 1992).

Consistency of phenotype

In the 18-24 month old cohort, EEG and EMG recordings collected across a 14-day recording period were classified into sleep/wake states using an automated scoring system. This methodology contrasts with the manner in which most sleep/wake studies are conducted, including previous studies of App^NL-G-F mice (Calafate et al., 2023; Maezono et al., 2020; Yao et al., 2023) as well as our study of Cohort 1. Those studies utilized the traditional sleep phenotyping approach in which EEG/EMG recordings are collected over a single 24-hour period and subsequently analyzed manually. We were interested to know how consistent sleep/wake patterns were over a longer period. While we observed some day-to-day variability in the amounts of wake, NREM, and REM sleep across the 14 days (Figure 10), the overall phenotypes of the 4 sex x genotype groups studied were remarkably consistent and clearly demonstrate that female App^NL-G-F mice exhibited the most wakefulness and least amount of sleep each day. This approach may allow more robust results to be obtained from smaller cohorts and should be considered as a means to reduce animal use in sleep research.

Limitations

App^NL-G-F mice express humanized APP, exhibit a progressive increase in the accumulation of Aβ, a higher ratio of Aβ42 to Aβ40, amyloidosis, and neuroinflammation in several brain areas (Saito et al., 2014) and thus provide numerous advantages over other murine models that overexpress App. However, this model only recapitulates some of the AD pathology: tauopathy does not occur and the cognitive phenotype and severe memory deficit associated with clinical onset of AD are not observed in App^NL-G-F mice (Saito et al., 2014; Sakakibara et al., 2018). As stated above, App^NL-G-F mice do not exhibit the profound reductions in EEG Slow Wave Activity found in advanced AD patients (Prinz et al., 1982). As such, App^NL-G-F mice should be viewed not as a model of advanced AD, but rather as a model of early AD that recapitulates limited pathological components of the disorder. App^NL-G-F mice may thus provide a translation tool providing insights into the phenotypic changes induced by amyloid pathology, which occurs primarily during the prodromal phase of AD.

Analysis of the 14-day EEG recording and classification into sleep stages required the use of automated scoring methods. Although the supervised machine learning algorithms used to segment the data analyzed in the present study are highly accurate, as in other automated sleep/wake methods, the accuracy for identification of REM sleep is lower than for other states. In our previous publications, we have mitigated this limitation by extracting the REM episodes that were most likely to have been misidentified for manual review (Sun et al., 2022; Tisdale et al., 2024). Given the duration of the recordings in the present study, such a manual review was impractical. Our approach instead was to apply rules that we used in our previous studies to identify epochs of REM sleep that were likely to be misidentified arousals or episodes of quiet wakefulness. As shown in Figure 14B, spectral analyses of the epochs classified as REM sleep revealed the typical spectral profile of REM sleep with a large peak in the theta frequency range (4-9 Hz), suggesting that our results using Somnivore are reasonable.

Conclusions

App^NL-G-F mice exhibit several characteristics that resemble the sleep phenotype of human AD patients: more wakefulness, less NREM and REM sleep and a slowing of the EEG. However, these mice do not exhibit the reductions in EEG Slow Wave Activity nor the tauopathy and cognitive deficits found in patients with advanced AD. As such, App^NL-G-F mice may be a better model of the prodromal phase of AD during which sleep interventions may be designed to slow disease progression. The longer Wake bout durations and reduced REM sleep characteristic of App^NL-G-F mice suggest a deficit in the mechanism underlying the transition between states and may be related to the recently-reported delayed REM onset in AD patients. Together, these results support the use of App^NL-G-F mice as a model to investigate sleep-related interventions to mitigate AD burden. Furthermore, these results further support the assertion that sex is an important determinant in disease severity in AD. Specifically, our results suggest that sleep measures, state-specific brain rhythms, and activity are more severely impacted by amyloid pathology in the female sex than in males.

Materials and Methods

Animals

Male and female App^NL-G-F mice that expressed humanized Aβ and the familial Alzheimer’s disease Swedish (K670N, M671L), Arctic (E693G), and Iberian (717, I) mutations on the C57BL/6J background (Saito et al., 2014) were obtained from Drs. Takaomi Saido and Takashi Saito (RIKEN Brain Science Institute, Japan). To obtain wildtype (WT) littermate controls, we crossed heterozygous App^NL-G-F/+ mice to produce homozygous App^{NL-G-F/NL-G-F} mice and App^+/+ (WT) littermate controls of both sexes. For simplicity, homozygous App^{NL-G-F/NL-G-F} mice are referred to here as App^NL-G-F mice; heterozygous App^NL-G-F/+ mice were not studied. A total of 31 mice were used in this study, including 17 App^NL-G-F mice (males: n = 9; females: n = 8) and 14 age-matched WT littermates (males: n = 8; females: n = 6). Mice were maintained on a LD12:12 light:dark cycle at room temperature (22 ± 2°C; 50 ± 20% relative humidity) and had access to food and water ad libitum. Due to hardware limitations, mice were studied in two cohorts of N=16. Each cohort was initially balanced at N=4 with respect to sex and genotype but became imbalanced over time due to mortality and subsequent substitution of replacement mice. The experimental procedures used here were approved by the Institutional Animal Care and Use Committees at SRI International and at the Gladstone Institute.

Surgical procedures

Mice were anesthetized with isoflurane and sterile telemetry transmitters (HD-X02, Data Sciences International., St Paul, MN) were placed subcutaneously on the left dorsum. At the time of surgery, App^NL-G-F mice from Cohort 1 ranged in age from 14.0-17.8 mos (16.0 ± 0.5 mos) and WT mice were from 14.3-17.9 mos (15.6±0.6 mos) whereas Cohort 2 App^NL-G-F mice ranged from 20.1-23.4 mos (21.7±0.4 mos) and WT littermates were 20.2-23.4 mos (22.0±0.4 mos). Biopotential leads were routed subcutaneously to the head and EMG leads were positioned in the right nuchal muscle. Cranial holes were drilled through the skull at -2.0 mm AP from bregma and 2.0 mm ML at −1.0 mm AP and -1.0 mm ML from lambda. The two biopotential leads used as EEG electrodes were inserted into these holes and affixed to the skull using dental acrylic. The placement of HD-X02 telemetry transmitters also allowed measurement of subcutaneous body temperature (T_sc) as well as locomotor activity. The incision was closed with absorbable suture. Analgesia was managed with ketoprofen (5 mg/kg, s.c.) and buprenorphine (0.05 mg/kg, s.c.) upon emergence from anesthesia and for the first day post-surgery. Meloxicam (5 mg/kg, s.c., q.d.) was continued for 2 d post-surgery. Mice were singly housed post-surgery and for the duration of the experiment.

Experimental design

Mice were recorded in a general mouse housing room at the Gladstone Institutes. DSI receiver boards were placed within two conventional animal housing racks with 8 receiver boards per rack in which each rack accommodated 16 cages, front and back. The singly-housed mice were thus recorded surrounded by cages in which there were usually multiple mice per cage. Food and water were available ad libitum and lighting was maintained on a 12:12 light/dark cycle in a temperature- and humidity-controlled environment.

Experiment 1 – Longitudinal comparison of App^NL-G-F vs. WT mice at 14-18 vs. 18-22 months of age

Cohort 1 mice underwent the sequence of procedures illustrated in Figure 1A. Beginning no sooner than 1 week post-surgery when the mice were 14.5-18.2 months of age (App^NL-G-F: 16.3 ± 0.5 mos; WT: 15.9 ± 0.6 mos), EEG and EMG were continuously recorded for 14 days. Mice were again recorded continuously for another 14 days when they ranged in age from 18-22 (App^NL-G-F: 20.4 ± 0.5 mos; WT: 19.8 ± 0.7) months. Due to mortality of one female App^NL-G-F and one female WT mouse in the intervening 16 weeks between recordings, a male App^NL-G-F mouse of comparable age was substituted and was recorded only at the older age. Upon conclusion of the second 14-day recording, Cohort 1 mice underwent a murine Multiple Sleep Latency Test (MSLT) to evaluate their basal level of sleepiness (Veasey et al., 2004). During the following week, Cohort 1 mice underwent a 6-h sleep deprivation (SD) beginning at light onset (ZT0) followed by an 18-h EEG/EMG recording to determine the response to perturbation of sleep homeostasis. Six weeks later, when the remaining 8 mice in Cohort 1 ranged from 20-24 months of age, they were deeply anesthetized, perfused with heparinized saline and 4% paraformaldehyde, post-fixed and frozen for subsequent histology along with 6 age-matched males (4 App^NL-G-F; 2 WT) that had not been implanted.

Experiment 2 – Comparison of App^NL-G-F vs. WT mice at 21-24 months of age

Cohort 2 mice underwent a 14-day continuous EEG/EMG recording when they ranged in age from 21.0-24.3 mos (App^NL-G-F: 22.3 ± 0.4 mos, N=8; WT: 22.8 ± 0.4 mos, N=8) to complement the data collected for Cohort 1 at 18-22 months (Figure 1B). At the conclusion of the 14-day recording, all 16 mice were perfused and brains removed for subsequent histology. For Experiment 2, the data collected from Cohort 1 at 18-22 months were combined with the data collected from Cohort 2 at 21.0-24.3 months so that analyses were conducted on N=31 mice as described in Animals above.

Classification of arousal states

EEG/EMG recordings were scored in 10-s epochs as either Wake, non-Rapid Eye Movement (NREM) or REM sleep using Somnivore (Allocca et al., 2019). The 14-day recording periods for each mouse were provided to Somnivore (version 1.1.7.0) in EDF format, from which 100 10-s epochs of each state were selected by an expert human scorer as training data. The selected data were used to train a classifier for each mouse and applied in a supervised machine learning model to automatically score each state for the duration of each recording. Following this initial scoring, the accuracy of the 3-4 autoscored hour-long bins was assessed visually. If the automated scoring was determined by the expert human scorer not to generalize well, additional epochs were provided to the training dataset and automatic scoring of the recording was repeated. The following scoring rules were applied to each recording: (1) single epochs of REM sleep were rescored as the same state as the previous epoch (i.e., a minimum of two consecutive 10-s epochs were required for a bout of REM sleep to be scored); and (2) REM sleep directly following wake was rescored as wake (i.e., sleep onset REM periods were excluded). For the MSLT recordings conducted in Cohort 1 mice, the same classifier used to score the baseline recordings was implemented, and manual review and adjustment of the MSLT period was performed.

Data Analysis and Statistics

General Procedures

In addition to the time spent in each state, the duration and the number of bouts for each state were determined. A “bout” of a particular state was defined as two or more consecutive epochs of that state and ended with a single epoch of any other state. The EEG power spectrum (0.5–60 Hz) during NREM and REM sleep was analyzed by fast Fourier transform algorithm on all epochs of each state. For spectral analyses, the first 2 transition epochs of each state were excluded. EEG spectra for each state were analyzed in 0.061 Hz bins and in standard frequency bands rounded to the nearest half-Hz value (delta: 0.5–4 Hz, theta: 6–9 Hz, alpha: 9–12 Hz, beta: 12–30 Hz and low gamma: 30–60 Hz). Sleep/wake state amounts and sleep/wake architecture measures (bout durations, number of bouts) were analyzed by ANOVA as described below. Statistics were calculated using GraphPad Prism (ver. 10.0.2). For the EEG spectra data, statistical comparisons for standard frequency bands were conducted as described for sleep architecture measures. Additionally, unpaired t-tests with corrections for multiple comparisons using the Holm-Šídák method were performed for each 0.061 Hz bin from 0.5-60 Hz between genotypes and within each sex.

Experiment 1 data analysis

As stated above, EEG/EMG recordings were conducted in a general mouse housing room at the Gladstone Institute. To obtain relatively undisturbed representative recordings of baseline sleep/wake in Cohort 1 App^NL-G-F vs. WT mice, 24-h periods recorded on a Sunday during the 14-day data capture when the mice were 14-18 months of age and 4 months later when Cohort 1 mice were 18-22 months old were selected for analysis and comparison between strains. After the recordings were scored as described above, a mixed-effects model analysis of variance (ANOVA) was conducted with Genotype, Age, and Time as factors. When ANOVA indicated statistical significance, Fisher’s least significant difference (LSD) tests were performed post hoc to determine specific differences.

Experiment 2 data analysis

Recordings collected throughout the entire 14-day period when Cohort 2 App^NL-G-F and WT mice were 21.0-24.3 months of age were scored, combined with the data from the recordings of Cohort 1 mice at 18-22 months of age, and the combined data from the App^NL-G-F (N=17) and WT (N=14) strains were then compared. Data were analyzed as time spent in each state as well as during the dark and light phases across the recording period. The larger sample size in Experiment 2 (N=31 mice) allowed a mixed-effects model ANOVA to be conducted with Genotype, Sex, and Time as factors followed by between group post hoc comparisons using Fisher’s LSD when appropriate.

Histology and immunohistochemistry

App^NL-G-F mice were anesthetized and transcardially perfused with 0.1 M phosphate buffer saline (PBS), and brains were extracted and post-fixed in 4% phosphate-buffered paraformaldehyde at 4°C for 48 h. After rinsing with PBS, brains were transferred to 30% sucrose in PBS at 4°C for 24 h and coronally sectioned with a sliding microtome. Ten subseries of floating sections (30 µm) were collected per mouse and kept at 20°C in cryoprotectant medium until use. Each subseries contained sections throughout the rostrocaudal extent of the forebrain. Brain sections were washed 3x for 10 min with PBS to remove cryoprotectant medium and once with 0.5% Tween-20 (PBS-T) to permeabilize the tissue. Endogenous peroxidases were blocked with a 15-min incubation of 3% hydrogen peroxide and 10% methanol in PBS. Sections were subsequently washed 3x for 10 min each in PBS-T. Nonspecific binding was blocked with a blocking solution containing 10% normal donkey serum (Jackson ImmunoResearch, 017-000-121) and 0.2% gelatin (Sigma-Aldrich, G2500) in PBS-T for 1 h. Brain sections were incubated overnight with biotinylated 82E1 mouse anti-Aβ antibody (IBL, 10326) at 1:500 and 1:1000 dilution for rabbit anti-Iba1 antibody (Wako, 019–19741) in 3% normal donkey serum 0.2% gelatin and PBS-T at 4°C. For anti-Iba1 staining, after washes, brain sections were incubated with biotinylated donkey anti-mouse antibody (Jackson ImmunoResearch) at 1:500 dilution for 2.5 h at room temperature (RT). Unbound antibodies were removed with three washes of 10 min with 0.5% PBS-T and one wash with PBS for 10 min. Iba1 and 82E1 stainings were developed with an avidin-biotin complex (ABC) kit (Vector Laboratories, PK-6100) according to the manufacturer’s protocol and sections were incubated for 1 h at RT. Brain sections were subsequently washed three times for 10 min each with PBS at RT and incubated with diaminobenzidine (Vector Labs) for colorimetric development. Sections were then washed 3x for 10 min each with PBS and mounted in 1.2% gelatin in H₂O and allowed to dry. After two 5-min washes in xylene, brain sections were permanently mounted on slides and coverslipped for analysis.

Quantification of immunostaining

DAB-stained sections (82E1 and Iba1) were imaged using a Leica Versa 200 Slide Scanner under identical exposure settings across samples. Area coverage (%) for plaque burden (82E1) and microglial coverage (Iba1) was quantified using ImageJ. Regions of interest included the entire cerebral cortex (CTX), retrosplenial cortex (RSC), posterior parietal cortex (PPC), somatosensory cortex (SSC), and hippocampus (HPC), which were manually outlined based on anatomical boundaries defined by the Allen Mouse Brain Atlas. For each region, stained signal was auto-thresholded, binarized, and measured, and values are reported as percent area coverage relative to the total region area. Iba1 measurements were normalized to WT controls, such that the mean Iba1 area coverage for WT mice within each region of interest was set to 1.

Supplementary Figures

Figure S1.

Figure S2.

Data availability

Upon acceptance of this manuscript for publication, all data will be deposited at the National Sleep Research Resource (https://sleepdata.org) as we have done previously.

Acknowledgements

We thank Takashi Saito of the Nagoya University Graduate School of Medical Science and Takaomi Saido of the RIKEN Center for Brain Science for providing the mice and Jasmine Heu and Charlene Sykes of SRI International for providing technical support.

Additional information

Author Contributions

RKT, SRM, JJP and TSK designed the study. RKT, SRM, SML and JS executed the study. Data collection occurred in laboratory space assigned to JJP at the Gladstone Institute. YS, SP, JS and TSK analyzed the results. GA provided analytical software. RKT, YS, JS and TSK wrote the manuscript.

Funding

Research was supported by NIH R01 HL059658 and R01 NS136808 to T.S.K. and K01-AG083732 to S.R.M.

Funding

HHS | NIH | National Heart, Lung, and Blood Institute (NHLBI) (R01HL059658)

• Thomas S Kilduff

HHS | NIH | National Institute of Neurological Disorders and Stroke (NINDS) (R01NS136808)

• Thomas S Kilduff