Natural Age-Related Slow-Wave Sleep Alterations Onset Prematurely in the Tg2576 Mouse Model of Alzheimer’s Disease

Introduction: Sleep insufficiency or decreased quality have been associated with Alzheimer’s disease (AD) already in its preclinical stages. Whether such traits are also present in rodent models of the disease has been poorly addressed, somewhat disabling the preclinical exploration of sleep-based therapeutic interventions for AD. Methods: We investigated age-dependent sleep-wake phenotype of a widely used mouse model of AD, the Tg2576 line. We implanted electroencephalography/electromyography headpieces into 6-month-old (plaque-free, n = 10) and 11-month-old (moderate plaque-burdened, n = 10) Tg2576 mice and age-matched wild-type (WT, 6 months old n = 10, 11 months old n = 10) mice and recorded vigilance states for 24 h. Results: Tg2576 mice exhibited significantly increased wakefulness and decreased non-rapid eye movement sleep over a 24-h period compared to WT mice at 6 but not at 11 months of age. Concomitantly, power in the delta frequency was decreased in 6-month old Tg2576 mice in comparison to age-matched WT controls, rendering a reduced slow-wave energy phenotype in the young mutants. Lack of genotype-related differences over 24 h in the overall sleep-wake phenotype at 11 months of age appears to be the result of changes in sleep-wake characteristics accompanying the healthy aging of WT mice. Conclusion: Therefore, our results indicate that at the plaque-free disease stage, diminished sleep quality is present in Tg2576 mice which resembles aged healthy controls, suggesting an early-onset of sleep-wake deterioration in murine AD. Whether such disturbances in the natural patterns of sleep could in turn worsen disease progression warrants further exploration.


Introduction
Altered sleep-wake patterns and reduced quality of sleep are associated with Alzheimer's disease (AD), the most common neurodegenerative disease worldwide, already in its preclinical stages. Sleep-wake disturbances in AD include circadian shifts in the sleep-wake cycle, excessive daytime sleepiness, a significant decrease in time spent in both non-rapid eye movement (NREM) sleepalso known as slow-wave sleep (SWS) -and rapid eye movement (REM) sleep, as well as increased sleep fragmentation [1][2][3].
Human studies suggest that AD pathology targets circuits involved in normal sleep-wake behavior. For instance, degeneration of locus coeruleus, which produces norepinephrine and plays a crucial role in the sleep/wake switch, is already present in the early stages of AD [4,5]. Likewise, functional impairments in the hippocampus, a brain area known to harbor sleep-mediated memory consolidation, were detected in healthy people at genetic risk for AD [6,7]. On the other hand, evidence also suggests that poor sleep is a risk factor for AD. Functional imaging and cerebrospinal fluid profiling in healthy elderly volunteers revealed that greater neurotoxic burden and higher amyloid pathology biomarker levels are associated with shorter sleep duration and poorer sleep quality, respectively [8,9]. These findings were further corroborated by a recent meta-analysis indicating that individuals with sleep disturbances had a ∼4 times higher risk of preclinical AD than those without sleep complaints [10]. Furthermore, a study in a murine transgenic AD model has shown that sleep deprivation increases the accumulation of amyloid plaques, while sleep induction reduces it [11], hinting sleep as a plausible target of interest in AD. Altogether, although sleep-wake disturbances have long been believed to be symptoms of AD, increasing evidence suggests a bidirectional relationship between sleep and AD. Overall, these findings emphasize the importance of studies focusing on sleep enhancement in at-risk populations, even before the neuropathological hallmarks of AD (i.e., amyloid plaques and tau tangles) set in.
Several studies have also examined the sleep-wake phenotype of AD mouse models. Researchers investigated sleep-wake states in PDAPP mice by employing behavioral and electrophysiological techniques and showed an increased motor activity during the dark phase in young mice but not in aged mice and circadian rhythm irregularities that are affected differently for young and old mice [12]. Zhang and colleagues (2005) observed a disruption of REM sleep related to plaque formation around the mesopontine cholinergic neurons in the Tg2576 line [13]. In the same model, the circadian period assessed by measuring wheel-running rhythms in constant dark was also found to be disrupted compared to the wild-type (WT) controls [14]. In another mouse model of AD, APP/PS1, expressing human APPswe and PS1-A264E mutations, the authors demonstrated a spectral shift which was not age related and therefore, independent of Aβ accumula-tion [15]. Others in the same model displayed an agedependent increase in the time spent awake in the light period of the day, as well as an increase of theta power together with a decrease of delta frequencies during wakefulness in comparison to PSEN1 and WT mice [16]. Quantitative EEG analyses yielded AD-like shift toward lower frequencies in PLB1 Triple mice expressing low levels of mutant human APP, tau, and presenilin-1 [17,18]. A more recent study examining 3 different mouse models (3xTgAD, Tg2576, and APP/PS1) demonstrated that both Tg2576 and APP/PS1 mice showed stage-dependent shifts in EEG power spectrum despite the time spent in each state being similar between the AD mice and their WT controls [19]. However, the study did not focus on age-dependent changes and, therefore, disease progression as a potentially determinant factor on sleep-wake phenotype was not examined. Taken together, only a few studies have explored progressive sleep-wake alterations in mouse models of AD with no consensus regarding the nature of the relationship between sleep disturbances and disease progression (i.e., plaque-free vs. plaque-burdened stages). Moreover, the current knowledge on the effect of light-dark cycle on EEG-defined SWS characteristics is rather scarce in murine AD.
In this study, we investigated the sleep-wake behavior of Tg2576 mice [20,21] via 24-h electroencephalography/ electromyography (EEG/EMG) recordings at two early stages of disease progression: (1) plaque-free and (2) moderately plaque-burdened disease stage and compared both groups to age-matched WT controls. We aimed to provide insights into the specific changes in sleep-wake behavior and characteristics in relation to (i) circadian aspects (light vs. dark periods), (ii) genotype, and (iii) age (i.e., stage of disease progression), in a well-characterized mouse model of AD.

Animals
We examined Tg2576 [20] mice overexpressing a mutant form of amyloid precursor protein (APP), APPK670/671L, linked to the early-onset familial AD and their non-transgenic WT littermates (Taconic Biosciences; Cologne, Germany). We excluded the mice carrying the Pde6brd1 retinal degeneration allele, which may lead to retinal degeneration resulting in light sensitivity and/or blindness. The study design included male and female Tg2576 mice at pre-morbid, i.e., plaque-free (online suppl. Fig. 1a, c, d; for all online suppl. material, see www.karger.com/doi/10.1159/000527786), and moderate, i.e., plaque-burdened (online suppl. Fig. 1b, e, f), disease stages, and age-matched WT controls (for group size and sex distribution please see online suppl. er was blinded to the genotype of the mice during the study execution. We group-housed female mice but kept males individually caged due to aggressive behavior leading to injuries that require experimental exclusion. During recordings, all animals were single-housed. The animal room temperature was constant at 21-23°C, with a 12:12 h light:dark cycle starting at 8.00 or 9:00 a.m., according to the season. Mice had access to food and water ad libitum and had daily routine health checks throughout the study. All experiments were approved by the veterinary office of the Canton Zurich and conducted according to the local and federal guidelines for care and use of laboratory animals under license ZH210/17.

Surgery
EEG/EMG implantation surgeries were performed as described previously (online suppl. Fig. 2) [22]. For details, please refer to online supplementary materials.

Statistical Analysis
Statistical analyses were conducted using IBM SPSS ® software, and the figures were prepared in GraphPad Prism 8 (GraphPad Software, Inc., San Diego, CA, USA). We detected the outliers by inspection of a boxplot for values greater than 1.5 box-lengths from the edge of the box, while the normal distribution of variables was determined by skewness and kurtosis (p > 0.05). Levene's test was used to assess the equality of variances.
For sleep proportion analysis, we conducted a two-way AN-COVA (genotype*age, covariate: sex). This was followed by pairwise comparisons. Due to abnormal distribution, we performed a square root transformation for the REM sleep proportions.
Due to lack of normality, we performed a square root transformation on the total number of bouts and ran an independent t-test to analyze sleep fragmentation in three vigilance states for each age group separately. Subsequently, we examined the distribution of WAKE, NREM, and REM bouts in the light and dark periods, respectively, by dividing the bout durations into two categories: short bout durations (0-40 s) and long bout durations (>40 s). We calculated the relative number of bouts by dividing the number of bouts within these two categories by the total number of bouts for each mouse. Mixed ANOVA (genotype*bout duration) was used to analyze the differences in bout distributions between the genotypes.
NREM delta power was analyzed with a mixed ANOVA (genotype*time or age*time). Because of the multiple missing data in the first hour of the day, we excluded the first hour and thus ran the analysis for the next 23 h.
To examine the differences in slow-wave energy (SWE), which was calculated as a cumulative sum of delta power in NREM sleep across 24 h recording, we ran a two-way mixed ANOVA (genotype*time or age*time). Mauchly's test of sphericity indicated that the assumption of sphericity was violated for the two-way interaction. Therefore, we interpreted the results by using the Greenhouse-Geisser correction. Slow oscillation analyses were done with a two-way ANCOVA (genotype*age, covariate: sex), and followed by pairwise comparisons.

Increased Wakefulness during Dark Period in Tg2576 Mice
Sleep-wake disturbances are highly common in AD patients and appear to lead to deterioration of cognitive symptoms [28]. Therefore, we first determine whether alterations in sleep-wake proportions are present in Tg2576 compared to their WT controls. We analyzed 24-h EEG/ EMG recordings for percentage of time spent in three vigilance states (WAKE, NREM, and REM) in the light and dark periods separately, as well as per 24 h. We observed no significant differences between Tg2576 and WT mice, regardless of age, for the percentage of time spent in WAKE, NREM sleep, and REM sleep during the light period (Fig. 1a, c).
We observed, however, a statistically significant main effect of genotype in WAKE (F(1, 35) = 20.54, p < 0.001, η p 2 = 0.370) and NREM (F(1, 35) = 20.42, p < 0.001, η p 2 = 0.368) proportions during the dark period, after controlling for sex. At 6 months of age (6-month), the genderadjusted mean WAKE proportion for the dark period in Tg2576 mice was higher than in WT controls (95% CI [3.276, 10.588], p < 0.001, Fig 1d). The result persisted at 11 months of age (11- Fig. 1g, h), but not in 11-month mice. REM sleep did not differ between genotypes in the 24 h period either (Fig. 1i). In order to eliminate a potential unequal estimation of sleep-wake proportions in Tg2576 compared to WT mice with automatic scoring, we compared manual scoring to automatic scoring. We found that the REM sleep proportion was undervalued with automatic scoring in both genotypes; however, this underestimation was indifferent between genotypes, and both types of scoring led to the same conclusion (online suppl. Fig. 3).

NREM Sleep Stage Continuity Appears to Be Impaired in Tg2576 Mice
Since malfunction in sleep stability is considered just as hazardous to several aspects of health as short sleep duration [29], we further analyzed sleep fragmentation, a common sleep-wake disturbance associated with cognitive decline in AD patients [30], in Tg2576 mice. We determined the existence of sleep-wake instability by calculating the number of bouts within each vigilance state. Our results showed no significant differences in these measures between genotypes in both age groups (online suppl. Fig. 4).
We further analyzed sleep stage continuity by focusing on the bout-length distributions in WAKE, NREM, and REM states. Our results indicated a trend toward higher amount of short and lower amount of long bouts only in NREM in 6-month Tg2576 mice both in light (F(1, 18) = 4.260, p = 0.054, η p 2 = 0.191) and dark (F(1, 18) = 3.602, p = 0.074, η p 2 = 0.167) periods when compared to agematched WT littermates (Fig. 2a). There were no differences in 11-month mice (Fig. 2b) in the bout-length distributions of all three states between genotypes.

Age-Related Spectral Alterations in Tg2576 Mice
We evaluated the EEG power spectrum in WAKE, NREM, and REM in 6-and 11-month Tg2576 and WT mice (Table 1). We observed a trend toward higher beta power both

Progressive Alteration in Delta Power in Tg2576 Mice
Sleep depth, for which power in the delta frequency range (0.5-4 Hz) is a thoroughly validated proxy [31], has been indicated as a highly predictive measure of the solute clearance rate in the rodent and human brain [32,33]. Therefore, we compared the delta power in SWS between Tg2576 and WT mice in both age groups.
We demonstrated that the change in delta power was not a result of the interaction between genotype and time both  (Fig. 3d) and expressed as a lower delta power in 11-month mice during the light period. The main effect of age on this change in the delta rhythm throughout the day was a trend with a large effect size (F(1, 18) = 3.437, p = 0.080, η p 2 = 0.160).

Differences in Slow Oscillations in Tg2576 Mice
One of the major EEG oscillatory events occurring during slow-wave sleep is the slow oscillation (SO). To evaluate whether the differences in NREM delta power between genotypes are mediated by the number of SOs or their properties, we characterized the number of SOs, average duration of a SO, and average amplitude of negative and positive peaks (Fig. 5). Our analyses revealed a significant interaction between age and genotype (F(1, 35) = 4.381, p = 0.044, η p 2 = 0.111) and a significant main effect of genotype (F(1, 35) = 4.398, p = 0.043, η p 2 = 0.112) on the number of SOs. Following pairwise comparisons showed that SOs were higher in 6-month Tg2576 mice in comparison to age-matched WTs (95% CI [2,618.16, 14,405.72], p = 0.006), however, there was no difference between genotypes in 11-month mice (95% CI [−5,761.19, 5,778.19], p = 0.998). There was no difference in the aver-a c b d Fig. 3. Delta power during SWS in 6-and 11-month Tg2576 and WT mice. Hourly distribution of delta power did not display a significant interaction between genotype and time in 6-month (a) and 11-month (b) mice. There was no interaction between age and time on delta power in WT mice (c), whereas this interaction was significant in Tg2576 mice (d). All data are expressed as mean ± SE. **p < 0.01, mixed ANOVA. Times 0-12: light period; times 12-24: dark period. 6/11-mo, 6/11 months old. DOI: 10.1159/000527786 age SO length between genotypes or ages. The analyses of average amplitude of the negative and positive peaks of the SOs revealed a trend toward higher amplitude in 6-month Tg2576 mice than age-matched WT mice (positive peak: 95% CI [−43,332.07, 3,263.17], p = 0.090; negative peak: 95% CI [−2,208.43, 47,061.35], p = 0.073) and not in 11-month mice.

Discussion
Sleep-wake abnormalities are prevalent in AD and other neurodegenerative dementias [35,36] with symptoms including insomnia, circadian disorders, excessive daytime sleepiness [37], and sleep fragmentation [38]. Already at the early stages of the disease, AD patients have decreased SWS, experience longer awakenings, and more pronounced sleep fragmentation [1]. In this study, we report sleep-wake disturbances in the frequently used Tg2576 mouse model of AD, exhibiting many of the agerelated disease characteristics observed in humans. Similar to previous studies with different transgenic mouse models of AD, PDAPP (12), and TgCRND8 [39], we found increased wakefulness and decreased NREM sleep during the active dark period in Tg2576 mice already at an early age (plaque-free pre-morbid stage) as well as in older mice (plaque-burdened stage) compared to agematched WT controls. The fact that these differences were most profound in the active period than in the resting period conflicts with the findings in AD patients [40]. These contradictory results could be explained by the increased body temperature and motor activity observed during the dark period in the Tg2576 line [41]. Moreover, in contrast with the sleep-wake properties of healthy el-a b c d Fig. 4. Effects of genotype and age on cumulative SWE. Cumulative SWE differ during the dark period between genotypes in 6-month mice (a), while there is no difference between genotypes in 11-month mice (b). Analysis within the same genotype revealed that SWE decreased in 11-month (c) WT mice compared to 6-month animals. However, SWE in Tg2576 mice (d) did not differ between the two ages. Comparisons were carried out by a twoway mixed ANOVA (genotype*time or age*time). All data are expressed as mean ± SE. *p < 0.05, **p < 0.01. SWE, slow-wave energy; 6/11-mo, 6/11 months old. derly humans [42,43], studies with aged mice also demonstrated the differences predominantly in the dark period [44][45][46][47][48]. All in all, genotype and/or species specific sleep-wake traits may be behind the discrepant findings between AD patients and Tg2576 mice, which warrants caution regarding over interpretation of rodent-based data.
Another key finding of our study is that during the total (24 h) recording, sleep-wake proportion alterations in Tg2576 were prominent at 6-month but not at 11-month, compared to WT mice. The reason for the disappearance of genotype-related differences in the 24 h sleep-wake proportions in the aged animals could be due to the natural aging of WT mice. These findings may indicate premature onset of age-related sleep alterations in Tg2576 mice and explain why previous studies in 12-month Tg2576 animals could not find differences in time spent in each state compared to WT controls [19]. Why alterations in sleep-wake proportions do not appear to progress (i.e., worsen with age) in Tg2576 mice as they do in WT animals remains speculative. Whether evaluation of geriatric cohorts would depict a further altered sleep-wake pattern than at 11 months of age shall be determined in future studies.
One of the major effects of aging on sleep-wake patterns is to limit the ability to sustain NREM sleep [49]. Although not significant, visual inspection of our results with short and long NREM bouts as well as the large effect sizes suggest an altered sustainability of NREM sleep in 6-month Tg2576 mice. Considering the large effect size of this trend, our study was likely underpowered, as confirmed by post-hoc power analysis with G*power (esti- mated group size required, n = 18). These findings suggesting destabilization of NREM sleep in young Tg2576 may herald a premature age-related sleep phenotype. Moreover, our SWE findings support the notion of an early aging sleep phenotype in Tg2576. Specifically, the cumulative SWE was lower in the 6-month Tg2576 mice than in the 6-month WT mice. This difference was especially visible in the dark period. However, there was no difference in SWE between the 11-month Tg2576 and WT mice, due to age-related reduction of SWE in WT mice. Altogether, our findings indicating an altered sleep quality and quantity already at the pre-morbid, plaquefree stage may have important implications over protein handling and accumulation in AD. In fact, Aβ levels have been reported to be higher during the dark period in mice, and an interrelation between Aβ and sleep quality in both humans and animal models [11,[50][51][52].
Unlike patients with advanced AD [53][54][55][56], in which a "slowing" (i.e., spectral power shifting to low frequency bands) of power spectra has been confirmed, our aged Tg2576 mice showed a decline in delta power and a rise in beta power in the NREM power spectrum at 11-month compared to younger transgenic mice. Similar to our findings, others also observed a not-AD-like spectral shift in Tg2576 mice [13,14,19]. It is conceivable that the age groups (6-month; plaque-free and 11-month; moderate) included in our study do not represent the sleep characteristics of patients in advanced stages of AD. In fact, others observed a trend toward higher delta power in 17-month Tg2576 mice [14], suggesting a "slowing" in the EEG spectrum similar to that of AD patients in the advanced stages of the disease. Moreover, studies investigating spectral power in patients with mild cognitive impairment detected a decrease in delta power [57][58][59], similar to our results, allowing to speculate the decline in the lower frequency bands as an early sign of AD.
SWS is associated with a more efficient clearance of toxic proteins [11,60,61] and with specific regulation of pathways related to protein and cellular homeostasis [62]. Therefore, we investigated SWS delta power and determined that the oscillation of delta power throughout 23 h is different between 6-month and 11-month Tg2576 mice, evident with 11-month mice expressing lower delta power, especially during the light period. This result together with the NREM proportion results suggest that quantity of sleep is already altered in a pre-morbid stage and sleep quality continues to deteriorate with the progression of the disease.
We found no intrinsic changes in SO characteristics in Tg2576 mice but, instead, only the number of oscillations was significantly reduced in the mutants. Interestingly, we determined trends toward larger amplitude of positive and negative peaks in Tg2576 mice in the both age groups. Since SOs are considered to underlie the restorative function of sleep, this increase in the amplitude could be a result of increased sleep pressure owing to increased wakefulness. In fact, studies with human volunteers detected adaptations in the SOs in order to recover the sleep after a prolonged wakefulness [63].
Moreover, our analysis confirmed a greater plaque burden in the hippocampus of 11-month Tg2576 in comparison to age-matched WT mice, whereas 6-month Tg2576 did not present Congo Red-positive amyloid plaques. These results together with our analyses of sleep indicate that, similar to AD patients [64], sleep-wake alterations start before plaque formation in murine AD.
A methodological limitation of this study is the staging accuracy of REM sleep, which is highly dependent on the EMG signal quality. Consequently, abnormalities in the EMG signal caused high variability in the REM sleep scoring in our animals. However, we demonstrated that SPINDLE sub-estimation of REM sleep proportions in relation to manual scoring is of similar magnitude in WT and Tg2576 animals. Another limitation of our report is the lack of purely chronobiologic assessments (i.e., circadian rhythms); hence, we cannot rule out circadian misalignments in this model of murine AD. Finally, the lack of male mice in the 11-month group due to excessive aggressive behavior and consequent exclusion from the experiment may be noted as a limitation of our study that hampers evaluation of potential sex contributions toward both age-related progression of amyloid pathology and associated sleep-wake alterations. Additionally, the exploration of early disruptions of sleep-wake patterns in AD models could provide useful information to the field of pre-symptomatic biomarkers of AD.

Conclusion
Our findings indicate that Tg2576 mice show some of the sleep-wake disturbances observed in AD patients. However, it is to be noted that interpreting the results with any mouse model requires caution as they do not reflect the entire pathology of the disease. Nevertheless, our research provides fruitful information to design future studies regarding age-related sleep alterations in murine AD. Finally, our findings highlight the importance of Neurodegener Dis 2022;22:55-67 DOI: 10.1159/000527786 designing preclinical and clinical research focusing on improving sleep-based therapeutics targeting early stages of the disease to potentially reduce the prevalence of symptomatic AD.