Resting After Learning Facilitates Memory Consolidation and Reverses Spatial Reorientation Impairments in Female 3xTg-AD Mice

Abstract

Background.

Sleep is an essential component of memory consolidation and waste clearance, including pathology associated with Alzheimer’s disease (AD). Facilitation of sleep decreases amyloid-β (Aβ) and tau accumulation and is important for spatial memory consolidation.

Objective.

We previously found that 6-month female 3xTg-AD mice were impaired at spatial reorientation. Given the association between sleep and AD, we assessed the impact of added rest on impaired spatial reorientation that we previously observed.

Methods.

We randomly assigned 3xTg-AD mice to a sleep (n = 7; 50 min pre- & post-task induced rest) or a non-sleep group (n = 7; remained in home cage pre- & post- task). Mice in both groups were compared to non-Tg, age-matched, non-sleep controls (n = 6). To confirm that our rest condition induced sleep, we performed the same experiment with rest sessions for both 3xTg-AD and non-Tg mice (n = 5/group) implanted with recording electrodes to capture local field potentials (LFPs), which were used to classify sleep states. Markers of pathology were also assessed in the parietal-hippocampal network, where we previously showed pTau (AT8) positive cell density predicted spatial reorientation ability (AT8, 6E10, M78, and M22).

Results.

We found that 3xTg-AD sleep mice were unimpaired at spatial reorientation compared to non-Tg mice and performed better than 3xTg-AD non-sleep mice (replicating our previous work). This recovered behavior persisted despite no change in the density of pathology-positive cells. Conclusions. Improving sleep in early stages of AD pathology offers a promising approach for facilitating memory consolidation and improving cognition.

1. Introduction

Alzheimer’s disease (AD) is a debilitating neurodegenerative disorder characterized by amyloid beta (Aβ) and tau protein aggregation.^{2, 3} in addition to other pathological processes.^{4, 5} There is a bidirectional relationship between AD and sleep, with AD pathology exacerbating sleep-wake cycle disturbances and impaired sleep furthering accumulation of Aβ.^{5, 6} Sleep disruption is a risk factor for AD.^5–9 Sleep also has homeostatic roles, such as promoting the removal of weak synaptic connections between neurons and decreasing cellular stress, which allows restoration of energy reserves in the brain, and is necessary due to the high energy cost during wakefulness.¹⁰ Pruning of weak synapses may have the dual role of priming the brain for building new connections that represent new memories or expanding on old ones.

Sleep plays a crucial role in consolidating the information we acquire while awake and is essential for renormalizing the strength of synaptic connections, a process that is not conducive to occurring during wakefulness.¹⁰ In addition to the homeostatic role of sleep, it also plays a critical role in memory formation. During sleep, behaviorally relevant neural activity is replayed in the hippocampus in coordination with the neocortex, which facilitates memory consolidation.^11–15 We previously observed that female 3xTg-AD mice were impaired at a behavioral paradigm that required them to repeatedly reorient in space using distal cues.¹⁶ Consistent with a role of sleep in memory facilitation, subsequent work in our lab found that impaired hippocampal-cortical coupling during sleep may at least partially explain the impaired spatial reorientation we observed.¹ However, impaired sleep in 3xTg-AD mice typically occurs at later time points than those examined here, after plaques have begun to form around 9-months of age, 3-months older than the mice in the present study.^{17, 18}

Another important function of sleep is to help restore the brain by clearing waste metabolites that may be neurotoxic.^19–21 This is believed to be largely accomplished by the glymphatic system, a macroscopic waste-clearance system which uses Aquaporin-4 (AQP4) water channels on astrocyte end-feet to promote removal of soluble proteins from the central nervous system.²² The two hallmarks of AD, Aβ and tau pathologies, have been shown to fluctuate diurnally throughout the sleep-wake cycle, with increases during wakefulness and decreases during sleep,^{20, 23, 24} suggesting that sleep may be capable of decreasing Aβ and tau pathologies. The glymphatic system is dependent on circadian timing and has been observed to be more effective during sleep.²² Recent work has questioned the increased glymphatic flow during sleep.²⁵ However, a number of shortcomings of this study, particularly in relation to what is known about the glymphatic system and Alzheimer’s disease, mean that while the results are provocative alone, they are insufficient to change current thinking about glymphatic flow during sleep.⁵ In humans with AD, an increase of Aβ in cerebrospinal fluid has been shown to emerge with the onset of sleep impairments, such as increased sleep fragmentation.^{1, 26–28} AD is associated with disrupted sleep, poor sleep quality, and reduced sleep duration, leading to a cycle of increased Aβ and tau accumulation.²⁶ More specifically, disrupted sleep and slow-wave sleep (SWS) deficits are associated with increased Aβ,^28–30 possibly because the glymphatic system is not able to effectively clear pathology.³¹ Animal models of AD also show sleep alterations occurring with Aβ deposition.^{1, 31–36} Decreased sleep has been shown to appear in some mouse models after the formation of Aβ plaques.^{31, 33, 37} Interestingly, an amyloidosis-only mouse model (APP/PSEN) exhibited reduced sleep even before plaque formation,³⁴ suggesting that sleep deficits may result from smaller Aβ aggregates. Further, in APP/PS1 mice, the sleep-wake cycle markedly deteriorated, and diurnal fluctuation of interstitial fluid Aβ dissipated following plaque formation. Clearing amyloid restored sleep and diurnal Aβ fluctuations, suggesting that plaque formation may lead to impaired clearance during sleep, thereby exacerbating amyloid aggregation³¹. Regardless of which comes first (sleep deficits or amyloid and tau aggregation), the bidirectional relationship between sleep and clearance acts as a positive feedback loop, further contributing to the development of neuroinflammation and pathology, thereby advancing disease progression.³⁸

A great deal of research examining the relationship between sleep and AD has focused on improving 24h circadian sleep because a common disorder in AD patients is irregularity of the sleep-wake cycle, with reduced amplitude of circadian rhythmicity, phase delay, and sundowning, which has been reproduced in animal models by numerous studies, including our own.^39–41 However, much less work, and none in rodents (to our knowledge), has examined the possibility of using rest sessions to improve cognitive symptoms associated with AD. We also focused on mice that had intracellular accumulation of Aβ and tau but no plaques or tangles (3xTg-AD mouse model), coinciding with the emergence of deficits on our sensitive cognitive test, unlike previous studies which have focused on time points after plaques and/or tangles have formed.

To assess the impact of sleep on early cognitive change in AD, we used a novel spatial navigation task. In AD, pathological processes develop and progress significantly, well before cognitive symptoms manifest.⁴² However, getting lost, particularly in new surroundings, is an early cognitive impairment in humans who develop AD.^{43, 44} Understanding the impairments in spatial navigation, or the frequency with which people with early symptoms of AD get lost, can help push the diagnostic timeline forward and possibly start treatments sooner. To this end, we applied a novel task, the spatial reorientation task, in mice with tau and amyloid aggregation. This task is designed to mimic some of the neural processing related to the conditions in which humans get lost in new surroundings. This task requires rodents to begin each trial in a new location and take in information from their surroundings to figure out their current spatial location. Recent work, including our work with this task, suggests that this early impairment could represent a failure to use distal cues to reorient in space,¹⁶ impaired consolidation of memories during sleep,¹ or likely both. Finally, while previous work in rats has shown that this task engages the hippocampus and leads to realignment of place cells to room-centered coordinates during each trial,⁴⁵ our work shows that dysfunction in an extended network, including the hippocampus and cortical structures (i.e., many of the default mode network structures in humans),⁴⁶ that are also dysfunctional in human AD.^47–50 Finally, impairments on this task emerge early, when tau and amyloid aggregation are low and prior to the formation of plaques and tangles.¹⁶ Therefore, we employed the spatial reorientation task here.

This task requires mice to repeatedly reorient in space using distal cues, but we also included pre- and post-task rest sessions. We focused on 6-month-old female 3xTg-AD mice because they previously exhibited spatial reorientation learning and memory deficits.¹⁶ Additionally, 6-month female 3xTg-AD mice showed a pTau (AT8) positive cell density profile across a parietal-hippocampal network which predicted their spatial reorientation ability.⁴⁶ Here, we assess spatial reorientation alongside markers of AD pathology to determine whether improving sleep early in disease progression enhances cognition and whether such improvements correlate with Aβ and tau accumulation. The novelty of the present study lies in its assessment of sleep’s impact on measures previously shown to be altered in 6-month-old female 3xTg-AD mice: spatial reorientation, Aβ, and pTau (AT8)–positive cell density. While this study includes a replication of our earlier findings—which we view as a mark of rigor and a strength—it centers on the unexpected discovery that a simple sleep-promoting intervention yields a robust improvement in cognitive performance.

2. Methods

2.1. Animals

We assessed the effects of 1h rest sessions on spatial reorientation ability in two experiments. In the first experiment, sleep was manipulated under conditions nearly identical to those in which we previously observed cognitive deficits in this spatial learning and memory task with a small brain stimulation implant.¹⁶ In the second experiment, we directly assessed sleep with a larger recording-array implant. Overall, for both experiments, 6-month female 3xTg-AD (n = 19) and age-matched non-Tg control mice from the same background strain (129X1SvJ/C57Bl6 hybrid; n = 11) were group-housed (2–4/cage) in 12:12 h light/dark cycles until the beginning of the experiment. We used only female mice because we found impaired cognition in 6-month 3xTg-AD females, but not males.^{16, 46} These mice underwent behavioral testing during the 12h light phase of the light/dark cycle in a dimly lit room. For Experiment 1, mice (3xTg-AD, n = 14; non-Tg, n = 6) did not have a recording-array implant to ensure that the effects we observed in Experiment 2 were not a consequence of the recording-array. The 3xTg-AD mice in Experiment 1 were subdivided into no-sleep (n = 7) and sleep (n = 7) groups. For Experiment 2, mice (n = 10) had a recording-array implant, some data from a different training phase was published previously,⁵¹ which allowed us to measure the quantity of SWS and rapid eye movement (REM) sleep. The groups in Experiment 2 were 3xTg-AD (n = 5) and non-Tg controls (n = 5). Both groups in this experiment had pre- and post-task rest sessions. For simplicity, all mice which underwent rest sessions are in the ‘sleep’ group. All experimental procedures were carried out in accordance with the NIH Guide for the Care and Use of Laboratory Animals and approved by the Florida State University Animal Care and Use Committee.

2.2. Behavioral Assessment

The behavioral assessments for pre-training, stimulation parameters, spatial reorientation task, and probe testing were performed using the same methods as in Stimmell et al.¹⁶ so that two groups in this study were a replication of our previous work (3xTg-AD no-sleep mice and age-matched non-Tg controls). The protocols for these training phases are described in the respective sections below.

2.3. Pre-training

Mice were water deprived to no less than 80% of their initial body weight before starting water restriction and given food ad libitum. They were then trained to shuttle to a black barrier at the end of a linear track and back for a water reward (pre-training; see Fig. 1a–b for experimental timeline). This barrier was positioned so that there was a black background behind it (from the view of the mouse) on the wall. The entire track was moved to different locations in the room leading to different starting positions, so the length of the track varied (Fig. 1c). The starting position was randomly selected from 9 possible start locations (spaced over a range of 56–76 cm from the center of the goal zone). Note, all calculations were performed in pixel coordinates. The distances listed in cm throughout the paper were estimated for visualization purposes. After reaching criterion (either asymptote, ±6 total runs, 3 out of 4 days, or a total of 50 or more runs down the track and back), the mouse was scheduled for surgery to implant stimulating electrodes and continued with pre-training every other day until the day before surgery, when water deprivation ceased. In an abundance of caution, we report the number of days mice in each group were maintained on pre-training every other day (Mean±SD; Experiment 1 No Recording-Array: Non-Tg 4.5±1.0 days; 3xTg-AD No-Sleep 4.1±1.7 days; 3xTg-AD Sleep 3.0±1.4 days; Experiment 2 Recording-Array; Non-Tg Sleep 5.8±2.5 days; 3xTg-AD Sleep 5.6±1.9 days). Finally, we also compared the number of days mice in each group were maintained on pre-training every other day and found no significant differences between groups for either experiment (Experiment 1: F_{(2, 17)} = 2.03, p = 0.16; Experiment 2: t₍₈₎ = 0.14, p = 0.89).

Figure 1. Rest sessions ameliorate impaired spatial reorientation in 6-month 3xTg-AD female mice.

a. Illustration of the timeline for the full experiment. 1-week of pre-training was followed by surgery, 1–2 weeks of recovery, and then 1–2 weeks of operant conditioning to determine optimal MFB Stimulation Settings. Sleep or home cage sessions began with the task, and the spatial reorientation task then lasted approximately 1 month. b. Each daily recording session consisted of 60 min in a pre-sleep or home cage session, followed by a 20 min task session, then a 60 min post-sleep or home cage session. c. The maze consists of a linear track with a start box at one end and an unmarked reward zone near the opposite end (blue bar). The reward zone remains fixed in the room; however, the start box and track move between trials. As the mouse traverses the track, it uses stable distal cues in the room to locate the reward zone and slow down to obtain a MFB stimulation reward. This panel was adapted from.^1,2 d. Mean (± SEM) Z-scored speed from the first half of the reward zone (left 1/2 of the blue bar in panel e) for each reward delay (0.5 – 2.5 s) in 6-month female non-Tg (blue) and 3xTg-AD (red) mice, as well as 6-month female 3xTg-AD sleep mice (yellow). 3xTg-AD sleep and non-Tg mice slow significantly more than 3xTg-AD mice with no-sleep sessions (*ps < 0.05). e. Mean (± SEM) Z-scored velocity plotted along the distance of the track during the 1 s reward delay for 6-month female non-Tg, 3xTg-AD, and 3xTg-AD sleep mice. Both non-Tg and 3xTg-AD sleep mice slowed in the reward zone (blue bar). f: Same as panel b, but for the 1.5 s reward delay. * 3xTg-AD no-sleep vs. sleep mice, ps <0.05; # 3xTg-AD no-sleep vs. non-Tg no-sleep mice ps <0.05; $ 3xTg-AD sleep vs. non-Tg no-sleep mice ps < 0.05

2.4. Stimulating Electrode Only Implantation

The surgical procedure for Experiment 1 was described previously.¹⁶ Briefly, once criteria were met on the pre-training task, mice underwent surgery during which bilateral stimulating electrodes were implanted targeting the left and right medial forebrain bundle (MFB; 1.9 mm posterior to bregma, 0.8 mm lateral, 4.8 mm below dura).

2.5. Recording-array Implantation

The surgical procedure for the subset of animals with a recording-array in Experiment 2 was described previously¹. For this subset, two bipolar stimulating electrodes were implanted unilaterally to accommodate the recording-array in the contralateral hemisphere, targeting the left MFB pathway (1.9 mm & 1.4 mm posterior to bregma, ±0.8 mm lateral, 4.8 mm below dura). The implantation of these stimulating electrodes allowed for intracranial stimulation of the MFB. A 16-tetrode recording-array⁵² was implanted, targeting the parietal cortex and dorsal hippocampus (2.2 mm posterior to bregma, 2.0 mm lateral). Animals recovered for one week, during which tetrodes were turned at the following rate: Tetrodes targeting the hippocampus were lowered approximately 124 μm daily for the first three days, and then 31 μm every other day, until reaching the hippocampus. Positioning was indicated by depth records, as well as the characteristic hippocampal LFP. Data from tetrodes targeting the parietal cortex are not reported here. We selected hippocampal tetrodes due to their more robust local field potential (LFP) features for sleep classification (e.g., theta power), as opposed to noisier signals from other sites, which would have required additional animals. Thus, the parietal cortex tetrodes used previously were not suitable for this study. Tetrode locations were histologically verified post-mortem, based on the presence of damage tracks and/or an electrolytic lesion corresponding to the tetrode tip. Mice implanted with recording-arrays underwent extensive habituation via daily tetrode adjustment over 7 days, following the turning rate schedule described above. Tetrode turning continued during 7–10 days of stim-box training, which included rest sessions on at least the final few days. This was followed by an additional day of pre-training for further habituation.

2.6. Stimulation Parameters

Following a 1-week recovery period, mice were placed in a custom 44 × 44 × 44 cm box with a circular nose poke port (Med Associates) mounted on the left half of one wall of the chamber. Using manual stimulations, mice were first shaped to approach the nose poke port, then to touch the port, and finally to poke their nose into the port to break an infrared beam. Once trained to nose poke, the animal automatically triggered each brain stimulation reward. Brain stimulation rewards lasting 500 ms were delivered using a custom-written MATLAB (The MathWorks Inc.) program attached to a Stimulus Isolator (World Precision Instruments). Over the course of one week, settings were adjusted (171–201 Hz frequency, 30–70 μA current, electrode wire combinations) to achieve maximal response rate. No attempt was made to balance responding across genotype; however, responding (nose poke) rate was compared across genotype to ensure that differences in reward strength were not likely to contribute to the observed effects. The ideal measure of stimulation effectiveness is use of a running wheel to trigger brain stimulation because operant conditioning is less effective in mice.⁵³ However, since our task requires mice to stop running for a reward, we did not want to confound the study by training mice to run for brain stimulation and then training them on the reverse (stopping for brain stimulation, i.e., reversal learning). Thus, we first assessed brain stimulation responding with an operant task, and this worked well for nearly all the mice in the study. For low responding mice, we assessed stimulation responsiveness with a task that required alternating for stimulations at either end of the track. This was assessed after spatial reorientation training, to prevent a confound of reversal learning. We then combined responses for stimulation across these two tasks, which produced a continuous distribution with all stim values falling below operant training values.

2.7. Spatial Reorientation Training

Once mice were responding maximally, one additional refresher pre-training session was conducted, followed by spatial reorientation training on a mouse version of a task we published previously (as in¹⁶ and¹) and adapted from the Rosenzweig et al.⁴⁵ task for rats. Throughout the remainder of training and testing, mice shuttled to the end of the track and back to the start box for a water reward delivered there, which they consumed while the entire track was moved slowly and smoothly to the next randomly selected starting location. The running for water reward is designed to create a baseline of regular motion, to increase confidence that the key spatial test (stopping) is due to intentional stopping by the mouse. There were 9 possible start locations, spaced over a range of 56–76 cm from the center of the goal zone, decided using random lists generated by Random.org (random with repeats). Next, an unmarked reward zone was added to the task (28 cm from the barrier at the end of the track) that could automatically trigger a single brain stimulation reward lasting 500 ms. This reward was only marked in the closed-loop software which automatically delivered a reward if the mouse remained in the zone for a sufficient period of time before progressing to the end of the track (Fig. 1c). This zone was fixed in relation to the room and cues positioned around the periphery of the room and was marked in camera coordinates only. Thus, there were no visible markings on the track indicating the location of the reward zone. Further, since the track was moved to new positions following each trial, any olfactory or other cues on the track could not signal the location of the reward zone. Thus, the reward zone occurred at a variety of physical locations on the track, but always in a fixed location within the room. If the mouse remained in the zone for the duration of a delay period (starting at 0.5 s), a brain stimulation reward was delivered. The delay was fixed for a given day and was increased by 0.5 s (up to 2.5 s) each time the mouse met a criterion of similar percent correct (i.e., ±15 %), 3 out of 4 days. This means that the mouse must use the distal cues distributed around the room to figure out the location of this unmarked reward zone on each trial. Distal cues consisted of black poster-board polygons—stars, rectangles, squares, arrows, and other random shapes—each about 1–2 ft tall or wide. They were affixed to the walls around the maze at or above maze level, within 4 ft of the track on both sides, with 4 inches to 2 ft spacing between cues. Cabinets and a bench along one wall served as additional salient landmarks. The use of two different rewards, brain stimulation and water, is necessary because we are teaching mice to do two opposite behaviors, to run back and forth and also to stop. To get mice to perform both behaviors, it is necessary to use different reinforcers for each: water reward for running back and forth and brain stimulation to stop in an unmarked reward location after being started from a randomly selected position in the room. Custom MATLAB software was used for timestamping the key events of each trial (start, reward zone, and end of track), automatically delivering the brain stimulation reward after the reward delay and generating a record of integrated events and the tracked position of the mouse for offline analysis. All behavior sessions were timed to last 20 min. The dependent variable is the velocity during the approach to the reward zone which we have found to be most sensitive to performance. This is the most sensitive measure likely because the mouse can accidentally obtain rewards for the smaller reward delays. We have previously made the software for running this task freely available.⁵⁴

There were pre- and post-task rest sessions before and after each daily session, excluding pre-training. These sessions were included while determining optimal stimulation parameters, and throughout the spatial reorientation task and probe testing. Each day consisted of a 50 min pre-task rest session, followed by 20 min of task, and then a 50 min post-task rest session. Sleep was induced by placing mice in a 17 × 16.5 × 16 cm box while connected to the recording system, as in.¹ As we have done previously (e.g.,¹⁶), we assessed the average stimulation response rate to verify that a particular group was not responding better than the other. We found no significant effects of group in Experiment 1 (3xTg-AD sleep vs. non-Tg vs. 3xTg-AD no-sleep) on response rate (F_{(2, 17)} = 1.65, p = 0.22). Though unlikely to influence the results given that responding did not differ across groups, it is possible that greater MFB stimulation was needed to produce equivalent responding in some mice. Therefore, we also performed the same comparison for the two brain stimulation parameters that we adjusted: frequency and current. Non-recording-array mice showed no variation across genotype for frequency (F_{(2, 17)} = 0.43, p = 0.66) or current (F_{(2, 17)} = 0.85, p = 0.44). Similarly, the two recording-array mice groups from Experiment 2 (3xTg-AD vs. non-Tg) showed no differences across genotype in response rate (t₍₁₀₎ = 0.69, p = 0.51). There was also no difference in frequency or current across genotype (t₍₁₀₎ = 0.52, p = 0.61; t₍₁₀₎ = 0.15, p = 0.88; as reported in Cushing et al. 2020).

2.8. Sleep Recording Procedures

As described in Cushing et al.¹, an electrode interface board (EIB-72-QC-Small or EIB-36–16TT, Neuralynx) was attached with a custom adapter to the recording-array utilizing independently drivable tetrodes (Chang et al., 2013) connected via a pair of unity-gain headstages (HS-36, Neuralynx) to the recording system (Digital Lynx SX, Neuralynx). Tetrodes were referenced to a tetrode wire in the corpus callosum and one wire from each tetrode was selected for LFP recording. Adjustments were made to each day’s recording to allow for stabilization overnight. A continuous trace was collected for processing as LFP from one of the tetrode wires (bandpass-filtered 0.1 – 1000 Hz and digitized at 6400 Hz). Mouse position was tracked using a colored dome of reflective tape for the spatial reorientation task and online position information was used to trigger MFB stimulation rewards. Video-tracking data were collected at 30 Hz and co-registered with LFPs and event timestamps. LFP analyses were performed using custom-written MATLAB code or Freely Moving Animal (FMA) Toolbox (https://fmatoolbox.sourceforge.net/). The LFP signal was collected at 6400 Hz and subsequently resampled to 2000 Hz for further analysis using the MATLAB resample function.

2.9. Sleep Quantification

First, still periods for both recording-array mice and non-recording-array sleep mice were extracted from the rest sessions.^{1, 11, 14} Rest sessions lasted 50 min and occurred immediately before and after the 20 min behavioral testing session (Fig. 1b). Only mice in sleep groups underwent these rest sessions. Rest sessions consisted of placing the mouse inside a small black enclosure with tall walls, so that the room is almost entirely obscured except for a small portion of the ceiling while tethered to the commutator via either the brain stimulation cables or the recording-array cables. We have found that placing mice in this “boring” enclosure encourages them to sleep.¹ The raw position data from each video frame were smoothed by convolution of both x and y position data with a normalized Gaussian function (standard deviation of 120 video frames). After smoothing, the instantaneous velocity was calculated by taking the difference in position between successive video frames. An epoch during which the velocity was constantly below 0.78 pixels/s (~0.19 cm/s) for more than 2 min was considered a stillness period. We find that this definition of sleep has the advantage of being very reliable for identifying true sleep periods (based on manual sleep classification of LFP) and in fact find that this criterion has the opposite problem of excluding short periods of “true sleep”. We found that 7% of sleep is excluded, which is comprised of these short sleep periods excluded by the measure used here, based on combined LFP/EEG/Accelerometer recordings in other mice (not included in this manuscript) where a shorter 4s minimum still period was used. Note, in the literature, 4s and 10s minimum still periods are frequently used so this is a “worst case” value.^55–58 All analyses of rest sessions were limited to these still periods. Still data were also collected in an identical manner for Experiment 1. For Experiment 2 only, SWS and REM sleep were identified during these still periods using K-means clustering of the theta/delta power ratio extracted from the CA1 pyramidal layer LFP recorded during stillness,^59–64 which has been validated.⁶⁵

2.9. Cross-frequency analyses

Phase-amplitude coupling (PAC).

To assess the degree to which gamma amplitude was modulated by theta phase throughout SWS, we considered 6 phase-frequency bands (theta, 7–10 Hz; 0.6-Hz increments, 1.2-Hz bandwidth) and 91 amplitude-frequency bands (gamma, 30–120 Hz; 1-Hz increments, 2-Hz bandwidth). Hippocampal LFP extracted from all SWS or REM sleep epochs (assessed separately) was bandpass-filtered, and the corresponding phase ($\phi$) and amplitude envelopes (A) were obtained using the Hilbert transform. As described in prior literature,^{66, 67} PAC was calculated as

$$PAC=\left|\frac{1}{n}\sum _{t=1}^{n}A*e^{i\phi }\right|$$

for each cross-frequency bin. PAC matrices were z-scored to control for variation (PAC_z). Representative comodulograms were generated from the average PAC_z matrix across all SWS sessions collected from each experimental task.

Permutation Testing. To determine significant differences in comodulation during the real world and virtual version of the task, we performed cluster-mass permutation testing,⁶⁸ using the average t-value within a significant cluster to evaluate significance. Observed t-values were first calculated for each cross-frequency bin. Half of the data was then randomly shuffled across experimental task groups, and unpaired t-tests were rerun for each bin. This randomization was repeated for 500 iterations to generate a null distribution of the largest cluster-mass values from each permutation. Observed clusters were considered significant if mass was less than the 12th or greater than the 488th values of the constructed null distribution (i.e., α = 0.05).^{67, 69, 70}

2.10. Histology and Imaging

After the conclusion of the experiment, mice (3xTg-AD no-sleep, n = 7; 3xTg-AD sleep, n = 7) were euthanized by administering an intraperitoneal injection of Euthasol and transcardially perfused with 1X phosphate-buffered saline (PBS), followed by 4% paraformaldehyde (PFA) in 1X PBS. Whole heads were post-fixed in PFA for 24 h to identify the placement of stimulating electrodes in the medial forebrain bundle and tetrode tracks when appropriate. After brain extraction, an additional 24 h post-fixation period was carried out, followed by cryoprotection of the brain in a 30% sucrose solution. Coronal frozen sections were cut at a thickness of 40 μm using a sliding microtome. Sections were split into 6 evenly spaced series for histological analysis. One series was processed for c-Fos for a subset of the data for a different larger study not reported here. Whole slides were imaged using a Zeiss Axio Imager M2 with a 10x objective (~100x magnification) and stitched together using Zeiss ZEN Blue software. Histology procedures were conducted as previously described.^{1, 16, 46} In summary:

2.10.1. M22 and M78.

Free-floating sections were blocked then incubated for two days in either primary antibody anti-MOC22 (monoclonal, rabbit, Abcam 205339) or anti-MOC78 (monoclonal, rabbit, Abcam 205341). Following a wash, sections were then incubated overnight in secondary antibody Anti-rabbit-alexa-488 (goat). Sections were stained for NeuN by utilizing fluorescently-labeled primary antibody anti-NeuN-Cy3 (polyclonal, rabbit, Millipore, ABN78C). Sections were mounted with DAPI-containing antifade mounting media, coverslipped, and imaged.

2.10.2. Phosphorylated Tau (AT8).

Free-floating sections were blocked then incubated overnight in primary antibodies anti-phosphorylated tau AT8 (monoclonal, mouse, MN1020; recognizes the Ser202 and Thr205 phosphorylation sites on tau) and anti-NeuN (polyclonal, rabbit, Millipore, ABN78). Following a wash, sections were then incubated 6 h in secondary antibodies anti-mouse-alexa-488 and anti-rabbit-alexa-594. Sections were mounted with antifade mounting media, coverslipped, and imaged.

2.10.3. 6E10.

Sections were mounted prior to incubation in 4% PFA for 4 min, followed by 70% formic acid for 2–4 min. Sections were then blocked and incubated for two days in primary antibodies anti-β-amyloid 1–16 (mouse, clone 6E10, Biolegend) and anti-NeuN (polyclonal, rabbit, Millipore, ABN78). Sections were washed, then incubated for 5–6 h in secondary antibodies anti-mouse-alexa-488 and anti-rabbit-alexa-594. Whole slides were rinsed, coverslipped with DAPI-containing antifade mounting media, and imaged.

2.11. Region of Interest Analyses

The density of positive cells for M22, M78, phosphorylated tau, and 6E10 intracellular pathology was assessed in four predefined regions of interest: retrosplenial cortex, CA1 field of the hippocampus, subiculum, and parietal cortex. These regions were further subdivided into dorsal (RSCd and SUBd) and ventral (RSCv and SUBv) parts for the retrosplenial cortex and subiculum. Dorsal CA1 (CA1d) encompassed all sections rostral to a point 2.55 mm posterior to bregma, while ventral CA1 (CA1v) included all sections caudal to that same point. Seven regions were examined in total. Using the Allen Brain Atlas, regions of interest (ROI) were manually drawn based on cytoarchitecture to locate regional boundaries. Automated cell counts were completed using Zeiss Zen 3.6 (Blue edition). Images were taken with a 10x objective (approximately 100x magnification) using a Zeiss Axio Imager M2 microscope. Outlines were manually drawn around the ROI in a 1:12 evenly spaced series throughout the brain using manual selection tool in Zen. The experimenter remained blind to groups. The number of M22, M78, pTau (AT8), and 6E10 labeled cells were then expressed as the proportion per unit of area.

2.12. Statistics

Following tests confirming that the data did not significantly deviate from a normal distribution (Shapiro-Wilk test, which we selected in part because it did not violate the sample size rules for the range of sample sizes used here), statistical comparisons for recording-array mice (Experiment 2) were performed using two-way repeated measures ANOVAs (genotype x delay or genotype x position on the track) and unpaired t-tests (for two-group or planned comparisons). A similar statistical approach was applied to Experiment 1 groups (non-Tg, 3xTg-AD, and 3xTg-AD sleep mice). Statistical comparisons of sleep data were conducted using one-way ANOVA and unpaired t-test if it passed Shapiro-Wilk test or Kruskal-Wallis and Mann-Whitney tests if it did not. Box plots instead of bar plots were used for all data on a given figure if there was at least one data set that significantly deviated from a normal distribution for consistency. Additionally, statistical comparisons of histology data between groups were performed for each set of brain regions with similar levels of pathology (CA1, SUB, and remaining cortical regions) using two-way repeated measures ANOVAs (group x ROI), followed by post hoc testing when appropriate (e.g., comparing sleep and no-sleep groups for each brain region). For some data the two-way ANOVA residuals deviated from normal distribution; therefore, we plotted all data using box and whisker plots. We explored transformations to improve these normality violating data sets and successfully improved two data sets. Finally, for the remaining three data sets, because non-parametric two-way tests have low power and ANOVA is robust against non-normality,⁷¹ and ultimately, we found a lack of differences we did not want to be attributed to a lower power test, we retained ANOVA analyses for the remaining non-normal data. p < 0.05 was considered significant. Statistical analyses were performed using StatView, GraphPad Prism, and R.

3. Results

3.1. 6-month 3xTg-AD no-sleep mice are impaired at spatial reorientation compared to the 3xTg-AD sleep group

Experiment 1.

Previously, we found that 6-month (but not 3-month) female 3xTg-AD mice were impaired at spatial reorientation learning and memory.¹⁶ For Experiment 1, we assessed mice without recording-arrays (i.e., under conditions identical to our prior study) but included an additional group of 3xTg-AD mice encouraged to sleep by being placed in a “sleep box” (a minimally stimulating “boring” chamber) before and after the spatial reorientation learning and memory behavioral session. We assessed spatial reorientation ability in 3xTg-AD sleep mice compared to non-Tg and 3xTg-AD no-sleep mice. The non-Tg and 3xTg-AD no-sleep mice do not undergo the pre- and post-task rest sessions, replicating our previous experimental conditions. In Figure 1d, we isolated velocity data (Z-scored) for the first half of the reward zone. This measure allows for comparison of performance across reward delays (i.e., across the range of task difficulty; 0.5 – 2.5 s) between groups. We found that 6-month 3xTg-AD sleep mice appear to perform similarly to non-Tg control mice, slowing more in the approach into the reward zone compared to 3xTg-AD mice, particularly at the 1 and 1.5 s reward delays. The velocity data did not deviate significantly from a normal distribution (W = 0.97, p = 0.06). Velocity varied significantly across groups (Fig. 1d; F_{(2, 68)} = 3.77, p < 0.05), but not across delays, and there was no significant interaction between delay and genotype (Fs_{(4–8, 68)} < 1.71, ps > 0.16). Planned comparisons to investigate the cause of the group main effect revealed that 3xTg-AD no-sleep mice performed significantly worse than non-Tg mice (F_{(1, 44)} = 6.54, p < 0.05). 3xTg-AD no-sleep mice also performed significantly worse than 3xTg-AD sleep mice (F_{(1, 48)} = 5.03, p < 0.05). Non-Tg mice and 3xTg-AD sleep mice did not differ from each other (F_{(1, 44)} = 0.017, p = 0.90).

Next, we assessed the velocity profile over the full length of the track (and not just the first half of the reward zone) during two intermediate reward delays with the greatest separation between groups on Figure 1a, the 1 and 1.5 s reward delays. During the 1 s delay, velocity varied across the length of the track (Fig. 1e; F_{(60, 1020)} = 3.63, p < 0.0001), but not across groups (F_{(2, 1020)} = 1.16, p = 0.34). The velocity profile also varied as a function of group and position on the track (genotype x position interaction: F_{(120, 1020)} = 4.75, p < 0.0001). Planned comparisons showed that 6-month female 3xTg-AD sleep mice slowed more than 6-month female 3xTg-AD no-sleep mice for multiple locations on the track, close to and in the reward zone and after the reward zone (ts< −2.22, ps < 0.05; *s in Fig. 1e), but towards the beginning of the track, an opposite effect was shown with 6-month female 3xTg-AD sleep mice running faster than 6-month female 3xTg-AD no-sleep mice (ts > 2.22, ps < 0.05; *s in Fig. 1e). Similarly, 6-month female non-Tg mice also slowed more than 6-month female 3xTg-AD no-sleep mice for multiple locations on the track prior to the reward zone, in the reward zone, and after the reward zone (ts > 2.22, ps < 0.05; #s in Fig. 1e). An opposite effect also appears towards the beginning of the track, with 6-month female non-Tg mice running faster than 6-month female 3xTg-AD mice (ts < −2.58, ps < 0.03; #s in Fig. 1e). In addition, 6-month female 3xTg-AD sleep mice ran more slowly than 6-month female non-Tg mice for multiple locations towards the end of the track and just prior to the reward zone (ts < −2.26; ps < 0.05, $s in Fig. 1e), but the opposite was seen in the middle of the track (ts > 2.32, p < 0.04; $s in Fig. 1e). Together, this pattern of data suggests that 6-month female 3xTg-AD sleep mice are slowing in the reward zone and are performing similarly or slightly better than 6-month female non-Tg mice, while 6-month female 3xTg-AD no-sleep mice are not slowing as well for the reward zone. 3xTg-AD no-sleep mice are also running more slowly over the earlier portions of the track, possibly indicating a lack of knowledge about the reward location.

During the 1.5 s delay, velocity again varied across the length of the track (Fig. 1f; F_{(2, 969)} = 2.61, p < 0.0001) but not across groups or as a function of group and position on the track (Fs_{(2–114, 969)} < 1.16, ps > 0.14). This suggests that all animals are adjusting their speed as a function of position on the track (potentially indicating some knowledge of the reward zone), but there are no differences between groups when considering data from the full length of the track.

To ensure that differences in running speed between the groups were not contributing to the observed effects, we assessed mean raw velocities from the portion of the track where differences in Z-scored velocities were not expected (i.e., from 104 cm to 39 cm in front of the reward zone). We looked at average velocity from the portion of the track because it is the portion of the track right after leaving the start box and well before the reward zone and thus running speed should not vary in order to perform the task (i.e., baseline velocity). This baseline velocity value did not vary between groups (F_(2,14) = 2.80, p = 0.09; Mean ± SEM velocity for 6-month non-Tg: 9.54 ± 0.98; 6-month 3xTg-AD sleep: 11.08 ± 0.56; 6-month 3xTg-AD no-sleep: 9.95 ± 0.53).

Finally, the mice in this experiment were able to meet criterion at each reward delay by performing similarly for three out of four days. The behavior sessions were always 20 min and the number of trials the mice would run each day varied. For the 3xTg-AD sleep mice, the average number of trials on the three days where the mice met criterion at the 0.5, 1.0, and 2.5 s reward delays was 10 trials and for 1.5 and 2.0 s delays it was 11 trials. For the 3xTg-AD no-sleep mice at the 0.5 and 1.5 s delays the average number of trials was 10, at the 1.0 and 2.0 s delays the average number of trials was 11, and at the 2.5 s delay the average number of trials was 9. For the non-Tg no-sleep mice, the average trials at the 0.5 and 1.5 s reward delays was 9 trials and for the 1.0, 2.0, and 2.5 s delays it was 10 trials.

3.2. 6-month female 3xTg-AD sleep mice are still for a significant portion of the daily rest sessions

We use a “boring” box to encourage mice to sleep, and our previous studies suggest that mice do spend significant time sleeping under these conditions.¹ However, those mice have a larger implant to allow for recording of brain activity patterns, which may further encourage immobility and sleep. Therefore, we next assessed stillness, defined as the amount of time they remained motionless for periods lasting at least 2 min, as a proxy for sleep in 3xTg-AD sleep mice (Fig. 2). We found that these mice were still¹ approximately 40% of the time they were in the sleep box (i.e., about 40 min per day). Furthermore, there was no significant difference in the proportion of stillness during pre-task and post-task rest sessions (F_{(1, 12)} = 2.04, p = 0.18). There was also no significant effect of reward delay (F_{(4, 48)} = 1.80, p = 0.15) and no significant interaction between the rest session and reward delay (F_{(4, 48)} = 1.02, p = 0.41) on sleep duration, suggesting that sleep did not change for the more difficult phases of the task. These data show that the mice were still for the same proportions during both rest sessions and likely spent a significant amount of time sleeping during these rest sessions. One hour may be a critical time point for sleep-dependent memory consolidation.⁷² In both rest sessions, approaching one hour, we found that these mice were still for approximately 80% of the last ten minutes at the 1.0 s and 1.5 s reward delays (1.0 s Average = 0.81, SEM = 0.04; 1.5 s Average = 0.84, SEM = 0.03).

Figure 2. Sleep in 3xTg-AD mice is increased during rest sessions immediately before and after a task.

a. Box and whiskers (min to max) reflect the proportion of stillness for non-Tg mice with a recording-array (dark blue), 3xTg-AD mice with a recording-array (light green), and 3xTg-AD non-recording-array mice (light purple). 3xTg-AD non-recording-array mice spend significantly less time still during sleep sessions than 3xTg-AD mice with a recording-array. b. 3xTg-AD mice with a recording-array had significantly fewer episodes of sleep but significantly higher proportions of sleep. Sleep in 3xTg-AD mice is similar to non-Tg recording-array mice. Box and whiskers (min to max) proportion of stillness, slow-wave sleep (SWS), and SWS episode length for non-Tg (dark blue) and 3xTg-AD (light green) mice. c. Time to sleep onset and wakefulness is reduced in 3xTg-AD mice. The time spent in SWS is increased in 3xTg-AD mice, but REM sleep is unchanged. d. Same as panel c but for proportions instead of time. e. The number of episodes of REM and SWS did not differ between groups. f. The SWS, but not REM, episode length was greater in 3xTg-AD mice. * p < 0.05, ** p < 0.01, *** p < 0.001.

3.3. 6-month 3xTg-AD sleep mice spend less time still than 3xTg-AD and non-Tg mice with a recording-array

Next, we compared results across experiments (Experiment 1 mice without recording-arrays versus Experiment 2 mice with recording-arrays) to explore potential impacts of the different experimental conditions (presence or absence of a recording-array). First, we compared stillness in mice with recording-arrays (Experiment 2; 3xTg-AD and non-Tg mice) to 3xTg-AD sleep mice without recording-arrays from Experiment 1 (Fig. 2a). Stillness varied significantly across groups (H = 13.66, p < 0.0001). Planned comparisons between each group showed that 3xTg-AD sleep mice with no recording-array were still significantly less than 3xTg-AD recording-array mice (Z = 3.67, p < 0.001). These data suggest that the larger recording-array leads to more sleep, but that even the lower amount of sleep induced in 3xTg-AD sleep mice without a recording-array is sufficient to reverse impairments in spatial reorientation learning and memory.

3.4. 6-month 3xTg-AD mice sleep more than non-Tg mice

Experiment 2.

There were two groups of mice which both had recording-arrays and rest sessions (i.e., were placed in the “boring” sleep box before and after the task): 3xTg-AD and non-Tg mice. This experiment was designed to confirm and extend findings from Experiment 1 by allowing us to also break down sleep periods into SWS and REM sleep.

We tested normality using Shapiro-Wilk test and found for most comparisons they did not deviate from a normal distribution (Ws > 0.84, ps > 0.17) except latency and REM (Fig 2c–d; Ws < 0.75, ps < 0.03). Thus, we used t-test for those that did not deviate from normal distribution and Mann-Whitney test for those that did. However, for consistency box and whiskers plots were used for all data.

First, we assessed sleep characteristics for 3xTg-AD and non-Tg mice with a recording-array. We found that 3xTg-AD mice had fewer sleep intervals using the original stillness criteria (Fig 2b; t₍₈₎ = 2.31, p < 0.05) and, consistent with previous work in our laboratory,¹ spent an increased proportion of time sleeping (t₍₈₎ = 3.68, p < 0.01). We also found that 3xTg-AD mice had fewer intervals of stillness with only the 0.78 pixels/s (~0.19 cm/s) threshold for movement (t₍₈₎ = 3.34, p < 0.05) and increased proportion of time still (t₍₈₎ = 3.10, p < 0.05). Finally, 3xTg-AD mice spent significantly less time awake while still (not shown; t₍₈₎ = 3.67, p < 0.01). Together, these findings suggest more and consolidated sleep in 3xTg-AD mice.

3xTg-AD mice also took significantly less time to fall asleep (U = 1, p < 0.05) and spent significantly less time awake (t₍₈₎ = 2.97, p < 0.05; Fig 2c). They spent significantly more time in SWS (t₍₈₎ = 3.94, p < 0.01; Fig 2c) and a larger proportion of their rest session in SWS (t₍₈₎ = 3.49, p < 0.01; Fig 2d). There was no significant difference between the groups in the number of SWS episodes (t₍₈₎ = 0.25, p = 0.81; Fig 2e); however, 3xTg-AD mice had a significantly longer average SWS episode length (t₍₈₎ = 2.53, p < 0.05; Fig 2f). There was no significant difference between the groups in the amount of time spent in REM sleep, the proportion of time in the session spent in REM sleep, the number of REM episodes or the average REM episode length (Fig 2c–f; ts₍₈₎ < 0.74, ps > 0.48). The findings of increased sleep in 3xTg-AD mice at this early pre-plaque/tangle stage of amyloid and tau aggregation is consistent with our previous observations that these sleep characteristics are also apparent later when these mice are learning a new task.¹ Thus, the sleep benefit for 3xTg-AD mice is not due to differences in sleep characteristics between 3xTg-AD and non-Tg mice.

Given the increase in SWS, we next examined whether post-task sleep predicted performance the following day (velocity in the reward zone) using multilevel correlations with animal ID as a factor. We also accessed REM sleep to determine whether REM vs. SWS explained the improved performance. Neither duration, proportion, episode number, nor episode length significantly predicted subsequent performance for SWS (rs_(22–28)<0.24, ps>0.27) or REM sleep (rs_(7–9)<0.16, ps>0.65). These results suggest that while group-level differences implicate SWS as critical for improved performance in 3xTg-AD mice (Fig. 2c), within-group variation in SWS does not account for within group task performance differences.

Finally, to assess the relationship between brain stimulation rewards and post-task sleep, we performed correlations between a number of sleep measures (duration, proportion, intervals and episode length for both SWS and REM sleep) with the number of brain stimulations obtained that day. In non-Tg mice, the number of stimulations (potentially reflecting physical exertion or other variables) predicted increased REM sleep for all measures (rs < 0.22, ps > 0.02) and decreased SWS (rs < −0.29, ps > 0.001). The only exceptions were SWS and REM intervals, which were positively correlated for both measures (rs < 0.24, ps > 0.008), and REM episode length, which was not significantly correlated with the number of MFB stimulations (r = −0.10, p = 0.24). In 3xTg-AD mice, only one measure was significantly correlated with stimulation count: REM episode length increased with more stimulations (r = 0.16, p = 0.05). No other sleep measures were significantly correlated in 3xTg-AD mice (rs < 0.11, ps > 0.14). These findings suggest that rewards may increase REM sleep (and potentially decrease SWS, although this may simply reflect the reciprocal classification of rodent sleep stages) in non-Tg mice. However, the mechanisms driving increased SWS in 3xTg-AD mice may overcome the effect of stimulation, preventing alterations in sleep characteristics in this mouse model.

3.5. 6-month female 3xTg-AD sleep mice with a recording-array are not impaired at spatial reorientation learning and memory

As a next step, we assessed spatial reorientation ability in mice implanted with a recording-array for Experiment 2, in which rest sessions took place just before and after the behavioral session for all mice. We assessed spatial reorientation ability in the 3xTg-AD sleep mice with a recording-array compared to non-Tg sleep mice with a recording-array. As in Experiment 1, we isolated Z-scored velocity data for the first half of the reward zone (Fig. 3a). We again found that the data did not significantly deviate from a normal distribution (W = 0.99, p = 0.95). We found that velocity for 6-month female mice varied significantly across delays (Fig. 3a; F_{(4, 32)} = 5.63, p < 0.05), indicating that task performance changed with increasing difficulty. However, velocity did not vary across genotype, and there was no significant interaction between delay and genotype (Fs_{(1–4, 32)} < 1.71, ps > 0.17). These findings suggest that the addition of rest sessions helped reverse the impairment previously observed in 3xTg-AD mice.

Figure 3. 6-month 3xTg-AD female recording-array mice are not impaired at spatial reorientation.

a. Mean (± SEM) Z-scored velocity from the first half of the reward zone (blue bar in b) for each reward delay (0.5–2.5 s) for 6-month female non-Tg (green) and 3xTg-AD (purple) recording-array mice. 3xTg-AD recording-array sleep mice did not perform significantly different than non-Tg recording-array sleep mice. b. Mean (± SEM) Z-scored velocity plotted for position on the track during the 1 s reward delay for non-Tg and 3xTg-AD recording-array sleep mice. Both non-Tg and 3xTg-AD recording-array sleep mice slowed into the reward zone (blue bar). c: Same as panel b, but for the 1.5 s reward delay. 3xTg-AD recording-array sleep mice did not significantly differ from their non-TG sleep controls.

Next, we examined the mean Z-scored velocities for the full length of the track individually for two reward delays where we had observed group differences in Experiment 1 (the 1 and 1.5 s delays; Fig. 1b). During the 1 s delay, velocity did not vary across the length of the track, genotype, and as a function of genotype and position in the room (Fig. 3b; Fs_{(1–53, 424)} < 4.20, ps > 0.07). During the 1.5 s delay, velocity varied across the length of the track (Fig. 3c; F_{(61, 488)} = 2.29, p < 0.0001), but it did not vary across genotype and as a function of genotype and position in the room (Fs_{(1–61, 488)} <0.95, ps > 0.59). This pattern of data further suggests that 6-month non-Tg and 3xTg-AD female sleep mice with recording-arrays were also performing similarly.

3.6. 6-month non-Tg mice perform similarly with or without sleep

Next, we again compared results across experiments (Experiment 1 mice without recording-arrays versus Experiment 2 mice with recording-arrays) to better understand the pattern of these data across experiments. To assess whether sleep benefited 6-month female non-Tg mice with a recording-array in Experiment 2, we compared their velocity to that of 6-month female non-Tg mice without a recording-array from Experiment 1, who did not experience pre- and post-task rest sessions. We first compared the Z-scored velocity isolated from the first half of the reward zone between the two groups during each reward delay. Velocity did not significantly vary across groups and delays, and there was no interaction present between delay and group (Fs_{(1–4, 36)} < 2.51, ps > 0.06). Although no significant difference was observed in the velocity in the first half of the reward zone, we also evaluated the velocity profile along the entire track during the two reward delays (1 and 1.5 s delays), where group differences were previously observed in Experiment 1. For both delays, velocity varied across the length of the track (Fs_{(54–59, 531)} > 1.43, ps < 0.05), but did not vary significantly across groups and there was no interaction between group and position on the track (Fs_{(1–59,498–531)} < 0.66, ps > 0.68). Finally, to ensure that differences in movement velocity (i.e., behavior) did not account for the lack of sleep-related improvement in control mice, we assessed velocity on portion of the track covered in the shortest track configuration (ensuring even sampling across trials). We then focuses on the half of this segment furthest from the reward zone (i.e., when task dependent velocity should not vary). Non-Tg sleep and no-sleep groups did not differ in velocity for this portion of the track (t₍₉₎=0.96, p=0.36), suggesting that behavioral differences are unlikely to explain the lack of sleep-related benefits in non-Tg mice. Overall, this suggests that while 3xTg-AD mice did benefit from sleep, non-Tg mice did not.

Further, to ensure that differences in the number of trials to criterion were not likely to account for the results we compared these values across groups. For the 3xTg-AD sleep mice with recording-arrays in Experiment 2, the average number of trials on the three days when the mice met criteria at 0.5, 2.0, and 2.5 s reward delays was 9 trials; at the 1.0 s delay and 1.5 s delay the average was 8 and 10 trials respectively. For the non-Tg mice with recording-arrays in this experiment, the average number of trials on the days that met criteria for the 0.5 and 2.5 s reward delays was nine trials; at the 1.5 and 2.0 s reward delays the average was 11; and at the 1.0 s reward delay it was 10 trials.

3.7. No differences in spatial learning and memory for 3xTg-AD sleep mice with recording-arrays (Experiment 2), which performed significantly better than no-sleep non-recording-array 3xTg-AD mice, but not differently from non-recording-array 3xTg-AD sleep mice (Experiment 1)

To ensure that these groups did not differ because the sleep mice had a recording-array while the no-sleep mice did not, we performed two additional comparisons: 1) 3xTg-AD sleep mice with recording-array versus without a recording-array; 2) 3xTg-AD no-sleep mice without a recording-array versus 3xTg-AD sleep mice with a recording-array. If the presence of a recording-array does not impact these comparisons significantly, then comparison 1 should not differ but comparison 2 should. Velocity did not vary across groups, and there was also no significant interaction between delay and group (Fs_(1–4,40) < 1.63, ps > 0.19). We found that the velocity varied significantly across delay (F_(4,40) = 4.012, p < 0.01). We did the same comparison between the 3xTg-AD recording-array sleep mice from Experiment 2 and the 3xTg-AD non-recording-array mice without sleep from Experiment 1. We found a significant effect between groups (F_(1,48) = 5.03, p < 0.05), with the 3xTg-AD sleep mice from Experiment 2 slowing more to obtain the reward stimulation. Velocity did not vary across delay, and there was also no interaction between delay and group (Fs_(4,48) < 1.45, ps > 0.23). This suggests that the presence of a recording-array was not a key factor in the results we observed.

Finally, we had previously characterized sleep and memory-related activity patterns during sleep in these mice, including delta waves (DW) in parietal cortex and interactions with sharp-wave ripples in the hippocampus (SWRs).¹ Thus, we looked for evidence whether these measures might explain variability in the improved performance of 3xTg-AD sleep mice. First, we examined the relationship between SWR×DW coupling during sleep and spatial reorientation performance the next day (slowing in the reward zone). SWR×DW coupling did not predict performance during the reward delays highlighted in Fig. 1 (1.0 s; r₍₂₆₎=−0.16, p=0.430). This lack of correlation is not surprising because SWR-DW coupling is impaired in 3xTg-AD mice at the 1.0 s reward delay. Specifically coupling increases in post-task sleep in non-Tg mice (mixed models to account for animal identity with data set as the sample: F(_1,3)=14.793, p=0.035) but not 3xTg-AD mice (F_(1,28)=1.587, p=0.218). Thus this coupling is impaired at this early timepoint and remains impaired in the same mice one month later as we reported previously.¹

Given that SWR-DW coupling is not a likely mechanism, we sought a different mechanism underlying the restored cognition. Theta-gamma coupling is thought to reflect synchrony between long-range communication and local processing,⁷³ so we next explored theta-gamma phase-amplitude coupling in the hippocampus. Prior work has focused on theta-gamma coupling during REM sleep, when theta power is high,⁷⁴ therefore, we examined theta-gamma cross-frequency coupling during REM sleep. We found that theta-gamma comodulation was apparent in all mice (Supplementary Fig. 1 Left & Middle); however, none of the correlations between theta-gamma comodulation and spatial reorientation task performance were significant (multilevel correlation accounting for animal identity: rs_(8–9)≤ |−0.39|, ps>0.166) for the 1.0 s and 1.5 s reward delays. REM sleep was relatively rare, and several datasets lacked REM entirely. Because extensive prior work (including our own) has implicated SWS in learning and memory, including spatial learning and memory^{14, 75–79}, we looked to see if theta-gamma cross-frequency coupling was also apparent during SWS. Although theta power did drop during SWS (which is classified by a theta-delta ratio reflecting higher delta and lower theta relative to REM), theta power remained relatively strong (Supplementary Fig. 1b Right). Which made it less surprising that cross-frequency coupling was strong during SWS in non-Tg mice (Fig. 4). 3xTg-AD sleep mice also exhibited strong theta-gamma comodulation during SWS that did not differ from control mice (Fig. 4a; ps > 0.05). However, approximately 1 month later – when the same mice were impaired on a virtual version of the task – theta-gamma coupling during SWS was now significantly reduced in 3xTg-AD sleep mice compared to Non-Tg sleep mice (Fig. 4b). Thus, we looked for a relationship between theta-gamma comodulation during SWS and spatial reorientation performance the next day. Maximum theta-gamma comodulation during SWS approached but did not reach significance for predicting performance in 3xTg-AD sleep mice at the 1.0s delay (multilevel correlation accounting for animal identity: r₍₂₈₎=−0.34 p=0.067). Comodulation also was not predictive at the 1.5s reward delay (r₍₃₉₎=−0.10 p=0.530).

Figure 4. Theta-gamma comodulation during SWS is preserved in 3xTg-AD sleep mice and is marginally predictive of spatial reorientation performance the following day.

a. During SWS adjacent to the spatial reorientation task, cross-frequency (phase-amplitude) coupling between theta phase and low-gamma amplitude in HPC was strong and did not significantly differ between non-Tg sleep mice (Left) and 3xTg-AD sleep mice (Right). b. In the same mice one month later, during a virtual version of the same task (behavioral data previously published¹; comodulation data shown here for the first time), theta–gamma comodulation was reduced in 3xTg-AD sleep mice and was significantly lower than in non-Tg sleep mice (see contour on the 3xTg-AD comodulation heatmap).

3.8. Pathology is not reduced in female 6-month 3xTg-AD mice with rest sessions

The mice in Experiment 2 were derived from a previous publication that included an additional virtual maze experiment following the spatial reorientation task data; thus, pathology was assessed 1 month after behavioral data collection. In contrast, the mice in Experiment 1 were euthanized immediately following the spatial reorientation task, therefore we only examined pathology in these animals (Fig. 5). The age at perfusion for the two groups of 6-month female 3xTg-AD mice (sleep vs. no-sleep; t₍₁₂₎ = 1.11, p = 0.29; sleep mean = 8.8 months; no-sleep mean = 8.2 months) did not differ significantly. We found few, if any, extracellular plaques were present at this age. Therefore, we quantified the density of the proportion of neurons (NeuN positive cells) that were also positive for intracellular accumulation of amyloid or pTau (AT8), measures we have previously shown to be associated with behavioral task performance.⁴⁶ We quantified pathology in the regions previously associated with task performance and their importance in spatial navigation, as well as in several control regions: the CA1 field of the hippocampus, parietal cortex (PC), and retrosplenial cortex (RSC).^{80, 81} The dorsal region of the CA1 field (CA1d) was counted separately from the ventral region (CA1v) because the dorsal region is thought to be critical for precise spatial navigation.^82–85 Finally, because we were measuring density of the proportion of pathology positive NeuN labeled cells we analyzed sets of regions with similar packing density separately: CA1, SUB and cortex (RSC and PC).

Figure 5. Histological markers of AD pathology do not show a significant decrease after sleep.

Box and whisker plots (Min to Max) of cell density for: a. Aβ 1–16 specific antibody (6E10); b. amyloid fibril conformation-specific antibody (mOC78; M78); c. amyloid fibril specific antibody (mOC22; M22), d. antibody specific to hyperphosphorylated isoforms of tau protein (AT8, pTau). While several brain regions showed significant differences, only three instances emerged in the sleep group: reduced Aβ staining (M78 in SUBv), increased Aβ staining (6E10 in SUBv), and increased pTau (AT8 in dCA1). Otherwise, sleep did not significantly impact tau or Aβ aggregation. For all stains, significant group differences are shown above (* p < 0.05, ** p < 0.01, *** p < 0.001). e. Representative image showing 6E10 (green) and NeuN (red) staining in retrosplenial cortex dorsal and ventral (RSCd/v), motor cortex secondary and primary (MOs/p), dorsal subiculum (Sub), and the CA1 field of the dorsal hippocampus (CA1). f. Same as panel e, but for M78 (green). g. Same as panel f, but for M22, showing parietal cortex posterior (PTLp) instead of MO. h. Same as panel e, but for pTau, also showing agranular RSC (RSCagl). Note: The DAPI (blue) channel is turned off in panels e-h for visualization purposes. Additional Abbreviations not shown on histological sections (e-h): ventral CA1 (CA1v). ventral subiculum (SUBv). Scale bar = 100 mm.

3.8.1. 6E10.

We found that pathology positive cell density did not deviate significantly from a normal distribution for CA1 and cortex (Ws > 0.98, ps > 0.82) but did for SUB (W = 0.84, p = 0.0007). However, after lognormal transformation the SUB did not deviate significantly from a normal distribution anymore (W = 0.97, p = 0.71) and so we performed statistical analyses on the log transformed data for SUB. For consistency we used non-parametric plotting approaches for all pathology data and parametric or nonparametric statistics depending on the outcome of the normality test (i.e., non-parametric statistics were used if distribution significantly deviated from a normal one and otherwise parametric statistics were used). There was an effect of brain region (Fig. 5a; RSCd, RSCv, and PC: F_{(1.425, 17.10)} = 12.07, p = 0.0013) but no brain region x group interaction and no main effect of group (Fs_{(1–2, 12–24)} < 0.13, ps > 0.12). Further, planned comparisons indicated that in SUBv the sleep group had a significantly higher 6E10 density than the no-sleep group (t₍₂₄₎ = 2.27, p = 0.03), but there were no effects of sleep for any other region (t₍₂₄₎s < 1.18, ps > 0.25).

3.8.2. M78.

We found that pathology positive cell density did not deviate significantly from a normal distribution for CA1 and SUB brain regions (Ws > 0.94, ps > 0.13) but did for cortex (RSC and PC; W = 0.94, p = 0.02). However, for cortex it still deviates from a normal distribution after the lognormality transformation (W = 0.76, p < 0.0001). Therefore, as described in Methods, we retained the ANOVA data on non-transformed cortical data because it is better than the alternative of a low powered non-parametric test. There was again an effect of brain region in CA1 and subiculum (Fig. 5b; CA1: F_{(1, 12)} = 8.28, p = 0.01; subiculum: F_{(1, 12)} = 19.34, p < 0.001) but not in cortex (RSC and PC; F_{(1.451, 17.41)} = 2.15, p = 0.16). There was also brain region x group interaction in CA1 (F_{(1, 12)} = 5.859, p = 0.03) but not in subiculum or RSC/PC (Fs_{(1–2, 12–24)} < 0.57, ps > 0.52). There were also no group effects in any of the brain regions (Fs_{(1–2, 12–24)} < 0.003, ps > 0.95). Planned comparisons showed that in SUBv the sleep group had a significantly lower density of M78 than the no-sleep group (t₍₂₄₎ = 2.08, p = 0.049), but there were no effects of sleep for any other brain region (t₍₂₄₎s < 1.48, ps > 0.15).

3.8.3. M22.

We found that pathology positive cell density deviated significantly from a normal distribution for all brain regions (Fig. 5e; Ws < 0.52, ps < 0.005). After lognormality transformation, cortex no longer deviated from a normal distribution (W = 0.98, p = 0.49) but the remaining two brain regions still did (Ws < 0.91, ps < 0.02). Therefore, we continued with log transformed cortex data only. There was no effect of brain region (Fs_{(1–2, 12–24)} < 1.50, ps > 0.07) or group (Fs_{(1–2, 12–24)} < 0.16, ps > 0.50), and group differences did not vary across brain regions (group x brain region interaction; Fs_{(1–2, 12–24)} < 0.35, ps > 0.09).

3.8.4. pTau (AT8).

We found that pathology positive cell density deviated significantly from a normal distribution for all brain regions (Fig. 5f; Ws < 0.92, ps < 0.0001). After lognormality transformation, both CA1 and SUB did not deviate from a normal distribution anymore (Ws > 0.95, ps > 0.22) but cortex still did (W = 0.87, p = 0.003). Thus, we did statistical analysis on lognormality transformation of CA1 and SUB but non-transformed data for cortex. There was an effect of brain region in SUB (F_{(1.447, 17.37)} = 4.54, p = 0.04) but not in CA1 or cortex (RSC and PC; Fs_{(1–2, 12–24)} < 0.47, ps > 0.50). There was also an effect of group in CA1 (F_{(1, 12)} = 5.15, p = 0.04) but not in subiculum or cortex (Fs_{(1–2, 12–24)} < 0.007, ps > 0.93). There was also no brain region x group interaction (Fs_{(1–2, 12–24)} < 0.67, ps > 0.52). Planned comparisons showed that in CA1d the sleep group had a significantly higher density of pTau than the no-sleep group (t₍₂₄₎ = 2.29, p = 0.03), but there were no effects of sleep for any other brain region (t₍₂₄₎s < 1.73, ps > 0.09).

4. Discussion

We found that the impairment in spatial reorientation learning and memory previously identified in 6-month female 3xTg-AD mice¹⁶ (and replicated in this study) is reversed when daily 50 min rest sessions are positioned immediately before and after the task. To confirm that actual sleep, rather than rest, may explain the lack of a spatial reorientation learning and memory impairment in 3xTg-AD mice, we assessed stillness and SWS during the pre- and post-task rest sessions in recording-array mice. We found that mice spent nearly all of these still bouts in either SWS or REM sleep. Interestingly, and consistent with our prior work,¹ 3xTg-AD mice slept more than non-Tg mice. This additional sleep may benefit 3xTg-AD mice by enhancing memory consolidation during sleep^86–88 or due to the clearance of pathology that has been shown to occur during sleep.^{21, 89, 90} Or improved cognition could result from improving sleep alone.⁹¹ Thus, we assessed Aβ and tau aggregation in sleep versus no-sleep 3xTg-AD mice and found few differences in Aβ (two regions with more and one region with less in the sleep group) or tau aggregation (except that there was significantly more tau in the dorsal CA1 fields of the hippocampus of the 3xTg-AD sleep group). Taken together, our data suggest that sleep may be capable of reversing cognitive impairments that occur as a consequence of tau and Aβ aggregation by enhancing memory consolidation, independent of sleep-induced reductions in tau or Aβ levels. Thus, when our results are combined with the extensive literature showing that sleep reduces Aβ and pTau aggregation, our findings suggest that sleep may have a dual benefit: both reducing pathology and independently facilitating cognition.

Previous studies have shown that undergoing a sleep session after learning enhances the retention of declarative information, as well as improving performance of procedural skills.^{86–88, 92} SWS specifically has been associated with facilitating the consolidation of declarative memories.^{86, 88, 93, 94} Further, post-learning sleep deprivation has been shown to lead to impaired memory in both humans and animals.^{92, 93, 95–97} Hippocampal-cortical interactions during similar sleep sessions have been shown to be important for memory consolidation and navigation.^{1, 12, 94, 98–101} Impaired hippocampal-cortical interactions in 3xTg-AD mice may underlie difficulties in learning a virtual spatial reorientation task.¹ Pattern reactivation, which occurs during sleep, may be crucial for memory consolidation, including spatial memories.¹⁰² For example, in rats, hippocampal neuronal activity observed during spatial tasks was shown to re-activate during subsequent sleep in the same spatio-temporal pattern.^{15, 86, 103, 104} Pattern reactivation has primarily been observed during SWS.^{103, 105} There is also evidence for pattern reactivation in humans.¹⁰⁵ Human neuroimaging studies have shown a similar phenomenon where hippocampal reactivation during sleep was shown to be similar to activity present during prior learning.^{103, 106} During SWS, hippocampal activity in humans was similar to activity present during a prior spatial navigation task.¹⁰⁵ Interestingly, the 6-month female non-Tg mice that underwent pre- and post-task rest sessions were shown to perform similarly to 6-month female non-Tg mice that did not undergo rest sessions. Thus, additional sleep may not be as beneficial for non-Tg animals. Finally, while sleep-facilitating learning-related activity patterns is the most parsimonious explanation for our present findings, because we did not measure sleep in home cage (no-sleep) mice, we cannot fully rule out other explanations with the present experimental design which did not involve additional groups with recordings in the home cage or colony room. It is possible that some aspect of remaining in the testing room facilitated learning in these mice. If so, visual stimuli are unlikely to be the source of this facilitation, since the sleep chamber obscured nearly all visual input. It is possible that exposure to some other sensory cue from the testing room (e.g., smell) contributed, but the sliding track renders olfactory cues uninformative. Therefore, sleep remains the most likely factor explaining this facilitation in learning and memory in 3xTg-AD mice.

AD patients experience disrupted sleep, which includes increased wakefulness after sleep onset, as well as increased latency to sleep.¹⁰⁷ These sleep disorders can be present early in disease progression and worsen over time.¹⁰⁷ Patients experiencing more SWS in post-learning sleep have been shown to perform better on recall of episodic memories.¹⁰⁸ People with AD have been shown to spend less time in SWS, and animal models of AD, including the 3xTg-AD mouse model, have been shown to have impaired SWS.¹⁰⁹ However, impaired sleep in 3xTg-AD mice typically occurs at later time points than those examined here, after plaques have begun to form around 9-months of age, 3-months older than the mice in the present study.^{17, 18} Interestingly, in the present study the 6-month female 3xTg-AD mice with low levels of tau and amyloid aggregation actually fell asleep more quickly and had more SWS. One month later, while performing a virtual version of this task, these same mice had increased time spent still and increased SWS bout length, suggesting that subtle sleep changes may begin to occur around 6-months of age.¹ Given that 6-month female 3xTg-AD sleep mice were not impaired on the spatial reorientation task, sleep, while short, may still be benefiting 3xTg-AD mice by assisting with the consolidation of spatial learning and memory. Interestingly though, 3xTg-AD non-recording-array sleep mice spent significantly less time still during their rest sessions than both non-Tg and 3xTg-AD recording-array mice, suggesting 3xTg-AD non-recording-array sleep mice spent less time overall in SWS during these sessions. This smaller amount of sleep could potentially be due to having a smaller and lighter implant. However, the 3xTg-AD non-recording-array sleep mice are still not impaired at spatial reorientation learning and memory, suggesting that even a smaller amount of sleep can still be beneficial and, while there may be an important threshold for the amount of sleep needed to restore cognition, these mice were able to gain enough to benefit behavior and restore performance to the level of non-Tg mice. What may be more important is when sleep occurs, for example, immediately after a behavioral task, which has been shown to be beneficial in rodents.^{86–88, 92} Further, 3xTg-AD sleep mice spent more time in SWS, but not REM sleep, than non-Tg mice, suggesting that SWS may be particularly important for the improved spatial learning and memory observed in 3xTg-AD mice. However, the amount of SWS was not significantly correlated with performance the following day, indicating that within-group variation in sleep does not predict variation in behavior. Combined with the observation that pathology levels did not change, these findings support the idea that even a small amount of sleep is sufficient to facilitate cognitive performance. Finally, theta-gamma comodulation during SWS was comparable in 3xTg-AD sleep and non-Tg mice immediately adjacent to the spatial reorientation task. One month later, however, this comodulation was impaired in 3xTg-AD mice during a virtual maze task for which they were behaviorally impaired – an age at which sleep no longer recovered cognition. This finding is also consistent with the idea that activity patterns during SWS are critical for the improved performance observed in younger 3xTg-AD mice. In other words, theta gamma coupling during sleep was capable of restoring cognition; however, once this coupling broke down as the mice aged the mechanism for restored cognition was gone and behavioral impairments emerged.

Sleep has also been linked to the removal of waste, including amyloid and tau aggregates. Conversely, decreased sleep has been shown to lead to a decrease in the clearance of both Aβ and tau, as well as an increase in phosphorylated tau.²³ Both the glymphatic system and the blood-brain barrier (BBB) are thought to participate in the clearance of waste from the interstitial space of the brain.⁹⁰ During sleep, the volume of the interstitial space increases, and decreases again upon waking.²¹ Changes in volume here influence the influx of cerebrospinal fluid (CSF), and the exchange with interstitial fluid (ISF) is enhanced.⁹⁰ Aβ accumulation occurs early in disease progression, and once enough has accumulated, insoluble plaques form. These plaques can contribute to sleeping problems, which in turn will facilitate further Aβ accumulation.⁶ Enhancing sleep quality and the amount of sleep may help with both decreasing the risk of progression to AD and/or slow or halt AD symptom progression by assisting in the clearance of Aβ and tau.⁵ Previously, we showed that 6-month female 3xTg-AD mice are impaired at spatial reorientation and that these mice had a tau pathology profile across the parietal-hippocampal network that was correlated with spatial reorientation behavior, suggesting a role of tau pathology affecting this behavior.^{16, 46} Thus, in the present study, there may be a potential increase in the clearance of these pathological products, potentially during SWS, which could enhance spatial learning and memory. However, this hypothesis was not supported by the results, as we found one instance of a reduction and two instances of an increase in amyloid aggregation with no instances of reduced tau phosphorylation in the parietal-hippocampal network. It should be noted that we did not measure the levels of small aggregates that are present early in AD and have been shown to be cleared by sleep.^{110, 111} However, given the short duration of the rest sessions and lack of impact on the measures employed here, it seems unlikely that the impact of sleep was due to reduced Aβ and pTau levels. Inflammation is increased by sleep deprivation,¹¹² so it is possible that the sleep mice had reduced inflammation (which we did not measure); however, given the brief rest sessions and the fact that 3xTg-AD mice sleep more – not less (Fig. 2) – this explanation is less likely. Surprisingly, the only group difference in tau aggregation was in the opposite direction: 3xTg-AD sleep mice had increased AT8-positive cell density in dorsal CA1 compared to no-sleep 3xTg-AD mice. One speculative possibility is that increased neural activity during the rest session¹ led to increased tau, since neural activity has been shown to increase tau production.¹¹³ Thus, sleep-facilitated memory formation remains the most likely explanation for the improved cognition we observed in 3xTg-AD sleep mice.

Finally, the high correspondence between LFP recordings and behavioral sleep classification underscores the reliability of our sleep behavioral sleep metrics. However, as noted in the Results, this does not imply that behavioral metrics are superior to LFP-based methods. Achieving high accuracy with behavioral sleep metrics requires a long stillness interval, which, as we demonstrate, fails to capture a small number of short bouts of true sleep. Each method therefore has distinct advantages and limitations: the behavioral approach is less invasive, whereas the LFP-based method is better suited for detecting brief sleep episodes. Because we have not observed impairments in male mice on the task employed here up to 12 months of age, we restricted our analysis to female 3xTg-AD mice. Similarly, the sophisticated spatial reorientation task we used – designed to mimic one of the earliest cognitive impairments in humans progressing toward AD (i.e., getting lost in new surroundings) – requires more than one month to complete, making additional testing impractical. We also focused on young mice with low levels of tau and amyloid aggregation. This approach aligns with our goal of targeting early interventions to preserve function when subtle cognitive impairments first emerge. Nonetheless, it remains important to determine whether older mice can also recover cognition when post-task rest sessions are employed.

The younger mice used here, with only pronounced intracellular aggregation, exhibit more sleep, particularly SWS, during short rest sessions compared to non-Tg controls. In contrast, older mice with extracellular plaques show reduced and fragmented sleep,^{4, 31, 114, 115} suggesting they may benefit less from rest sessions. Nevertheless, our findings demonstrate that even short rest sessions were sufficient to recover cognition in younger mice. Therefore, as long as learning-related activity patterns during sleep remain intact, some recovery of function in older mice could be possible; however, our present data suggests that a key activity pattern during sleep (theta-gamma comodulation) would be degraded in older mice potentially preventing the ability of sleep to recover cognition. Still directly testing this hypothesis in future studies is important. Future work should also examine sleep in male mice. This would require work with other rodent models where both sexes have cognitive impairments. Finally, future studies should use sleep with a standard battery of cognitive tests (e.g., Morris Water maze, Y-maze, etc.) to ensure the robustness of the sleep effects we observed here.

In summary, we previously demonstrated that 6-month female 3xTg-AD mice exhibit impaired spatial reorientation and that tau pathology in the parietal-hippocampal network is indicative of this behavior. This study shows that introducing pre- and post-task rest sessions to 3xTg-AD mice can reverse their spatial reorientation learning and memory impairments, bringing their performance back to non-Tg-like levels. Interestingly, despite the 6-month female 3xTg-AD non-recording-array mice spending less time still during rest sessions compared to their recording-array counterparts, this limited rest was sufficient to reverse their cognitive impairment. This finding suggests that even a relatively brief amount of sleep may benefit spatial reorientation learning and memory—provided that a critical threshold is met or, importantly, that sleep occurs post-task. Additionally, the lack of reduced pathology in 3xTg-AD sleep mice implies that the cognitive benefits of sleep are likely driven by memory consolidation during sleep rather than the clearance of amyloid or tau. These findings distinguish the effects of sleep-related memory consolidation from those of amyloid/tau clearance, suggesting that extending sleep duration in the early stages of Alzheimer’s disease could simultaneously facilitate memory consolidation (based on the present results) and promote clearance of pathology (based on a large body of prior work), offering dual benefits to alleviate cognitive impairment.

Data Availability:

Data and Matlab Code (beyond that which has already been shared in our publication cited in this manuscript) will be made available upon request.