Abstract
The impact of tau pathology on sleep microarchitecture features, including slow oscillations, spindles, and their coupling, has been understudied, despite the proposed importance of these electrophysiological features toward learning and memory. Dual orexin receptor antagonists (DORAs) are known to promote sleep, but whether and how they affect sleep microarchitecture in the setting of tauopathy is unknown. In the PS19 mouse model of tauopathy MAPT (microtubule-associated protein tau) P301S (both male and female), young PS19 mice 2–3 months old show a sleep electrophysiology signature with markedly reduced spindle duration and power and elevated slow oscillation (SO) density compared with littermate controls, although there is no significant tau hyperphosphorylation, tangle formation, or neurodegeneration at this age. With aging, there is evidence for sleep disruption in PS19 mice, characterized by reduced REM duration, increased non-REM and REM fragmentation, and more frequent brief arousals at the macrolevel and reduced spindle density, SO density, and spindle-SO coupling at the microlevel. In ∼33% of aged PS19 mice, we unexpectedly observed abnormal goal-directed behaviors in REM, including mastication, paw grasp, and forelimb/hindlimb extension, seemingly consistent with REM behavior disorder (RBD). Oral administration of DORA-12 in aged PS19 mice increased non-REM and REM duration, albeit with shorter bout lengths, and increased spindle density, spindle duration, and SO density without change to spindle–SO coupling, power in either the SO or spindle bands, or the arousal index. We observed a significant effect of DORA-12 on objective measures of RBD, thereby encouraging future exploration of DORA effects on sleep-mediated cognition and RBD treatment.
SIGNIFICANCE STATEMENT The specific effect of tauopathy on sleep macroarchitecture and microarchitecture throughout aging remains unknown. Our key findings include the following: (1) the identification of a sleep EEG signature constituting an early biomarker of impending tauopathy; (2) sleep physiology deteriorates with aging that are also markers of off-line cognitive processing; (3) the novel observation that dream enactment behaviors reminiscent of RBD occur, likely the first such observation in a tauopathy model; and (4) a dual orexin receptor antagonist is capable of restoring several of the sleep macroarchitecture and microarchitecture abnormalities.
Introduction
Sleep is thought to have a bidirectional relationship with Alzheimer's disease (AD) pathology such that sleep disruptions can accelerate the accumulation of AD pathology, and worsening AD pathology can lead to the breakdown of various aspects of sleep physiology (Ju et al., 2014; Cedernaes et al., 2017). Several studies have suggested that both amyloid and tau can have a negative impact on sleep duration and continuity (Varga et al., 2016; Dufort-Gervais et al., 2019; Wang and Holtzman, 2020). The impact of AD pathology on sleep microarchitecture features, including slow oscillations, spindles, and their coupling, has been less studied, despite the proposed importance of these electrophysiological features toward learning and memory processes (Mölle et al., 2009; Oyanedel et al., 2014; Latchoumane et al., 2017; Kam et al., 2019a; Schreiner et al., 2021). Further, because some of the earliest tau pathology is expressed in the locus ceruleus (Braak and Del Tredici, 2011; Stratmann et al., 2016) and can spread to neighboring brainstem structures important in sleep/wake regulation (Braak and Del Tredici, 2011; Iba et al., 2015), tau pathology may represent the earliest driver of sleep physiology changes even in preclinical AD and may carry disproportionate weight toward sleep disturbances observed in AD dementia versus amyloid. In cognitively normal older human subjects, measures of increased tau load by either CSF biomarkers or tau PET imaging have been associated with reduced sleep spindle density (Kam et al., 2019b), reduced coupling between spindles and slow oscillations (Winer et al., 2019), and reduced slow-wave activity (Lucey et al., 2019).
To isolate the effects of tau in the current study, we used the PS19 mouse model of tauopathy, in which a human tau transgene containing the P301S mutation is present in all neurons. This mutation is found in human subjects who develop frontotemporal dementia (Bugiani et al., 1999; Sperfeld et al., 1999) and increases the aggregation of tau into neurofibrillary tangles. PS19 mice have a reduced lifespan of ∼12 ± 2 months compared with littermate controls. Prior work evaluating sleep in PS19 mice revealed reduced REM duration at ∼9 months of age, with reduced non-REM duration and reduced slow-wave activity at 11 months of age (Holth et al., 2017). By evaluating PS19 mice and littermate controls at 2–3 months of age when no tau pathology is present and 10–14 months of age when both neurofibrillary tangles and neurodegeneration are present, we aimed to characterize sleep macroarchitecture and microarchitecture as a function of aging in this tauopathy model.
Orexin, also known as hypocretin, is a neurotransmitter crucially involved in the regulation of wake and sleep (de Lecea, 2021), and autoimmune loss of orexin neurons results in narcolepsy (Mahoney et al., 2019), characterized by excessive daytime sleepiness. The primary source of orexigenic neurons is in the lateral hypothalamus, and they project diffusely to other neurons in the brainstem, diencephalon, and cortex. Dual orexin receptor antagonists (DORAs) have been developed to promote sleep and are used clinically for the treatment of insomnia, including in mild-to-moderate probable Alzheimer's disease (Herring et al., 2020). By evaluating the response to a DORA versus vehicle control in aged PS19 mice, we aimed to evaluate the extent to which DORAs promote sleep in tauopathy and DORA's effects on slow oscillations, spindles, and their coupling.
Materials and Methods
Animals/surgery
All experimental procedures were approved by the Animal Care and Use Committee at the Icahn School of Medicine at Mount Sinai. PS19 mice, also known as MAPTP301S (microtubule-associated protein tauP301S) were a gift from the Ehrlich/Gandy lab (Audrain et al., 2019) and kept on a C57BL/6J background (The Jackson Laboratory). Animals were housed on a 12 h light/dark schedule and provided ad libitum access to food and water. Implantation of EEG/electromyogram (EMG) electrodes was performed as previously described (Kam et al., 2019a). Mice were anesthetized with inhaled isoflurane and placed in a stereotaxic apparatus (David Kopf). Two subdural electrodes (2.5-mm-diameter screws with tapered tips, Pinnacle Technologies), symmetrically placed over left and right primary motor cortices (1.5 mm anterior to bregma, ±2.0 mm lateral to the midline) served as EEG electrodes. Two epidural screw electrodes were placed above the cerebellum to serve as reference and ground. A bipolar, twisted stainless steel electrode (California Fine Wire) inserted into the nuchal muscles served as an EMG site. After implantation, a six-pin connector (Mill-Max) was centered over the skull with dental cement (Dentsply Sirona), and the animal was placed in its home cage on top of a heating pad set to 37°C (Harvard Apparatus) until fully ambulatory. All animals were supplied with subcutaneous hydration and pain control (buprenorphine) following surgery. Twenty-four-hour recordings (12 h light/dark cycle) were performed in mice of both PS19 genotype and nontransgenic littermates at either 2–3 months (termed young; WT, n = 12, 6 male/6 female; PS19, n = 14, 7 male/7 female), when little to no overt tau pathology is present, or 10–14 months (termed old; WT, n = 10, 5 male/5 female; PS19, n = 14, 6 male/female), when tau hyperphosphorylation and NFTs are present throughout many cortical and brainstem areas.
Acute DORA-12 administration
DORAs are a class of hypnotic currently in use to treat insomnia. DORA-12 (Merck Research Laboratories) at 100 mg/kg or vehicle (peanut oil) was administered to PS19 mice (n = 9, 6 male/3 female) twice daily at both zeitgeber time (ZT)0 and ZT9 by oral gavage during periods of 24 h video EEG/EMG recording (12/12 h light/dark cycle). Two doses per day were selected to promote sleep in both the light and dark phases of a 24 h cycle. To avoid prolonged washout periods, vehicle was always administered before DORA-12.
Sleep analysis
All 24 h video EEG/EMG analyses (young vs old PS19 and DORA-12 vs vehicle in aged PS19 mice) were performed in MATLAB using the FieldTrip toolbox (Oostenveld et al., 2011). Sleep/wake scoring was performed as previously described in a semiautomated way (Kam et al., 2019a). Video EEG/EMG was analyzed continuously to characterize behavioral states in 1 s epochs. Behavioral states [wakefulness, non-REM (NREM) and REM sleep] were classified in an epoch-free approach based on the following criteria: time-varying ratio of theta over delta power (theta, 5–10 Hz; delta, 1–4 Hz) using both the right and left primary motor cortex lead, presence of slow waves (delta power, 1–4 Hz) defined as segments with greater than one z score analyzed from both the right and left primary motor cortex lead across the entire recording, and movement detected by EMG and confirmed by a simultaneous manual review of the video.
Typical REM sleep was defined by a high ratio of theta/delta power (ratio >2.5), and little or no movement of the body (based on EMG <1 z score). In addition, a criterion for REM sleep was that the prior behavioral state was NREM sleep (which is the normal pattern for sleep in rodents). REM sleep segments separated by <3 s were merged because these were periods when small twitches or slight posture changes appeared to interrupt an otherwise continuous period of REM sleep. If movement was <1 z score, but other criteria for REM were not met, the behavioral state was classified as NREM sleep or quiet wakefulness. NREM was discriminated from quiet wakefulness based on power in the delta band and presence of putative spindles. Thus, NREM sleep showed a lower ratio of theta/delta power (<2.5) than quiet wakefulness. Sleep episodes were confirmed manually by reviewing video and finding that mice had a curled or resting body position. Periods with relatively low delta power (<1 z score) and minimal movement (<1 z score) were designated as quiet wakefulness. Periods with movement for >3 s were classified as active wakefulness and included exploration/walking, grooming, sniffing, consummatory behavior (eating/drinking), and spontaneous arousals from sleep, with the exception of RBD behaviors (see below, RBD scoring and quantification). Extended Data Figure 4-1 demonstrates example EEG/EMG tracings and spectral ratio analyses. All spectral thresholds were verified manually for each recording.
Figure 4-1
A, Example 1000 s clip of a PS19 mouse in REM with atonia. Top, Hypnogram. Right and left EEG spectrograms are plotted below hypnogram with high power in yellow and low to no power in blue. White trace indicates the ratio of theta/delta power, red trace indicates the ratio of spindle/gamma power, and the purple trace indicates the ratio of delta/gamma power across time. The EMG signal below hypnogram rectified and smoothed, and below the EMG signal is a 30 s clip of EEG centered on the white dashed line. Black arrowheads indicate bouts of REM with atonia. B, Example 1000 s clip of a PS19 mouse in REM without atonia with the same viewing parameters as described in A. Gray arrowheads indicate bouts of REM without atonia. Download Figure 4-1, EPS file.
For the arousal index, arousals are defined as scored Wake bouts that occur after NREM or REM sleep and last for a duration of 1–60 s. The arousal index is the number of arousals divided by the total sleep time.
Sleep spindles in the 10–16 Hz range were detected by a two-EEG channel sparse low rank optimizer (Parekh et al., 2017) or for each EEG channel individually by a normalized squared signal algorithm that has been previously described (Khodagholy et al., 2017). Spindle duration and sigma power (10–16 Hz) were determined using the McSleep detector (Parekh et al., 2017). Slow oscillations were detected from bandpass-filtered EEG channels between 0.5 and 1.5 Hz (zero-phase shift, third-order infinite impulse response filter) using published scripts (Cox et al., 2018). Coupling of spindles to the slow oscillation (spindle–SO coupling) was computed using phase-amplitude modulation of detected SO events (Cohen, 2008; Staresina et al., 2015). For every detected SO, the raw EEG signal was filtered into the two corresponding frequencies with lower and higher frequencies ranging from 0.5 to 30 Hz (0.5 Hz increments, with a 1 Hz filter bandwidth) and 10–30 Hz (0.5 Hz increments, with a 1 Hz filter bandwidth), respectively. The coupling between the low-frequency phase values and the phase of the amplitude envelope obtained by Hilbert transform was quantified by calculating the median phase-locking value synchronization index between the SO range of the lower phase modulated signal (0.5–1.5 Hz) and the spindle range (10–16 Hz) of the higher-amplitude modulated signal (Cohen, 2008).
RBD scoring and quantification
In old PS19 mice, we observed apparent dream enactment behavior including mastication, paw grasp, and forelimb or hindlimb extension during REM in PS19 mice undergoing 24 h sleep recordings, not typically observed in younger PS19 or age-matched nontransgenic mice of either sex, suggestive of a possible RBD phenotype. Wake-like behaviors occurred during theta-dominant EEG that immediately followed slow-wave sleep.
Atonia within REM sleep was measured by computing the EMG area, a metric that accounts for both the amplitude and duration of EMG activity during REM sleep. First, a high-pass (10 Hz) Butterworth filter was applied to the raw EMG signal. This filtered trace was rectified and smoothed using a 1 s convolution then z score standardized across the recording. Periods of EMG activity >0 z score were defined as EMG active (above the mean centered distribution). To characterize REM without atonia, we quantified the EMG area (area of suprathresholded EMG activity) during REM episodes in five ways as follows: duration of suprathreshold EMG area in REM sleep, cumulative EMG activity during REM sleep across the entire 24 h recording; cumulative suprathreshold EMG area during REM sleep across the entire 24 h recording; normalized suprathreshold EMG area during REM sleep, (suprathreshold EMG area divided by the duration of REM sleep); percentage of REM sleep without atonia (duration of REM without atonia divided by duration of REM sleep both across the entire 24 h recording); and in consolidated REM bouts lasting >20 s (with the first 10 s and last 5 s excluded), we computed the cumulative EMG area. We also computed the EMG area in the immediately preceding NREM bout. We then computed the ratio of the suprathreshold EMG area in REM/NREM as described by Valencia Garcia et al. (2018) in rats to compare muscle tone between sleep stages.
Immunohistochemistry
Mouse brains were prepared for immunohistochemistry as described previously with minor modifications (Varga et al., 2000). Phosphate-buffered saline-perfused brains were fixed overnight in 4% paraformaldehyde and embedded in paraffin. Antigen retrieval was completed by immersing slides in citrate buffer, pH 6, for 15 min at 90–95°C. Immunohistochemistry was performed on 8–10 μm coronal sections of the dorsal hippocampus using a synaptophysin primary antibody (1:1000; catalog #17785-1-AP, Proteintech) and horseradish peroxidase conjugated secondary antibody (1:2000; catalog #PK-4001, Vector Laboratories) followed by diaminobenzidine reaction (Vector Laboratories). Sections were counterstained with hematoxylin. Digital whole-slide images were acquired with a Hamamatsu NanoZoomer S60. Images were quantified using QuPath open-source pathology image analysis software (Bankhead et al., 2017). The CA3 region of the hippocampus was manually delineated. The area of positive synaptophysin staining was determined by a QuPath thresholder applied to all images, which was designed to exclude both hematoxylin and diffuse staining.
Novel object recognition task
The novel object arena was a white rectangular open field (40 cm × 40 cm × 30 cm). Mice were habituated to the arena three times for 5 min each in the absence of objects the day before testing. On the testing day, mice were placed in the arena with two identical objects in opposite quadrants and allowed to explore for 10 min (training). Mice were returned to the home cage for 60 min. Thereafter, mice were returned to the arena, where one of the two objects was replaced with a novel one, and allowed to explore for 10 min (recall). All objects were cleaned with 70% ethanol before the first trial of each session and between all trials. All training and recall sessions were evaluated by an experimenter blind to genotype who recorded object exploration time with a stopwatch. Object exploration time during both the training session and recall session was tallied whenever a mouse was touching the object or facing/sniffing the object within 3 cm of it. Preference for the novel object during the recall session was expressed as the percentage of time exploring the novel object in relation to the total duration of time spent exploring both the novel and familiar objects. Object identity and spatial location within the arena was consistent between the PS19 and wild-type groups. We ascertained that no mice showed significant preference for one of the objects during the training session (>66% time with one object).
Statistical analysis
All data are reported as mean ± SEM with the p criterion set to 0.05. For comparison of two means, either paired t tests or two-sample t tests were used. Wilcoxon rank sum tests were used to compare data that were not normally distributed. The Kolmogorov–Smirnov (KS) test was used to infer differences in NREM and REM sleep continuity via the cumulative distribution of bout lengths. Two-way ANOVA with main effect of genotype or age was used to assess differences in sleep measures between groups with the Dunn–Sidak post hoc test.
Results
Sleep macrostructure and continuity
We observed a significant interaction between genotype and age for non-REM sleep duration, such that non-REM was higher in young PS19 mice and declined with aging, whereas non-REM duration was lower in young littermate controls and increased with aging (Fig. 1A; genotype effect, F(1,46) = 1.11, pgenotype = 0.29; age effect, F(1,46) = 0.05, page = 0.82; interaction effect, F(1,46) = 9.1, pgenotype×age = 0.004). We observed a significant main effect of age and significant interaction between age and genotype for REM sleep, such that the decline in REM sleep duration with aging was greater for PS19 mice than littermate controls (Fig. 1B; genotype effect, F(1,46) = 0.21, pgenotype = 0.64; age effect, F(1,46) = 6.45, page = 0.01; interaction effect, F(1,46) = 8.72, pgenotype×age = 0.005). We observed a significant main effect of age and significant interaction between age and genotype for non-REM bout length, such that the decline in non-REM bout length with aging was greater for PS19 mice than littermate controls (Fig. 1C; genotype effect, F(1,46) = 2.44, pgenotype p = 0.12; age effect, F(1,46) = 12.11, page = 0.001; interaction effect, F(1,46) = 4.78, pgenotype×age = 0.03). We observed a significant main effect of age for REM bout length, such that the decline in REM bout length occurred similarly in both genotypes (Fig. 1D; genotype effect, F(1,46) = 0.24, pgenotype = 0.05; age effect, F(1,46) = 0.56, page = 0.004; interaction effect, F(1,46) = 0.12, pgenotype×age = 0.15). We observed significant main effects of genotype and age with a significant interaction between genotype and age for arousal index, such that the arousal index in aged PS19 mice was greater than all other groups (Fig. 1E; genotype effect, F(1,46) = 14.44, pgenotype < 0.001; age effect, F(1,46) = 9.95, page < 0.001; interaction effect, F(1,46) = 1.16, pgenotype×age = 0.009).
A, NREM duration across a 24 h recording period was variable (pgenotype = 0.29, page = 0.82, pgenotype×age = 0.004; post hoc, *p < 0.05 for young WT vs young PS19). B, REM duration across a 24 h period decreased in old PS19 mice (pgenotype = 0.64, page = 0.01, pgenotype×age = 0.005; post hoc, **p < 0.001 PS19 young vs old). C, NREM bout length was shortest in old PS19 mice (pgenotype = 0.12, page = 0.001, pgenotype×age = 0.03; post hoc, *p < 0.05 for PS19 old vs young PS19). D, REM bout length was shortest in old PS19 mice (pgenotype = 0.05, page = 0.004, pgenotype×age = 0.15; post hoc: **p < 0.001 for PS19 old vs young PS19). E, Arousal index was highest in the aged PS19 mice compared with the other groups (pgenotype < 0.001, page < 0.001, pgenotype×age = 0.009; post hoc, **p < 0.001 for PS19 old vs young PS19 and old WT). For this figure and subsequent figures, two-way ANOVA with Dunn–Sidak post hoc test was used; *p < 0.05, **p < 0.001, ***p < 0.0001. For this figure, nWTyoung = 12, nPS19young = 14, nWTold = 10, nPS19old = 14. F, Continuity curves (cumulative probability distributions) of NREM sleep runs in young PS19 and WT mice demonstrate more fragmented NREM sleep in the PS19 genotype (**p < 0.001, KS test). G, In old mice, PS19 genotype was also more fragmented compared with WT (***p < 0.0001, KS test). H, No differences were observed in REM sleep continuity for PS19 young and WT young mice (p = 0.118, KS test). I, In old mice, PS19 genotype showed more fragmentation in the bouts of REM sleep longer than 1 min in duration (***p < 0.0001, KS test).
Continuity of non-REM and REM sleep was further assessed by the cumulative distribution of sleep bout lengths. The cumulative bout distribution for non-REM sleep across 24 h recordings was significantly left shifted in both young PS19 mice compared with young WT (Fig. 1F; p < 0.001) and old PS19 mice compared with old WT (Fig. 1G; p < 0.0001), suggesting increased sleep fragmentation in non-REM. The cumulative bout distribution for REM sleep was not different in young mice (Fig. 1H; p = 0.12) but was significantly different in old mice such that PS19 mice displayed increased probability for short REM bout lengths less than ∼1 min and decreased probability for longer REM bout lengths greater than ∼1 min (Fig. 1I; p < 0.0001).
Sleep microstructure and coupling
We observed significant main effects of both age and genotype without interaction for spindle density by the McSleep detector (Parekh et al., 2017), with aged PS19 mice showing the lowest spindle density (Fig. 2A; genotype effect, F(1,46) = 5.10, pgenotype = 0.03; age effect, F(1,46) = 10.69, page = 0.002; interaction effect, F(1,46) = 0.97, pgenotype×age = 0.33). We observed a similar pattern with a separate normalized-squared-spindle power detector (Kam et al., 2019a; Fig. 2B; genotype effect, F(1,46) = 23.05, pgenotype < 0.001; age effect, F(1,46) = 8.77, page = 0.005; interaction effect, F(1,46) = 0.00, pgenotype×age = 0.97). For sleep spindle duration and power, there were striking significant effects of genotype, such that each was markedly reduced in PS19 mice, with an additional significant effect of age without interaction (spindle duration, Fig. 2C; genotype effect, F(1,46) = 151.99, pgenotype < 0.001; age effect, F(1,46) = 6.17, page = 0.02; interaction effect, F(1,46) < 0.001, pgenotype×age = 0.98); (spindle power, Fig. 2D; genotype effect, F(1,46) = 114.38, pgenotype < 0.001; age effect, F(1,46) = 5.68, page = 0.02; interaction effect, F(1,46) = 1.67, pgenotype×age = 0.20).
A, Spindle density was lowest in old PS19 mice with detector 1 (pgenotype = 0.03, page = 0.002, pgenotype×age = 0.33; post hoc, *p < 0.05 for old PS19 vs young PS19). B, Spindle density was also lowest in old PS19 mice with detector 2 (pgenotype < 0.001, page = 0.005, pgenotype×age = 0.97; post hoc, **p < 0.001 for old PS19 vs old WT). C, Spindle duration was lowest in old PS19 mice (pgenotype < 0.001, page = 0.02, pgenotype×age = 0.98; post hoc, *p < 0.05 for old PS19 vs old WT). D, Spindle power was lowest in old PS19 mice (pgenotype < 0.001, page = 0.02, pgenotype×age = 0.20; post hoc: *p < 0.05 for old PS19 vs old WT). E, SO density was higher in young PS19 mice than in young WT mice (pgenotype = 0.16, page = 0.85, pgenotype×age = 0.005; post hoc, *p < 0.05 for young PS19 greater than young WT). F, SO power was not different across groups (pgenotype = 0.24, page = 0.43, pgenotype×age = 0.06). No post hoc comparisons. G, Phase-locking value of the spindle to the SO was lowest in the old PS19 mice (pgenotype = 0.45, page < 0.0001, pgenotype×age = 0.08; post hoc, **p < 0.001 for young PS19 vs old PS19). H, WT young phase-amplitude coupling grandaveraged heat map. I, PS19 phase-amplitude coupling grandaveraged heat map. J, WT old phase-amplitude coupling grandaveraged heat map. K, PS19 Old phase-amplitude coupling grandaveraged heat map. For this figure, nWTyoung = 12, nPS19young = 14, nWTold = 10, nPS19old = 14.
We observed a significant interaction between genotype and age for SO density such that SO density was higher in young PS19 mice and declined with aging, whereas SO density began lower in young littermate controls and increased with aging (Fig. 2E; genotype effect, F(1,46) = 2.02, pgenotype = 0.16; age effect, F(1,46) = 0.04, page = 0.85; interaction effect, F(1,46) = 8.70, pgenotype×age = 0.005). There was no difference in SO power (Fig. 2F). We explored coupling of SOs and spindles and found a a significant main effect of age, without significant effect of genotype or interaction such that aged mice of either genotype displayed reduced spindle–SO coupling (Fig. 2G–K; genotype effect, F(1,46) = 0.59, pgenotype = 0.45; age effect, F(1,46) = 20.10, page < 0.0001; interaction effect, F(1,46) = 3.27, pgenotype×age = 0.08).
Synaptic markers and behavior
Having observed these sleep EEG microstructure changes in young PS19 mice before significant tau hyperphosphorylation or neurodegeneration, we sought to evaluate synaptic integrity by performing immunohistochemistry against synaptophysin, a presynaptic density marker. We measured synaptophysin in both young and old PS19 mice and wild-type littermates (Fig. 3A–D). In the CA3 subregion of the hippocampus, we found a significant effect of age, genotype, and interaction between the two (genotype effect, F(1,26) = 8.83, pgenotype = 0.006; age effect, F(1,26) = 11.52, page = 0.002; interaction effect, F(1,26) = 6.78, pgenotype×age = 0.01). In post hoc comparisons, we observed a significant reduction in synaptophysin intensity in old PS19 mice compared with young PS19 mice (p = 0.001) and old PS19 mice compared with old WT mice (p = 0.01) but notably not between young PS19 mice and young WT mice (p = 0.99; Fig. 3E).
A, Representative section of dorsal hippocampus tissue from a young (2–3 months old) PS19 mouse. B, An old (10–14 months old) PS19 mouse. C, A young (2–3 months old) WT mouse. D, An old (10–14 months old) WT mouse. E, Quantification of the percentage area of the CA3 hippocampal region stained with synaptophysin [pgenotype = 0.006, page = 0.002, pgenotype×age = 0.01; post hoc, not significant (ns) for young PS19 vs young WT, **p < 0.001 for young PS19 vs old PS19, *p < 0.01 for old PS19 vs old WT]. F, Novel object recognition memory in 2- to 3-month-old PS19 and age-matched WT controls (p = 0.64, n = 10 WT, n = 5 PS19). Immunohistochemistry data for this figure, nWTyoung = 10, nPS19young = 10, nWTold = 5, nPS19old = 5.
The changes in early-life sleep architecture could have implications on the off-line processing of memories. To evaluate this, we compared performance on the novel object recognition task in young (2–3 months old) PS19 and wild-type mice. We observed no significant difference in performance (p = 0.64), with both genotypes demonstrating preference for the novel object during the recall session suggestive of appropriate memory for the familiar object (Fig. 3F).
Dream enactment behavior
In addition to changes in sleep architecture and non-REM oscillations with aging in PS19 mice, we observed apparent dream enactment behavior including video recording of mastication, paw grasp, and forelimb/hindlimb extension during REM resembling RBD in aged PS19 mice with video EEG/EMG recordings but not in young PS19 mice or WT mice at any age (Movie 1). Prevalence of dream enactment behavior depended on age. It was never observed before 10 months of age, and among 27 PS19 mice age 10–14 months screened with 24 h video recording, 9 mice displayed subsequent video EEG/EMG-confirmed dream enactment, suggesting a prevalence of ∼33%. REM sleep was typically scored with the presence of a theta-dominant spectral frequency in the EEG and the absence of EMG tone following a bout of non-REM sleep (Fig. 4A, example of normal REM sleep with atonia). However, in the subset of mice exhibiting apparent dream enactment, REM without atonia was observed (Fig. 4B). Quantitatively, the absolute duration of REM without atonia and the percentage of REM without atonia (cumulative duration of suprathreshold EMG in REM), EMG area within REM sleep (cumulative suprathreshold EMG area during REM sleep, both absolute and normalized to REM duration), and the ratio of REM to NREM EMG area were all significantly increased in mice displaying dream enactment (Fig. 4C–G). REM sleep scoring with and without atonia was used to classify bouts of REM that may contain dream enactment-like behavior (Extended Data Fig. 4-1).
Still from a video of an aged PS19 mouse experiencing REM Behavior Disorder.
A, Example REM epoch in a mouse without RBD. EEG from the right and left parietal cortex and the nuchal EMG is shown. Extended Data Figure 4-1 contains additional spectral information for scoring REM with atonia. B, Example REM epoch in a mouse with RBD. Movements within REM sleep are detected as nuchal muscle deflections in the EMG. Extended Data Figure 4-1 contains additional spectral information for scoring REM without atonia. C, Suprathreshold EMG duration during REM was increased in aged mice with RBD (Wilcoxon rank sum test, ***p = 0.005). D, Cumulative EMG area in REM was increased in aged mice with RBD (Wilcoxon rank sum test, ***p = 0.005). E, Normalized cumulative EMG area in REM was increased in aged mice with RBD (Wilcoxon rank sum test, ***p = 0.005). F, Percentage of REM without atonia was increased in aged mice with RBD (Wilcoxon rank sum test, ***p = 0.005). G, Ratio of EMG area in REM divided by the preceding NREM bout was increased in aged mice with RBD (Wilcoxon rank sum test, ***p = 0.005). For this figure, nPS19RBD− = 8, nPS19RBD+ = 3.
Dual orexin receptor antagonist effect in aging
Having compared sleep architecture and spindle/SO measures in PS19 mice and WT controls at two ages, we sought to examine the effect of DORA-12 at 100 mg/kg twice daily on these sleep measures in aged PS19 mice. Across the 24 h recording period of the vehicle-treated or DORA-treated sessions, there was a significantly reduced latency to non-REM sleep with DORA compared with vehicle (Fig. 5A; p = 0.002). Non-REM sleep (Fig. 5B; p = 0.01) and REM sleep duration increased in response to DORA-12 (Fig. 5C; p = 0.01) whereas non-REM bout length (Fig. 5D; p = 0.02) and REM bout length both decreased (Fig. 5E; p = 0.03). Arousal index was not significantly different after treatment with DORA-12 (Fig. 5F). Continuity of sleep, assessed by the cumulative distribution of sleep bout lengths, was reduced for both non-REM sleep (Fig. 5G; p < 0.0001) and REM sleep (Fig. 5H; p < 0.001) with acute DORA-12 treatment. DORA-12 treatment resulted in a significant increase in spindle density (Fig. 6A; p = 0.007), with a similar trend albeit with slightly reduced effect size using a second spindle detector (Fig. 6B; p = 0.07). Spindle duration increased (Fig. 6C; p = 0.01) in response to DORA-12 with no change to spindle power (Fig. 6D; p = 0.29). SO density increased with acute DORA-12 (Fig. 6E; p = 0.03) without significant difference in SO power (Fig. 6F; p = 0.41) or spindle–SO coupling (Fig. 6G; p = 0.98).
A, Non-REM latency to the first bout of a 3 min or greater sleep episode was decreased with acute DORA-12 (*p = 0.002). B, Non-REM duration was increased with acute DORA-12 (*p = 0.01). C, REM duration was also increased with acute DORA-12 (*p = 0.01). D, Non-REM bout length was decreased with acute DORA-12 (*p = 0.02). E, REM bout length was decreased with acute DORA-12 (*p = 0.03). F, Arousal index was not different (not significant (n.s.)) between vehicle and DORA-12 (p = 0.12). G, DORA-12 reduced continuity of non-REM sleep (***p < 0.0001, KS test). H, DORA-12 reduced continuity of REM sleep (***p < 0.0001, KS test) versus vehicle (Veh) in aged PS19 mice as assessed via cumulative probability distribution curves. For this figure, n = 11.
A, Spindle density by detector 1 was increased with acute DORA-12 treatment compared with vehicle (**p = 0.007). B, Spindle density by detector 2 trended higher compared with vehicle (p = 0.07). C, Spindle duration was increased with acute DORA-12 (*p = 0.01). D, Spindle power was not changed with DORA-12 (p = 0.29). E, Slow oscillation density increased with acute DORA-12 (*p = 0.03). F, There was no change to SO power (p = 0.41). G, There was no change to phase-amplitude coupling of spindles with the slow oscillation (p = 0.98). For this figure, n = 11. n.s. = not significant.
Notably, we observed significant differences in objective measures of dream enactment with DORA-12 versus vehicle in PS19 mice manifesting the RBD phenotype (n = 9, 6 male/3 female), including significant decreases in normalized EMG area in REM sleep, percentage of REM sleep without atonia, and the ratio of REM over NREM EMG area in response to DORA-12 (Fig. 7A–C).
DORA-12 reduces quantitative measures of REM sleep without atonia in aged PS19 mice. A, Normalized cumulative EMG area in REM (paired t test, *p = 0.02). B, Percentage of REM without atonia (paired t test, **p = 0.004). C, Ratio of EMG area in REM divided by the preceding NREM bout (paired t test, **p = 0.002). For this figure, n = 9.
Discussion
We began by evaluating sleep macroarchitecture with aging in PS19 tauopathy mice versus littermate controls. With aging, we observed decreased non-REM and REM sleep duration and increased markers of sleep fragmentation, including decreased non-REM and REM sleep bout length and increased arousal index. These observations are largely consistent with prior observations of sleep macroarchitecture in PS19 mice (Holth et al., 2017); however, this prior work reported increased, rather than decreased, non-REM bout length at 9 months in PS19 mice, with no significant difference at 11 months. These differences could be a by-product of differences in sleep scoring, but given that wake and REM sleep were shown to be consistently fragmented, it seems likely to expect that non-REM sleep would also be fragmented.
Next, we evaluated several sleep microarchitecture properties such as biophysical features of sleep spindles, including density, duration, and sigma-band power. Spindle density was reduced with both advancing age and PS19 genotype without a significant interaction, suggesting that the slope of change in spindle density with aging was similar between genotypes. We observed more marked differences between genotypes in spindle duration and power, with PS19 mice displaying significantly reduced spindle duration and power, even at young ages, with further declines with aging in both genotypes. These findings are consistent with observations in humans showing a decline in spindle density with normal aging (Purcell et al., 2017), as well as with observations suggesting that greater tau burden with aging in cognitively normal adults correlates with reduced overall spindle density (11–16 Hz), fast spindle density (13–16 Hz), sigma-band power, and duration (Kam et al., 2019b; Mander et al., 2022).
Although we observed no differences in SO power among ages and genotypes, we observed a significant interaction between genotype and age for SO density. In early life, SO density is elevated in PS19 mice compared with controls and then declines with aging, whereas the opposite occurs in control mice, with SO density increasing with aging. The etiology of increased slow oscillations with aging in wild-type mice is not clear; however, this phenomenon has been observed previously (Campos-Beltrán and Marshall, 2021) and may be specific to increases in non-REM sleep during the active (dark) phase (Soltani et al., 2019). We suspect the reduction in slow oscillations in aged PS19 mice is likely because of cortical atrophy as opposed to the tau burden itself. Several studies have demonstrated reduced measures of slow oscillations with increasing cortical atrophy (Mander et al., 2014; Dubé et al., 2015; Varga et al., 2016), particularly in prefrontal cortical areas, in cognitively normal older subjects. In contrast, increasing tau burden by CSF fluid biomarkers (Kam et al., 2019b) or tau PET imaging (Winer et al., 2019) was not associated with measures of slow oscillations in cognitively normal older individuals (although tau load by PET imaging did inversely correlate with slow oscillations in an admixed population of cognitively normal subjects and subjects with mild cognitive impairment (Lucey et al., 2019)).
Reductions in spindle–slow oscillation coupling were largely driven by age, consistent with observations in human subjects (Helfrich et al., 2018). The degree of spindle–SO coupling decline with aging was somewhat greater in PS19 mice, but the significance of the age times genotype interaction was only at the trend level (p = 0.078). Overall, these observations are probably consistent with the inverse associations observed between medial temporal lobe tau burden by PET imaging and spindle–slow oscillation coupling in cognitively normal older individuals (Winer et al., 2019).
These combined sleep microarchitecture changes have several implications, particularly in the young PS19 mice. Young PS19 mice were recorded at an age when no tau hyperphosphorylation, tangle formation, or neurodegeneration is present, and yet there is a clear sleep EEG signature with elevated slow oscillation density, reduced spindle density, and markedly reduced spindle duration and power. These findings may be reflective of early synaptic loss, although we did not observe differences in synaptophysin immunohistochemistry between PS19 and wild-type mice at the age of 2–3 months. We acknowledge that loss or dysfunction of synapses at this early age could be reflected in alternative synaptic markers. Although synapse loss as reflected by decreased synaptophysin has been consistently reported with aging in PS19 mice, the precise age at which this becomes evident has been variably reported. Several synaptic markers including synaptophysin show reductions in staining at dentate gyrus synapses onto CA3 hippocampal neurons in PS19 mice anywhere from 3 months (with normal staining at 1 month; Yoshiyama et al., 2007) to 6 months (with normal staining at 3 months; Crescenzi et al., 2017). The early-life sleep microarchitecture phenotype observed here is consistent with the cortical hypoexcitability observed in rTg4510 mice (Menkes-Caspi et al., 2015; Busche et al., 2019) in which a similar MAPT P301L mutation is restricted to neocortical neurons, including reduced spindle/delta power ratio in sleep (Menkes-Caspi et al., 2015), and stands in contrast to the cortical hyperexcitability phenotype observed in amyloid mouse models lacking tauopathy (Busche et al., 2008), including early expression of interictal spikes during REM sleep (Kam et al., 2016). The early-life sleep architecture phenotype may not only serve as a noninvasive marker of potential future tauopathy and neurodegeneration but also predicts deficits in sleep-dependent memory tasks as measures of sleep spindles have been shown to predict off-line gains in motor (Kam et al., 2019a), object-place recognition (Yuan et al., 2021), and odor-reward pair memory (Eschenko et al., 2006; Mölle et al., 2009). Deficits in a novel object recognition memory task in young PS19 mice versus wild-type mice were not observed using a 1 h delay (Fig. 3F) or using a 24 h delay (Zampar and Wirths, 2021); however, long-term novel object recognition memory has been associated with spindle–slow oscillation coupling (Sawangjit et al., 2022), which is unaffected in young PS19 mice, rather than individual biophysical properties of spindles, such as density or mean duration.
In addition to changes in sleep macroarchitecture and microarchitecture with aging in PS19 mice, we also unexpectedly observed dream enactment behaviors reminiscent of RBD in a subset of mice. We were able to capture dream enactment behaviors with EEG/EMG recordings in recorded older PS19 mice and quantitatively demonstrated significant increases in cumulative normalized EMG tone during REM sleep, percentage of REM sleep without atonia, and ratio of REM to the non-REM EMG area. Although RBD is most commonly associated with the development of a subsequent synucleinopathy (Miglis et al., 2021), many rodent models of parkinsonism do not recapitulate RBD (Hunt et al., 2022), and only one transgenic mouse model using bacterial artificial chromosome–mediated alpha-synuclein overexpression has been shown to develop spontaneous RBD (Taguchi et al., 2020). To our knowledge, the PS19 mouse is only the second transgenic model and the only known tauopathy model to develop spontaneous RBD. There are several known causes of RBD outside synuclein pathology (Keir and Breen, 2020), including several tauopathies such as Guadeloupean parkinsonism (De Cock et al., 2007), progressive supranuclear palsy (Arnulf et al., 2005; Compta et al., 2009), and Alzheimer's disease (Schenck et al., 1996).
Dual orexin receptor antagonists are used clinically to promote sleep and improve insomnia, and we demonstrate here that DORA-12 reduces sleep latency and increases non-REM and REM duration across a 24 h light/dark cycle in aged PS19 mice. DORA-12 also decreased non-REM and REM bout length, similar to what is observed in narcoleptic patients who lack endogenous orexin signaling (Christensen et al., 2015; Kaushik et al., 2021; Maski et al., 2021). Notably, DORA-12 increased sleep spindle density and duration and increased slow oscillation density without having an impact on power in the spindle or slow oscillation bands in aged PS19 mice. A related DORA (DORA-22) was shown to improve sleep spindle count (by virtue of increased sleep time) but not spindle density in rat models of stress/insomnia (Gamble et al., 2020). In addition to the inherent benefits on sleep quality, the increased spindle and slow oscillation density could have implications for processing of memories. Although not all forms of sleep disruption are identical, acute administration of DORAs improved forms of spatial memory in rodent models of insomnia (Gamble et al., 2020) and optogenetically induced sleep fragmentation (Li et al., 2018), and chronic DORA administration improved memory performance in a mouse amyloid model (Zhou et al., 2020). Importantly, we also observed a significant reduction in several quantitative measures of RBD in response to DORA-12 that were not observed with vehicle in those older PS19 mice displaying the RBD phenotype. Currently, clinical treatment of RBD is largely limited to melatonin and clonazepam, so the possibility that dual orexin receptor antagonists could represent a novel therapeutic approach for RBD warrants further investigation.
In conclusion, we have demonstrated an early-life sleep EEG signature in the PS19 mouse model of tauopathy, consisting of reduced spindle density, duration, and power and elevated slow oscillation density. These changes occur at an age when there is no known neurodegeneration, tangle formation, or tau hyperphosphorylation, and sleep EEG may serve as one of the earliest noninvasive tests for future tauopathy. With aging, there are additional degradations to sleep characterized at both the macroarchitecture and microarchitecture levels, including the novel observation of an RBD phenotype. In aged PS19 mice, DORA-12 reduced sleep latency, increased non-REM and REM sleep duration, and increased sleep spindle density and duration and slow oscillation density. Future work is needed to evaluate potential benefits of DORAs on sleep-dependent cognitive processes, measures of RBD, and long-term consequences on markers of neurodegeneration.
Footnotes
This work was supported by Alzheimer's Association Grant 2018-AARG-589632, National Institutes of Health Grants R01 AG066870 and R01 AG056682, and the Merck Investigator Studies Program. We thank the laboratories of Drs. Cameron McAlpine and Ana Pereira for discussion and Drs. David Rapoport, Indu Ayappa, Ricardo Osorio, Ankit Parekh, Michael Bubu, and Anna Mullins for sleep physiology insight.
The authors declare no competing financial interests.
- Correspondence should be addressed to Andrew W. Varga at andrew.varga{at}mssm.edu
