Abstract
Disrupted sleep is a major feature of Alzheimer’s Disease (AD), often arising years before symptoms of cognitive decline. Prolonged wakefulness exacerbates the production of toxic amyloid-beta (Aβ) species, suggesting that sleep loss further accelerates AD progression. However, the mechanism by which Aβ affects sleep is unknown. We demonstrate here that Aβ acutely and reversibly enhances or suppresses sleep as a function of oligomer length. Genetic and pharmacological disruption revealed that short Aβ oligomers induce acute wakefulness through Adrenergic receptor b2 (Adrb2) and Progesterone membrane receptor component 1 (Pgrmc1), while longer Aβ forms induce sleep through Prion Protein (PrP) signaling. We suggest that changes to the brain’s balance of Aβ oligomeric species during AD progression disrupts the stability of a bi-directional, Aβ-sensitive sleep/wake switch.
One Sentence Summary Amyloid beta acutely and size-dependently affects behavior through distinct wake- and sleep-promoting pathways.
Main Text
The first known event in Alzheimer’s disease (AD) is an increase in soluble amyloid-beta (Aβ) with small changes in Aβ levels detectable years before the development of Aβ plaques and the onset of AD (1–2). Sleep disruption is also an early symptom of AD (3), and patients have sleep-wake abnormalities, including an increase in the amount of wakefulness at night and increased daytime napping (4–6). Multiple transgenic AD mouse models that overproduce Aβ also show disrupted sleep phenotypes (7–9), often in the absence of neuronal loss and preceding impairments of learning and memory (10). Understanding the relationship between Aβ accumulation, sleep disruption, and AD pathogenesis may therefore lead to better early detection and prevention of disease.
Some physiological changes, such as sleep, can alter the levels of extracellular Aβ (11). Aβ levels in the cerebrospinal fluid (CSF) increase during wake and decrease during sleep (11, 12), as the generation and release of both Aβ40 and Aβ42 are controlled by electrical and synaptic activity (13, 14). These observations have led to the proposal that sleep and Aβ dynamics create a vicious cycle, wherein increases in wakefulness result in increased extracellular Aβ and aggregation, which then dysregulates sleep, further exacerbating pathogenic Aβ production (9). How increased Aβ burden leads to disruptions in sleep remains unknown, although AD-related cell death of critical sleep/wake regulatory neurons has been suggested as a possible mechanism (15–17).
We tested an alternative hypothesis that Aβ species directly signal to alter sleep/wake states. We took advantage of the ability to directly deliver small molecules and Aβ peptides to the brain of larval zebrafish, which unlike Drosophila have a conserved repertoire of amyloid precursor protein (APP) and its processing machinery. Moreover, sleep in zebrafish is diurnal and controlled by similar genetic, pharmacological, and neuronal mechanisms as mammals (18). We found that Aβ size-dependently and reversibly modulates behavior through two distinct genetic, pharmacologically tractable pathways that regulate sleep in opposing directions.
To test the acute signaling effects of Aβ on sleep, we developed a minimally invasive intra-cardiac injection assay in 5 days post fertilization (5dpf) larval zebrafish, which avoids direct damage to brain tissue (Fig. 1A). This technique rapidly (<1hr, peaking within 2-3hrs) and reversibly delivers Aβ to the larval brain, as assessed by injection of fluorescently tagged Aβ42 and subsequent confocal brain imaging (Fig. 1B and Fig. S1A-B). To generate different Aβ oligomeric species, we incubated Aβ42 overnight at increasing temperatures, which significantly varied the amount and lengths of Aβ oligomers as measured by Transmission Electron Microscopy (TEM) (Fig. 1C and S1C). Aβ42 incubated overnight at 4°C consists of fewer and shorter oligomers (Aβshort, mean 45±11 nm, median=39nm) than when incubated at 25°C (Aβlong, mean 75±10 nm, median=61nm) or at 37°C (Aβv-long, mean 121±10 nm, median=88nm) (Fig. 1C).
We then assessed how each Aβ preparation affected sleep and wake behavior in zebrafish relative to an Aβ42-1 “reverse” peptide control (Aβrev) using video-monitoring (Fig. 1A) (19,20). In zebrafish larvae, sleep is defined as a period of inactivity lasting longer than one minute, as this is associated with an increased arousal threshold and other features of behavioral sleep (19). To determine the appropriate Aβ injection dose, we first injected 1nL of a 1-1000 nM dose response series for both Aβshort and Aβlong and found each species had opposing effects on larval sleep and waking activity (Fig. S1D-E), with maximal differences between these preparations at 10 nM (p≤0.01 doseXprep interaction, two-way ANOVA). As 1 nL of a 10 nM monomeric concentration of Aβ will yield a final estimated brain concentration <50 pM (see Methods), which is within the reported concentration ranges of Aβ42 in human CSF of 100 pM-5nM (2), we chose this dose for all subsequent experiments.
We found that Aβ42 affected sleep and wake in opposing directions depending on its oligomeric state (Fig. 1D-J). In the day following injection, Aβshort significantly increased waking activity (+19±6% relative to Aβrev) compared to Aβlong oligomers, which reduced waking activity by −9±5% (Fig. 1D-E, G). In contrast, Aβlong significantly increased sleep during the post-injection day by +19±8% while Aβshort decreased total sleep by −12±9% (Fig. 1D’-E’, I). Increased sleep induced by Aβlong was due to a significant increase in the average sleep bout length (Fig. 1I) but also due to a modest increase in the number of sleep bouts (Fig. 1J), indicating higher sleep initiation and consolidation. Injection of longer Aβ fibers (Aβv_long) had neither the wake-promoting effect of Aβshort nor the sleep-promoting effect of Aβlong (Fig. 1F-J), demonstrating that Aβ’s biological activity in vivo is narrowly size-dependent. There were no significant differences in sleep or waking activity among the Aβ preparations the night following injection (Fig. S1J-K).
Several lines of evidence suggest the dual-effects of Aβ on sleep and wake are due to acute signalling and not to long-term toxic effects from neuronal damage or death. First, injected animals returned to baseline sleep and activity levels within 24 hours (Fig. S1F-I). Second, injection with Aβ42 had no effect on apoptosis, as detected by staining for activation of Caspase-3 (Fig. S2A-C). Finally, Aβ injected animals raised to adulthood showed no major differences in their general health or in their survival rates (Fig. S2D). We therefore sought to identify the neuronal and molecule substrates through which Aβ signals to modulate sleep/wake behavior.
If Aβ oligomers alter behavior through acute signalling in the brain, the differential effects of Aβshort and Aβlong should be reflected at the level of neuronal activity. In situ hybridization (ISH) for expression of the immediate early gene, c-fos, identified several areas of the larval brain that are upregulated after injection of Aβshort relative to Aβrev, including the posterior hypothalamus and the dorsal and ventral telencephalon (Fig. 2A), which are also upregulated in hyperactive zebrafish mutants21. In contrast, c-fos expression following Aβlong injections was globally dampened relative to Aβrev (Fig. 2B). Because immediate early gene expression is an imperfect readout of changes in neuronal activity, we also used the MAP-Mapping technique to visualize the relative brain-wide changes in phosphorylated ERK levels, which strongly correlates with neuronal activity levels in zebrafish larvae (22). Consistent with c-fos, Aβshort upregulated P-ERK in the ventral telencephalon and posterior hypothalamus (Fig. 2C, C’ and Supplementary Data 1), while Aβlong resulted in a widespread reduction in P-ERK levels throughout most of the brain (Fig. 2D, D’ and Supplementary Data 2). These neuronal activity maps are consistent with the sleep and wake behavioral changes induced by Aβshort and Aβlong.
Many putative Aβ binding partners have been implicated in mediating the toxic signalling effects of Aβ on synapses, with some targets showing preferences for Aβ dimers, such as adrenergic receptor β2 (Adrb2) (23), or low molecular weight (50-75 kDa) species, such as the progesterone membrane receptor component 1 (PGRMC1) (24), while other targets preferentially bind to longer oligomers/protofibrils, such as the Prion Protein (PrP) (25,26). We therefore made genetic lesions in candidate Aβ receptors (Fig S3, S5) and tested whether these had any effect on Aβshort-induced waking activity. Although adrb2a−/− and pgrmc1−/− exhibited only minor changes in baseline sleep and wake on a 14hr:10hr light:dark cycle (Fig S4A-F’), they had profoundly altered responses after injection of Aβshort : Neither mutant increased waking activity (Fig. 3A-C,E) or suppressed sleep as observed in wild type controls (Fig. 3A’-C’,F). Injection of Aβshort into adrb2a−/− animals even significantly increased sleep instead of reducing it as in wild type larvae (Fig. 3B’,F). In contrast, Aβshort injected into mutants that lack both zebrafish Prp orthologs (prp1−/−; prp2−/−) (27,28) elicited even stronger increases in waking activity and reductions in sleep (Fig. 3D-F). Thus, the wake-promoting activity of Aβshort requires intact Adrb2a and Pgrmc1 but not functional Prp1 and Prp2.
Because the oligomeric species in the Aβlong preparation falls into the range that interacts with Prion Protein (PrP) with high affinity to modulate synapses (25,26,29,30), we tested whether PrP is instead required for Aβlong-induced sleep. Under baseline conditions, we detected no significant differences in day or night sleep and waking activity in prp mutants compared to prp+/+ siblings generated from either prp1+/−; prp2+/− or prp1+/−;prp2−/− in-crosses (Fig S6A-F’). However, after injection of Aβlong, prp1−/−;prp2−/− null mutants failed to increase sleep (−12±8% vs. +64±17% change in wild type sleep, Fig. 4A-D). Zebrafish have a third PrP ortholog (Prp3) with conserved GPI and repetitive prion protein domains but lacking all recognizable Aβ binding regions (Fig. S5A,B). To ensure all Prp function was eliminated, we made genetic lesions in prp3 (Fig. S5C,D) and found that the blockade of Aβlong–induced sleep in prp1−/−;prp2−/− double mutants was not additionally modulated by prp3−/− (Fig. S7B,D). Thus, while Prps are not required for the wake-inducing effects of Aβshort, functional Prp1 and Prp2 are essential for sleep induced by Aβlong.
If Aβlong interacts directly with Prp to drive changes in sleep, pharmacological manipulations known to disrupt this signalling cascade should similarly block the sleep-promoting effect of Aβlong (Fig. 4E). Consistent with a role for direct Aβlong-Prp interactions in sleep, soaking the larvae in Chicago Sky Blue 6B, a small molecule reported to disrupt Aβ-PrP interactions (31), significantly abolished the sleep-inducing effect of Aβlong (Fig. 4F-G, S7E-F). Similarly, pharmacological inhibition of the putative Aβ-Prp downstream signalling components mGluR5 and Fyn kinase (30,32) significantly blocked the sleep-inducing properties of Aβlong (Fig. 4H-K, S7G-H). Both the mGluR5 inhibitor MPEP and the Src-kinase inhibitor saracatinib even resulted in significant sleep reductions after exposure to Aβlong, consistent with Aβlong-Prp activating this pathway to signal sleep and oppose the wake-promoting effects of Aβshort. Thus, both genetic and pharmacological interference with Aβ-Prp signalling prevents the ability of Aβlong to increase sleep behavior.
Previous studies have suggested that changes in sleep during AD may further accelerate Aβ accumulation and neuronal damage, creating a viscous cycle that leads to further neuronal dysregulation and increased sleep-wake cycle abnormalities (9). Our results show that Aβ42 can acutely change sleep and wake depending on its oligomeric form, independently of neuronal death. Not only does our data identify behaviorally relevant targets of Aβ, but our proposed bi-directional model for Aβ sleep/wake regulation (Fig. 4L) predicts that alterations to the relative concentrations of Aβ oligomers during healthy aging and AD disease progression will have dual consequences on sleep and wake states.
Many potential receptors of Aβ have been proposed, but controversy surrounds both the nature of the oligomeric species that interacts with each target and their relative importance during progressive cognitive decline and other phenotypes associated with AD (33). We identified a wake-promoting pathway requiring Adrb2a and Pgrmc1 that is activated by shorter Aβ species and a distinct, Prp-dependent sleep-promoting pathway activated by longer oligomers. Since AD brain extracts have been shown to activate both PGRMC1 (24) and Prp-signalling (34), both signalling pathways may be directly relevant in AD sleep/wake regulation. However, our identification of Aβ targets that are activated in vivo at low concentrations to alter behavior also reveals potential roles for Aβ signalling in non-diseased contexts.
The Aβ assembly process is highly complex and meta-stability is an intrinsic property of Aβ solutions35. While the precise makeup of Aβ species present in healthy and AD brains remains difficult to determine, several studies have demonstrated that the amount and type of Aβ oligomeric species (including dimer, trimers, and 56 kDa oligomers) in healthy brains change across the human life cycle (36) and are heterogeneous and elevated in AD patients (34,37). Our data suggests that changes in the relative concentration of sleep- and wake- promoting Aβ oligomeric species over the course of both healthy aging and AD will signal either increases or decreases in sleep (Fig. 4L). On shorter time-scales, sleep deprivation acutely increases Aβ burden (38), fever promotes amyloidogenic Aβ generation (39), and microbial infection can trigger Aβ fibrilization (40). Considering that infection and fever are also potent drivers of sleep (41), the Aβ-Prp sleep signalling module identified here could mediate recovery sleep during illness and should be the subject of future studies. On longer timescales, AD progression is associated with increasingly large disruptions in sleep patterns, with patients exhibiting high levels of sleep fragmentation, a lack of circadian rhythm, night-time insomnia and irregular daytime napping throughout the day (42). Based on our results, the sleep symptoms associated with normal aging and AD may reflect changes in Aβ burden that lead to altered sleep signalling via molecular cascades that are potential therapeutic targets for treating disrupted sleep early in AD progression.
Funding
This was supported by operating funds to WTA from the Alzheimer Society of Alberta & Northwest Territories and Alberta Prion Research Institute of Alberta Innovates. PLAL was supported by Alzheimer Society of Canada, and JR was funded by an Interdisciplinary Research Grant from the Alzheimer’s Research UK, an ERC starting grant, and a UCL Excellence Fellowship.
Author Contributions
GGO and JR conceived and designed the experiments. GGO performed all experiments with assistance from SL. GGO, JR, and SL analysed the data. GGO and SL generated the prp3, adrb2a and pgrmc1 mutants. PLAL and WTA generated the prp1 and prp2 mutants. GGO and JR wrote the manuscript with input from all authors.
Competing interests
The authors declare no competing interests.
Data and materials availability
All data is available in the manuscript or the supplementary materials.
Acknowledgements
We thank Dr. Tom Hawkins and Mark Turmaine for assistance with TEM, Dr. Gaia Gestri for assistance with heart injections and Dr. Marcus Ghosh for assistance with tERK/pERK experiments. We also thank Dr. Dervis Salih, Prof. Steve Wilson and Prof. John Hardy for their comments and all first-floor fish lab members for their input throughout the project.