Sleep is bi-directionally modified by amyloid beta oligomers

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 amyloid-beta (Aβ) species, a major driver of AD progression, suggesting that sleep loss further accelerates AD through a vicious cycle. However, the mechanisms by which Aβ affects sleep are unknown. We demonstrate in zebrafish that Aβ acutely and reversibly enhances or suppresses sleep as a function of oligomer length. Genetic disruptions 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 a pharmacologically tractable Prion Protein (PrP) signaling cascade. Our data indicate that Aβ can trigger a bi-directional sleep/wake switch. Alterations to the brain’s Aβ oligomeric milieu, such as during the progression of AD, may therefore disrupt sleep via changes in acute signaling events.


Introduction
Accumulation of amyloid-beta (Ab) in plaques, along with tau tangles, is one of the two pathological hallmarks of Alzheimer's disease (AD). Change in Ab levels in the brain is one of the earliest known pathological events in AD and is detectable years before the development of Ab plaques and decades before the clinical onset of AD (Bateman et al., 2007;Jack et al., 2013). Because of its importance in AD progression, Ab has been mostly characterized as a functionless, pathological, intrinsically neurotoxic peptide (Moir and Tanzi, 2019). However, Ab is an ancient neuropeptide conserved across vertebrates through at least 400 million years of evolution (Moir and Tanzi, 2019). Ab's cleavage from amyloid precursor protein (APP) is tightly regulated by multiple enzymatic reactions (O'Brien and Wong, 2011), and its release from neurons is carefully controlled (Kamenetz et al., 2003). Ab interacts with numerous surface receptors and can activate intrinsic cellular signalling cascades to alter neuronal and synaptic function (Jarosz-Griffiths et al., 2016). More recently, Ab has been suggested to act as an antimicrobial peptide (Soscia et al., 2010), and the deposition of Ab may be induced as an innate immune defence mechanism against microbial pathogens (Kumar et al., 2016). However, the various biological effects of Ab in health or disease remain obscure.
One of the earliest symptoms of AD is the disruption of sleep, and AD patients have sleep-wake abnormalities, including insomnia at night and increased napping during the day (Allen et al., 1987;Loewenstein et al., 1982;Moran et al., 2005;Prinz et al., 1982). Multiple transgenic AD mouse models that overproduce Ab also show disrupted sleep phenotypes (Roh et al., 2012;Sterniczuk et al., 2010;Wang et al., 2002), often in the absence of neuronal loss and preceding impairments of learning and memory (Irizarry et al., 1997). In non-pathological conditions, Ab levels in the cerebrospinal fluid (CSF) are modulated by the sleep-wake cycle (Kang et al., 2009;Xie et al., 2013). Ab generation and release are controlled by electrical and synaptic activity (Cirrito et al., 2005;Kamenetz et al., 2003), leading to increased extracellular Ab levels during wakefulness and decreased levels during sleep (Kang et al., 2009;Xie et al., 2013). These observations have led to the proposal that sleep and Ab dynamics create a vicious feed-forward cycle, wherein increases in wakefulness result in increased extracellular Ab and aggregation, which then dysregulates sleep, further exacerbating pathogenic Ab production (Roh et al., 2012). How increased Ab 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 (Fronczek et al., 2012;Lim et al., 2014;Manaye et al., 2013).
Given the relationship between Ab and sleep, we hypothesized that Ab may directly modulate sleep-regulatory pathways independently of neuronal cell death. To test this, we took advantage of the ability to directly deliver small molecules and Ab peptides to the brain of larval zebrafish, which have conserved APP processing machinery and Ab peptides (Newman et al., 2014) and share genetic, pharmacological, and neuronal sleep-regulatory mechanisms with mammals (Barlow and Rihel, 2017). We found that Ab size-dependently and reversibly modulates behavior through two distinct genetic, pharmacologically tractable pathways that regulate sleep in opposing directions.

Ab dose-dependently modifies zebrafish sleep and wake behavior
Isolating the specific biological effects of Ab has been experimentally difficult. One challenge is that Ab is processed from a series of complex cleavage steps of a longer transmembrane protein, APP, which also produces other protein products with a variety of functions (O'Brien and Wong, 2011). This restricts the utility of genetic manipulations to tease out Ab-specific roles from the other APP components. Another challenge is that Ab forms, in vitro and in vivo, a variety of oligomeric species (e.g. dimers, longer oligomers, or large fibrils) with diverse structures, binding affinities, and signalling properties (Benilova et al., 2012;Jarosz-Griffiths et al., 2016). Teasing out the biological signalling capabilities of these diverse oligomeric species requires selective manipulation of Ab oligomeric states, which is difficult in vitro and is currently nearly impossible endogenously in vivo.
To overcome some of these barriers, we developed an injection assay in which the amount and type of the Ab oligomers can be controlled and then tested the acute signaling effects of Ab on sleep and wake behavior. Our minimally invasive intra-cardiac injection assay in 5 days post fertilization (5 dpf) larval zebrafish avoids direct damage to brain tissue ( Figure 1A and B). This technique rapidly (<1 hr, peaking within 2-3 hr) and reversibly delivers Ab to the larval brain, as assessed by injection of fluorescently tagged Ab42 and subsequent confocal brain imaging ( Figure 1B To generate different Ab oligomeric species, we modified previously established in vitro monomeric Ab incubation protocols (see Extended Methods) that enrich for Ab with different oligomeric sizes and opposing effects on rat neuronal excitability (Kusumoto et al., 1998;Orbán et al., 2010;Whitcomb et al., 2015). By incubating Ab42 overnight at increasing temperatures, we generated Ab oligomeric pools with significantly different lengths, as measured by transmission electron microscopy (TEM) ( Figure 1C and Figure 1-figure supplement 1C). Ab42 incubated overnight at 4˚C consisted of fewer and shorter oligomers (Ab short , mean 45 ± 11 nm, median = 39 nm) than when incubated at 25˚C (Ab long , mean 75 ± 10 nm, median = 61 nm) or at 37˚C (Ab v-long , mean 121 ± 10 nm, median = 88 nm) ( Figure 1C).
We then assessed how each Ab preparation affected sleep and wake behavior in zebrafish relative to an Ab42-1 'reverse' peptide control (Ab rev ) using automated video-monitoring (Prober et al., 2006;Rihel et al., 2010). In initial experiments, we determined the appropriate Ab injection dose by injecting 1 nL of a 1-1000 nM dose series for both Ab short and Ab long and assessing subsequent waking activity and sleep, which is defined in zebrafish larvae as a period of inactivity lasting longer than one minute, which are associated with an increased arousal threshold and other features of behavioral sleep (Prober et al., 2006). Unexpectedly, these oligomeric species had opposing behavioral effects (Figure 1-figure supplement 1D-G). Ab short increased waking activity and decreased sleep relative to Ab rev peptide, while Ab long decreased waking and increased sleep (prep waking effect, p<0.001; prep sleep effect p<0.05, two-way ANOVA). These effects were generally consistent across doses, although some dose-responses elicited stronger differential effects than others Figure 1. Ab oligomers bi-directionally affect sleep and wake in zebrafish larvae. (A) Experimental schematic. Ab was injected into the heart of 5 dpf larvae in the morning (ZT2 = zeitgeber time 2, that is 2hr after lights on). Behavior was then monitored in a square-welled 96-well plate for 24-48 hr on a 14 hr:10 hr light:dark cycle. (B) Heart-injected HiLyteTM Fluor 647-labeled Ab42 (fAb) penetrated the whole larval brain as visualized by confocal microscopy (optical sections, dorsal view) taken 2 hr after injection. Anterior is to the top. (C) Ab prepared under increasing temperatures adopted Figure 1 continued on next page (Figure 1-figure supplement 1D,E), with the maximal difference between the Ab short and Ab long preparations at 10 nM (p<0.01 doseXprep interaction, two-way ANOVA). We estimate this dose yields a final concentration that falls within the lower range of physiological concentrations reported for Ab42 in human CSF of 100 pM-5nM (Bateman et al., 2007). For example, assuming that all injected Ab goes into the brain, the highest possible concentration would be 1500 pg/ml or 300 pM (45 ng/ml x 1 nl in 30.4 nl brain = 1.5 ng/ml = 1500 pg/ml). At the lower end, assuming equal distribution of Ab over the whole body yields a final concentration estimate of 150 pg/ml or 30 pM (45 ng/ml x1 nl in 300 nl of body = 150 pg/ml). We therefore continued with 10 nM injections for all subsequent experiments, as it combines the maximal differentially observed behavioral effects between Ab short and Ab long with physiologically reasonable concentrations.
Ab affects sleep and wake in opposing directions as a function of oligomer size and independently of neural death To explore the effect of Ab oligomeric size on sleep, we then systematically tested the behavior effects of each Ab species relative to a Ab rev control for n = 3-5 independent experiments each. As in the dose response experiments, Ab affected sleep and wake in opposing directions depending on its oligomeric state ( Figure 1D-I'). In the day following injection, Ab short significantly increased waking activity by +12.8% and reduced total sleep relative to Ab rev by 15.5% ( Figure 1D-E'). The magnitude of the sleep effect is likely partially masked by a flooring effect due to generally reduced sleep during the day; we therefore favor reporting effect sizes and confidence intervals as recommended (Amrhein et al., 2019;Ho et al., 2019). Indeed, if we were to combine all the additional control Ab short experiments subsequently reported in this manuscript (n = 160 Ab rev n = 164 Ab short , see Figure 3G and H), the effect size remains robust at À15.9% and the result is statistically significant (p<0.05, one-way ANOVA). These effects were reversible, as there were no significant differences in sleep ( Figure 1E, black bar) or waking activity ( Figure 1D, black bar) between Ab short and reverse peptide in the night following injection, and the behavior of Ab short -injected larvae returned to baseline levels in the subsequent day ( Figure 1-figure supplement 3A).
In contrast, while injection of longer Ab fibers (Ab v_long ) had no effect on behavior, ( Figure 1H-I'), injection of the intermediate Ab long oligomers significantly increased sleep during the post-injection day by +47.2% and reduced waking activity by 11.3% ( Figure 1F-G'). The increased sleep induced by Ab long was due to a significant increase in the average number of sleep bouts but not an increase in sleep bout length ( Figure 1J), indicating higher sleep initiation is responsible for the change in sleep rather than an increased sleep consolidation. This increased sleep effect by Ab long was not Figure 1 continued longer oligomeric lengths, as measured by transmission electron microscopy. Each dot is a single oligomer (N = number measured), and the bars show the median. Data was taken from five randomly selected micrographs from two independent experiments. **p 0.01, ****p 1Â10-7 Kruskal-Wallis, Tukey-Kramer post-hoc test. (D, E) Exemplar 24 hr traces post-injection comparing the effect of Ab short (blue) on average waking activity (D) and sleep (E) versus Ab rev controls (grey). Ribbons represent ±the standard error of the mean (SEM). Light and dark bars indicate the lights ON and lights OFF periods, respectively. N = the number of larvae in each condition. (D', E') The effect of Ab short relative to Ab rev on waking (D') and sleep (E') during the first day is shown, pooled from n = 5 independent experiments. Each dot represents a single larva normalized to the mean of the Ab rev control, and error bars indicate ± SEM. The mean difference effect size and 95% confidence interval is plotted to the right. *p<0.05, T p <0.1, one-way ANOVA. (F, G) Exemplar 24 hr traces post-injection comparing the effect of Ab long (green) on average waking activity (F) and sleep (G) versus Ab rev controls (grey). (F', G') The effect of Ab long relative to Ab rev on waking (F') and sleep (G') during the first day is shown, pooled from n = 4 independent experiments. *p<0.05, **p<0.01, one-way ANOVA. (H, I) Exemplar 24 hr traces post-injection comparing the effect of Ab v_long (magenta) on average waking activity (H) and sleep (I) versus Ab rev peptide controls (grey). (H', I') The effect of Ab v_long relative to Ab rev on waking (H') and sleep (I') during the first day is shown, pooled from n = 3 independent experiments. (J) The effect of different Ab preparations on the number of sleep bouts relative to Ab rev controls. The difference effect size and 95% confidence interval is plotted below. The asterisks indicate statistically significant different effects among the preps (***p<0.001, one-way ANOVA). See also     observed in the night following injection ( Figure 1F and G, black bar), and behavior returned to baseline by the following morning ( Figure 1-figure supplement 3B).
This data is consistent with Ab short increasing wakefulness and Ab long decreasing wakefulness and increasing sleep. Additional control experiments ruled out experimental artefacts, as larvae undergoing no treatment, anesthesia only, mock injection, or PBS only injections had indistinguishable effects on sleep/wake (Figure 1-figure supplement 1H-J). Next, we recalculated the behavioral analysis only for the evening period before lights off, when vehicle-injected larvae were statistically indistinguishable from larvae that had been acclimated to the tracking rig for 24 hr (Figure 1-figure supplement 1J). Except for an even more severe flooring effect in the Ab short injection experiments, the results from evening-only analysis were indistinguishable from calculations across the whole day (Figure 1-figure supplement 1K). We therefore used full day analysis for all subsequent experiments.
We next considered if the dual-effects of Ab on sleep and wake are due to either neuronal damage or generalized toxic effects, such as the induction of seizure, paralysis, or sickness behavior.
First, injection with either long or short forms of Ab had no effect on apoptosis, as detected by staining for activation of Caspase-3 (Figure 1-figure supplement 2A-C). In addition, Ab injected animals raised to adulthood showed no major differences in their general health or in their survival rates ( Figure 1-figure supplement 2D). Moreover, injected animals recovered fully in the long term, returning to baseline sleep and activity levels within 24 hr (Figure 1-figure supplement 3A, B). Second, both Ab short and Ab long injected larvae responded normally to salient stimuli such as a light:dark pulse, demonstrating that these larvae were not paralyzed, in a coma, or undergoing sickness behavior (Figure 1-figure supplement 3C). Finally, we considered if the changes in motility in Ab-injected larvae were seizure-like behaviors. Wild type (WT) zebrafish larvae display 'burst-andglide' movements characterized by single short forward or turn movement followed by a short pause (Figure 1-figure supplement 3D and Figure 1-video 1). In contrast, epileptogenic drugs like the GABA-receptor antagonist PTZ induce electrophysiological and behavioral seizures (Baraban et al., 2005), which are observed as dramatic rearrangements in zebrafish bout structure (Figure 1-figure supplement 3D). The bout structure of Ab rev , Ab short , and Ab long injected fish was highly similar to WT behavior (Figure 1-figure supplement 3D,E and Figure 1-video 1), and the high-frequency bouts (HFB) indicative of seizures (Reichert et al., 2019) were only found in PTZ exposed fish but not Ab injected larvae (Figure 1-figure supplement 3D,E). Together these experiments indicate that exposure to Ab modulates normal sleep/wake behavior without inducing toxic states.
We conclude that the changes in behavior after Ab exposure are due to acute signalling events and therefore sought to identify the neuronal and molecular substrates through which Ab signals to modulate sleep/wake behavior.

Ab short and Ab long induce opposing changes in neuronal activity and differentially engage sleep-promoting neurons
If Ab oligomers alter behavior through acute signaling in the brain, the differential effects of Ab short and Ab long should be reflected at the level of neuronal activity. In situ hybridization (ISH) for expression of the immediate early gene, c-fos, identified several discrete areas of the larval brain that are upregulated after injection of Ab short relative to Ab rev , including the posterior hypothalamus and the dorsal and ventral telencephalon (Figure 2A), areas that are also upregulated in mutants with extended wakefulness (Ashlin et al., 2018). Comparing the Ab short induced c-fos patterns to WT brains collected at zeitgeber time 1 (ZT1, ZT0 = lights ON), when larvae are maximally awake, reveals at least nine populations of c-fos-positive neurons in both Ab short and waking brains (Figure 2A,C). In contrast, c-fos expression following Ab long injections was globally dampened relative to Ab rev ( Figure 2B) in a manner consistent with the low expression of c-fos in WT brains collected at ZT19, when larvae are maximally asleep ( Figure 2C).
Immediate early gene expression is an imperfect readout of changes in neuronal activity and brain state, as baseline c-fos is expressed in low amounts in zebrafish and has a relatively slow time course of 15-30 min for transcription of mRNA (Baraban et al., 2005). We therefore also quantified changes in the more rapid (<5 min) neuronal activity marker, phosphorylated ERK (p-ERK), using the larval zebrafish MAP-Mapping technique (Randlett et al., 2015). This method identifies the relative quantitative changes in brain region-specific levels of p-ERK relative to total ERK between Ab injections and reverse peptide control conditions. Consistent with c-fos induction, Ab short upregulated P-ERK in the ventral telencephalon and posterior hypothalamus ( Figure 2D  . Ab oligomers differentially alter neuronal activity in the larval zebrafish brain. (A) As detected by ISH, the immediate early gene c-fos is upregulated in many larval brain areas following Ab short injection, including the dorsal and ventral telencephalon (tel) and the posterior hypothalamus (black arrowheads), relative to Ab rev control injections. Other upregulated areas in the midbrain and hindbrain are indicated (white arrowheads). hyphypothalamus; hb-hindbrain. D = dorsal, p=Posterior, R = Right. n = blind counts of brains with the shown expression pattern/total brains. 24/43 Figure 2 continued on next page source data 1), while Ab long resulted in a widespread reduction in p-ERK levels throughout most of the brain ( Figure 2E and E', Figure 2-source data 2). These brain activity states are consistent with the induction of wakefulness by Ab short and sleep by Ab long .
Finally, if the behavioral states induced by Ab are bona fide sleep/wake states, we reasoned that known zebrafish sleep/wake regulatory neurons should be engaged. Galanin-expressing neurons of the preoptic area and hypothalamus are active and upregulate galanin transcription during zebrafish sleep (Reichert et al., 2019). Similarly, ISH for galanin 4-6 hr post-injection of Ab oligomers revealed that wake-promoting Ab short slightly decreased (À6%, blinded counts), while sleep-promoting Ab long slightly increased (+12%, blinded counts), the number of galanin-positive cells in the hypothalamus compared to Ab rev injected larvae ( Figure 2F-G). The differential effects on galanin neurons are consistent with that the induction of wakefulness by Ab short and sleep by Ab long .

Ab binding targets are required for behavioral responses to Ab
Many candidate Ab binding partners have been implicated in mediating the signalling effects of Ab on synapses, with some targets showing preferences for Ab dimers, such as Adrenergic Receptor b2 (ADRB2) (Wang et al., 2010), or low molecular weight (50-75 kDa) species, such as the Progesterone Membrane Receptor Component 1 (PGRMC1) (Izzo et al., 2014b), while other targets preferentially bind to longer oligomers/protofibrils, such as the Prion Protein (PrP) (Laurén et al., 2009;Nicoll et al., 2013). We therefore used Crispr/Cas9 to make genetic lesions in several zebrafish candidate Ab receptors, choosing examples with reported affinities for various sized Ab oligomers ( We also isolated an adrb2a allele with an 8 bp deletion that leads to a severely truncated protein lacking all transmembrane domains (Figure 3-figure supplement 1E-G). We also obtained a prp1 -/-;prp2 -/double mutant (Fleisch et al., 2013;Leighton et al., 2018) that lacks both zebrafish Prp proteins with conserved Ab binding sites (  stringently counts only brains with the major areas upregulated. (B) Compared to Ab rev injections, Ab long oligomers induce less c-fos expression. The Ab rev and Ab long treated brains were stained longer than in (A) to ensure detection of weaker c-fos expression. n = blind counts of number of brains with the shown expression/total brains. (C) c-fos is upregulated in many larval brain areas at 10 am (ZT1) awake fish, including the dorsal and ventral telencephalon and the posterior hypothalamus (black arrowheads), and other discrete regions of the mid and hindbrain (white arrowheads). c-fos expression is downregulated in later timepoints (ZT13) and is very low in ZT19 brains, when larvae are predominantly asleep. N = 10 fish/timepoint. (D, D') Brain expression of the neuronal activity correlate pERK/tERK comparing Ab short (n = 6) to Ab rev (n = 5) injected larvae identified areas upregulated (green) and downregulated (magenta) by Ab short . Data are shown as a thresholded maximum projection overlaid on the Z-Brain Atlas tERK reference (gray). White arrowheads indicate regions in the ventral telencephalon and posterior hypothalamus that are upregulated similar to c-fos in (A). Dorsal view in (D), lateral view in (D'). (E, E') pERK/tERK expression after Ab long injections (n = 7) shows widespread downregulation of neuronal activity (magenta) compared to Ab rev controls (n = 7), consistent with c-fos data in (B). Dorsal view in (E), lateral view in (E'). (F) As detected by ISH, the number and intensity of hypothalamic galanin-positive neurons are downregulated following Ab short injection and upregulated following Ab long injection, relative to Ab rev control injections. Representative images from N = 22-24 per condition. (G) Normalized, blinded counts of hypothalamic galaninpositive cell numbers 4-6 hr after Ab short and Ab long injections, relative to Ab rev . Error bars indicate ± SEM. The mean difference effect size and 95% confidence interval is plotted at the bottom. **p<0.01, one-way ANOVA. See also Figure 2-source datas 1 and 2. The online version of this article includes the following source data for figure 2: Source data 1. MAP-Mapping of brain areas that are significantly up-and down-regulated in P-ERK levels in response to Ab short . Source data 2. MAP-Mapping of brain areas that are significantly up-and down-regulated in P-ERK levels in response to Ab long .
We first tested the effects of Ab short injection on mutant behavior. Unlike the wild type controls, neither the adrb2a -/nor the pgrmc1 -/mutants increased waking activity ( Figure 3A-C and E) or suppressed sleep as observed in wild-type controls ( Figure 3A'-C' and G). Injection of Ab short into adrb2a -/animals even significantly increased sleep (+83.7%) instead of reducing it as in wild-type larvae ( Figure 3B' and G). In contrast, Ab short injected into mutants that lack both zebrafish Prp orthologs (prp1 -/-; prp2 -/-) elicited slightly stronger increases in waking activity and significantly large (À45%) reductions in sleep ( Figure 3D,D', F and H). Thus, the wake-promoting activity of Ab short requires intact Adrb2a and Pgrmc1 but not functional Prp1 and Prp2.
Because the size of oligomeric species in our Ab long preparation (20-400 nm) falls into the size range that exhibits high-affinity binding to mammalian Prion Protein (PrP) (20-200 nm) and thereby acts to modulate synapses (Gimbel et al., 2010;Laurén et al., 2009;Nicoll et al., 2013;Um et al., 2012), we tested whether PrP is instead required for Ab long -induced sleep. After injection of Ab long , prp1 -/-;prp2 -/null mutants failed to increase sleep compared to wild-type controls ( Figure 4A-D). The modest reduction of wakefulness induced by Ab long was even reversed in prp1 -/-;prp2 -/mutants, with Ab long instead significantly increasing wakefulness ( Figure 4C). Thus, while Prps are not required for the wake-inducing effects of Ab short , functional Prp1 and Prp2 are essential for sleep induced by Ab long . Moreover, since prp double mutants have exacerbated wakefulness in response to Ab short injections, the sleep-inducing Prp pathway is likely co-activated along with the wake-promoting pathway by Ab to partially dampen the behavioral response of wild-type larvae ( Figure 3D and H).

Mutants lacking Ab targets have altered brain activity in response to Ab consistent with behavioral effects
If Ab short interacts with Adrb2a and Pgrmc1 to drive changes in wakefulness, the increased neuronal activity we observed in wild-type larvae after Ab short injections (Figure 2A) should also be abolished in the adrb2a -/and pgrmc1 -/mutant backgrounds. Consistently, lack of either Adrb2a or Pgrmc1 abolished the neuronal activity-inducing effect of Ab short ( Figure 5A), as detected by in situ hybridization for c-fos. In particular, the neuronal activity observed in the posterior hypothalamus and the dorsal and ventral telencephalon after Ab short into WT controls was not detected after injection into either adrb2a -/or pgrmc1 -/mutants ( Figure 5A). This result is consistent with Ab short failing to induce wakefulness in these mutants. Similarly, although Ab long dampens neuronal activity when injected into wild-type larvae, Ab long injections into the prp1 -/-; prp2 -/double mutants elicited no reduction in c-fos expression ( Figure 5B). Instead, c-fos expression in the telencephalon and hypothalamus was upregulated relative to reverse injected controls ( Figure 5B), consistent with the increased behavioral wakefulness observed in the prp mutants ( Figure 4C). Together, these data are consistent with Ab short acutely upregulating neuronal activity and behavioral wakefulness through interactions with Adrb2a and Pgrmc1, while Ab long interactions with Prp drive increased sleep and a global reduction in neuronal activity.

Pharmacological blockade of the Prp-mGluR5-Fyn kinase signaling cascade prevents sleep induction by Ab long
One of the advantages of using the zebrafish model system is the ability to perturb Ab signalling cascades with small molecule inhibitors added directly to the water Rihel et al., 2010). To further dissect the Ab long -PrP sleep-inducing pathway, we focused on disrupting the putative signalling cascade downstream of Ab-Prp interactions that lead to synaptic changes in neuronal culture (Um et al., 2012; Figure 6A). Consistent with a role for direct Ab long -Prp interactions in sleep, soaking the larvae in Chicago Sky Blue 6B, a small molecule reported to disrupt Ab-PrP interactions (Risse et al., 2015), significantly abolished the sleep-inducing effect of Ab long ( Figure 6B and C, S7A and Figure 6-figure supplement 1A,B). Similarly, pharmacological inhibition of either of the putative Ab-Prp downstream signalling components Metabotropic Glutamate Receptor 5 (mGluR5) or Fyn kinase (Um et al., 2013;Um et al., 2012) significantly blocked the sleep-inducing properties of Ab long ( Figure 6D and E, Figure 6-figure supplement 1C,D). Both the mGluR5 inhibitor MPEP and the Src-kinase inhibitor saracatinib even resulted in significant sleep reductions after Ab short and Ab long affect sleep through distinct neuronal/molecular pathways Although long and short oligomers require different receptors to affect behavior, they may act within the same neuronal circuit signalling cascade. If so, one phenotype should predominate over the other when the two oligomers are co-administered. Alternatively, oligomers may signal through parallel signalling circuits to bi-directionally modulate sleep in an additive manner. To test this, we co-injected Ab short and Ab long in a 1:1 ratio and compared the sleep phenotype to injection of either oligomer alone. As expected, Ab short alone reduced sleep and Ab long alone increased sleep. In contrast, co-injection of both Ab short and Ab long resulted in an intermediate phenotype that is indistinguishable from control injections of Ab rev ( Figure 6F), suggesting that the effects of Ab short and Ab long are additive and likely act through distinct neuronal circuits or signaling cascades to modulate sleep.
Considering this result is aligned with the genetic and pharmacological data, we propose a bidirectional model of Ab sleep regulation in which Ab short and Ab long act through distinct receptors and neuronal pathways to independently modulate behavioral state ( Figure 6G). In this model, the presence of both oligomers leads a balance of signaling through sleep-promoting and sleep-inhibiting pathways, resulting in little or no change in behavior. Tipping the balance from one oligomeric state to the other leads to either the sleep activating or the sleep inhibiting pathway to predominate.

Discussion
Previous studies have suggested that changes in sleep during AD may further accelerate Ab accumulation and neuronal damage, creating a vicious cycle that leads to further neuronal dysregulation and increased sleep-wake cycle abnormalities (Roh et al., 2012). Our results show that, depending on its oligomeric form, Ab can acutely increase or decrease sleep and wake behaviors and brain states through behaviorally relevant molecular targets and independently of neuronal cell death. The exogenous application of Ab oligomers in our experiments limit the conclusions we can draw about endogenous functions of Ab, which in vivo may present with different structure, local concentrations, and kinetics. However, the bi-directional Ab modulation of sleep and wakefulness ( Figure 6G) predicts that alterations to the relative concentrations of different Ab oligomeric forms during healthy aging and AD disease progression will have opposing consequences on sleep and wake behavior.

Distinct molecular pathways for Ab sleep-wake regulation
We found that Ab short -triggered wakefulness required intact Adrb2a and Pgrmc1, while Ab longinduced sleep required functional Prp signalling. These data are consistent with a model in which Ab directly binds to these targets to modulate downstream signaling cascades that ultimately affect neuronal circuits that regulate behavioral state. Our results match well with previous reports demonstrating binding preferences of Ab dimers, trimers, and 56 kDa oligomers for different targets in vitro. For example, Ab dimers, which have been detected in the brains of AD patients ( Vázquez de la Torre et al., 2018), have been shown to directly bind human ADRB2 with high-affinity, causing increased calcium influx and neuronal hyper-activation in rat prefrontal cortical slices (Wang et al., 2010). PGRMC1 can be activated by AD brain extracts (Izzo et al., 2014b) and also has shown preferential binding for 50-75 kDa Ab species in vitro (Izzo et al., 2014a). Both types of shorter Ab species that bind to ADRB2 and PGRMC1 fall within the size ranges that induce Adrb2a-and Pgrmc1-    (Nicoll et al., 2013), the size range of our sleep-inducing Ab long preparation. In neuronal culture and slice preparations, longer Ab-oligomers trigger reduction of synaptic strength (Laurén et al., 2009) via a Prp signaling cascade through mGluR5 and Fyn kinase activation (Um et al., 2012). Similarly, we found that pharmacological blockade of either the direct Ab-Prp interaction (with Chicago Sky Blue 6B), mGluR5 signalling (with MPEP), Fyn kinase activity (with saracatinib), or by mutation of the Prp receptors prevented the widespread reduction of neuronal activity and increase in sleep that was induced by longer Ab-oligomers.
Although triggering neuronal and behavioral changes through distinct molecular pathways, several aspects of Ab's effects on sleep-wake regulation remain to be elucidated. For example, the elimination of either Adrb2a or Pgrmc1 is sufficient to fully prevent Ab short -induced wakefulness. This suggests that Adrb2a and Pgrmc1 function in the same molecular pathway, and signaling by Ab on either alone is insufficient to modulate behavior. Not much is known about how these two receptors interact with one another, but at least one study  has suggested they can directly physically interact. Whether Ab short binds both receptors to affect behavior or whether one receptor is an obligate component of the other's ability to transmit Ab signals is currently unclear. It also remains unclear if the Ab short -Adrb2a/Pgrmc1 wake pathway and the Ab long -Prp sleep pathway occur in the same or different sets of neurons to modulate behavior, as these receptors have widespread expression in zebrafish (Cotto et al., 2005;Málaga-Trillo et al., 2009;Steele et al., 2011;Thisse and Thisse, 2004;Wang et al., 2009), although our co-injection experiment suggests they act on parallel neuronal circuits. Numerous wake-and sleep-promoting neuronal populations that Figure 5. Neuronal activity after exposure to Ab preparations is altered in mutants of Ab binding targets. (A) After Ab short injection into WT larvae (top right), c-fos is detected in many larval brain areas, including the dorsal and ventral telencephalon and posterior hypothalamus (black arrowheads), but not after injection of Ab reverse controls (left). In contrast, Ab short injections into either adrb2a -/-(middle right) or pgrmc1 -/mutants (bottom right) do not induce c-fos expression. The brains in the middle and lower panels were stained longer than the WT (+/+) brains to ensure detection of weaker expression. D = dorsal, p=Posterior. n = blinded counts of brains with expression pattern/total brains. (B) Compared to Ab rev injections, Ab long oligomers induce less c-fos expression in WT larvae (top panels). In contrast, Ab long induced relatively increased c-fos expression in the telencephalon and posterior hypothalamus (black arrows) in the prp1 -/-; prp2 -/double mutants. These Ab rev and Ab long injectedbrains were stained longer to ensure detection of weaker c-fos expression. D = dorsal, p=Posterior.  (Nygaard et al., 2014). Small molecules that block each step in the pathway are indicated. (B) Representative traces of sleep behavior after Ab long versus Ab rev injections in the absence (left) or presence (right) of the Ab-Prion binding disruptor, Chicago Sky Blue 6B (3 nM). Ribbons represent ± SEM. (C) The effect of Ab long relative to Ab rev on normalized sleep Figure 6 continued on next page could serve as neuronal targets for these signalling cascades to drive changes in behavioral state have been uncovered in zebrafish (Barlow and Rihel, 2017). Future experiments will be needed to tease out the neuronal components involved, for example by replacing functional receptors into candidate neurons in otherwise mutant animals and rescuing the responses to Ab, or by mutating receptors selectively in cell types with conditional genetics.

Altered Ab oligomeric ratio-implications for sleep in health and AD
Our model investigates alterations in sleep/wake behavior due to acute changes in exogenously delivered Ab levels. Thus, it is possible that the sleep/wake effects observed in our study may be different from those observed when Ab fluctuates over 24 hr or when it is chronically accumulating as in AD. However, our model predicts that alterations to the ratio of Ab oligomeric forms present in the brain could have differential effects on sleep-wake regulation, as the balance between sleepand wake-promoting Ab signals is tilted to favour one pathway over the other ( Figure 6G).
Given the natural daily increase in Ab secretion during wakefulness and increased levels of Ab clearance during sleep (Xie et al., 2013), changes in extracellular Ab levels could sculpt behavior over the normal 24 hr circadian cycle. As Ab burden is acutely increased by sleep deprivation (Shokri-Kojori et al., 2018), perturbations to the normal sleep-wake cycle may feedback on behavior through altered Ab signaling. Other phenomena have been reported to alter Ab generation and fibrilization over short time-scales. For example, temperature changes in the physiological range (35-42˚C) have been reported to significantly affect Ab oligomerization (Ghavami et al., 2013), suggesting that either the natural daily fluctuation in body temperature (in humans, up to 2˚C) or the induction of a fever can promote changes in amyloidogenic Ab generation (Szaruga et al., 2017). In addition, Ab can act as an antimicrobial peptide (Kumar et al., 2016;Soscia et al., 2010), and microbial infection can trigger Ab fibrilization (Eimer et al., 2018). Considering that infection and fever are also potent drivers of sleep (Imeri and Opp, 2009), the sleep-inducing Ab-Prp signaling pathway we identified here could mediate recovery sleep during illness-a hypothesis for future investigation.
On longer timescales, the amount and type of Ab oligomeric species (including dimers, crosslinked dimers, trimers, and 56 kDa oligomers) found in healthy brains change across the human life cycle (Lesné et al., 2013) and are heterogeneous and elevated in AD patients (Izzo et al., 2014a;Kostylev et al., 2015). Although the precise makeup of Ab species present in healthy and AD brains has remained difficult to quantify (Benilova et al., 2012), some studies have indicated that short (dimers, trimers) Ab oligomers are more enriched in the early, mild cognitive impairment (MCI) stages of AD, while longer oligomers predominate in the CSF at later clinical stages of AD (De et al., 2019). Similarly, 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 (Videnovic et al., 2014). One possibility consistent with our data is that sleep symptoms of both normal aging and AD may reflect changes in Ab burden that lead to an altered balance in sleep-and wake-promoting signaling Figure 6 continued during the first day in the in the absence or presence of 3 nM Chicago Sky Blue 6B. The data is pooled from n = 2 independent experiments **p 0.01, one-way ANOVA. (D) Representative traces of sleep behavior after Ab long versus Ab rev injections in the presence of mGluR5 inhibitor MPEP (5 uM,left) and Fyn Kinase inhibitor saracatinib (300 nM, right). Ribbons represent ± SEM. (E) The effect of Ab long relative to Ab rev on normalized sleep during the first day in the absence or presence of 5 uM MPEP (left) and 300 nM saracatinib (right). Each dot represents a single larva normalized to the mean Ab rev . Data is pooled from two independent experiments. **p 0.01, ****p 10-5 one-way ANOVA. (F) The effect of a 1:1 mixture of Ab long to Ab short relative to single injections of Ab rev , Ab short , and Ab long on normalized sleep during the first day. The data is pooled from n = 4 independent experiments. (G) A bi-directional model for sleep/wake regulation by Ab. In wild-type animals (centre), injection of Ab short species signal through Adrb2a/Pgrmc1 to drive wakefulness while Ab long oligomers signal via Prp to induce sleep. In mutants that lack Prp (left), only Ab short species (as shown by the overlapping distributions) remain to inhibit sleep with no residual Ab long oligomers to stimulate the sleep-inducing pathway to counteract wake-inducing signals. Thus prp1 -/-; prp2 -/mutants have enhanced wakefulness in response to Ab. Conversely, mutants that lack Adrb2a/Pgrmc1 (right), retain only the sleeppromoting Ab pathway and fail to increase wakefulness in response to Ab short . See also Figure 6- cascades. These signaling molecules might therefore be potential therapeutic targets for treating disrupted sleep early in AD progression, which may in turn slow disease progression. Table (Appendix 1-key resources table) for details of reagents.

Zebrafish strains and husbandry
Zebrafish (Danio rerio) were raised under standard conditions at 28˚C in a 14:10 light:dark cycle and all zebrafish experiments and husbandry followed standard protocols of the UCL Zebrafish Facility. AB, TL and ABxTup wild-type strains were used in this study. prp1 (ua5003/ua5003), prp2 (ua5001/ 5001), adrb2a (u511/u511) and pgrmc1 (u512/u512) mutants were outcrossed multiple times, and pgrmc1 F2 and adrb2a F2 and later generations were used for behavior. Ethical approval for zebrafish experiments was obtained from the Home Office UK under the Animal Scientific Procedures Act 1986 with Project licence numbers 70/7612 and PA8D4D0E5 to JR.

Ab preparations
HFIP treated Ab42 peptide (JPT Peptide Technologies) and Ab42-1 reversed peptide (Sigma) were dissolved in DMSO, vortexed occasionally for 12 min at room temperature and sonicated for 5 min to obtain 100 mM solution. The stock solutions were aliquoted as 5 ml in individual tubes and are kept are À80˚C. 1 ml of the 100 mM stock was diluted in (Phosphate buffered saline) PBS to yield 10 mM solutions which were incubated at 4˚C, 25˚C or 37˚C for 24 hr (Kusumoto et al., 1998;Orbán et al., 2010;Whitcomb et al., 2015). 1 hr before injecting, this stock was diluted to 10 nM using 1:10 serial dilutions in PBS and kept at the respective temperature (4˚C, 25˚C, 37˚C) until injecting.

Transmission Electron Microscopy (TEM)
1 ml of (4˚C, 25˚C or 37˚C incubated) 10 mM Ab solution was loaded onto formvar/carbon coated 300 mesh grids from Agar Scientific. The grid was washed twice in 20 mM phosphate buffer for 10 s and negatively stained in 2% aqueous uranyl acetate for 30 s. After drying for 2-3 days, samples were imaged using a Phillips TEM. At least five micrographs were used for each condition to blindly measure the length of the Ab42 oligomeric structures using at least 30 measurements/condition. Using FIJI, 30-50 measurements were taken for each condition by drawing a free-hand line on the fibril, which was then scaled using the scale bar.

Heart injections
Injections were carried out blindly with a Pneumatic PicoPump (WPI) and glass capillary needles (Science Products Gmbh) prepared with a Micropipette Puller (Shutter Instruments). Five dpf larvae were anesthetized using 4% Tricaine (42 mg/L, Sigma) 30 min before injections. Larvae were immobilized in 1.8% low melting point agarose (ThermoFischer) in fish water on their sides on a slide. 1 nL of Ab (10 nM starting concentration) was injected into the heart chamber of the fish along with a high molecular weight fluorescent dye (2000 kDa dextran-conjugated FITC (3 mg/ml, Sigma). We estimate that 1 nl of a 10 nM Ab injection into a~3.01 (±0.16) x10 8 mm 3 (285-317 nL) 5dpf larva yields a final monomeric brain/CSF concentration of~28-32 pM. The success of the injection was checked under a standard fluorescent scope by the presence of fluorescence in the heart of the animal. Larvae were transferred to fresh fish water for 20 min to recover from Tricaine and transferred to sleep/wake behavior box. For drug blocking experiments, zebrafish larvae were soaked into 3 nM Chicago Sky Blue 6B (Sigma), 5 mM MPEP (Cambridge Biosciences), or 300 nM Saracatinib (Generon) 1 day before the injections (from 4 to 5 dpf). Fluorescently tagged HiLyte Fluor 647-labeled Ab42 (Eurogentech LTD) was injected at 10 mM.

Behavioral experiments
Larval zebrafish behavioral experiments were performed and analysed as described (Rihel et al., 2010). Briefly, 5 dpf larvae were transferred to 96 square-well plates and continuously illuminated with IR and white lights from 9 am to 11 pm in a Zebrabox (Viewpoint life sciences) for 24-48 hr. The movement of each larva was measured and duration of movement was recorded with an integration time of 10 s. Data were processed according to Rihel et al., 2010, and statistical tests were performed using MATLAB. Mutant larval zebrafish experiments were performed on siblings from heterozygous in-crosses, differing only in the mutation of the specific gene and genotyped at the end of the experiment.

Dark pulse experiments
Larvae were placed in the behavior tracking boxes, and two or three dark pulses for 10 min with a 2-4 hr interval were introduced in four independent experiments. For data analysis, only the dark pulses after the acclimatization period in the late afternoon were combined for each genotype.
In situ hybridization RNA in situ hybridization (ISH) to detect c-fos and galanin was performed as described (Thisse and Thisse, 2008). Zebrafish larvae were fixed in 4% paraformaldehyde in PBS at 4˚C overnight. A template for in vitro transcription was generated by PCR using a reverse primer that contains a T7 promoter sequence 5'-TAATACGACTCACTATAGGG-3' from cDNA. A digoxygenin (DIG)-labelled antisense RNA probe was synthesized using the DIG labelling kit (Roche) and T7 RNA polymerase according to the manufacturer's recommendations. The probe was detected with anti-DIG-AP antibody (1:2,000, Roche) and nitro-blue tetrazolium chloride (NBT)/5-bromo-4-chloro-3 0 -indolyphosphate (BCIP) substrate (Roche) according to published protocols. To detect the differences in expression between the mutant backgrounds and WTs after Ab short and Ab long injection, larvae were incubated and washed in the same tubes throughout the ISH procedure to avoid staining artefacts. To do this, larvae from different genotypes were marked by cutting the tail to allow identification after ISH. The brains were exposed by dissection, keeping the brain and spinal cord intact. Embryos were stored in 60% glycerol/PBS for imaging.

Baseline c-fos ISH
Zebrafish larval siblings were kept in 14:10 day/night normal or reverse-cycle incubators. 50 larvae were collected at each time point (ZT1, ZT13, and ZT19) and fixed in 4% paraformaldehyde in PBS overnight. RNA in situ hybridization (ISH) to detect c-fos was performed as described (Thisse and Thisse, 2008).

KASP genotyping
For rapid genotyping of mutant zebrafish harbouring the adrb2aD8 and pgrmc1D16 alleles, a mutant allele-specific forward primer, a wild-type allele-specific forward primer and a common reverse primer were used (LGC Genomics). The primer sequences were targeted against the following: where [x/-] indicates the indel difference in [WT/mutant]. PCR amplification was performed using KASP Master mix (LGC Genomics) according to the manufacturer's instructions. Fluorescence was read on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad) and the allelic discrimination plot generated using Bio-Rad CFX Manager Software.

Time lapse confocal microscopy
Three Casper larvae at five dpf were mounted dorsally on a slide in 1.5% agarose. 10 nl of Ab42-Hi488 was intra-cardiac injected to embryos, control fish were untreated. Fish were imaged for 6 min taking 2-mm-thick stacks through the whole brain using a confocal microscope (Leica SP8). Each fish was imaged every 20 min for 8 hr.
Larvae were mounted in 1.5% low melt agarose and imaged with a custom two-photon microscope (Bruker; Prairie View software) with a 203 water immersion objective (Olympus).
Images were noise filtered using a custom MATLAB (The MathWorks) scripts and registered into Z-Brain using the Computational Morphometry Toolkit (http://www.nitrc.org/projects/cmtk/) with the command string: -a -w -r 0102 l af -X 52 C 8 G 80 R 3 -A '-accuracy 0.4 -auto-multi-levels 4' -W '-accuracy 1.6' -T 4. Registered images were prepared using a custom MATLAB/MIJ (http:// bigwww.epfl.ch/sage/soft/mij/) script to downsize, blur, and adjust the maximum brightness of each stack to the top 0.1% of pixel intensities to preserve dynamic range. Activity maps were generated using MATLAB scripts (Randlett et al., 2015).

Crispr/Cas9 mutant generation
The CRISPR design tool CHOPCHOP (http://chopchop.cbu.uib.no) was used to identify a target region in zebrafish adrb2a and pgrmc1 (Labun et al., 2019). The gene-specific oligomers were ordered from Thermofisher including the 5' and 3' tags: For adrb2a: 5'ATTTAGGTGACACTATAGTTTGGACAGATAAGATCTTGTTTTAGAGCTAGAAATAGCAAG-3' For pgrmc1: 5'ATTTAGGTGACACTATATGCAGACTATGGCCCGGTTGGTTTTAGAGCTAGAAATAGCAAG-3' Constant oligomer: 5'AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGC TATTTCT AGCTCTAAAAC-3' The constant oligomer and the gene-specific oligomer were annealed on a PCR machine and filled in using T4 DNA polymerase (NEB) (Gagnon et al., 2014). The template was cleaned up using a PCR clean-up column (Qiaquick) and the product was verified on a 2% agarose gel. The sgRNA was transcribed from this DNA template using Ambion MEGAscript SP6 kit (Gagnon et al., 2014). Cas9 mRNA and the purified sgRNA were co-injected into one-cell stage embryos at a concentration of 200 ng and 100 ng per embryo, respectively.
Caspase-3 staining 5dpf larvae that were injected with Ab oligomers or soaked in Camptothecin (1 mM; Sigma Aldrich) were fixed 5 hr after injection/drug treatment in 4% PFA and kept overnight at 4˚C. Brains were dissected and dehydrated the next day and were washed three times in PDT buffer (0.3 Triton-X in PBST with 1% DMSO) and incubated with Caspase-3 antibody (1:500; BD Biosciences) at 4˚C. The brains were incubated with Alexa Fluor 568 goat anti-rabbit antibody (1:200; Invitrogen) next day at 4˚C overnight and imaged using a confocal microscope.

Statistical analyses
For data analyses, we used a Gardner-Altman estimation plot, which visualizes the effect size and displays an experimental dataset's complete statistical information (Ho et al., 2019). The bootstrapped 95% confidence interval (CI) was calculated from 10,000 bootstrapped resamples (Ho et al., 2019).
Details of statistics used in each panel are also described in the figure legend. For multiple comparisons, data was first tested for normality by the Kogloromov-Smirnov (KS) statistic and extreme outliers were discarded by Grubb's test (p 0.01). Those that violated normality were analysed with the non-parametric Kruskal-Wallis test, with either Tukey-Kramer or Bonferonni post hoc testing; otherwise, data was analysed with one-way ANOVA followed by Tukey's post-hoc testing. For dose response curves, a two-way ANOVA was performed to test the interaction effects between preparation type and dose. For return to baseline statistics, paired t-tests were performed. Survival curves were analysed with Kaplan-Meier log rank test. All statistical tests and graphs were generated in MATLAB (R2015a) loaded with the Statistical Toolbox. Injection and tracking experiments were performed blinded.