Genetic and chemical disruption of amyloid precursor protein processing impairs zebrafish sleep maintenance

Summary Amyloid precursor protein (APP) is a brain-rich, single pass transmembrane protein that is proteolytically processed into multiple products, including amyloid-beta (Aβ), a major driver of Alzheimer disease (AD). Although both overexpression of APP and exogenously delivered Aβ lead to changes in sleep, whether APP processing plays an endogenous role in regulating sleep is unknown. Here, we demonstrate that APP processing into Aβ40 and Aβ42 is conserved in zebrafish and then describe sleep/wake phenotypes in loss-of-function appa and appb mutants. Larvae with mutations in appa had reduced waking activity, whereas larvae that lacked appb had shortened sleep bout durations at night. Treatment with the γ-secretase inhibitor DAPT also shortened night sleep bouts, whereas the BACE-1 inhibitor lanabecestat lengthened sleep bouts. Intraventricular injection of P3 also shortened night sleep bouts, suggesting that the proper balance of Appb proteolytic processing is required for normal sleep maintenance in zebrafish.


INTRODUCTION
Sleep disturbances are prevalent in neuropsychiatric and neurodegenerative disorders such as Alzheimer disease (AD).][7][8][9][10][11] Recently, human studies have found that harboring a high genetic risk for AD correlates with sleep changes, such as increased sleep rebound following sleep loss, even in young adults. 12These observations suggest that there may be underlying early biological processes important for sleep regulation that are governed by AD susceptibility genes and contribute to disease progression.
One of the risk genes for AD encodes the amyloid precursor protein (APP), a transmembrane protein that is proteolytically processed into multiple smaller fragments, including AD-associated fragments such as Ab40 and Ab42.Mutations in App are associated with both early-and late-onset AD 13,14 including some alleles that are dominant and fully penetrant for early-onset AD. 15 Duplications of the App gene, including those associated with trisomy 21 (Down syndrome), also confer increased AD risk. 168][19] Additionally, APP is processed into fragments other than Ab, including P3, sAPPa, sAPPb, and the APP intracellular domain (AICD), but the endogenous roles of these peptides are poorly understood.Whether APP and its proteolytic fragments play a role in sleep regulation is unknown.
To explore the role of APP and its derivatives in sleep regulation, we introduced loss-of-function mutations into both zebrafish orthologs of the human APP gene, appa and appb, and evaluated their sleep-wake behaviors using a high-throughput behavioral assay. 202][23][24][25] In addition, our previous work had revealed that depending on its structural configuration, Ab can either increase or decrease sleep duration in larval fish by signaling through distinct receptors, 26 suggesting that App-derived products may act as sleep signals in zebrafish.In Drosophila, elevated Black, dark gray, and light gray boxes indicate strictly, highly, and moderately conserved amino acid residues, respectively.(C) As detected by multiplexed hybridization chain reaction (HCR), appa (green) and appb (red) are both expressed widely but in non-overlapping regions of the 5dpf larval brain, including the cerebellum and nuclei in the hindbrain (C, cerebellum; H, hindbrain).Shown is a representative image from a single brain taken at production of Ab leads to sleep fragmentation, memory loss, and neuronal hyperexcitability, [27][28][29] even though flies do not make Ab natively.Thus, it is very important to disentangle the impact of toxic Ab causing neuronal dysfunction, which can alter sleep even in species that do not produce Ab, and the role of endogenous APP signaling events.By combining genetic analysis of app mutants with pharmacological interventions that block g-secretaseand BACE-1-dependent cleavage of APP, we provide evidence that proteolytic processing of appb is required for maintaining sleep at night in zebrafish larvae.

Characterizing zebrafish appa and appb and generating mutants
There are two app genes in zebrafish, appa and appb, that raise the question whether they have redundant functions.We first examined their relationship to the human APP isoforms, because gene duplications often take on isoform-or tissue-specific roles. 30Zebrafish Appa is more similar to the Kunitz-type protease inhibitor (KPI)-domain-containing APP751 and 770 isoforms, whereas the Appb protein lacks the KPI domain and is more similar to the APP695 isoform 31,32 (Figures 1A and 1B).Both Appa and Appb have respectively an 80% and 71% conserved identity in the Ab42 region compared with human APP, with conserved proteolytic cleavage sites for processing by a-, b-, and g-secretases (Figure 1B).Both appa and appb genes are abundantly expressed in the zebrafish brain, although with non-overlapping expression patterns (Figure 1C).For example, while appa is expressed strongly in the cerebellum caudal lobe, olfactory bulb, and torus longitudinalis, appb is more strongly expressed in nuclei in the hindbrain (Figures 1C, S1, and S2).Examination of the brain expression pattern and levels of appa and appb during the day (4 h post lights on) and night (4 h post lights off) failed to detect any time-of-day differences, either globally or region specifically (Figures S2B and S3D).appb is also highly expressed in the very early stages of zebrafish development, indicating that it is maternally deposited (Figure S3A).Together, the gene expression patterns and structural homology differences of zebrafish Appa and Appb are consistent, with these proteins possibly having both isoform-and brain-tissue-specific functions.
We next generated zebrafish mutants with deletions in the appa and appb genes.To isolate mutations in appa, we used CRISPR/Cas9. 33he CRISPR design tool CHOPCHOP (http://chopchop.cbu.uib.no) was used to identify candidate gRNAs to target the conserved Ab region (amino acid residues 25-35) of appa, 34 which was then coinjected with Cas9 mRNA into zebrafish eggs at the one-cell stage.Injected animals (F0s) that harbored frameshift mutations were then identified by Illumina sequencing and outcrossed to wild-type animals to generate mutant families (see STAR Methods).One family was isolated that harbors a 5 base pair frameshift deletion (appa D5 ) that leads to an early stop codon within the Ab domain.The appa D5 allele therefore lacks the conserved residues 26-42 of the Ab and the entire intracellular C-terminus of Appa (Figures 1B and 1D).
To generate a loss-of-function mutation in the appb gene, two transcription activator-like effector nuclease (TALEN) arms targeting a conserved region within the first exon of the zebrafish appb gene were designed using the Zifit software (http://zifit.partners.org/ZiFiT/). 35hese TALEN arms were each fused to one-half of a Fok1 heterodimer to generate mutagenic double-strand breaks within the first exon of appb (Figure 1D).F0 fish that had been coinjected at the one-cell stage with mRNA encoding the two TALEN arms were Illumina sequenced to identify a founder that contains an appb allele (appb D14+4 ) with a 14 base pair deletion and a 4 base pair insertion that generates a frameshift followed by an early stop codon.This founder was used to generate a stable heterozygous family (herein called appb À/+ ) for subsequent behavioral analysis (Figure 1D).
To confirm that these appa and appb alleles represent loss-of-function mutations, we performed western blot analysis on brain homogenates from adult double homozygous appa D5/D5 ; appb À/À mutants using an anti-APP antibody (22C11) that recognizes both zebrafish Appa and Appb.APP protein was detectable as a $100 kD band in wild type (WT) but not in appa D5/D5 ; appb À/À double mutants, confirming that neither Appa nor Appb are made (Figures 1E and S3C).We also observed by RT-qPCR that mutant appb transcripts are only 25% of WT levels, consistent with non-sense-mediated decay that is often observed in transcripts harboring an early termination codon (Figure S3B). 36lthough zebrafish have the machinery to process App into Ab fragments, whether Ab40/42 are actually generated in zebrafish has not been formally demonstrated.To measure endogenous Ab levels, we used a highly sensitive electrochemiluminescence-based ELISA kit and detected in adult brains $0.02 pg Ab42/mg total protein and $0.2 pg Ab40/mg total protein, yielding an Ab42:Ab40 ratio of 1:15-1:20 (Figure 1F), similar to the range observed in various mammalian species. 37Ab40 and Ab42 were completely undetectable in brains following overnight exposure of WT adults to the g-secretase inhibitor DAPT, confirming the efficacy of this drug to block Ab40/42 production in zebrafish (Figure 1F).Moreover, levels of both Ab40 and Ab42 in either appa D5/D5 or appb À/À mutants were significantly lower than in WT animals, Figure 1.Continued one z-plane (z140/420) (dorsal view, above) and through the midline of the same brain (lateral view, below) from an experiment with a total of n = 22 fish.A, anterior; D, dorsal; L, left.(D) CRISPR/Cas9 targeting of zebrafish appa resulted in a 5-bp deletion.The target guide RNA (gRNA) sequence is shown in bold, and the obligatory PAM sequence (AGG) is in red.The predicted translation of appaD5 leads to a premature stop codon within the Ab region (Appa amino acid 665).TALEN targeting of zebrafish appb resulted in a 14-bp deletion (dashes) and 4-bp insertion (red).The left and right sites targeted by the TALENs are highlighted, and the spacer sequence, where cleavage occurs, is bolded.The predicted translation of appbD14+4 leads to a frameshift and a premature stop codon.(E) Western blot analysis of APP in brain homogenates from wild-type (WT) and appa D5/D5 ; appb À/À double mutants.(F) Elisa detection of Ab42 (left) and Ab40 (right) levels in adult brain homogenates from WT controls, WT animals treated with the g-secretase inhibitor DAPT for 24 h, appa D5/D5 and appb À/À mutants quantified as picograms (pg) of Ab per mg total protein extracted.Each dot is an independent biological replicate, and error bars represent the mean G SEM.The bottom panel plots the ratio of Ab40 to Ab42 in WT zebrafish adult brain homogenates, n = 5. *p % 0.05, **p % 0.01, Dunnett's test.
suggesting that Ab is made from both Appa and Appb (Figure 1F).The reduction of Ab levels in appa D5/D5 mutants (À89% and À90% for Ab40 and Ab42, respectively) was stronger than in appb À/À mutants, which may suggest that more Ab is generated from Appa than from Appb; however, the Ab40/42 epitopes detected by this kit are better matched in the Appa sequence than Appb, so the relative difference in detected Ab levels may be due to slight differences in antibody affinity/detection between Appa and Appb.Nevertheless, these results are consistent with Ab being produced from both Appa and Appb in zebrafish and demonstrate that both mutants have disruptions in Ab production.
appa D5/D5 and appb D14+4 (appb À/À ) have distinct sleep-wake profiles Zebrafish appa D5/D5 mutants do not have any obvious morphological abnormalities during development, have normal survival rates to adulthood, and are generally healthy and fertile.To examine whether appa D5/D5 mutant larvae have sleep or wake phenotypes, we used automated video monitoring to track larvae from in-crosses of appa D5/+ parents over several days on a 14h:10h light:dark cycle (Figure 2).Sleep in zebrafish is defined as a period of inactivity lasting longer than 1 min, as quiescent periods lasting at least this long are associated with an increased arousal threshold and other features of behavioral sleep, including circadian and homeostatic regulation. 38Zebrafish sleep is organized into bouts, with the sleep bout length describing the duration of consecutive, uninterrupted minutes of sleep.We also measured the vigor of their movements during the active bouts, quantified as the average waking activity.Sleep and waking activity are therefore not merely mirror-images and can be selectively and differentially modulated by drugs 20 or mutation. 39Assessing these parameters across the day and night for appa D5/D5 mutants and their WT siblings uncovered subtle differences in mutant behavior.The appa D5/D5 mutant had a reduction of 9.0% (lower bound, À15.0%; upper bound, À3%, 95% confidence interval [CI]) in waking activity during the day compared with appa +/+ siblings (Figures 2A and 2B).At night, appa D5/D5 mutants also had slightly lower waking activity levels (À4.7%, [À10.0;À0.5, 95%CI]) (Figure 2C).In contrast, neither the total sleep (Figures 2E and 2F) nor the structure of sleep, such as the number and duration of sleep bouts (Figure S4), were statistically different across genotypes during either the night or the day.
Together, these data show Appa is not required for normal sleep states in zebrafish larvae, although it influences locomotor drive during the waking day.We also did not observe any obvious developmental delays or morphological abnormalities in appb À/À mutant larvae or adults and therefore assessed whether Appb might play a non-redundant role in larval sleep regulation.Similar to appa D5/D5 mutants, appb À/À larvae had a reduction in day-time waking activity of 13.0% [À20.2;À5.6, 95%CI] relative to appb +/+ siblings (Figures 3A and 3B).However, unlike appa D5/D5 animals, appb À/À larvae had an increase in activity of 8.2% [3.3; 13.0, 95%CI] specifically at night (Figure 3C).We also observed that although appb À/À larvae had unaffected sleep during the day (Figures 3D and 3E), they had a 7.9% (À14.0;À7.0, 95%CI) reduction in sleep at night (Figure 3F), which corresponds to $30 min less sleep per night.Thus, both Appa and Appb regulate daytime waking activity levels but have non-overlapping roles in regulating nighttime activity and sleep, with only appb mutants exhibiting sleep phenotypes.
To further investigate the nature of the decreased night sleep in appb À/À mutants, we compared the sleep architecture of these mutants with their WT and heterozygous siblings.We specifically examined whether the change in total sleep was due to alterations in the number of sleep bouts (i.e., how often sleep is initiated) or in the average lengths of sleep bouts (i.e., once sleep is initiated, how long it is maintained).The average sleep bout length was shorter by 12.1% (À21.2;À3.2, 95%CI) in appb À/À mutants during the day and by 14.9% (À23.9;À5.9, 95% CI) at night compared with their WT siblings (Figures 3G and 3H).The number of sleep bouts during either the day or night were not significantly different between appb À/À mutants and WT animals (Figures 3I and 3J).These results show that the appb À/À mutants initiate sleep normally but cannot sustain continuous sleep as long as WT, indicating a defect in sleep maintenance.

g-and b-secretase inhibitors modulate sleep maintenance in an appb-dependent manner
Because App undergoes complex proteolytic processing, we decided to test whether drugs that inhibit App cleavage also modulate sleep in an Appb-dependent manner.We first tested whether the g-secretase inhibitor DAPT, which effectively prevented Ab production in zebrafish after 24 h (Figure 1F), alters sleep and waking activity in ether WT or appb homozygous mutants (Figures 4A-4D).Unlike either appa or appb mutants, which had lower waking activity during the day, DAPT significantly increased WT waking activity, with no effect at night (Figures 4A,  S5A, and S5E).In contrast, DAPT significantly reduced daytime waking activity in appb À/À larvae (Figures 4B and S5A).Together, these results indicate that g-secretase-dependent cleavage products of Appb increase daytime waking, whereas other g-secretase targets have a net effect of reducing wake activity.
DAPT also significantly reduced total nighttime sleep of WT larvae by 7.29% (À13.38;À1.06, 95%CI) and trended to lowered daytime sleep by 18.46% (À38.58;+0.87, %95CI) (Figures 4C-4E and S5B).This overall nighttime sleep reduction was due to a shortening in the average length of sleep bouts (À19.6% [-33.83;À6.55, %95CI], an effect size similar to appb À/À alone), even though the number of sleep bouts was slightly increased by DAPT (Figures 4F and S5F).However, when DAPT was tested on appb À/À mutants, there was no effect on total sleep or sleep bout lengths (Figures 4D-4F).This significant, non-additive effect (genotype 3 drug interaction, p < 0.05 for night sleep and p < 0.01 for sleep bout length at night, two-way ANOVA) cannot be explained by a flooring effect, as wake-promoting drugs can reduce zebrafish larval sleep much more than observed in appb À/À mutants alone. 20,40Instead, this demonstrates that DAPT requires the presence of Appb to influence sleep length at night and further suggests that the short sleeping phenotype of appb mutants is due to the loss of Appb-derived g-secretase cleavage products, such as Ab, P3, or AICD.
To further dissect which Appb cleavage products modulate sleep and wakefulness, we next tested the effects of the BACE-1 inhibitor, lanabecestat, on WT and appb À/À mutant behavior.Like g-secretase inhibitors, blocking b-secretase will prevent Ab production; however, other products blocked by g-secretase inhibitors will remain unaffected, such as AICD, or even enhanced, such as P3, allowing us to test which Appb-derived products are responsible for shortening sleep (Table 1).Unlike DAPT but similar to appa and appb mutants, lanabecestat slightly reduced daytime waking activity of WT larvae (Figures 5A and S6A) and similar to appa mutants, reduced daytime waking activity at night (Figures S6A and S6E).However, when appb mutants were exposed to lanabecestat, there was no longer an effect on daytime waking levels, whereas nighttime waking activity was even slightly increased (Figures 5B and S6E).Thus, Appb must be present for b-secretase inhibition to exert an effect on daytime waking activity.

P3 brain injections reduce the length of sleep bouts at night
Blocking either b-secretase or g-secretase resulted in Appb-dependent, but opposing, sleep phenotypes at night.Because band g-secretase inhibition differentially alter the formation of App cleavage products, of which all are lost in appb À/À mutants, we hypothesized that the nighttime sleep phenotypes might be explained by fragments that are blocked by g-secretase inhibitors but enhanced or unchanged by b-secretase inhibition (Table 1).We therefore focused on P3, a partial Ab fragment (Ab 17-42 ) that is boosted by BACE-1 inhibitors and absent when g-secretase is blocked or appb is mutated.Injection of P3 into the brain ventricle of WT larvae had no effect on waking activity in either the day or the night (Figures 6A-6C) but caused a significant 10.03% (À15.05;À5.17, %95CI) decrease in night sleep relative to vehicle-injected controls (Figures 6D and 6F).As in other manipulations of App processing, this nighttime reduction in sleep was caused by shortened sleep bout lengths (À16.50% [À26.60;À6.30, %95 CI]) rather than a change in the number of sleep bouts (Figures 6G-6J).Although this demonstrates that App cleavage products such as P3 can have acute effects on sleep maintenance at night, because this effect is in the same direction as g-secretase inhibition, which blocks the formation of P3, and in the opposite direction from BACE-1 inhibition, which enhances P3 production, alterations to P3 levels alone cannot explain the sleep phenotypes seen in appb mutants or drug manipulations that affect APP processing.

Comparison to other App loss-of-function studies
We found that both appa D5/D5 and appb À/À zebrafish had reduced locomotor activity during the day, but only appb À/À animals had a reduction in sleep maintenance at night.5][46][47][48][49][50][51][52][53] However, the lack of major morphological and developmental phenotypes in both appa and appb mutants stands in stark contrast to several reports that investigated zebrafish App function with morpholino knockdowns.For example, morpholino knockdown of zebrafish Appb has variously been reported to affect convergent-extension during gastrulation, 54 axon outgrowth of spinal motor neurons, 55,56 hindbrain neurogenesis, 57 or cerebrovascular development. 580][61] Because appb À/À larvae from appb À/À mothers also are morphologically normal (Figures 4 and 5), the difference in phenotype compared with morpholino studies cannot be explained by differential effects on maternally deposited transcripts of appb (Figure S3).Although sleep has not been investigated in zebrafish app mutants before, two other studies have examined the role of appb on larval locomotor activity, coming to different conclusions than what we draw here.One appb morpholino study found that knockdown resulted in hyperlocomotion between 28 and 45 hpf, 55 whereas we find both appa D5/D5 and appb À/À larvae at older stages (4-7dpf) are less active than their WT siblings.These contrasting results are likely due to differences in locomotor behavior regulation at different stages of development, methodological differences (morpholino vs. knockout), or both.Another study of appb À/À larvae tracked locomotor behavior for 60 min at a similar developmental age to our study (6 dpf) but found no differences in locomotion when comparing appb mutants with non-sibling WT animals. 59Given that we found appa and appb mutants are only 8%-10% less active than sibling-matched WT animals during the day, the short, 60-min observation window may not have been sufficient to capture this difference; alternatively, the time of day of observation might affect the ability to detect locomotor phenotypes (e.g., see Figures 2A and 3A).Indeed, we found appb mutants were significantly more, not less, active at night (Figure 3C).
Overall, our sleep and wake analysis of appa and appb mutants are broadly consistent with other rodent and zebrafish studies and expands the known phenotypes associated with App loss-of-function mutations.These results also demonstrate that Appa and Appb play at least partially non-redundant roles in the regulation of sleep and locomotor activity in larval zebrafish, which may explain why the zebrafish phenotypes of single mutants are somewhat milder than that reported for rodent App knockouts.Differences in the phenotypes between appa and appb mutants may reflect their differential expression patterns in the brain (Figures 1C, S1, and S2A), overall expression levels, or possibly different sensitivities or exposure to App processing enzymes, yielding different ratios of cleavage products from Appa versus Appb.For example, we found that Appa may be a larger source of Ab in larvae, as appa D5/D5 mutants had lower detectable levels of Ab compared with appb À/À mutants (Figure 1F).

Sleep maintenance and appb proteolytic cleavage
We observed no sleep phenotypes in appa D5/D5 mutants, but appb À/À mutants had reduced nighttime sleep due to an inability to maintain longer sleep bout durations.Inhibition of g-secretase also shortened the length of nighttime sleep bouts in an Appb-dependent manner, suggesting that g-secretase-dependent cleavage products of Appb such as Ab, P3, or AICD may act as a signal for maintaining nighttime sleep that is lost in appb À/À mutants.Previous work has demonstrated that exogenously delivered Ab does have both sleep-promoting and -inhibiting properties when injected into zebrafish larvae, 26 engaging many areas of the brain including the sleep-promoting galanin-positive neurons of the preoptic area and hypothalamus. 62However, that work found that longer Ab oligomers promoted sleep predominantly by increasing sleep initiation rather than altering sleep bout maintenance, whereas shorter forms of Ab promoted wakefulness. 26Moreover, inhibition of BACE-1, which also prevents the formation of Ab, instead led to increased sleep maintenance in an Appb-dependent manner.This suggests that loss of Appb-cleavage products other than Ab is responsible for the short-sleeping phenotype of appb À/À mutants.
One candidate App cleavage product that we tested was P3, as production of this peptide requires g-secretase but is boosted by inhibition of BACE-1.Injection of P3 also reduced nighttime sleep by affecting sleep maintenance (Figure 6).Although this result underscores the potential of App-cleavage products to act as acute signals that regulate sleep, this contradicts the straightforward hypothesis that loss of P3 signaling is the cause of shortened nighttime sleep in appb À/À mutants.However, because injection experiments do not recapitulate the precise timing or localization of P3 release, its local concentration, or (possibly) its structure, a role for P3, alone or in complex combinations with other App products like Ab or AICD, in the observed appb mutant phenotypes cannot be completely ruled out.Future studies could examine whether mutations in components of g-secretase, such as Presenilin-1 or Presenilin-2, or b-secretase, such as BACE-1, also ][43] have reduced sleep bout lengths at night.Zebrafish bace1 À/À mutants have been reported to have hypomyelination in the peripheral nervous system, 25 but to date no sleep phenotypes have been described.As we performed here for appb and DAPT/lanabecestat, the examination of sleep phenotypes in appb À/À ; bace1 À/À or presenilin1/2 À/À double mutants could be used to tease out which phenotypes are due to the specific cleavage of Appb.Another possibility by which Appb could contribute to sleep regulation is raised by the recent observation that in zebrafish both Appa and Appb are colocalized to cilia and cells lining the ventricles at 30 h postfertilization. 60The appa À/À ; appb À/À double mutants were reported to have morphologically abnormal ependymal cilia and smaller brain ventricles. 60It would be interesting to see if the localization of App proteins to cilia and ventricles is important for sleep and locomotion, as the coordinated periodic beating of the cilia is involved in the generation of cerebrospinal fluid (CSF) flow within ventricle cavities, 63 and CSF circulation is believed to facilitate transfer of signaling molecules and removal of metabolic waste products important for behavior. 64,65hat emerges from these results is a complex picture of endogenous App-derived signals that can regulate sleep and wake in a bidirectional manner, with some signals boosting sleep and some inhibiting sleep.Changes in the relative composition of App-derived molecules, including both Ab42 and P3 peptides (Ab 17-40 and Ab 17-42 ) over the progression of preclinical and clinical AD makes for even more ) are typically found in the diffuse plaques of individuals with AD, 66 and cell culture experiments suggest that these peptides may be produced in even greater quantities than Ab42 and Ab40. 67Additionally, microglia in AD patients harbor various N-terminally truncated Ab species, 68 and the presence of P3 peptides in CSF shows a positive correlation with cognitive decline in AD patients, 69 indicating a potential role of these peptides in AD pathogenesis.Our results suggest that in addition to Ab, 26,70 P3 might also interfere with sleep and wakefulness in AD and should be further investigated in rodent AD models.
2][73][74] Furthermore, generation and release of Ab42 into the interstitial fluid (ISF) is controlled by synaptic activity 75 and even one night of sleep disruption can increase Ab42 levels. 76As sleep can directly alter Ab levels, sleep history over one's lifetime may be a significant contributor to AD risk and progression.Our results are consistent with the idea that alterations in App gene products may be a direct contributor to sleep phenotypes associated with preclinical and clinical AD.The specificity of the effect of Appb loss on sleep architecture also suggests that specific changes in sleep patterns that could serve as a useful AD biomarker may yet be discovered.

Limitations of the study
We have interpreted the lack of sleep effect of the g-secretase (or b-secretase) inhibitors in the appb À/À mutant background to be due to the loss of Appb proteolytic cleavage; however, it is technically possible that loss of Appb affects expression or localization of other g-secretase targets (such as Notch), and the lack of drug-induced sleep alterations is due to this indirect effect.Another limitation of our study is that the exogenous ventricle P3 injections might not recapitulate the actual localization or structure of P3, and further experiments would be needed to dissect the roles of the App fragments.binding site sequence: 5 0 -TATGGACCGCACGGTATT-3 0 , Right TALEN binding site sequence: 5 0 -CGACTTTGTCCCTCGCCA-3 0 and Spacer sequence: 5 0 -TTAATGCTGACGA-3'.TALENs were generated using the FLASH assembly method following the protocol of 80. Starting with a library consisting of 376 plasmids that encode one to four TAL effector repeats consisting of all possible combinations of the NI, NN, HD or NG repeat variable di-residues (RVDs), the four 130 bp a-unit DNA fragments were amplified from each a-unit plasmid using the Herculase II Fusion DNA polymerase (Agilent) and oJS2581 and oJS2582 primers. 80The resulting 5 0 biotinylated PCR products were digested with BsaI-HF (NEB) to generate four basepair overhangs.To generate the DNA fragments encoding the bgdε (extension fragment) and bgd (termination fragment) repeats, each of these plasmids was digested with BbsI followed by serial restriction digests of XbaI, BamHI-HF and SalI-HF (New England Biolabs) to cleave the plasmid backbone.The four TALEN expression vectors encoding one of four possible RVDs were linearised with BsmBI (NEB).The biotinylated a unit fragments were ligated to the first bgdε fragments using Quick T4 DNA ligase and bound to Dynabeads MyOne C1 streptavidin-coated magnetic beads (Life Technologies).The bead bound a-bgdε fragments were digested with BsaI-HF (NEB) to prepare the 3 0 end of the DNA fragments for the subsequent ligation step.Each extension and termination fragment was then ligated to assemble the complete DNA fragment encoding the TALE repeat array by repeated digestion and ligation steps, and a final digestion with BbsI (NEB) released the full length fragments.The purified DNA fragments were ligated into one of four BsmBI (NEB) digested TALEN expression vectors encoding one of four possible RVDs using Quick T4 DNA ligase.Ligation products were transformed into chemically competent XL-10 Gold E. coli cells and clones grown on LB Agar plates containing Ampicillin at 37 C overnight.Bacterial colonies of each TALEN arm were selected and screened by colony PCR using primers oSQT34 (5 0 -GACGGTGGCTGTCAAATACCAAGATATG-3 0 ) and oSQT35 (5 0 -TCTCCTCCAGTTCACTTTTGACT AGTTGGG-3 0 ).Clones showing a correct sized band were cultured in LB medium containing Ampicillin at 37 C overnight.Following plasmid mini-preparation the inserts were sequenced using primers oSQT1 (5 0 -AGTAACAGCGGTAGAGGCAG-3 0 ), oSQT3 (5 0 -ATTGGGCTACGATG GACTCC-3 0 ) and oJS2980 (5 0 -TTAATTCAATATATTCATGAGGCAC-3 0 .mRNA was synthesised using the mMESSAGE mMACHINE T7 and polyA tailing kit. 100 pg of each of the TALEN mRNAs are injected into the cytoplasm of one-cell stage embryos, which were raised to adulthood and sequenced (below).
Sequencing/genotyping pipeline F0 embryos were raised to adulthood, fin-clipped and deep-sequenced by Illumina Sequencing (MiSeq Reagent Nano Kit v2 (300 Cycles) (MS-103-1001)) to identify founders.Fin-clipping was done by anesthetizing the fish by immersion in 0.02% MS-222 (Tricaine) at neutral pH (final concentration 168 mg/ml MS-222).DNA was extracted by HotSHOT 81 by lysing a small piece of the fin in 50 ml of base solution (25 mM KOH, 0.2 mM EDTA in water), incubated at 95 C for 30 min, then cooled to room temperature before 50 ml of neutralisation solution (40 mM Tris-HCL in water) was added.For appa, a 214 base pair fragment surrounding the conserved 25-35 th amino acid region within appa was PCR amplified using gene-specific primers with miSeq adaptors (forward primer, 5 0 -TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTGC AGGAATAAAGCTGATCT-3'; reverse primer, 5 0 -GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG ATGGACGTGTACTGCTTCTTCC- The PCR product's concentration was quantified with Qubit (dsDNA High Sensitivity Assay), then excess primers and dNTPs were removed by ExoSAP-IT (ThermoFisher) following the manufacturer's instructions.The samples were then sequenced by Illumina MiSeq to assess the presence of insertion/deletions.The mutant F0 fish containing a 5 base pair deletion in the Ab region resulting in a stop codon on the 26 th residue of Ab (human numbering) was chosen to generate a stable mutant line appa D5 (u539).An appb mutant carrier containing a 14 bp deletion and a 4 bp insertion appb D14+4 (u537) that is predicted to generate a frameshift and early stop codon was selected to make stable mutant lines for further analysis.F0 fish with indels were then outcrossed to wild-types and 10 one day old F1 embryos from each pairing were screened by Sanger sequencing to assess the nature of the mutations that passed into the germline.To minimize potential off-target mutations, mutant fish were crossed to ABxTL and TL WT strains for 3 generations before performing any behaviour experiments.

DNA extraction
Zebrafish DNA was extracted by the HotSHOT method. 8150 ml of 13 base solution (25 mM KOH, 0.2 mM EDTA in water) was added to finclips in individual wells.Plates were sealed and incubated at 95 C for 30 min, cooled to room temperature and neutralised by adding 50 ml of 13 neutralisation solution (40 mM Tris-HCL in water).Genomic DNA was then stored at 4 C.
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.

Behavioural experiments
Behavioural tracking of larval zebrafish was performed as described in 20,82 with the following adjustments.Zebrafish larvae were raised on a 14hr:10hr light:dark cycle at 28.5 C and at were placed into individual wells of a square-well 96-well plate (Whatman) containing 650 mL of standard embryo water (0.3 g/L Instant Ocean, 1 mg/L methylene blue, pH 7.0) at 4-5 dpf.Locomotor activity was monitored using an automated video tracking system (Zebrabox, Viewpoint LifeSciences) in a temperature-regulated room (26.5 C) and exposed to a 14hr:10hr white light:dark schedule with constant infrared illumination (Viewpoint Life Sciences).Larval movement was recorded using the Videotrack quantization mode.The movement of each larva was measured, and duration of movement was recorded with an integration time of 60 sec.Data were processed using custom PERL and MATLAB (The Mathworks, R2019a) scripts, and statistical tests were performed using MATLAB (The Mathworks, R2019a).
Any one-minute period of inactivity was defined as one minute of sleep. 82Sleep bout length describes the duration of consecutive, uninterrupted minutes of sleep whereas sleep bout number is the number of such sleep events in a given time interval.Average waking activity represents activity only during active periods.
All mutant larval zebrafish experiments were performed on siblings from appa +/D5 or appb +/D14+4 heterozygous incrosses, except for drug experiments, which were simultaneously performed on larvae from WT and appb D14+4/D14+4 incrosses from different parents.DAPT (Cell Guidance Systems, SM15-10) was dissolved in DMSO to make a stock concentration of 10 mM and diluted further to a working concentration of 10 mM in water in 1:10 serial dilutions.6 ml of the 10 mM DAPT stock or 6 ml of 0.1% DMSO control was added individually to the wells in the behaviour plate to make a 100 nM final concentration the second day at Zeitgeber time 0 (Lights ON) of the video-tracking in each experiment.Lanabecestat (Cambridge Bioscience, HY-100740-2mg) was dissolved in DMSO to make a to make a stock concentration of 10 mM and diluted further to a working concentration of 100 mM in water in 1:10 serial dilutions.1.8 ml of the 100 mM Lanabecestat stock or 1.8 ml of 1% DMSO control was added individually to the wells in the behaviour plate to make a 300 nM final concentration.The drug was added on the second day at Zeitgeber time 0 (Lights ON), when the larvae are 6 dpf.

Zebrafish in situ hybridization (ISH)
RNA was extracted from 30 WT embryos (5dpf) by snap freezing in liquid nitrogen and TRIzol RNA extraction (Ambion 15596026).1 mg of RNA was reverse transcribed with AffinityScript (Agilent, 600559) to make cDNA following the manufacturer's protocol.

Western blots
Protein was extracted from adult appa D5/ D5 ; appb -/-mutants or WT larvae.Fish were euthanized, brains were dissected out and immediately frozen in liquid nitrogen prior to use and stored at À 80 C. Samples were homogenized in an ice-cold lysis buffer (10 mM Tris-HCl pH 8.0, 2% sodium deoxycholate, 2% SDS, 1 mM EDTA, 0.5 M NaCl, 15% glycerol) supplemented with protease inhibitors cocktail (Calbiochem protease inhibitors cocktail III) using a syringe needle (BD microlance, Ireland; 27G ½'' 0.4 x 13 mm) on ice.Samples were then incubated 20 min on ice, sonicated for 10 min at 70% amplitude with a pulse of 30s on and off and then centrifuged at 10,0003g at 4 C. Supernatants were collected and kept on ice and protein concentration measured with a Qubitä protein BR assay kit (Thermo Fisher Scientific, Waltham, MA) and samples stored at À 80 C. Protein samples (40-50ug) were then diluted in a denaturing lysis buffer (1X NuPAGE LDS Sample Buffer (Thermo Fisher Scientific, Waltham, MA), 0.05 M DTT (Sigma-Aldrich, St. Louis, MO), lysis buffer completed with protease inhibitors) and then boiled for 5 min at 95 C. Proteins were then separated on a NuPAGE NOVEX 4-12% gradient Bis-TRIS pre-cast gel (Thermo Fisher Scientific, Waltham, MA) at 150V for 1 hour and transferred onto a 0.2 mm Amershamä Portranä nitrocellulose membrane at 400 mA for 50 minutes on ice.The membrane was incubated in a blocking solution (5% milk) for 2 hours at RT and then immunoblotted overnight at 4 C with the primary mouse anti-amyloid precursor protein A4 antibody (clone 22C11) (Sigma MAB348-AF647) (1:3000) and with a loading concentration control mouse anti-g-tubulin monoclonal (1:10,000) (Sigma, St. Louis, MO).The membrane was then washed in TBS-Tween three times for 10 min at RT and incubated with the secondary antibody anti-mouse-HRP (1:5000) (Sigma Aldrich) for one hour at RT.The signal was developed using SuperSignal West Dura Extended Duration Substrate kit (Thermo Fisher Scientific, Waltham, MA) and imaged using ChemiDoc Imaging (Bio-Rad, Hercules, CA).Western blot images were processed using ImageJ (NIH, USA).qPCR Total RNA was isolated from 4-6 dpf zebrafish larvae using the RNeasy Plus Micro Kit (Qiagen).cDNA was synthesized using the SuperScript III First-Strand Synthesis System (Invitrogen).qPCR was performed using a CFX96 machine (Bio-Rad) and accompanying BioRad CFX Manager (v3.1) using GoTaq qPCR master mix (Promega, A6001) with the primers qPCR_appb_F2 (5 0 -CGTGGTCATCGCTACTGTCA) and qPCR_appb_R2 (5 0 -CTGCCGCATCCACCTCAATA) at 60 C resulting in a 98 bp product.ef1a was chosen as the reference gene as it has been validated in the zebrafish for qPCR normalisation. 87Reactions were performed up to a total volume of 10ml per reaction with primer concentrations of 10mM for ef1a, (eef1a1_qRT-PCR Forward: 5 0 -TGCTGTGCGTGACATGAGGCAG-3 0 and eef1a1_qRT-PCR Reverse: 5 0 -CCGCAACCTTTGGAACGGTGT-3 0 ) or 20mM for appb primers.Efficiency of ef1a and appb primers were established to be within acceptable efficiency thresholds through a serial dilution series (90-110% efficiency).Threshold cycle values (Cq) were obtained for each gene in each sample in technical replicates.Two replicate experiments were performed per gene with 3 technical replicates.

qPCR analysis
Analysis was undertaken using the 'delta-delta Cq' method to compare the relative gene expression of the target gene (appb) to the reference gene (ef1a). 88Further analysis used the Qiagen REST programme (2009)(v2.0.13).This software compares treated versus untreated samples using provided serial dilution data for efficiency calculations.REST then performs a pairwise fixed reallocation randomisation test (permutations = 2000) to determine p values between samples, as well as giving confidence intervals and standard error of the permutation analysis.Melt curve analysis was conducted through the BioRad CFX manager software.Boxplots were generated using GraphPad Prism (Dotmatics).

Endogenous Ab measurements in adult zebrafish brains
For the DAPT treatment, WT adult fish were treated with a final concentration 25 mM of DAPT for 24 hours in a small tank.Adult zebrafish brains were dissected, weighed and were mechanically homogenized in 100 mL TBS (50mM Tris-HCL, pH 8.0) containing Calbiochem protease inhibitor cocktail set III (1:200).Whole brain homogenates were centrifuged at 16,000 g at 4 C for 30 min, the supernatant was aliquoted and stored at -80 C. Total protein concentration of the samples was determined using a Pierce Detergent Compatible Bradford assay kit according the manufacturer's instructions (Thermofisher, 23246).Ab40 and Ab42 measurements of the samples were done according to the manufacturers protocols using Mesoscale Discovery V-plex Plus Ab42 (4G8) or V-plex Plus Ab peptide panel 1 (4G8) kits on the Meso Scale Discovery platform (MSD, Rockville, Maryland) in technical duplicates or triplicates.Standard curves were created using the MSD Mesoscale Discovery Workbench Toolbox to benchmark Ab concentrations in the samples.A 4-parameter logistic curve was used to fit standards and calculate the concentration for unknowns and Ab controls.Ab standards for the calibration curve were measured in duplicate and were set in serial in 1:4 dilutions.The upper and lower limits of detection were set as 2.5 standard deviations from the bottom and top calibrator.The calculated Ab amounts were then normalized to the total extracted protein levels from each sample were determined using the Bradford assay.

Figure 1 .
Figure 1.Zebrafish App protein organization, gene expression, and mutant generation (A) There are 2 App orthologs, appa and appb, in zebrafish.Appa contains the Kunitz-type protease inhibitor (KPI) domain and thus has a similar gene organization to the human APP770 isoform.Zebrafish Appb lacks the KPI domain similar to the brain-enriched human APP695 isoform.Both Appa and Appb have the functional App domains, including the heparin-binding domain (HBD), copper-binding domain (CuBD), extracellular E2 domain (E2), the conserved YENTPY motif, and amyloid-beta region (AB).(B) Alignment of Ab regions of zebrafish Appa and Appb to human APP695 and APP770 shows high conservation within the Ab region and the proteolytic cleavage sites (indicated with black arrows): a-secretase cleavage site (a), b-secretase cleavage site (b), g-secretase cleavage sites (g), and ε-cleavage sites (ε).Black, dark gray, and light gray boxes indicate strictly, highly, and moderately conserved amino acid residues, respectively.(C) As detected by multiplexed hybridization chain reaction (HCR), appa (green) and appb (red) are both expressed widely but in non-overlapping regions of the 5dpf larval brain, including the cerebellum and nuclei in the hindbrain (C, cerebellum; H, hindbrain).Shown is a representative image from a single brain taken at

Figure 2 .
Figure 2. appa D5/D5 mutants have reduced day waking activity but no sleep phenotype Exemplar 48 h traces of average waking activity taken from a single experiment of appa D5/D5 mutants, heterozygous, and wild-type siblings (5-7dpf) on a 14 h:10 h light:dark cycle.Each line and shaded ribbon show the mean G SEM. (B) Day waking activity and C) night waking activity for appa D5/D5 mutants and siblings from appa +/D5 incrosses, combined across N = 5 experiments.Each dot is a single larva, normalized to the mean of their experimentally matched WTs.At the bottom are plotted the effect sizes (G95% confidence interval [CI]) relative to WT. (D) Exemplar 48 h traces of average sleep for the same experiment shown in (A).(E) Day sleep and (F) night sleep of WT, heterozygous, and appa D5/D5 mutants normalized to their WT siblings, as in (B) and (C).ns p > 0.05, *p % 0.05, Kruskal-Wallis, Tukey's post hoc test.n = number of larvae.

Figure 3 .
Figure 3. appb À/À mutants have altered waking activity and sleep across the day-night cycle Exemplar 48 h traces of average waking activity taken from a single experiment of appa D5 mutants, heterozygous, and wild-type siblings (5-7dpf) on a 14h:10h light:dark cycle.Each line and shaded ribbon show the mean G SEM. (B) Day waking activity and (C) night waking activity for appb À/À mutants and siblings from appb +/À incrosses, combined across N = 5 experiments.(D) Exemplar 48 h traces of average sleep for the same experiment shown in (A).(E) Day sleep, (F) night sleep, (G) day sleep length, (H) night sleep length, (I) day sleep bout number, and (J) night sleep bout number of WT, heterozygous, and appb À/À mutants normalized to WT siblings as in (B) and (C).At top, each dot is an animal normalized to the mean of their experimentally matched WT.At bottom, shown are the effect size G95%CI relative to WT. ns p > 0.05, *p % 0.05, **p % 0.01, Kruskal-Wallis, Tukey's post hoc test.n = number of larvae.

Figure 4 .
Figure 4.The g-secretase inhibitor DAPT shortens sleep bout lengths at night in WT but not in appb À/À mutants (A and B) Exemplar 48 h traces on a 14h:10h light:dark cycle of the average waking activity of WT (A) and appb À/À mutants (B) continuously exposed to either 100 mM DAPT or DMSO vehicle control.(C and D) Exemplar 48 h traces of the average sleep from the same experiment shown in (A) and (B).(E) Night sleep and (F) night sleep bout length of WTs and appb À/À mutants exposed to either 100 mM DAPT or DMSO vehicle.At top, each dot represents a single larva normalized to its experiment-matched WT (-DAPT) mean; error bars indicate GSEM.At bottom, the within-genotype effect size and 95%CIs of DAPT treatment are plotted.n = the number of larvae.Data are pooled from N = 4 independent experiments, omitting the first day and night to account for any delay in drug action.ns p > 0.05, *p % 0.05, **p % 0.01, two-way ANOVA, Tukey's post hoc test.n = number of larvae.

Figure 5 .Figure 6 .
Figure 5.The b-secretase inhibitor lanabecestat increases sleep at night in WT but not in appb À/À mutants (A and B) Exemplar 48 h traces on a 14h:10h light:dark cycle of the average waking activity of WT (A) and appb À/À mutants (B) continuously exposed to either 0.3 mM lanabecestat or DMSO vehicle control.(C and D) Exemplar 48 h traces of the average sleep from the same experiment shown in A and B. (E) Night sleep and (F) night sleep bout length of WTs and appb À/À mutants exposed to either 0.3 mM lanabecestat or DMSO vehicle.At top, each dot represents a single larva normalized to its experiment-matched WT (-lanabecestat) mean, and error bars indicate GSEM.At bottom, the within-genotype effect size and 95% CIs of 0.3 mM lanabecestat treatment are plotted.n = the number of larvae.Data are pooled from N = 4 independent experiments, omitting the first day and night to account for any delay in drug action.ns p > 0.05, *p % 0.05, **p % 0.01, two-way ANOVA, Tukey's post hoc test.n = number of larvae.

Table 1 .
Effects of blocking g-secretase or b-secretase on proteolytic cleavage products on App and effects on night sleep