The localization of amyloid precursor protein to ependymal cilia in vertebrates and its role in ciliogenesis and brain development in zebrafish

Amyloid precursor protein (APP) is expressed in many tissues in human, mice and in zebrafish. In zebrafish, there are two orthologues, Appa and Appb. Interestingly, some cellular processes associated with APP overlap with cilia-mediated functions. Whereas the localization of APP to primary cilia of in vitro-cultured cells has been reported, we addressed the presence of APP in motile and in non-motile sensory cilia and its potential implication for ciliogenesis using zebrafish, mouse, and human samples. We report that Appa and Appb are expressed by ciliated cells and become localized at the membrane of cilia in the olfactory epithelium, otic vesicle and in the brain ventricles of zebrafish embryos. App in ependymal cilia persisted in adult zebrafish and was also detected in mouse and human brain. Finally, we found morphologically abnormal ependymal cilia and smaller brain ventricles in appa−/−appb−/− mutant zebrafish. Our findings demonstrate an evolutionary conserved localisation of APP to cilia and suggest a role of App in ciliogenesis and cilia-related functions.


Results
Appa and appb mRNA expression patterns at the brain ventricular limits. The zebrafish app genes, appa and appb, are expressed in the CNS, and have both distinct and shared expression patterns 1,14 . Due to the lack of specific antibodies, we used fluorescent whole mount in situ hybridization to increase the cellular resolution of appa and appb mRNA expression in areas with motile cilia on 30 hpf wild-type larvae zebrafish (Fig. 1). Consistent with previous studies, we observed appa mRNA expression in the lens, the olfactory bulb and epithelium, in the trigeminal ganglia and in the otic vesicle. (Fig. 1C). Similarly, the appb mRNA expression signal corroborated previous data on appb mRNA expression 1 in the olfactory and otic vesicle epithelia (Fig. 1H).
In addition, both appa ( Fig. 1C-G and high magnification Fig. 1N,O) and appb ( Fig. 1H-L,P,Q) mRNA signals labelled cells lining the diencephalic ventricle both in the dorsal and ventral areas. Negative controls did not show any specific signal (Supplementary file 1). Together, these results show expression of appa and appb in areas with ciliated cells, including cells lining brain ventricles, otic vesicle and olfactory organ, thus suggesting a possible role of App in cilia formation and function.
App protein is localized to cilia of the olfactory sensory neurons and otic vesicle in zebrafish larvae. The expression of both appa and appb in ciliated cells made us ask if the proteins become distributed out to the cilia. The zebrafish olfactory epithelium and the otic vesicle comprise ciliated cells and are regions where both appa and appb mRNAs are expressed. To address if Appa and Appb become localized to these cilia, we performed immunofluorescent staining on zebrafish larvae.
Olfactory sensory neuron cilia. We used the Y188 antibody, binding to a conserved epitope in the C-terminal end of human, mouse, and zebrafish App (Fig. 6C), in combination with the anti-acetylated tubulin antibody, labelling microtubule structures of cilia. Immunofluorescent co-labelling detected a punctate App signal in the heavily ciliated olfactory epithelium area at 30 hpf ( Fig. 2A). However, while the resolution of the images did not allow distinction between each cilium, App signal seemed to localize to most of them. In addition to the cilium, App expression was also found at the base of these motile cilia ( Fig. 2A').
Otic vesicle cilia. Similar to the olfactory neurons, high accumulation of App was noted at the base of the cilia in the otic vesicle. In zebrafish, hair cells of the otic vesicle have two types of cilia, a long single kinocilium and a bundle of shorter stereocilia 43 . The immunofluorescent staining revealed App expression in both types of cilia at early time points in the larvae development (Fig. 2B,C). Staining of 24 hpf larvae with glutamylated tubulin, highlighting the cilia basal bodies, clearly showed an App signal within the hair cells and close to the basal body (Fig. 2B,B' ,B''). App expression became more distinct at 30 hpf (Fig. 2C,C'). Plots of the intensity profile of App (green) and acetylated tubulin (magenta) showed a punctate distribution of App throughout the kinocilium (Fig. 2D), which supports that App localizes to the cilium membrane (Fig. 2E). No signal was detected in the intensity profile in the absence of App puncta (Fig. 2F). Together, these data show expression of App in cilia and ciliated cells of the otic vesicle and olfactory bulb and indicate that App is located at the cilium membrane.
App localizes to cilia decorating the brain ventricle surface of zebrafish. As APP was previously shown to be expressed by the ependymal cells in rodents and in humans [44][45][46] , we explored App expression by ependymal cells and App localization at their cilia in larvae and adult zebrafish (Fig. 3). At 30 hpf, the brain ventricles are inflated and the differentiation of motile cilia in the most ventral and dorsal regions have just started but do not yet contribute to directional CSF flow 47 . This facilitates whole-mount imaging and measure- www.nature.com/scientificreports/ ment of single cilium. Using the same combination of antibodies (anti-APP (Y188) and anti-acetylated tubulin) as above, we could detect App-positive puncta along the acetylated tubulin signal in most cilia (Fig. 3B,C). To address if App localization to cilia is maintained into adulthood, we performed immunofluorescent staining on coronal sections of adult zebrafish brains using antibodies detecting App (Y188) and acetylated tubulin to label cilia. Our results showed that consistent to larvae, App was distributed to ependymal cilia in the adult brain. In contrast to larvae, ependymal cells in adult individuals were covered with multiple motile cilia. Cryosections  www.nature.com/scientificreports/  www.nature.com/scientificreports/  www.nature.com/scientificreports/ of adult zebrafish brain revealed dense cilia tufts with App-positive staining at the apical side of the ependymal cells (Fig. 3F,G). Furthermore, App was also expressed by ependymal cells, similarly to what has previously been described in rodents and humans (Fig. 3G). Negative controls did not show any cilia-specific staining (Fig. 3D).
No signal was detected in adult zebrafish samples for App in the negative control staining (absence of primary antibody Y188) or in the presence of rabbit IgG and secondary antibody (Supplementary file 2).

Conserved localization of APP in ependymal cilia in mouse and human brains.
APP is also localized to ependymal cilia in mice and humans. We performed immunostaining on mouse brain sections using two antibodies directed to APP, Y188 binding to the C-terminal intra-cellular domain and 22C11 detecting the E1 domain of the N-terminal region (Fig. 6C), together with anti-acetylated tubulin. The ependymal motile cilia were easily observed in the third ventricle of the brain sagittal section (Fig. 4A,B). Congruent with our results on adult zebrafish brains, we detected strong APP expression with both antibodies throughout the ependymal cells layer and punctate APP staining (Y188 see Fig. 4C and 22C11 see Fig. 4D) overlapping with acetylated tubulin-positive cilia. Interestingly, APP expression by the choroid plexus cells was detectable (Fig. 4B). Negative controls for primary antibodies were performed and showed no or weak signal (Fig. 4E).
In the human brain, acetylated tubulin staining allowed separation of cellular layers of the caudate nucleus and identification of acetylated tubulin-positive cilia of the ependymal cell layer lining the lateral ventricle ( Fig. 5A,B,D). However, while many ependymal cells had intact cilia, many were found broken and dislocated from their cell (Supplementary file 3).
To support the presence of APP in ependymal cilia, brain serial sections of the caudate nucleus were incubated with horseradish peroxidase (HRP)-conjugated Y188 or A8717 antibodies, both recognizing the C-terminal domain of APP. Similarly to our results obtained in mouse and zebrafish brains, brightfield images confirmed strong APP expression in the ependymal cells and, upon higher magnification, in ependymal cilia (Fig. 5C,E, Supplementary file 3). In contrast to zebrafish and mouse, APP in human ependymal cilia was evenly distributed and was not detected as puncta.
In summary, these results show that the expression of APP in the ependymal cells and their cilia are conserved between species as far apart as zebrafish, mice, and humans.
Generation of appa and appb double mutant zebrafish. In contrast to humans and mice, zebrafish have two APP orthologues, appa and appb (together designated app). The zebrafish appb 26_2 mutant, carrying a frame shift deletion of five base pairs at the end of exon two, which introduce a premature stop codon, was generated and described by our lab previously 14 . However, to investigate the requirement of both App proteins in ciliogenesis, we used the CRISPR/Cas9 method to generate mutations in the zebrafish appa gene (Fig. 6A). A mutation was identified in exon 2 (Fig. 6A), and Sanger sequencing confirmed a deletion of 10 nucleotides (Fig. 6B). The mutation resulted in a premature stop codon predicted to give rise to a protein truncation at amino acid 109 (Fig. 6C). The appa mutant allele was outcrossed into the AB background until generation F4 and then bred with the appb −/− to generate the double mutant appa −/− appb −/− zebrafish line. The app mutant zebrafish were healthy and fertile and did not show any gross morphological phenotypes. qPCR analysis of genes expression showed very low appa and appb mRNA levels in the double mutant fish line (Fig. 6D). Western blot analysis using the Y188 and 22C11 antibodies with epitopes in the intracellular and extracellular domain, respectively, showed decreased protein levels in app double-mutant larvae ( Fig. 6E and Supplementary file 4). Both antibodies are likely cross-reacting with Aplp2 since the epitope sequences are highly similar. These data show that the introduced mutation in appa resulted in a significant decrease of both transcription and translation of the Appa protein indicating that the mutation gives rise to a loss-of-function mutation. Longer brain ventricle cilia in appa −/− appb −/− larvae. The conserved distribution of APP in brain ventricle cilia prompted us to address the requirement of App during ciliogenesis. We measured the length of cilia in the midbrain ventricle detected by acetylated tubulin immunostaining signal in both appa −/− appb −/− double mutants and wild-type larvae at 30 hpf. At this stage, the cilia delineating the dorsal and ventral parts of the diencephalic ventricles are not yet motile 47 . A 3D-region of interest (ROI) was used to measure cilium length. The ROI was established from the dorsal part of the midbrain ventricle to the ventricular space at a depth of around 25 μm. To our surprise, we found that the ependymal cilia in the ROI were significantly longer in appa −/− appb −/− mutants compared with wild-type larvae (Fig. 7), which was confirmed by frequency distribution (Supplementary file 5).
The appa −/− appb −/− double mutants exhibit smaller diencephalic ventricle. We then went on to address if defects in ependymal cilia affect brain ventricle formation. The gross morphology was determined by www.nature.com/scientificreports/ measuring the length between specific points and areas of the ventricles: rostral to caudal, diencephalon ventricle sagittal length, amplitude and height (Fig. 9A). However, no significant change was detected compared with wild-type larvae (Fig. 9B). We next analysed brain ventricle area and volume in 2dpf larvae (Fig. 9C) and found significant reductions in both area and volume of the ventricular space in appa −/− appb −/− compared with wildtype larvae (Fig. 9D). The diencephalic ventricle was smaller in both area and volume when (Fig. 9E) compared   www.nature.com/scientificreports/

Discussion
In this study, we show that App localizes to different non-motile and motile cilia in zebrafish larvae including the stereo-and kinocilia of the otic vesicle, motile cilia of olfactory sensory neurons in the olfactory epithelium, and cilia of the ependymal cells lining the brain ventricles. We also show an evolutionary conserved localization of APP to cilia of the ependymal cells lining the brain ventricles of adult zebrafish, mice and humans. As these results indicated a possible function of APP in ciliogenesis or cilium function, we used zebrafish lacking the two APP orthologues, Appa and Appb, and found longer ependymal cilia and smaller brain ventricles in larvae zebrafish. Thus, our results suggest that APP not only is distributed to cilia but also seems to have an important function in ciliogenesis and brain development. www.nature.com/scientificreports/ Using different antibodies, we found a punctate localization of App in cilia, indicating that the protein is randomly distributed within the cilium similar to other membrane receptors such as SSTR3 and Smoothened (Smo) 56,57 . In contrast, we observed a continuous rather than punctate localization of APP in human ependymal cilia suggesting that the distribution of APP may differ between species. The cilium membrane, although continuous with the plasma membrane, possesses a specific and conserved composition of proteins and lipids. This specification is established through an active transport of ciliary membrane proteins 58 that at least partly depends on specific CTSs within proteins 59 . The presence of several such CTSs and their conservation between zebrafish, mice, and human supports a motif-based transport of APP to cilia.
APP expression by ependymal cells was first reported in rodents and humans in the late 1980s and early 1990s 44,45,60,61 . In line with these findings, we here not only confirm the expression of App in adult zebrafish ependymal cells, but also show that App localizes to ependymal motile cilia in vertebrates as far apart as zebrafish, mice, and humans. The longer but structurally normal ependymal cilia of appa −/− appb −/− mutants suggests a role of App in ciliogenesis. However, the appa −/− appb −/− mutants gave rise to fertile adults without phenotypic changes associated with primary cilium defects, such as curved body and hydrocephalus [62][63][64][65] . During early development, motile cilia of Kupffer's vesicle are essential to establish laterality of organs 66,67 , while others maintain CSF flow within brain ventricles. Consequently, cilia-driven flow is crucial to form and maintain proper brain ventricles, as zebrafish, clawed frog and mouse ciliary mutants display ventricular defects 68  CSF circulation is thought to play an important role in removal of waste products from the brain parenchyma 69 . Thus, subtle changes in the coordinated beating of cilia may contribute to altered CSF flow, impair clearance and hence contributes to a slow build-up of waste products over time in the aging brain. Supporting this are findings that individuals with Down syndrome, expressing approximately 50% higher levels of APP, have changed CSF flow in the lateral ventricles 70 and develop Alzheimer's disease at early age 71 . Although the morphology of ependymal cilia of DS brains is unknown, in vitro cell cultures show decreased primary cilium length 39 . Further studies on App's effect on ciliary movement and CSF flow during development and aging are needed.
Analysis of APP fragments in CSF, currently used to diagnose individuals with cognitive impairments, are thought to represent pathological changes within the brain. However, protein fragments detected in CSF may not only originate from the brain parenchyma but may also be derived from APP processed within the ependymal cells and the protruding cilia as secretases needed for APP processing are present in cilia 72 . The release of APP from ependymal cells could be mediated through the release of extracellular cleavage products or by budding extracellular vesicles and ectosomes from the cilium 73 . Interestingly, APP-containing vesicles released into the CSF 74 were found to have lower levels of APP in AD patients compared to healthy individuals 75 . Furthermore, a well-established feature of normal pressure hydrocephalus, where ciliary function is impaired 76 , is decreased CSF levels of soluble APP and Aβ, which are restored upon successful shunt treatment of the condition [77][78][79] . The impact of ependymal integrity and the contribution of cilium-mediated APP release need further studies but could potentially change the present interpretation of biomarkers used to assess disease progression. The extent to which APP mediate other processes requiring the ependymal cells including neurogenesis 61 and migration of new-born neuronal cells 80 still remain to be investigated.
Finally, accumulation of App at the root of the basal body, as observed in the olfactory sensory neurons and otic vesicle cilia in larval zebrafish, correlates with the findings reported by Yang and Li on APP enrichment along ciliary rootlets 81 . The presence of App in cilia mediating sensory input, both olfaction and hearing, opens up the possibility that sensory changes may not only result from defects in the brain regions receiving input from these organs, but could also be due to direct effects on their cilia. Thus, further studies on cilia of sensory neurons may give insights into the mechanisms resulting in the sensory deficiencies observed in AD mouse models and neurodegenerative diseases 82 .
Altogether, our data show the presence of App in various cilia and at least in the ependymal cilia, a conserved distribution across vertebrates. The evolutionary conserved CTSs of APP and its expression throughout development and aging suggest a central role of APP within the ependymal in both ciliogenesis and brain ventricle formation. Further studies are required to fully understand the impact of App in cilia in our olfactory and auditory organs and to which extent defects in ependymal cell integrity and ciliation contribute to APP-related developmental processes and disease progression.

Methods
Animal care and ethics statement. The zebrafish (Danio rerio) facilities and maintenance were approved and follow the guidelines of the Swedish National Board for Laboratory Animals. This study was approved by the Animal Ethical Committee at the University of Gothenburg. All procedures for the experiments were performed in accordance with the animal welfare guidelines of the Swedish National Board for Laboratory Animals and followed the recommendations in the ARRIVE guidelines 83 . Zebrafish were maintained in Aquatic Housing Systems (Aquaneering, San Diego, CA) at 28.5 °C, under a 14:10 h (h) light:dark cycle at the Institute of Neuroscience and Physiology, University of Gothenburg. Fish were fed twice daily a diet of live-hatched brine shrimps and Gemma fish food (Skretting, Amersfoort, Netherlands). System water was created using reverse osmosis water kept at a pH of 7.2-7.6 with NaHCO 3 and coral sand and salt (Instant Ocean, Blacksburg, VA) to maintain the conductivity at 600μS. Breeding of fish was carried out in 1-2 L breeding tanks and embryos were collected in embryo medium (EM) ( www.nature.com/scientificreports/ The following fish lines were used in the present project; AB fish from the Zebrafish international resource centre (ZIRC) or was used for outbreeding and as wild-type background, appb 26_2/26_2 and appa −/− as described below 14 .
Human brain tissue samples were performed by Queen Square Brain Bank for Neurological Disorders, Department of Clinical and Movement Neurosciences, Institute of Neurology, University College London (UCL). Ethical approval for the use of human post-mortem samples was approved by a London Research Ethics Committee and tissue stored for research under a license from the Human Tissue Authority. Informed consent was obtained from each donor or obtained from the next of kin/or legal guardian(s) of the donors. Human brain tissues were used in accordance with the Helsinki declaration and the regional ethics committees at UCL and the University of Gothenburg have provided approval for the study.
Mutagenesis using the CRISPR/Cas9 system. Genetic mutations in the appa gene were introduced using the CRISPR/Cas9 system as previously described 85 . Briefly, gRNAs were generated with a target-specific DNA oligonucleotide (Integrated DNA Technologies, Leuven, Belgium) containing a T7 promoter sequence in the 5'-end and a 'generic' DNA oligonucleotide for the guide RNA. The two oligonucleotides were annealed and extended with Platinum Taq DNA polymerase (Thermo Fisher Scientific, Waltham, MA), in a final concentration of 1 × buffer, 0.25 mM dNTP, 0.5 μM of each oligonucleotide and 0.04U/ul Taq with one cycle at the following temperatures (98 °C 2 min; 50 °C 10 min, 72 °C 10 min). The resulting product was analyzed on a 2.5% agarose (Roche, Basel, Switzerland) gel to confirm a single fragment of 120 basepairs (bp) and used to transcribe RNA. In vitro transcription was performed with the T7 Quick High Yield RNA Synthesis kit (New England Biolabs, Ipswich, MA) and incubated at 37 °C for 16 h. DNA template was removed with RNase Free DNase at 37 °C for 15 min. After purification with the RNA clean & concentrator-5 (Zymo Research, Irvine, CA), gRNA was analyzed on a 2.5% agarose gel for integrity and diluted to 250 μg/μl with RNase free water and stored at − 80 °C. Cas9 protein was diluted to 500 nM in Hepes (20 mM HEPES, pH7.5; 150 mM KCl) and stored at − 80 °C. Embryos were co-injected with 50 pg gRNA and 300 pg Cas9 protein at the one to two cell stage using a microinjector apparatus FemtoJet express (Eppendorf AG, Hamburg, Germany). Injected embryos were screened for gRNA activity using the T7 endonuclease assay (New England Biolabs, Ipswich, MA). Ten embryos from each gRNA injection were pooled at 48 hpf and genomic DNA extracted with 50 mM NaOH at 95 °C for 30 min. M13-and PIG-tailed primers (IDT, Leuven, Belgium) were used to amplify a region surrounding the mutated site of each locus using 1 × buffer, 2.5 mM MgCl 2 , 0.2 mM dNTP, 0.2 μM primers, 1U Taq polymerase (Promega, Fitchburg, WI). The polymerase chain reaction (PCR) was separated on a 1% agarose gel with the QIAquick Gel Extraction Kit (Quiagen, Hilden, Germany), and then 200 ng of the purified PCR product was dissociated and reannealed (95 °C for 5 min, 95-85 °C at − 2 °C /s, 85-25 °C at 0.1 °C /s) in a reaction containing 1 × NEB buffer 2 (New England Biolabs, Ipswich, MA) and then digested with 5U T7 endonuclease I (New England Biolabs, Ipswich, MA) for one hour at 37 °C. Fragments were analyzed on a 2% agarose gel. The remaining embryos were raised to adulthood and outcrossed with AB wild-type fish. Sixteen embryos from each outcrossed pair were screened for mutations in the F1 generation using a three-primer fluorescence PCR method. A 300-450 bp region surrounding the target site was amplified using forward primers linked with a M13 sequence and a PIG-tailed reverse primer in combination with a generic M13-FAM primer. The appa C21_16 mutants, refer to as appa −/− , carry a deletion of 10 bp in exon 2. Sanger sequencing with BigDye Terminator v1.1 Cycle Sequencing Kit (Applied Biosystems, Waltham, MA) on an ABI3130xl sequencer (SeqGen Inc, Los Angeles, CA) revealed a deletion of ten nucleotides in exon 2 that likely introduce a frameshift mutations. Heterozygous mutant carriers were raised and subsequently outcrossed into the wild-type AB fish line until generation F4. Outcrossed adults were genotyped using M13-FAM primers and PCR reactions diluted in HiDi formamide (Applied Biosystems, Waltham, MA) with ROX 500 dye size ladder (Thermo Fisher Scientific, Waltham, MA) and analyzed for amplified fragment length polymorphism (AFLP) on an ABI3130xl sequencer. Offspring from heterozygous F4 inbreeds were inbred to generate homozygous wild-type and mutant lines. Generation of appa −/− appb −/− double mutants were obtain from mating single mutant appa −/− with single mutant appb −/− . Protein sequence alignment. Sequences of APP were obtained from the UniProt database 86  Immunostaining of human brain sections. Neurologically normal human post-mortem control tissue was obtained from Queen Square Brain Bank for Neurological Studies. Paraffin-embedded sections were cut from caudate nucleus brain region, which contains ependymal lining containing cilia. Sections were dewaxed www.nature.com/scientificreports/ in three changes of xylene and rehydrated using graded alcohols. Endogenous peroxidase activity was blocked using 0.3% H 2 O 2 in MeOH for 10 min followed by pressure cooker pre-treatment for 10 min in citrate buffer, pH 6.0. Non-specific binding was blocked using 10% non-fat dried milk ( . cDNA values from each sample was normalized with average C T 's of house-keeping genes (eef1a1l1 and actb1), then the relative quantity was determined using the ΔΔC T method 90 with the sample of wild-type sibling embryos (24 hpf) as the calibrator. TaqMan Gene Expression Assays (Applied Biosystems, Waltham, MA) were used for the following genes: amyloid beta (A4) precursor protein A (appa) (Dr03144365_ m1), eukaryotic translation elongation factor 1 alpha 1, like 1 (eef1a1l1) (Dr03432748_m1) and actin, beta 1 (actb1) (Dr03432610_m1).

Cilium length measurement in zebrafish larvae.
To compare the number of brain ependymal cilia and their length, 30 hpf AB wild-type and appa −/− appb −/− zebrafish larvae were used. The larvae were treated with PTU, fixed in 4% PFA and the immunostaining with antibody against acetylated tubulin was performed as describe in the section above. Stacks (of around 25 μm depending on the angle of the mounted sample) were taken in the region of interest (ROI) of the dorsal portion of the diencephalic ventricle using Zeiss LSM710 confocal microscope using inverted 40 × water immersion objective (Plan-Apochromat 40×/1.0). Images were then processed using Imaris (Bitplane, Belfast, United Kingdom) and the cilium length was measured with the acetylated tubulin signal using the ¨measuring points¨ tool of the program. Raw data of the measurement were exported to Microsoft Excel and compiled into GraphPad Prism 7 for statistical analysis.

Brain ventricles injection and size measurement.
To measure the size of the brain ventricles in live zebrafish, 2dpf PTU-treated zebrafish larvae were used. Rhodamine-Dextran injection protocol was performed as describe by Gutzman and Sive 91 . Briefly, the larvae were anesthetized with tricaine in the EM and transferred onto a Petri dish covered with 1% agarose, lined with rows moulded. The larvae were kept in EM complemented with tricaine during the whole procedure and place on a ventral position, with top of their head facing upwards. Injections were performed using borosilicate injection needles previously pulled (P-97 Flaming/Brown micropipette puller) (Sutter Instrument, Novato, CA). Using a microinjector apparatus, 2 nl of Rhodamine B www.nature.com/scientificreports/ isothiocyanate-Dextran (Sigma, St. Louis, MO) were injected in the hindbrain ventricle without perforating or hitting the brain tissue below. Larvae with non-effective injections were sorted out using a fluorescent stereomicroscope (Nikon Instruments, Melville, NY). Quickly after the sorting, the larvae were mounted in 1% low-melting point agarose on glass bottom 35 mm Petri dish. Confocal imaging stacks were acquired using an inverted Nikon A1 confocal system using a 20 × objective (Plan-Apochromat 20×/0,75). Image processing of the confocal stacks were done with Imaris program. The "surface" tool option was used for each sample. Data of the surface volume and area were automatically generated by the program. Length measurements of the areas of the ventricles were obtain manually with the "measuring tool". All data were exported into Microsoft Excel and GraphPad 7 Prism for statistical analysis.
Transmission electron microscopy. To evaluate the integrity of the internal structure of the axonemes and microtubules doublets of the brain motile cilia in older zebrafish, transmission electron microscopy was performed on fixed brains. Adult zebrafish were euthanized in tricaine and brains dissected, rinsed in ice-cold PBS and fixed in 2% PFA and 2% glutaraldehyde (Sigma, St. Louis, MO), in 0.042 M Millonig buffer (0.081 M Na 2 HPO 4 , 0.0183 M NaH 2 PO 4 , 0.086 M NaCl) pH 7.4 at least 24 h at 4 °C. After fixation, brains were cut in two halves and then treated in 2% osmium tetroxide (Sigma, St. Louis, MO) in 0.1 M Millonig buffer pH 7.4. Specimens were then rinsed and incubated overnight in 4% sucrose solution in 0.1 M Millonig buffer pH 7.4 after which they were dehydrated in series of ethanol and embedded in a mix of acetone and agar 100 resin plastic (TAAB Laboratories Equipment Ltd, Berks, United Kingdom) and allowed to polymerize for 48 h. Blocks were trimmed as semi-thin (1 μm) and ultra-thin (70 nm) sections collected with a commercial ultramicrotome (Leica EM UC7, Leica Microsystems, Wetzlar, Germany). Sections were post-stained with 5% uranyl acetate in distilled H 2 O during 40-60 min, rinsed in distilled H 2 O and then treated with 0.3% Lead Citrate (Thermo Fisher Scientific, Waltham, MA) for 30-60 s. Images were acquired using secondary electron detection. Images were acquired with a Tecnai Spirit BT transmission electron microscope (Field Electron and Ion Company, Hillsboro, OR).

Data availability
The datasets generated during the current study are available in the supplementary data file.