Ependymal cells-CSF flow regulates stress-induced depression

Major depressive disorder (MDD) is a severe, common mood disorder. While reduced cerebrospinal fluid (CSF) flow adversely affects brain metabolism and fluid balance in the aging population and during development, only indirect evidence links aberrant CSF circulation with many diseases including neurological, neurodegenerative, and psychiatric disorders, such as anxiety and depression. Here we show a very high concentration of p11 as a key molecular determinant for depression in ependymal cells, which is significantly decreased in patients with MDD, and in two mouse models of depression induced by chronic stress, such as restraint and social isolation. The loss of p11 in ependymal cells causes disoriented ependymal planar cell polarity (PCP), reduced CSF flow, and depression-like and anxiety-like behaviors. p11 intrinsically controls PCP core genes, which mediates CSF flow. Viral expression of p11 in ependymal cells specifically rescues the pathophysiological and behavioral deficits caused by loss of p11. Taken together, our results identify a new role and a key molecular determinant for ependymal cell-driven CSF flow in mood disorders and suggest a novel strategy for development of treatments for stress-associated neurological, neurodegenerative, and psychiatric disorders.

imipramine, fluoxetine and escitalopram were administered daily intraperitoneal injection (i.p.) injection (20 mg/kg/day) for 2 weeks. Imipramine, fluoxetine and escitalopram were purchased from Sigma-Aldrich. Fluoxetine and escitalopram were dissolved in dimethylsulfoxide (DMSO) and then diluted in saline. Imipramine was dissolved in saline. Each drug was finally diluted in 100 µl of 0.9% saline and administered at the dose indicated. Control groups were administered saline.

BacTRAP Translational profiling
Two or three male Dcdc2a-EGFP/L10a mice and Dcdcd2a-EGFP-L10a crossed with p11KO mice were used for independent TRAP replicates. The TRAP procedure was performed as described previously 31,32 . In brief, tissues that included lateral ventricles were dissected from individual mice. Tissue was immediately homogenized with a motor-driven Teflon glass homogenizer. Polyribosomes were immunoprecipitated by two monoclonal anti-EGFP antibodies (19C8 and 19F7) with coated protein L magnetic beads. RNAs from polyribosomes were extracted and further purified with a Rneasy Plus Micro Kit. RNA quantity and quality were determined with an Agilent 2100 Bioanalyzer. cDNA was synthesized from 5 ng of mRNA from IP and input samples and further amplified using a Oviation RNA-seq Kit. cDNA fragments of 200 bp were end-repaired and ligated with adapters for HiSeq 2000 (Ilumina Inc., San Diego, CA, USA) technology using TruSeq Nano DNA Sample kit (Illumina). Quality of libraries was assessed using a HT DNA High Sensitivity Chip (Agilent) and a 2100 Bioanalyzer.
RNA-seq reads were aligned to the UCSC mm10 reference genome using STAR (version 2.3.0e_r291). Aligned reads were quantified by htseq-count module, part of the 'HTSeq' framework (version 0.6.0). Differentially expressed genes were identified by performing a negative binomial test using DESeq2 (R-package version 1.4.5) with default settings. Significant p values were corrected to control for the false discovery rate of multiple testing at 0.05 threshold. All TRAP-seq data were subjected to the work flow described in https://gensat.rockefeller.edu/heintzp30/Bioinformatics_Flow_Chart.jsp

Quantitative RT-PCR
As described previously 10 , 10 ng of cDNA was used for each qPCR reaction and all samples were run in triplicate. Q-PCR was carried out using an Applied Biosystems 7900HT system.

4
Taqman assays were purchased from Applied Biosystems and performed as follows. Tagman assay ID for each gene; p11 (Mm00501457_m1). All data were normalized to TaqMan Rodent GAPDH Control, and relative expression levels between conditions were calculated by the comparative CT (2-ΔΔCT) method.

Magnetic resonance imaging (MRI)
Imaging. Imaging was performed on a 7.0 Tesla 70/30 Bruker Biospec small animal MRI system (Bruker Biospin, Billerica, MA) with 450 mT/m gradient amplitude and a 4500 T/m/s slew rate.
The animals were anesthetized with isoflurane in oxygen and fixed in the MRI using a nose cone and bite ring. A volume coil was used for transmission and a quadrature surface coil for reception.
Anatomy. Anatomical images were acquired using a T2 RARE sequence with a RARE factor of 16, 20 averages of 24 axial slices were acquired with a field of view of 20 × 20 mm, spatial resolution of 0.1 × 0.1 × 0.5 mm3, matrix size of 192 × 192, echo time TE of 9.4 ms, effective TE of 56 ms, and recovery time TR of 4.1 s. Fat and motion suppression parameters were used.
The covered volume consisted of the entire brain excepted the olfactory bulb. This resulted in a total scan time for the anatomy images of 16 min.
CSF Flow. Using a FLOWMAP phase contrast sequence, eight averages of nine axial slices, centered at the anatomical image volume, were acquired with the same field of view and in-plane resolution as the anatomical images, with slice thickness of 1.0 mm, TE = 5.64 ms, TR = 15.1 ms, and flip angle = 20 o . Velocity mapping with flow encoding in slice direction was used with a velocity encoding coefficient of 4 cm/s, directed away from the cerebellum. The total scan time was 7 min.
Analysis on slices of the cerebral aqueduct from cerebellum forwards where aqueduct looked more or less round and not yet inverted y shaped. Freehand ROIs on slightly smoothed images, tracing the darkened area, meaning negative flow, towards the cerebellum. The anatomical images of the all ventricles were quantified with Image J software (NIH).

Ultrastructural analysis
Electron microscopy and transmission electron microscopy. Mice were perfused in a fixative containing 2% paraformaldehyde and 2.5% glutaraldehyde (GA) in 0.075M sodium cacodylate buffer pH 7.4 and post-fixed overnight with the same fixative at 4°C. The mouse brain tissue, which is included the lateral ventricle was placed in a mold and sliced into 1 mm thickness. The selected slices with structure of interest were re-fixed with 2.5% GA and with 0.1% tannic acid for 1 hour and with 2.5% GA in the buffer overnight. The slices were post-fixed with 1% osmium tetra-oxide, 0.4% potassium ferrocyanide for 1 hour, followed by en bloc staining with 1% uranyl acetate overnight. Sections were subsequently dehydrated with a graded ethanol series, infiltrated with Eponate12 resin (Ted Pella) and embedded with the resin. Thin sections (70-nm thick) of the ventricular-subventricular zone (V-SVZ) of lateral ventricle were cut by an ultramicrotome (Ultracut E, Leica) and analyzed by JOEL 100CX TEM with a tungsten filament at 80 kV with a digital imaging system (XR41-C, Advantage Microscopy Technology).
Immuno-electron microscopy. Mice were perfused with a fixative, containing 4% paraformaldehyde in 0.075 M sodium cacodylate buffer pH 7.4, post-fixed overnight and then cut into 50-µm-thick sections using a vibrating microtome (Vibratome VT100, Leica). The lateral ventricle sections were processed for immunostaining with ABC method (VECTASTAIN® Elite® ABC-HRP Kit) by following the vendor's instruction and silver enhancement as described previously 47 . Subsequently, sections, which were included the V-SVZ of lateral ventricle, were processed for EM and analyzed by TEM as described above.
Scanning electron microscopy. Mice brains were fixed in a fixative, containing 2% paraformaldehyde and 2.5% glutaraldehyde in 0.075 M sodium cacodylate buffer pH 7.4. Tissue were dehydrated by a graded series of ethanol (50%, 75%, 95%, and 100% 3 times) and followed by a critical point dry (CPD), which was initiated in the ethanol-filled chamber and replaced with carbon dioxide liquid in the CPD chamber (Autosamdri A-815, Tousimis). 2 nm thickness conductive coating was applied over the dried samples with iridium to ~2 nm (ACE600, Leica).
The samples, which were included the V-SVZ of lateral ventricle, were examined under a SEM 6 (LEO 1550; Carl Zeiss) with a field-emission electron gun and operation/data acquisition software (Smart SEM version 5).
Analysis. TEM and SEM images were analyzed with Image J software (NIH) for the cilia morphology, such as cilia diameter (TEM), cilia number, length and direction (SEM) on the V-SVZ of LV ependymal cells and averaged on each group. To analysis cilia direction, eight radial lines, 45 degrees of each were draw from the middle part of cilia on each cell on SEM images.
The relative distribution of cilia within 45 degrees or the other 315 degrees were counted as the unidirectional or multidirectional cilia orientation, respectively, and were calculated as the percentage of cilia direction per cells and averaged on each group. All EM studies were conducted at The Rockefeller University Electron Microscopy Resource Center.

Immunohistochemistry
Mouse brains were perfused transcardially with cold PBS, followed by 4% paraformaldehyde (PFA) and post-fixed in the same solution overnight at 4°C, as described previously 10, 28 . The mouse brains were coronally cut into 40-µm-thick sections with a vibratome (VT 1000S, Leica).  Retrieval Reagent (Advanced Cell Diagnostics) for 5 min. Afterwards, the slides were washed two times with distilled water and dehydrated with 100% ethanol. Next, protease III Reagent (Advanced Cell Diagnostics) was applied for 30 min at 40°C. Human fresh frozen sections (14µm-thick), were post-fixed in 4% PFA for 15 min at 4°C, dehydrated in graded alcohols and exposed to Protease IV (Advanced Cell Diagnostics) for 30 min at room temperature.
Subsequently, both perfused mouse and fresh frozen human sections were hybridized with p11 probes (mouse, Mm-S100a10, 410901; human, Hs-S100a10, 506421; Advanced Cell Diagnostics) respectively, for 2 h at 40°C. The hybridization step was followed by standardized steps of amplification (Amp 1-FL 30 min at 40°C, Amp 2-FL 15 min at 40°C, Amp 3-FL 30 min at 40°C, Amp 4C-FL 15 min at 40°C). The last amplification step was followed by immunofluorescent staining with primary antibodies and Alexa-fluor-conjugated secondary antibodies. Then the slides were mounted with Dako fluorescent mounting medium (Agilent Technologies). Sections were imaged on a Carl Zeiss LSM 880 confocal microscope (Carl Zeiss AB) using a 63× oil immersion objective. Z-stacks of 7-10 µm thickness were obtained in each caption.

Cell intensity
The intensity of eGFP and p11-immunolabeled ependymal cells was quantified with ImageJ software (NIH). Three to five coronal sections that included lateral ventricles per mouse and human were quantified and averaged for each group. Fluorescence images for lateral ventricles from mice and humans, third ventricle and fourth ventricle from mice were acquired using a Zeiss LSM710 confocal microscope with a ×40/0.50 NA objective (45176.65 um 2 ; 262144 pixels). Background autofluorescence was accounted for by applying an equal cut-off threshold 8 to all images. All imaging and analyses were performed blind to the experimental conditions. Data were analyzed by ANOVA test or Student's t-tests and later graphed using Microsoft Excel or GraphPad Prism Software.

Viruses
As described previously 10 , for Cre-mediated recombination/inversion of the flanked p11 as DIO In the absence of Cre expression, the p11 or eYFP were not produced. In the presence of Cre expression, the transgene was FLEXed, leading to the expression of p11 or eYFP. The titers (genome copies per milliliter) of the AAVs were as follows: 2.16e 13 for AAV1-EF1a-DIO-eYFP-WPRE-hGH (AAV_eYFP) and 3.51e 12 for AAV1-EF1a-DIO-p11-WPRE-hGH (AAV_p11).

Stereotaxic surgery
All stereotaxic injections were carried out on an Angle Two stereotaxic frame for mouse with motorized nanoinjector (Leica). 10-week-old male mice were anesthetized with ketamine and xylazine and stereotaxically injected with AAV1-EF1a-DIO-eYFP-WPRE-hGH and AAV1-EF1a-DIO-p11-WPRE-hGH intracerebroventricularly (I.C.V., AP: 0.02 mm; ML: ± 0.89 mm; DV: -2.53 mm from bregma). The total injection volume was 1 µl. All injections were performed at a rate of 0.15 µl/min using Hamilton syringes (33 gauge) and the needle was kept in place for an additional 5 min. After 14 days of injection, depression-like behavioral tests and MRI were performed.

Behavioral assessments
All the behavioral tests were performed during the light cycle in a dedicated sound-proof behavioral facility by experimenters blind to treatment and genotype information, as described previously 10, 11 . Mice were brought to the testing room 30 min before the start of each behavioral test and remained in the same room through the test. At all times, sound was masked with 60-65 dB white noise.
Tail suspension test. Mice were suspended individually by their tails. The rod was fixed 50 cm above the surface of a table covered with a safety mat in a sound-isolated room. The tip of the tail was fixed using adhesive Scotch tape; the duration of the test was 5 min. The test session was videotaped and immobility scored by using automated TST/FST analysis software from Clever

Systems.
Forced swim test. Mice were placed in a glass cylinder (height: 30 cm, diameter: 16 cm) containing water at 24°C and a depth of 14 cm so that they could neither escape nor touch the bottom. Mice were forced to swim for 6 min. The animals were habituated for the first 1 min and behavior was monitored over the next 5 min. A 6 min test session was videotaped and immobility scored by using automated TST/FST analysis software from Clever Systems.

Novelty-suppressed feeding test.
After 24 hours food-deprivation (water was provided ad libitum), mice were tested in the NSF test, At the end of this time, a single 2 x 2 cm oval food

Bioinformatics
RNAseq analysis and geneset enrichment. Ependymal RNA-seq reads were aligned with STAR software 48 , and differential expression analysis was performed by EdgeR 49 . Gene set enrichment analysis was performed by PAGE (Parametric Analysis of Gene Set Enrichment) 50 and conducted on mouse Gene Ontology Biological Processes and expert-curated terms (only terms with 10-500 genes were included in the analysis). Annotated PCP genes were referred to a previous report 33 .  51,52 . Specifically, to represent lateral ventricular ependymal cell specific functions, lateral ventricular ependymal specific genes were determined by the union of genes highly expressed in lateral ventricular ependymal cells in data presented in this study, ependymal cell markers identified in previous study 52 , as well as expert curated markers; gene markers from the other major brain cell types (neuron, microglia, astrocyte, oligodendrocyte) were filtered out to ensure ependymal cell specificity. The network was evaluated by cross validation, as well as the ability to use network neighborhood to recapitulate the differential expression pattern of p11 knockout in this study. Mouse genes were mapped to human based on the combination of sequence and functional conservation using the Functional Knowledge Transfer (FKT) approach 53 .
Clustering the depression-associated ependymal cell network. In order to identify depressionassociated functional modules of the ependymal cells, we created a subset of the ependymal cellspecific functional network containing the top 1,000 depression-associated genes from our genome-wide ranking and all the edges between them. Then, we used an approach based on 11 shared k-nearest-neighbors (SKNN) and the Louvain community-finding algorithm to cluster the network into distinct modules of tightly connected genes. The SKNN-based strategy has the advantages of alleviating the effect of high-degree genes and accentuating local network structure by connecting genes that are likely to be functionally clustered together. Cytoscape used to layout the network in Supplementary Fig. S13. The entire table of enriched GO terms is provided in listed in Supplementary Table S7 depression-associated genes were clustered using a shared-nearest-neighbor-based communityfinding algorithm to elucidate several modules of genes. Eight of the clusters that contained 10 or more genes, labeled C1 through C8, were tested for functional enrichment using genes annotated to Gene Ontology biological process terms. Representative processes and pathways enriched within each cluster are presented here alongside the cluster label. The enriched functions provide a landscape of cellular functions potentially dysregulated by depression-associated ependymal genes from mice and humans. All processes are rained with FDR<0.05. number of cilia (e) from wild-type (WT; p11 f/f ) and ependymal cell conditional p11 knockout (p11 cKO; Tppp3-Cre x p11 f/f ) mice (n = 5 for b; n = 11 cells from 3 mice in each group for c-e).