OTX2 Homeoprotein Functions in Adult Choroid Plexus

The choroid plexus is an important blood barrier that secretes cerebrospinal fluid, which essential for embryonic brain development and adult brain homeostasis. The OTX2 homeoprotein is a transcription factor that is critical for choroid plexus development and remains highly expressed in adult choroid plexus. Through RNA sequencing analyses of constitutive and conditional knockdown adult mouse models, we reveal putative functional roles for OTX2 in adult choroid plexus function, including cell signaling and adhesion, and show that OTX2 regulates the expression of factors that are secreted into the cerebrospinal fluid, notably transthyretin. We also show that Otx2 expression impacts choroid plexus immune and stress responses, and affects splicing, leading to changes in the mRNA isoforms of proteins that are implicated in the oxidative stress response and DNA repair. Through mass spectrometry analysis of OTX2 protein partners in the choroid plexus, and in known non-cell-autonomous target regions, such as the visual cortex and subventricular zone, we identify putative targets that are involved in cell adhesion, chromatin structure, and RNA processing. Thus, OTX2 retains important roles for regulating choroid plexus function and brain homeostasis throughout life.


Introduction
The choroid plexus (ChP) epithelium is located in the brain ventricles and secretes cerebrospinal fluid (CSF) containing molecules that regulate embryonic brain development and adult brain homeostasis [1]. The ventricular system includes the two lateral ventricles (LVs) in each cerebral hemisphere, the central third ventricle of the forebrain diencephalon, and the central fourth ventricle (4V) in the hindbrain. This system is interconnected, allowing for CSF flow throughout, and is also connected via the 4V with the central canal of the spinal cord. The OTX2 homeoprotein transcription factor is critical for ChP embryonic development and functions [2]. Interestingly, temporal and spatial heterogeneity is evident, as the role of OTX2 evolves during development and differs between ChPs. For example, in late embryonic development, OTX2 is required for the maintenance of the 4V ChP, but not LV ChP [2]. Indeed, embryonic LV and 4V ChP show distinct gene expression patterns [3,4], suggesting different signaling properties. In the adult, OTX2 is still strongly expressed in the ChP [5], but its role has not been thoroughly investigated.
Homeoproteins are transcription factors that are important for embryonic development, and adult homeostasis and cell survival, and several homeoproteins have functions beyond transcription, including translation regulation, DNA repair, and signal transduction [6][7][8]. While several studies have explored the molecular partners and transcriptional targets of OTX2, they were typically restricted to embryonic contexts [9][10][11][12][13]. In the adult mouse, recent analyses of OTX2 protein and DNA targets have focused on the retina [14,15], visual cortex [16,17], and ventral tegmental area [18]. These studies revealed targets that are implicated in transcription, epigenetics, signal transduction, and homeostasis, and confirmed that OTX2 not only binds multiple sites across DNA, but also interacts with the machinery for RNA processing, export, and translation. To examine the role of OTX2 in adult ChP, we use a mouse model for constitutive heterozygous Otx2 knockdown, and a model for the ChP-specific conditional knockdown of Otx2. Through transcriptomic analysis of LV and 4V ChPs, we reveal dysregulation of cell adhesion and membrane proteins, secreted factors, signaling factors, immune response, and oxidative stress response. Through mass spectrometry analysis of OTX2 partners in ChP and non-cell-autonomous target regions [19,20], including the ventricular-subventricular zone (SVZ), rostral migratory stream (RMS), and visual cortex (VCx), we identified putative targets that are involved in cell adhesion, chromatin structure, and RNA processing. We also performed splice variant analysis and confirmed, by acute viral shRNA-Otx2 knockdown in the ChP of adult wildtype mice, that OTX2 can regulate the isoform distribution of genes involved in stress response and DNA repair. Taken together, our findings suggest that OTX2 has direct roles in ChP signaling, barrier, and surveillance functions.

Conditional and Constitutive Knockdown of Otx2 in Adult ChP
OTX2 is a key regulator of ChP and brain development, but its role in adult ChP is not well known. To gain an insight into its adult ChP functions, we performed RNA sequencing analysis with two mouse models. The first consisted of 3-month-old Otx2 lox/lox mice for the conditional knockdown of Otx2, specifically in the ChP, through intracerebroventricular (icv) injections of Cre-Tat recombinant protein, which leads to a~50% reduction in (mRNA) Otx2 and a >70% reduction in OTX2 protein levels [5]. The ChPs from LV and 4V were dissected separately from both Cre-Tat-injected (Cre + Otx2 lox/lox ) and control vehicle-injected mice (Veh + Otx2 lox/lox ). While the bilateral stereotaxic injections of vehicle or Cre-Tat are performed only in the lateral ventricles, we have previously shown, with this protocol, that the level of Otx2 knockdown in 4V ChP is proportional to that in LV ChP [19]. Indeed, we found a 48% decrease in (mRNA) Otx2 (exon 2) in the LV ChP (vehicle, 7924 mean reads; Cre-Tat 4143 mean reads), and a 39% decrease in 4V ChP (vehicle, 6898 mean reads; Cre-Tat 4169 mean reads). The second model consisted of Otx2 +/GFP mice as a constitutive heterozygous knockout mutant, with~50% Otx2 protein levels compared to the wildtype [21]. For this model, given the significant overlap in gene expression changes in the LV and 4V of Cre + Otx2 lox/lox mice (see below), only the 4V ChPs were dissected and pooled from 3-month-old wildtype and mutant mice. We found a 38% decrease in (mRNA) Otx2 (exon 2) in 4V ChP (wildtype, 6399 mean reads; Otx2 +/GFP 3957 mean reads).
The transcriptomics analysis of adult ChP showed highly expressed genes that are involved in energy metabolism, protein signaling, solute transport, cell adhesion, the cytoskeleton, and chaperone activity (Table 1). While not in the same order of gene expression level, this list compares favorably with those obtained from other ChP transcriptomics studies [4,22,23]. The conditional adult mouse knockdown of Otx2 led to significant changes in the expression of 375 genes in LV ChP and 808 genes in 4V ChP (p-adj < 0.05). The top ten upregulated and downregulated genes show a range of functions, including solute transport, signaling, immune response, and trafficking (Table 2). While there is a significant overlap in the altered gene expression between the ChPs ( Figure 1A), the 4V ChP seems to be more susceptible to loss of Otx2 activity. The response to Otx2 knockdown results in a rather even distribution of upregulation (522 genes) and downregulation (392 genes) when grouping both 4V and LV ChPs. However, ontological analysis reveals that the downregulated genes in both ChPs show higher levels of enrichment in significantly altered classes, suggesting that upregulated genes have more broadly distributed functions ( Figure 1B,C). Interestingly, both ChPs have similar ontology enrichment in downregulated genes, indicating that Otx2 is generally important for the expression of membrane proteins, glycoproteins, signaling proteins, and cell adhesion proteins. While some of these functions are recapitulated in the upregulated genes, there is much more heterogeneity between the LV and 4V ChP. The LV ChP shows more immune response ontology, while the 4V ChP shows more signaling-related ontology. This suggests that conditional knockdown of Otx2 leads to altered ChP barrier function and ChP signaling, and can impact immune responses.  The constitutive heterozygote Otx2 +/GFP adult mice showed significant expression changes in 528 genes of the 4V ChP (p-adj < 0.05), which is comparatively less than for conditional Otx2 knockdown in the 4V ChP ( Figure 2A). Given that fewer genes are deregulated in this constitutive model, this suggests that compensatory mechanisms for countering reduced OTX2 levels were activated during development. The changes in gene expression were relatively balanced between upregulation (273 genes) and downregulation (255 genes), and ontology analysis revealed shared terms, including glycoprotein, signal, membrane-related, and secreted proteins ( Figure 2B). The upregulated genes are also enriched for cell adhesion and alternative splicing, while downregulated genes are enriched for trafficking and transport. This suggests that brain-wide and life-long knockdown of Otx2 leads to altered ChP signaling, barrier functions, and brain homeostasis.  We hypothesized that genes that were deregulated in both conditional and constitutive models could be either direct targets of OTX2 transcription regulation or targets of important OTX2-dependent pathways. Comparison of gene expression changes in 4V ChP of these two models revealed an overlap of more than 80 genes, in both the upregulated and downregulated repertoires ( Figure 2A). This represented about half of the identified expression changes in Otx2 +/GFP mice, but less than a third of the changes in the conditional model. When genes from LV ChP conditional Otx2 knockdown are included in the analysis, we identified 42 genes that are globally upregulated and 34 genes that are globally downregulated (Table 3). To determine whether this list contains direct OTX2 transcription targets, we compared it with OTX2 chromatin immunoprecipitation experiments that were previously performed in mouse embryonic brain [9] or adult retina [14]. However, we found almost no overlap, with only Ttr being a common target. This suggests that the transcription-related activity of OTX2 is different in the adult choroid plexus and/or that these deregulated genes are downstream targets of OTX2-dependent pathways. It will be necessary to perform ChIPseq or CUT&RUN analysis of adult choroid plexus, to distinguish between these possibilities. Taken together, our analysis identifies new potential functions for Otx2 in the adult brain. We found upregulation of immune factors, specifically in the conditional Otx2 loss-of-function model, and deregulation of genes involved in cellular adhesion, trafficking, signaling, and secretion, in both knockdown models, suggesting altered ChP function and disruption of the ChP barriers. Table 3. Choroid plexus genes with significant expression changes in both Otx2 knockdown models, including lateral ventricle and fourth ventricle from Cre + Otx2 lox/lox mice and fourth ventricle from Otx2 +/GFP mice. Genes with * are also deregulated upon embryonic Otx2 knockdown in the hindbrain choroid plexus [2].

Altered Expression of ChP Secreted Factors
Given that our various ontology analyses often evoked secreted factors, we focused on ChP factors secreted in CSF and implicated in embryonic and/or adult neurogenesis (Table 4), which is one of the identified functions of adult ChP [19,23,24]. The factors implicated in embryonic neurogenesis include SHH, BMPs, and WNTs [25]. While Shh expression was not observed (mean reads < 1) in either ChPs of the wildtype mice, which is consistent with published data [3], the 4V ChP of Otx2 +/GFP mice (but not the ChPs of conditional Otx2 ChP knockdown mice) showed a significant increase in Shh expression. Between the various Bmp and Wnt family genes, only Bmp7 and Wnt2b were differentially expressed in Cre + Otx2 lox/lox mice, as compared to Veh + Otx2 lox/lox mice. Canonical Wnt signaling is perturbed in embryos with Otx2 4V ChP knockdown, which was attributed to the dysregulation of Wnt modulators, including Rspo1, Sfrp2, Sostdc1, Tgm2, and Wnt4, and to the increased levels of WNT4 and TGM2 in the CSF of mutant mice [2]. While Rspo1, Sfrp2, and Wnt4 were very poorly expressed and unchanged in both the LV and 4V ChP of Cre + Otx2 lox/lox adult mice, the expression of Sfrp1, Sostdc1, and Tgm2 were significantly changed in 4V ChP (Table 4), suggesting adult OTX2 retains some embryonic functions. To further explore this hypothesis, we compared our RNA sequencing analysis of 4V ChP Cre + Otx2 lox/lox to the previous microarray analysis of embryonic Otx2 knockdown, specifically in the hindbrain ChP, performed by Götz and colleagues [2]. They found 340 significantly (FDR < 10%, >2-fold change) expressed genes, with 135 genes upregulated and 225 genes downregulated. Compared to our adult knockdown, there was less than a 20% overlap with upregulated genes (24 of 135 genes) and an almost 30% overlap with downregulated genes (62 of 225 genes). Some of these genes are found among the top ten dysregulated 4V genes (Table 2) and the globally altered genes (Table 3), and they have functions related to cell adhesion, trafficking, and secretion. Given that Otx2 knockdown experiments in late embryonic development showed that OTX2 is necessary for 4V, but not LV, ChP maintenance [2], our results suggest that adult Otx2 expression could retain this maintenance function in 4V ChP.
We have previously shown that OTX2 that is secreted into the CSF from the ChP, can regulate adult neurogenesis non-cell autonomously, by transferring into the astrocytes in the SVZ and RMS, thereby affecting neuroblast migration [19]. This study also showed that ChP Otx2 knockdown in Cre + Otx2 lox/lox adult mice, which will impact both cell-and noncell-autonomous activity, also led to significantly reduced SVZ neurogenesis, suggesting that OTX2 may regulate secreted factors that are implicated in adult neurogenesis, through cell-autonomous effects in the ChP. IGF2 and SLIT1/2 regulate both embryonic and adult neurogenesis [24,[26][27][28]. While Slit1 is not expressed and Slit3 is only weakly expressed in adult ChP, Slit2 is highly expressed, but shows no significant change in expression in the ChP, with reduced Otx2. Admittedly, there is a trend towards increased Slit2 expression and we cannot exclude the potential for biological relevance. Igf2 was significantly downregulated, more than 2-fold, in all the ChPs, upon Otx2 knockdown. However, there was a concomitant downregulation in Igfbp2, which can inhibit IGF2, suggesting that the level of IGF2 activity could be maintained through the compensatory reduction in inhibiting factors. Other factors influencing adult neurogenesis include amphiregulin (AREG) [29], FGF2 [30][31][32], and TGF-α [33], yet we found no significant change in their expression (and no detectible expression of Areg). Finally, other factors show more change in gene expression after acute Otx2 knockdown compared to constitutive knockdown. TGF-β negatively regulates adult neurogenesis [34], and Tgf-β2 is downregulated in both the ChPs of Cre + Otx2 lox/lox mice (Table 4). Taken together, these minimal or compensatory changes in specific secreted signaling factors suggest that Otx2 expression in the ChP could have only a minor cell-autonomous role in regulating adult neurogenesis. This hypothesis is consistent with similar levels of decrease in adult neurogenesis, observed with both this ChP Otx2 knockdown model and the non-cell-autonomous-only OTX2 knockdown mouse model [19].

Altered Expression of Immune and Stress Factors
Given the altered expression of homeostasis and stress response-related factors in both the ChPs of conditional Otx2 knockdown mice, we turned to the viral expression of shRNA-Otx2 in LV and 4V ChPs. Intracerebroventricular-injected AAV5 results in ChPspecific expression [35,36], and provides a tool to acutely affect Otx2 expression in any mouse model. We validated this model by qPCR analysis, which showed a 69% decrease in (mRNA) Otx2 and a concomitant very large decrease in the expression of a known direct transcriptional target, transthyretin (Ttr) ( Figure 3A). Comparing models, the decrease in (mRNA) Ttr in Cre + Otx2 lox/lox mice was 45%, while it was 87% in the shRNA-Otx2 mice, which suggests that viral knockdown provides a more robust effect. TTR, the most highly expressed protein in ChP (Table 1), is secreted into CSF and transports thyroxin and retinolbinding protein, and has a role in regulating cognition and memory, psychological health, and emotions (for a recent review, see [37]), suggesting that OTX2 levels can potentially impact similar brain functions. Furthermore, the downregulation of aquaporin 1 (Aqp1) (Table 3, Figure 3A) was also confirmed, with Aqp4 as a negative control, suggesting that OTX2 can also regulate CSF water homeostasis. In keeping with roles in brain homeostasis and surveillance, we also chose targets from ontology analysis (Figure 1), in functions related to oxidative stress, immune response, and metal ion transport. A surprising finding was the over 100-fold increase in glutathione peroxidase 3 (Gpx3), an extracellular enzyme that catalyzes the reduction in peroxidases and protects cells from oxidative damage, suggesting that a loss of OTX2 has a significant impact on cell physiology ( Figure 3B). Other compensatory mechanisms against oxidative stress include decreased fatty acid oxidation (Scd1), increased peroxisome function (Acox2, Ddo), for countering oxidative stress and inflammation [38], and changes in the structural cell response (Vim) [39] (Figure 3C). Concerning the immune response ( Figure 3D), we tested a complement activation factor (CD55), an inflammatory response chemokine (Ccl9), and an innate immune response factor (Iigp1). The direction of change in the expression of all of these factors, upon acute viral Otx2 knockdown, was consistent with the constitutive and conditional mouse models. It remains unclear whether the loss of OTX2 provokes oxidative stress, and thus indirect activation of genes such as Gpx3, or whether OTX2 regulates the genes involved in reactive oxygen species signaling and/or stress response. Finally, given their role in brain homeostasis, we also quantified factors related to metal ion transport (Steap1 and Slc31a) ( Figure 3E), which had altered expression in the conditional Cre + Otx2 lox/lox mice. Only Slc31a, which transports copper ion, had concomitant reduced expression upon Otx2 knockdown. These findings suggest that ChP function is greatly impacted by Otx2 expression level, opening the question of whether Otx2 overexpression in ChP would also deregulate homeostasis and illicit immune responses in a wildtype context or, on the contrary, rescue deficits in homeostasis in an aged or diseased animal.

Otx2 Protein Interactions
To further analyze OTX2 function in adult ChP, we performed several OTX2 coimmunoprecipitation (co-IP) experiments with mass spectrometry (MS) analysis, to identify potential protein partners. We previously discovered that OTX2 protein is secreted by the ChP into CSF, and accumulates non-cell autonomously in SVZ and RMS astrocytes [19] and VCx parvalbumin cells [5,40]. The identification of alternate protein partners in cellautonomous and non-cell-autonomous contexts would suggest that OTX2 takes on specific roles after transferring between cells. To test this hypothesis, and to reinforce ChP analysis, we also performed OTX2 co-IP on lysates from adult mouse SVZ, RMS, and VCx.
We used three criteria to obtain a list of potential OTX2 protein interactions in the four brain structures (Table 5), as follows: (i) unique proteins with three or more peptides identified exclusively in OTX2 compared to IgG co-IP samples (unique protein, ≥3 peptides); (ii) proteins identified with three or more peptides in OTX2 co-IP samples and having a relative peptide difference greater than 50% compared to IgG co-IP (selected protein, ≥50% rel. ∆); and (iii) all small proteins (≤25 kDa) exclusive to OTX2 co-IP samples, regardless of peptide number (unique small protein, ≤25 kDa), given that they have fewer identifiable MS peptides. These lists were used for comparison between structures and for ontology analysis. We generated a list of 60 high-confidence protein partners of OTX2 in ChP that were common to all three ChP samples (Table 6). These proteins cover several functions, including cell adhesion, cell trafficking, cell signaling, metabolism, RNA binding, RNA processing, transcription, chromatin structure, and DNA repair. Interestingly, more than 10% (eight proteins) belong to the "tier 1" proteins identified in stress granules [41], which are involved in translational control and post-transcriptional regulation. Although this functional class was not identified by KEGG pathway analysis (see below), this is likely due to the absence of annotation, given the only recent emergence of updated comprehensive inventories of stress granule proteins. Thus, we can only hypothesize that OTX2 interacts with these granules, although this putative function is given weight by the presence of the PAX1 homeoprotein among the "tier 1" proteins, by the in vivo interaction between the EMX2 homeoprotein with the translation initiation factor eIF4E [42], and by the involvement of the PROX1 homeoprotein in liquid-liquid phase separation [43], which also underlies stress granule assembly [41]. Also of note are the putative partners MECP2 and MOV10, given that OTX2 has been shown to regulate MECP2 foci in the postnatal mouse visual and auditory cortex [16], and that the EN1 homeoprotein is involved in LINE-1 regulation [44].  Few OTX2 partners have been biochemically and functionally validated. One key partner during embryogenesis is MEIS2, which is a co-activator for mesencephalon specification [45]. Meis2 is expressed in ChP at low levels (170 mean reads in 4V ChP; 31 mean reads in LV ChP), as compared to OTX2 (5954 mean reads in 4V ChP and 6829 mean reads in LV ChP), and appears not to be a major partner of OTX2 in ChP. TLE4 is another identified protein partner of OTX2 during development, and allows repression of mesencephalon fate [46]. Despite its expression in ChP (798 mean reads in 4V ChP and 689 mean reads in LV ChP), TLE4 was not identified in our OTX2 co-IP, although this could be due to TLE4 being under the limit of detection or having peptides that are too hydrophobic or hydrophilic for MS detection. The potential absence of TLE4 suggests that OTX2 protein interactions depend strongly on the cell type and developmental context.
To identify novel OTX2 protein partners that are ubiquitous throughout the brain, we compared the lists from the four brain structures, as follows: ChP (pooled LV and 4V), SVZ, RMS, and VCx. Few high-confidence proteins (selected protein ≥ 50% rel. ∆), 14 in total, were common to the three non-cell-autonomous structures (Table 7). Of these 14 common proteins, 5 were also identified in the ChP. Interestingly, these top-ranked proteins include FIG4, VAC14, and PIKFYVE, which play a role in phosphatidylinositol (3,5)bisphosphate [PI(3,5)P2] regulation, in multivesicular body (MVB) biogenesis, and in endosome autophagy and trafficking [47], suggesting that OTX2 plays a role in vesicle transport or is carried via MVBs. Given that MVBs can give rise to extracellular vesicles, interaction with OTX2 may reinforce its role in regulating the pathways of extracellular protein expression that are identified in our RNA sequencing analysis. To identify potential differences between cell-autonomous and non-cell-autonomous partners, we performed KEGG pathway analysis on OTX2 protein partners, for all structures individually (Table 8). No dramatic differences were found between the structures, suggesting conserved roles of OTX2 in cell-autonomous and non-cell-autonomous OTX2 target structures. Common to nearly all structures are metabolic pathways, RNA transport, oxidative phosphorylation, RNA processing, and spliceosome. Pathways that are specific to the ChP pertain to the maintenance of tight junctions, protein processing, and actin cytoskeleton regulation. Downregulation of the cell adhesion class was also identified in the RNAseq analysis of conditional Cre + Otx2-lox mice (Figure 1), suggesting a direct involvement of OTX2 both in gene regulation and cellular functions for cell-autonomous ChP maintenance. Of the 14 proteins in common between VCx, SVZ, and RMS, eight of them are involved in RNA processing, suggesting a novel function for OTX2. Although the spliceosome pathway was also enriched in ChP, these proteins stand out for their involvement in the U5 snRNP complex, exon junction complex, or mRNA export complex, whereas the spliceosome proteins identified in the ChP are either splicing co-factors, part of the SMN complex, or part of the U2 snRNP complex. Interestingly, OTX2 has been shown to bind the initiation factor eIF4e in GST pull-down experiments [42], while other homeoproteins have been shown to bind translation machinery [6,42,48] that is implicated in RNA export, transport, and translation. Taking the high-confidence partners together with KEGG pathway analysis, cell-autonomous OTX2 is likely implicated in the regulation of genomic landscape, the regulation and processing of RNA, the trafficking of signals, and the maintenance of cellular adhesion, while non-cell-autonomous OTX2 is more implicated in the processing of RNA.

Splice Variant Analysis
Given the high confidence of OTX2 interaction with spliceosome pathway proteins, we extended the analysis of our transcriptomic data of LV ChP from Cre + Otx2 lox/lox mice to measure changes in splice variants. The isoform usage was found to be significantly changed in the coding transcripts for only four genes (Mcrs, Ldlr, Tspan12, and Daxx), and generally for only two isoforms among the splice variants ( Figure 4A,B). These genes showed no change in overall expression upon Otx2 knockdown ( Figure 4C). Through acute Otx2 knockdown by the viral expression of shRNA-Otx2, we confirmed a significant increase in the expression of only the Mcrs-209 and Daxx-204 isoforms, as other isoforms either did not change significantly or changed in the opposite direction ( Figure 4D). Interestingly, MCRS and DAXX interact within a protein complex with various nuclear functions, including transcription regulation, chromatin remodeling, and DNA repair. Further research is needed to determine the functional consequences of these changes in the distribution of transcript isoforms.  Homeoproteins have been postulated to regulate transcript splicing. The PAX6 homeoprotein has been shown to modulate splicing machinery, such that changes in Pax6 expression alter the population of tenascin-C splice variants without changing the total tenascin-C expression [49], while the CDX2 homeoprotein interacts with splicing machinery [50]. Regarding OTX2, its protein interactome in the adult retina revealed putative RNA processing partners, such as SFPQ and U2AF [15], while in the ChP and non-cellautonomous structures, we found the potential partners ACIN1, DDX41, DDX46, HNRPLL, and PRRC2A, RBM25, RBM39, SF3A1, SF3B1, SNRNP200, and SRRM2 (Tables 6 and 7). It remains to be determined whether OTX2 controls ChP splicing activity through direct interaction with splicing factors and/or by regulating their expression.

Animal Ethics Statement
All animal procedures, including housing, were carried out in accordance with the recommendations of the European Economic Community (2010/63/UE) and the French National Committee (2013/118). For surgical procedures, animals were anesthetized with xylazine (Rompun 2%, 5 mg/kg) and ketamine (Imalgene 500, 80 mg/kg). For biochemical analysis, mice either underwent transcardial perfusion or were sacrificed by cervical elongation.

Quantitative PCR Analysis
Total RNA from LV and 4V ChPs was extracted by using the RNeasy lipid tissue mini kit (Qiagen, Courtaboeuf, France) with DNA removal. Total RNA (10 to 20 ng) was retrotranscribed by using the QuantiTect reverse transcription kit (Qiagen, Courtaboeuf, France). Quantitative PCR (qPCR) analyses of cDNA (diluted at 1/10) were performed in triplicate with a LightCycler 480 II (Roche, Meylan, France) using the SYBR Green I master mix (Roche, Meylan, France). Gene-to-Hprt or gene-to-Gapdh ratios were determined by the 2 −∆∆Ct method. For Otx2 expression analysis, expression was compared to mean expression of vehicle-injected mice of the same experiment.

RNA Sequencing Analysis
For analysis of conditional knockdown mice, the RNA was extracted separately from LV and 4V ChPs of Cre-Tat and vehicle-injected mice. A small sample of each ChP was tested by qPCR to ensure Cre-Tat samples had less than 50% Otx2 expression on average compared to control mice. Duplicate samples were prepared by pooling ChP lysates from 2 × 5 Cre-Tat-injected mice and from 2 × 4 vehicle-injected mice. For analysis of constitutive knockout mice, the RNA was extracted from pooled 4V ChPs of four 3-month-old Otx2 +/GFP and five wildtype littermates, with duplicate samples (n = 2) of each genotype. PolyA+ mRNA purification, mRNA sequencing with technical replicates, and data normalization and quantification was performed by the Genomic Paris Center (IBENS, Paris, France) using Illumina HiSeq 1500 (Illumina, Evry, France).

Protein Co-Immunoprecipitation
For each ChP co-IP experiment, ChP from LV and 4V were pooled from four 3month-old mice and were lysed with 1 mL lysis buffer (100 mM Tris pH 7.5, 1 mM EDTA, 100 mM NaCl, 1% NP40, 1 mM MgCl 2 , 1X protease/phosphatase inhibitor (Roche, Meylan, France)) containing 1µL of benzonase (Roche, Meylan, France). ChP were dissociated using 26G syringe and incubated 30 min on ice. Tubes were centrifuged at 21,000× g for 10 min and supernatant was recovered and divided in two. Each half was incubated with 44 µg of either anti-OTX2 (ab21990, Abcam, Paris, France) or anti-IgG (ab27478, Abcam, Paris, France) antibodies coupled to magnetic beads (10 mg/mL with 9.5 µg of antibody per mg of beads, Dynabeads antibody coupling kit, Invitrogen, Vilnius, Lithuania) in lysis buffer at 4 • C on rotating wheel overnight. Using magnetic separation, beads were washed 5 times in 1 mL of cold lysis buffer. Pelleted beads were eluted in 20 µL of laemmli buffer by heating 5 min at 95 • C, and then stored at −20 • C.

Mass Spectrometry Analysis
Proteomics analyses were performed by the Protein Mass Spectrometry Laboratory (Institut Curie, Paris, France). Eluted samples in laemmli were processed and resulting peptides were analyzed by nano-LC-MS/MS using an Ultimate 3000 system (Dionex, Thermo Fisher Scientific, Paris, France) coupled to an Orbitrap Fusion mass spectrometer (Q-OT-qIT, Thermo Fisher Scientific, Paris, France). Data were acquired using Xcalibur software and the resulting Mascot files (v2.5.1) were further processed by using myProMS software (v3.9) [63]. Percolator [64] was used for FDR calculations set to 1% peptide level. For ChP proteomics, three experiments were performed (N = 3). For SVZ, RMS and VCx proteomics, one experiment was performed (N = 1).

Ontology Analysis
Genes with >10 mean reads in at least one of the ChP samples were selected for ontology analysis. Differentially expressed gene lists were generated by using threshold of p-adj < 0.05. Ontology term enrichment and KEGG pathways were analyzed with DAVID Bioinformatic resource v6.7 [65,66] and ontology terms were plotted as -log 10 scale of the enrichment p-values. UniProt (access date 21 August 2020, http://www.uniprot.org) was used for obtaining func-tional classes (Tables 1, 3, 6 and 7). Gene list comparisons and Venn diagram data were generated with web-based tools (http://www.bioinformatics.lu/venn.php).

Conclusions
The ChP has barrier functions for controlling what gets in and out of the brain, and homeostasis functions for controlling brain metabolites in the CSF. The data in this present study allow us to go a step further, by imparting important ChP endocrine functions that are putatively regulated by the cell-autonomous and non-cell-autonomous activities of OTX2. Indeed, in addition to regulating the expression and post-transcriptional modification of genes encoding signaling and hormone-transport proteins that are secreted into the CSF, such as Igf2 and Ttr, OTX2 itself is secreted by the ChP and exerts essential non-cellautonomous activities, such as the regulation of cerebral cortex plasticity or that of adult neurogenesis [5,19,67]. Transcriptomic analysis of different genetic Otx2 loss-of-function models, including conditional knockdown, specifically in the ChP, coupled with proteomic analysis, are the first steps towards a better understanding of the molecular biology of this traveling transcription factor in and out of its main cerebral source.