Introduction
The brain is bathed in cerebrospinal fluid (CSF) that occupies the ventricular system, the subarachnoid space, and the interstitial space between structures and cells in the brain. The CSF serves to create buoyancy for the brain, to protect it from mechanical insult, and as a route by which metabolites, nutrients, and hormones can disperse within the brain [
1]. The CSF is produced at a rate of 500 ml per day in adult humans [
2], and the majority of the CSF secretion takes place across the choroid plexus [
3], which is a specialized secretory tissue located in each of the ventricles. The choroid plexus consists of a monolayer of tight junction-connected epithelial cells, which rest on highly vascularized stroma with connective tissue [
4].
A range of cerebral pathologies, i.e. hydrocephalus, stroke and subarachnoid hemorrhage, associate with elevated intracranial pressure (ICP). If left untreated, the brain tissue and the vasculature within compress, further reducing blood flow to the affected areas. Elevated ICP can occur following brain fluid accumulation arising either by reduced drainage of CSF following an obstruction in the brain fluid exit pathways or by hypersecretion of CSF, the latter of which has been observed in conditions such as choroid plexus hyperplasia, choroid plexus papilloma, and in a rodent model of posthemorrhagic hydrocephalus [
5‐
7]. Elevated ICP is routinely treated by insertion of a ventriculo-peritoneal shunt or by a craniectomy [
8,
9]. Although these are life-saving procedures, they are highly invasive and associated with severe side effects. Targeted and efficient pharmaceutical treatment aimed at reducing CSF secretion, and thus balancing the brain fluid content, is a desired addition to the clinical toolbox. However, such pharmaceutical approaches have, so far, generally failed due to intolerable side effects or lack of efficiency [
10,
11].
Although the existence and production of CSF have been acknowledged for more than a century, the molecular mechanisms underlying this fluid secretion remain unresolved. Some choroidal transport mechanisms have been implicated in the CSF secretion, but their quantitative contribution and the molecular mechanisms by which the fluid is transported from the vascular compartment to the brain ventricles await determination [
2,
3,
12]. Importantly, a complete map of the choroidal transport proteins may reveal other fluid-secreting transport mechanisms that could serve as future choroid plexus-specific pharmaceutical targets aimed at reducing CSF secretion in pathological conditions that would benefit from such treatments.
Here we performed transcriptomic analysis of rat choroid plexus from male and female rats and created a searchable database on the obtained transcriptomic profiles. To reveal putative future pharmacological targets, transport mechanisms and regulatory pathways were identified, ranked according to expression levels, and tied together in association networks.
Materials and methods
Experimental rats
This study conformed to the European guidelines and ethical regulations for use of experimental animals. The study utilizes 9-week-old Sprague Dawley rats (Janvier Labs, France) of male and female sex. The rats were housed with 12:12 light cycle with access to water and food ad libitum in accordance with the guidelines of the Danish Veterinary and Food administration (Ministry of Environment and Food) and approved by the animal facility at the Faculty of Health and Medical Sciences, University of Copenhagen. The rats were anaesthetized with intraperitoneal injection of xylazine and ketamine (6 mg/ml and 60 mg/ml in sterile water, 0.17 ml per 100 g body weight (ScanVet, Fredensborg, Denmark)) prior to decapitation and tissue collection.
Isolation of choroid plexus and proximal tubules
Choroid plexus (from lateral and 4th ventricles) were isolated from five male and five female rats, pooled respectively and stored in RNAlater® (Sigma-Aldrich, St. Louis, Missouri, USA) at − 80 °C. Kidney tissue was collected from the male rats, minced, and subsequently digested for 25 min at 37 °C in a table shaker at 850 rpm in collagenase solution containing 1 mg/ml collagenase (type II, Gibco®, Grand Island, NY, USA) and 1 mg/ml pronase (Roche, Mannheim, Germany) in buffer solution containing (in mM): 140 NaCl, 0.4 KH2PO4, 1.6 K2HPO4, 1 MgSO4, 10 Na-acetate, 1 α-ketoglutarate, 1.3 Ca-gluconate, 5 glycin, in addition to 48 mg/l aprotinin (trypsin inhibitor, Sigma-Aldrich, St. Louis, Missouri, USA) and 25 mg/l DNase I (grade II, Roche, Mannheim, Germany), pH 7.56. In five-minute intervals, 1 ml of the solution containing the kidney tissue was transferred to an eppendorf tube containing 1 ml cold buffer solution with 0.5 mg/ml bovine serum albumin (Sigma-Aldrich, St. Louis, Missouri, USA) and replaced by 1 ml collagenase solution. Proximal tubules were manually collected under a microscope, centrifuged at 600×g for 5 min, and the pellet stored in RNAlater® at − 80 °C.
Fluorescence-activated cell sorting (FACS) of choroid plexus epithelial cells
Choroid plexus was isolated from 10 male rats, minced, and digested in collagenase (15 mg/ml collagenase (type II, Gibco®, Grand island, NY, USA) in artificial CSF (aCSF)-HEPES containing (in mM): 120 NaCl, 2.5 KCl, 3 CaCl2, 1.3 MgSO4, 1 NaH2PO4, 10 glucose, 17 Na-HEPES, pH 7.56 for 30 min at 37 °C in a table shaker at 800 rpm. The supernatant was removed after 5 min centrifugation (600×g) and the pelleted cells were resuspended in aCSF-HEPES. The cells were triturated 20 times with a 1000 µl pipette and filtered through a 70 μm filter (pluriStrainer, Mini 70 µm, PluriSelect, Leipzig, Germany), prior to incubation with an anti-NKCC1 antibody with an extracellular epitope (1:200 in aCSF-HEPES, #ANT-071, Alomone Labs™, Jerusalem, Israel) for 30 min at 4 °C. The cells were pelleted (600 × g, 5 min) and resuspended in secondary antibody (1:500 in aCSF-HEPES, Alexa Fluor® 647—A-21245, Invitrogen™, Carlsbad, California, USA), in which it was kept for 20 min at 4 °C prior to centrifugation (600×g, 5 min) and resuspension in cold aCSF-HEPES. Cells were analyzed and sorted on a FACSAria Fusion flow cytometer (BD Biosciences, Lyngby, Denmark).
Immunohistochemistry
15 μl FACS suspension was placed on poly-d-lysine-coated coverslips for 30 min at room temperature, after which excessive liquid was removed and the attached cells covered with 4% paraformaldehyde in PBS for 15 min at room temperature. Coverslips were washed 3 times with 0.02% tween-20 in PBS (PBST) and permeabilized with PBST for 10 min at room temperature. Cells were treated with a blocking solution (4% normal goat serum (NGS) in PBST) for 1 h at 4 °C prior to exposure to primary antibody against AQP1 (1:400, #AQP-001, Alomone Labs™, Jerusalem, Israel) at 4 °C O/N. Coverslips were washed with PBST and incubated with secondary antibody (1:700, A-11034, Alexa Fluor® 488, Invitrogen™, Carlsbad, California, USA) and Phalloidin (1:400, A22287, Alexa Fluor™ 647, Invitrogen™, Carlsbad, California, USA) for 2 h at room temperature. Cells were washed and mounted onto microscope glass slides with ProLong™ Gold Antifade Mountant with DAPI (P36935, Invitrogen™, Carlsbad, California, USA).
RNA extraction and sequencing
The RNA extraction and library preparation were performed by Novogene Company Limited, UK with NEB Next® Ultra™ RNA Library Prep Kit (NEB, USA) prior to their RNA sequencing (paired-end 150 bp, with 12 Gb output) on an Illumina NovaSeq 6000 (Illumina, USA).
Although it remains unknown whether a certain transcript level cut-off represents physiological relevance, all analyses exclude genes transcribed at levels below a cut-off at 0.5 transcripts per million (TPM) [
13]. Raw data are available at the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database (accession number: GSE194236).
RNA sequencing analysis
The 150 base paired-end reads were mapped to reference genome (Rattus norvegicus Rnor_6.0 v.103) using Spliced Transcripts Alignment to a Reference (STAR) RNA-seq aligner (v. 2.7.2a) [
14]. The mapped alignment by STAR was normalized to TPM with RSEM (RNA-Seq by Expectation Maximization v. 1.3.3) [
15]. Gene information was gathered with mygene (v3.1.0) python library [
16‐
18], from which gene symbol, alias, and Gene Ontology (GO) terms [
19‐
21] were extracted. Mitochondrial genes (seqname: MT) were omitted from the reference genome prior to comparison of the kidney proximal tubule transcriptional profiles with that of the choroid plexus, since the greater abundance of these genes in the proximal tubule would skew the plasma membrane transporter expression comparison.
Cross-species comparison
The human choroid plexus transcriptome was obtained from GEO database (GSE137619, SRR10134643-SRR10134648) [
22‐
24] and the mouse choroid plexus transcriptome was obtained from GEO (GSE66312, SRR1819706-SRR18197014) [
25]. All samples were quality controlled with fastqc [
26] and trimmed with Trimmomatic [
27] (Slidingwindow 4:20, minimum length of 35 bp). The human and mouse samples, together with rat sample 3 (male), were mapped to the human reference genome (Homo sapiens GRCh38 v.104), the mouse reference genome (Mus musculus GRCm39 v.104), and rat reference genome (Rattus norvegicus Rnor_6.0 v.103) with STAR (v. 2.7.2a). The reference genome for rat, mouse, and human for the cross-species analysis only contained gene/transcripts of biotype ‘protein coding’. Mapped alignments were normalized to TPM with RSEM (v. 1.3.3.) and the mean from the samples for human and mouse, respectively, were used for further analysis. Transcribed genes sharing gene name between the compared species (or their orthologue, collected from ensemble.org via biomart martview) were included in the cross-species comparison.
Category section
Gene lists of transporters, pumps, water and ion channels, and G protein-coupled receptors (GPCR) were collected from the ‘target and family list’ from Guide to Pharmacology [
28‐
31]. Genes annotated as ‘transporters’ were employed to generate the list of membrane transporters and pumps whereas genes annotated as ‘voltage-gated ion channels’, ‘ligand-gated ion channels’, and ‘other ion channels’ were employed to generate the list of water and ion channels. To filter for plasma membrane proteins, the transporter and pump gene list was initially filtered to exclude the mitochondrial and vacuolar transport families SLC25, ATP5, and ATP6V, after which the transporter and channel lists were filtered based on associated GO terms; ‘integral component of plasma membrane’ or ‘plasma membrane’, but only included genes annotated as ‘integral component of membrane’ or ‘transmembrane’, but not annotated as ‘lysosome’, ‘endosome membrane’, ‘lysosomal’, ‘mitochondrion’, ‘mitochondrial’, ‘golgi apparatus’, ‘vacuolar’, or ‘endoplasmic’. Genes annotated as ‘GPCR’ were employed to generate the list containing GPCRs. Receptor tyrosine kinases (RTK) were gathered from Human Genome Organisation (HUGO) Gene Nomenclature Committee (HGNC) database [
32], with the annotations ‘receptor tyrosine kinase’ including sub group ‘ephrin receptors’ and ‘ErbB family’. The list of kinases was obtained from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database [
33‐
36]for entries of ‘EC 2.7.10.2’ (non-specific protein-tyrosine kinase), ‘EC 2.7.12’ (Dual-specificity kinases) with the two sub-categories, and ‘EC 2.7.11’ (Protein-serine/threonine kinases) with the 33 sub-categories. These three entries were collected with organism specific “rno” (Rattus norvegicus) filter. These kinases were filtered for protein kinases by GO terms: ‘protein kinase activity’, ‘protein serine/threonine kinase activity’, ‘protein serine kinase activity’, ‘protein threonine kinase activity’, ‘protein tyrosine kinase activity’, ‘map kinase activity’. Kinases involved solely with transcription or cell cycle modulation were subsequently excluded. Phosphatases were gathered from the KEGG database with entries EC 3.1.3 (Phosphoric Monoester Hydrolases) with 108 subcategories. The phosphatases were filtered for protein-interacting phosphatases by GO terms: ‘phosphoprotein phosphatase activity’, ‘protein serine/threonine phosphatase activity’, ‘protein serine phosphatase activity’, ‘protein threonine phosphatase activity’, and ‘protein tyrosine phosphatase activity’. Phosphodiesterases (PDE) were collected based on gene name starting with ‘PDE’. Cyclases were obtained from the KEGG database [
33‐
36] entry numbers ‘EC 4.6.1.1’ (adenylate cyclase) and ‘EC 4.6.1.2’ (guanylate cyclase).
Network analysis
The network analysis was generated from protein–protein association tables from the String-database [
37,
38] as a plugin for Cytoscape (v. 3.8.2) [
39]. Firstly, full interaction tables were generated through a full protein query of every protein in the lists of transporters and pumps, water and ion channels, GPCRs, RTKs, kinases, phosphatases, PDEs and cyclases. Secondly, the tables were filtered for interaction between the ‘transporters and pumps’ and all the regulatory genes (lists of GPCRs, RTKs, kinases, phosphatases, PDEs, and cyclases). All interactions between regulatory proteins were discarded. The same was done for ‘water and ion channels’. We only included protein–protein associations that were curated in a database or were demonstrated to interact experimentally. Thirdly, these interaction tables were loaded back into Cytoscape (v. 3.8.2) and modulated: Confidence score (from string-db) of 0.6–1 was used for genes from the list ‘transporter and pumps’ interactions and 0.7–1 was used for genes from the list ‘water and ion channels’.
Discussion
Here we reveal the choroid plexus abundance of transcripts encoding plasma membrane transport mechanisms that could potentially be involved in CSF secretion as well as those encoding regulatory factors that could partake in regulation of the CSF secretion machinery. The transcriptomic profile of male and female rat choroid plexus was nearly identical [
68], as was the case for their transportome. The rat choroid plexus transcriptomic profile displayed high (~ 90%) similarity to that of human and mice. The similarity may be even been higher than here reported, due to limited gene information of some of the retrieved genes.
Despite the fact that the existence of CSF and its continuous secretion from the blood to the brain have been long acknowledged, the exact molecular mechanisms by which this secretory process occurs have remained elusive [
2,
12]. Several plasma membrane transport proteins are suggested to be implicated in the process, but their individual quantitative contribution remains unresolved. Several of these were detected amongst the 20 highest expressed membrane transport mechanisms revealed by filtering and curation of the RNAseq data obtained from rat choroid plexus; The Na
+/K
+-ATPase α1β1, NKCC1, TRPV4, AQP1, and the HCO
3− transporters NBCe2, NCBE, and AE2 [
12,
42]. The transporter ranking was near-identical in RNAseq of entire choroid plexus and in choroid plexus epithelial cells captured by FACS, in support of the choroid plexus consisting predominantly of cells of epithelial origin [
43]. These ranked lists of choroid plexus transport mechanisms, in addition, provided gene names of other highly expressed transport proteins, which could potentially contribute to CSF secretion, but may never have been investigated for such a function. Of interest could be the highly expressed cation and anion transporters (BOCT, BSAT1), both of which are detected in brain barrier tissues [
69,
70]. BOCT is expressed widely across many brain cell types, whereas the BSAT1 is expressed predominantly in choroid plexus and other barrier cell types, i.e. endothelial cells and pericytes [
71]. The transported substrate of the former remains elusive [
69], whereas the latter is involved in the transport of thyroxine, for which the choroid plexus is renowned [
70]. The Na
+-coupled sulfate transporter (SUT-1), ranked as number 6 highest expressed amongst transporters and pumps, is virtually exclusively expressed in choroid plexus amongst brain cells [
72] (with some expression in the vascular leptomeningeal cells [
71]). MCT8 and SNAT3 are both highly expressed in choroid plexus epithelial cells, but also detected in other brain cell types, albeit the latter predominantly in barrier-related cells, such as vascular endothelial cells, pericytes and ependymal cells [
71]. MCT8 is involved in thyroxine transport [
73] while SNAT3 contributes to the high glutamine content of the CSF [
74]. None of these has, as of yet, been investigated for a potential implication in CSF secretion.
Comparison of the choroid plexus transcriptomic profile to that of another isotonic fluid-transporting epithelium of similar capacity, the kidney proximal tubules [
2], revealed that choroid plexus expressed more than double the number of genes (~ 13,500 genes) compared to the proximal tubules (~ 6000 genes, [
75,
76]). 64% of the transport protein transcripts detected in the proximal tubules were also expressed in the choroid plexus, but these only amounted to 30% of those detected within this transcript category in the choroid plexus, supporting the notion that the choroid plexus serves a variety of tasks other than CSF secretion, some of which include transepithelial solute transport. Amongst the genes encoding transport proteins transcribed in both fluid-transporting epithelia, surprisingly few placed among highly expressed genes in both tissues: the Na
+/K
+-ATPase α1β1, AQP1, a voltage-gated Cl
− channel (ClC-4), in addition to two TRP channels (TRPML1 and TRPM7). Such similarity could suggest important roles of these particular transport proteins in the secretory processes (or regulation thereof) in these tissues [
77‐
80], although, clearly, each epithelium appears to employ additional tissue-specific transport pathways to provide the transepithelial fluid transport. Of interest, an isoform of the Na
+-coupled glucose transporter (SGLT2; SLC5A2), which was previously annotated to sole expression in the proximal tubules [
81], was here detected in the choroid plexus transcriptome, albeit at a lower relative expression level (6 TPM) than that observed in the proximal tubule sample (22 TPM). The choroidal transcript abundance of SGLT2 is confirmatory of its recently demonstrated protein expression in mouse and human choroid plexus [
82,
83]. The proposed selective SGLT2 expression in the proximal tubules led to development of SGLT2 inhibitors as a selective treatment option for type 2 diabetes mellitus [
84,
85]. Such approach may have to be reconsidered based on SGLT2 expression in choroid plexus (this study and [
82,
83]), where the transport protein could potentially partake in CSF secretion, like its homologue, SGLT1, participates in fluid transport across the small intestine [
86].
The families of solute carriers (the SLCs) were highly represented in the transportome of the choroid plexus (approximately 63% of all the plasma membrane transport and pump protein transcripts), with most of the existing families expressed in this tissue (52 out of 66). Choroid plexus SLC expression is developmentally regulated, with notable upregulation of amino acid transporter families during embryonic stages [
45,
87]. Grouping these SLC families into super categories defined by their transported substrate, we demonstrate that the electrolyte/HCO
3− and anion/cation transporters dominated at the transcript level. These were followed by amino acid, sugar, metal and vitamin transporters, in support of a role for choroid plexus in supplying the brain tissue with nutrients, micro-nutrients, and various co-factors [
87,
88]. Notably, the choroid plexus is enriched in transcripts encoding metal transport proteins, with 22 genes dispersed among five different families of metal ion transporters. The choroidal expression of 15 different genes encoding various Zn
2+ transporters of the efflux (SLC30; ZnT) and influx (SLC39, ZIP) types may serve to ensure transepithelial brain delivery of Zn
2+ to various biochemical processes that are instrumental for proper brain development and function [
87,
89,
90].
To reveal potential regulatory cascades involved in modulation of CSF secretion, we obtained lists of highly expressed plasma membrane receptors and signaling pathways expressed at the transcriptional level in choroid plexus. Amongst the GPCRs, the serotonin receptor of the 5-HT2C type was expressed at fourfold higher abundance that the second-highest expressed receptor. The 5-HT2C receptor is expressed on the luminal side of the choroid plexus epithelium [
91] and its activation leads to G
q-dependent Ca
2+ release from intracellular stores [
92,
93], which subsequently promotes release of the insulin produced within the choroid plexus epithelium [
93] and may modulate choroidal ion channel activity [
94] and the rate of CSF secretion [
95,
96]. The endothelin receptor B appears as the second highest expressed GPCRs, with subtype A further down the list, in support of their protein expression in the choroid plexus [
97]. Endothelin may reduce the rate of CSF secretion [
98], possibly in part via its action on the choroidal blood flow [
99]. Of the receptor tyrosine kinases, growth factor receptors dominate the list of highest expressers, with three members of the family of fibroblast growth factor (FGF) receptors on the top 20 list (and two of these at the top). FGF receptors are detected at the protein level in choroid plexus [
100] and FGFs may be implicated in brain fluid homeostasis by their ability to modulate NKCC1 activity [
101] and to induce ventriculomegaly in a rodent model upon prolonged intraventricular infusion, at least in part due to formation of fibrosis and collagen deposits in the CSF drainage paths [
102]. The latter observation aligns with the diminished foramen magnum area observed in hydrocephalic children bearing mutations in the gene encoding FGFR2 [
103].
Lists of highly expressed intracellular signaling molecules include various cyclases, phosphodiesterases, kinases and phosphates. The vast majority of these remains to be associated with CSF secretion or regulation thereof, but may provide valuable hints to pursue in future efforts to modulate CSF secretion pharmacologically without targeting the choroidal transporters, many of which are expressed in other cell types or epithelia in the body. The cyclase-coupled receptors for atrial natriuretic peptide (ANPR-A and ANPR-B) were both detected amongst the top 10 highest expressed cyclases in the choroid plexus, and previously demonstrated at the protein level in this tissue [
104]. ANP, via its induction of cGMP formation, may [
104] or may not [
105] cause decreased CSF secretion, and altered choroid plexus ANP receptor abundance in various forms of experimental hydrocephalus could indicate involvement in brain fluid dynamics [
106]. Also of interest is the placement of the Ste20-related proline/alanine-rich kinase, SPAK, as the second highest expressed kinase in choroid plexus. SPAK is implicated in regulation of the CSF-secreting NKCC1 [
6,
65,
66] and figures prominently in our network analysis as associated with NKCC1 as well as other cation, Cl
− cotransporters (CCCs). Other choroidal kinases and phosphatases associate with different transport mechanisms, e.g. various isoforms of the Na
+/K
+-ATPase and the TRPV4 channel, and may provide novel paths to investigate in future determinations of choroid plexus transport and its regulation.
As potential limitations to our study, we acknowledge the possibility that the lists of transporters, receptors, and intracellular signaling molecules are not absolute, as the information and annotation in the various databases, on which these are based, may be incomplete and with various levels of reliability. Accordingly, the volume activated anion channels of the LRRC8 family, which have been detected in rodent choroid plexus [
107], were not included in the ion channel list from Gene Ontocology in the first filtration step (see Methods), and are therefore not included in the list of choroidal ion channels. The filtration based on GO term annotation and manual curation was performed based on the available information and to the best of our knowledge. The network analysis is based on published work. Therefore, novel and unexplored connections between different transporters and their potential regulatory pathways are anticipated to be revealed by future research efforts. Lastly, transcript abundance may vary among the different choroid plexuses [
25] and may not mirror the quantitative expression at the protein level, some of which have previously been reported [
108]. In addition, it remains unresolved whether transport proteins with lesser mRNA abundance, at the protein level may remain functionally represented and physiologically relevant (e.g. K
v1.3 with low transcript abundance (this study and [
25]), but detected at the functional level in isolated rodent choroid plexus [
94]). Nevertheless, in the current study, we have created discovery tables of the transport mechanisms and regulatory pathways of the rat choroid plexus, and linked them via network analysis. We demonstrated high similarity between species (human and mouse) and sexes. The discovery tables provide semi-quantitative ranked lists of transport mechanisms that could participate in CSF secretion and suggest regulatory candidate genes that could be involved in their regulation. With these lists, we envision that researchers in the field may devise hypotheses regarding future quantification of transport mechanisms and their regulation, with the vision to obtain rational pharmaceutical targets for CSF production modulation in the pathologies involving disturbed brain water dynamics.