Aquaporin 1 and the Na⁺/K⁺/2Cl⁻ cotransporter 1 are present in the leptomeningeal vasculature of the adult rodent central nervous system

We used C57BL/6JRj mice (Janvier Labs, Le Genest-Saint-Isle, France) at postnatal days (P) 30, 60 and 90 (n = 20) of both sexes. Mice were housed in groups (4–5 mice per cage) under a 12-h light–dark cycle and had access to water and standard chow food ad libitum. Additionally, brains were obtained from FVB.Cg-Slc12a2^tm1Ges/Mmjax, a knockout mice strain for Slc12a2 gene encoding NKCC1 [32] (P60, n = 2). Paraffin embedded brain sections from 3-month-old Sprague–Dawley rats were also obtained from a previous study [33].

We applied different antibodies against AQP1, recognizing epitopes localized both in the intracellular (rabbit anti-AQP1, Alomone Labs, Jerusalem, Israel and rabbit anti-AQP1 Alpha Diagnostic, San Antonio, TX, USA) and in the extracellular (mouse anti-AQP1, Abcam, Cambridge, UK) domains of the protein. We aimed to improve sensitivity of the immunohistochemistry, minimizing the effects of conformational changes of the target proteins due to fixative cross-linking and treatments such as dehydration and freezing, which might influence epitope availability, especially at the extracellular domain.

The rabbit anti-AQP1 (Alomone Labs) detected bands in Western blot analysis corresponding to previously described molecular weights [34‐36]. Immunohistochemistry using three different antibodies targeting AQP1 (two rabbit anti-AQP1, from Alomone Labs and Alpha Diagnostic, and one mouse anti-AQP1, from Abcam) showed the characteristic apical expression of AQP1 in choroid plexus, consistent with previous publications [37, 38]. Two antibodies targeting NKCC1, rabbit anti-NKCC1 (Abcam and Alomone Labs) were used and showed expression in choroid plexus consistent with previous studies [31].

The blood vessels were identified with two primary antisera: 1) mouse anti-α-smooth muscle actin (α-SMA; Abcam) and 2) mouse anti-cluster of differentiation 31 (anti-CD31; Abcam), also known as platelet endothelial cell adhesion molecule, PECAM-1. Both antisera stained a pattern of cellular morphology entirely consistent with previous reports for these markers [39, 40].

The band corresponding to the housekeeping protein glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was detected by Western blot using a mouse anti-GAPDH antibody (Sigma-Aldrich, St. Louis, Missouri, USA), with an apparent molecular weight of 37 kDa, as previously described [41, 42].

Animals of both sexes (P90; n = 6; 3 males, 3 females) were anesthetized by intraperitoneal (i.p.) injection of a mixture of ketamine and xylazine (100/20 mg/kg) and perfused transcardially with 10 mL 0.01 M phosphate buffer saline (PBS, pH 7.4, Sigma-Aldrich) followed by 30 mL 4% paraformaldehyde solution (PFA, Sigma-Aldrich) diluted in PBS. To label the vascular endothelia, some animals (n = 2) were also perfused with lectin from Triticum vulgaris (wheat germ agglutinin, WGA, Sigma-Aldrich, 12.5 μg/mL diluted in PBS, pH 7.4), prior to 4% PFA. Brain, kidney and heart were harvested and post-fixed in 4% PFA overnight. Some mice (n = 4) were decapitated under deep anesthesia without perfusion, and their brains fixed by overnight immersion in 4% PFA at 4 °C. The samples were sectioned using a vibratome (50 or 100 µm thick sections; Leica VT1200S, Wetzlar, Germany). After PBS washes, histological sections were blocked for 1 h at room temperature (RT) in a solution containing 0.3% Triton X-100 (Sigma-Aldrich) and 5% normal donkey or goat serum (Gibco™; Thermo Fisher Scientific, Waltham, Massachusetts, USA) in PBS followed by incubation overnight at 4 °C with primary antibodies (Table 1) diluted in blocking solution. Immunolabeling was revealed by incubation with the appropriate secondary antibodies coupled to fluorophores (Alexa Fluor, 1:500; Invitrogen™ Molecular Probes™; Thermo Fisher Scientific) for 2 h at room temperature. DAPI (4′,6-diamidino-2-phenylindole, Thermo Fisher Scientific, 1 μg/mL diluted in PBS, pH 7.4) was used for nuclear counterstaining prior to mounting with Prolong Gold Antifade Reagent (Invitrogen/Thermo Fisher Scientific, Carlsbad, California, USA). Images of the immunolabeled sections were acquired on a confocal microscope (Nikon Eclipse Ti, Tokyo, Japan) with Plan Fluor 20×/0.75 Mlmm and 40×/1.30 oil objectives or an epifluorescence microscope (Nikon Ni-E) with Plan Apo λ 4×/0.20 objective. Acquired images were adjusted for brightness and contrast using FIJI/ImageJ software.

In addition, we used paraffin sections obtained from mice and rats, processed according to standard protocols. Endogenous peroxidase activity was first quenched by immersion in a 0.5% solution of hydrogen peroxide in TRIS buffered saline (TBS, 5 mM Tris–HCl, 146 mM NaCl, pH 7.6) for 15 min. After rinsing with TBS, non-specific binding was inhibited by incubation for 30 min with 10% goat serum (Biological Industries, Kibbutz Beit-Haemek, Israel) at room temperature. Next, sections were incubated overnight at 4 °C with the anti-AQP1 primary antibodies (Alpha Diagnostic; 1:800 and Alomone Labs; 1:400) diluted in 10% goat serum and washed with TBS. For bright field light microscopy analysis, the REAL™ EnVision™ Detection System, Peroxidase/Diaminobenzidine + (DAB+) rabbit/mouse (K5007, Dako, Glostrup, Denmark) was used for detecting the primary antibodies. The detection reagent consists of a dextran backbone coupled to peroxidase and polyclonal secondary antibody molecules. The sections were washed with TBS, followed by incubation for 10 min with the DAB + solution. Sections were counterstained with Mayer`s hematoxylin, dehydrated in graded alcohols and cover-slipped with Pertex mounting medium. Additionally, Bouin’s fixed paraffin embedded coronal sections (5 µm thick) of 3-month-old Sprague–Dawley rat brains were selected from serially sectioned tissue obtained from a previous study [33].

Mice of both sexes (P60; n = 4) were perfusion-fixed with 4% PFA, and the brains harvested as described above for immunohistochemistry. We used young adult mice, which have a less rigid skull, allowing a more efficient penetration of the antibodies into the brain parenchyma. The whole mice heads were immunolabeled and cleared using the uDISCO protocol [43]. In brief, fixed tissue was serially dehydrated in methanol (Sigma-Aldrich), and then bleached by immersion in ice-cold 5% H₂O₂ (Sigma-Aldrich) containing 20% dimethyl sulfoxide (DMSO, Sigma-Aldrich) in methanol at 4 °C overnight (o/n). After bleaching, samples were rehydrated for immunolabeling, with all steps carried out at 37 °C. First, whole heads were permeabilized by o/n incubation in 20% DMSO, 0.3 M glycine (Sigma-Aldrich) and 0.2% Triton X-100 in PBS. Samples were blocked in 20% DMSO, 6% normal donkey serum and 0.2% Triton X-100 in PBS for 1 day. Prior to primary antibody incubation, samples were washed with PBS added with Tween-20 and heparin (PTwH; 0.2% Tween-20, Sigma-Aldrich with 10 mg/mL heparin, Thermo Fisher Scientific diluted in PBS) o/n. Primary antibody (anti-AQP1int, Alomone; 1:400) was diluted in PTwH containing 5% DMSO and 3% normal donkey serum. Incubation was performed for 14 days and the primary antibody solution was refreshed every third or fourth day. After the primary antibody incubation period, samples were washed thoroughly with PTwH and incubated with appropriate secondary antibody diluted in 3% normal goat serum in PTwH for 8 days, refreshing every second day. Final washes were done with PTwH. For tissue clearing, a mixture of benzyl-alcohol and benzyl benzoate (BABB, dilution ratio 1:2; Sigma Aldrich) was applied, and the samples dehydrated by serial incubations with tert-butanol (Sigma Aldrich), diluted in distilled water when applicable. After dehydration, the samples were delipidated by incubation in dichloromethane (DCM; Sigma Aldrich) for 1 h followed by a 2 h incubation in BABB mixed with diphenyl ether (Alfa Aesar) 4:1 (BABB-D4) along with 0.4% d,l-α-tocopherol (Vitamin E; Thermo Fisher Scientific). Samples were then stored in BABB-D4 at room temperature in the dark until imaging with a light-sheet microscope (LaVision BioTec UltraMicroscope II, Göttingen, Germany), using an Olympus 2X/0.15 NA (WD 10 mm) objective. Image analysis was performed using Imaris software (BitPlane, Belfast, UK).

Animals of both sexes (P30; n = 6; 3 males and 3 females) were anesthetized by i.p. injection of a mixture of ketamine and xylazine (100/20 mg/kg) and perfused transcardially with 10 mL of Hank’s Balanced Salt Solution (HBSS, Gibco™; Thermo Fisher Scientific). Young mice (P30) were used in order to minimize the time of tissue harvesting, thus preventing protein degradation of the different regions analyzed in the present study. Brains were removed from the skull and the different regions analyzed in this study—brain stem, cerebellum, choroid plexus, cortex, hippocampus, hypothalamus and olfactory bulb—were quickly dissected under a stereomicroscope (SMZ1270; Nikon) and snap frozen in dry ice. The right kidney of each mouse was harvested and used as a positive control for the assessment of AQP1 protein levels. Samples were stored at − 80 °C until homogenization with 12.5% protease inhibitor solution (cOmplete, Mini, EDTA-free Protease Inhibitor cocktail; Roche, Basel, Switzerland) in lysis buffer (50 mM Tris–HCl, 1% NP-40, 0.5% sodium deoxycholate, 5 mM EDTA, 150 mM NaCl diluted in water; Thermo Fisher Scientific). Total protein concentrations were determined using BCA assay (Kit BCA™ Protein Assay Kit; Sigma-Aldrich). Proteins from the different tissue homogenates (80 µg/mL of total protein content) were resolved under reduced conditions (NuPAGE™ Sample Reducing Agent; Thermo Fisher Scientific) by SDS-PAGE (Invitrogen™ Novex™ NuPAGE™ 4 - 12% Bis–Tris Protein Gels; Thermo Fisher Scientific) using MOPS SDS running buffer (NuPAGE™, Thermo Fisher Scientific). Proteins were subsequently transferred onto nitrocellulose membranes (Amersham Protran Premium 0.45 µm NC; GE Healthcare Life Sciences, Chicago, Illinois, USA). For AQP1 and GAPDH detection, membranes were blocked using Pierce Clear Milk Blocking Buffer (Thermo Fisher Scientific) diluted in wash buffer (0.25 M Tris Base, 1.7 M NaCl, 0.5% Tween-20 diluted in water, pH adjusted to 7.6) for 2 h at room temperature under agitation, and then incubated with appropriately diluted primary antibodies at 4 °C o/n (see Table 1). The membranes were washed three times for 10 min with washing buffer before incubation with the secondary antibodies diluted in blocking buffer for 2 h at room temperature. After washing, the membranes were imaged with ChemiDoc™ MP (Bio-Rad, Hercules, California, USA) using 530/28 nm and 695/55 nm emission filter/bandpass. The intensity of the detected bands was determined using ImageJ gel analyzer by subtracting the background. Band intensities were normalized to those for GAPDH.

The AQP1/GAPDH ratios obtained from three independent Western blot experiments were analyzed using R 3.4.0 [44]. Each band, corresponding to the AQP1 protein content of different brain regions, was normalized by dividing by the sum of all bands on the same blot, as previously described [45]. Bar plot results are expressed as mean \( \pm \) SEM, and differences between multiple regions were assessed via Kruskal–Wallis test. A p value < 0.05 was considered significant for rejection of the null hypothesis.

Using light-sheet microscopy we assessed the distribution of AQP1 in the intact adult mouse brain (P60; n = 4) upon uDISCO tissue clearing technique. The use of an antibody that recognizes an epitope in the intracellular domain of AQP1 (AQP1int) revealed immunoreactive cells in the subarachnoid cisterns adjacent to the cerebellum, in the leptomeningeal vasculature, notably along the middle cerebral arteries (MCAs), and in the olfactory bulb (Fig. 1a). As previously described [46], AQP1⁺ cells were restricted to the outer layer of the olfactory bulbs (Fig. 1b, c), corresponding to olfactory ensheathing glia cells that surround the glomeruli. Also in accordance to previous studies [47‐49], AQP1⁺ epithelial cells were observed in the choroid plexus (Fig. 1b, d).

In order to confirm and extend these findings, we next employed two distinct antibodies to compare the regional profiles by which they label AQP1 in the adult mouse brain (P90; n = 6). In addition to AQP1int, we used an antibody that recognizes an epitope in the extracellular domain (AQP1ext). Our analysis showed that AQP1ext⁺ epithelial cells are present in the choroid plexus epithelium located in the fourth and the lateral ventricles (Fig. 1e–g). No immunoreactivity was detected in the glomerular layer of the olfactory bulb using the AQP1ext antibody (Fig. 1h). In contrast, anti-AQP1int immunolabeled the epithelial cells of the choroid plexus located in the third and lateral ventricles (Fig. 1i, j), and the ensheathing glial cells surrounding the glomerular layer of the olfactory bulb (Fig. 1k).

We next accessed AQP1 protein content in different brain regions by Western blotting. Using the anti-AQP1int antibody, we quantified the AQP1 protein levels in tissue homogenates obtained from the brain stem (BS), cerebellum (Cb), choroid plexus (ChP), cerebral cortex (Ctx), hippocampus (Hip), hypothalamus (Hyp) and olfactory bulb (OB). We used kidney homogenate (Kdy; P30; n = 2 animals, 1 male and 1 female) as a positive control [50]. The immunoblotting analysis detected two bands, corresponding to the non-glycosylated (28 kDa) and glycosylated (35 kDa) forms of AQP1 [34, 51] (Fig. 1l). Under reducing conditions, homogenates obtained from the different brain regions showed different levels of the glycosylated form of AQP1 (Fig. 1l, m). In sharp contrast, the non-glycosylated form of AQP1 was the prevalent form in homogenates obtained from the choroid plexus and kidney. AQP1 was detected in all brain regions analysed, but no significant regional differences were observed (BS: 0.10 ± 0.03; Cb 0.06 ± 0.01; ChP: 0.27 ± 0.05; Ctx: 0.05 ± 0.03; Hip: 0.05 ± 0.02; Hyp: 0.05 ± 0.02; Kdy: 0.24 ± 0.12; OB: 0.17 ± 0.06; mean ± SEM; Kruskal–Wallis test, H = 12.36, df = 7, p = 0.089; n = 3 independent experiments).

In contrast to what is observed in the peripheral vasculature, studies have reported that endothelial cells in the CNS are devoid of AQP1 expression [52‐54]. In view of our present results showing marked AQP1 distribution in the leptomeningeal vasculature, and given that we find that the glycosylated form predominates in brain regions other than the choroid plexus and olfactory bulbs, we sought to confirm the presence of AQP1⁺ cells in the CNS vasculature. Adult mouse brain sections (P60; n = 6 animals) were double-labelled for anti-AQP1int and anti-AQP1ext. Confocal microscopy imaging revealed AQP1ext⁺ blood vessel-like structures located next to the ventricular system (Fig. 2a, b), close to the AQP1ext⁺/AQP1int⁺ choroid plexus epithelial cells (Fig. 2a, c–f). Histological sections obtained from adult mouse kidney were immunolabeled for CD31 and AQP1. As previously described [55, 56], epithelial cells in the proximal tubules are AQP1⁺, as well as renal vascular endothelial cells (Fig. 2g–j). Endothelial cells distributed in the adult mouse heart vasculature were also immunoreactive for AQP1int (Fig. 2k, l). AQP1⁺ blood vessels were also observed in the adult rat brain, lining the hippocampus, along with AQP1⁺ epithelial cells and blood vessels of the choroid plexus located in the third ventricle (Fig. 2m–o). Hence, we find AQP1 expression in cerebral blood vessels, notably those in close proximity to the ventricular system and lining the hippocampus, in addition to expression in the leptomeningeal vasculature in the SAS observed along the surface of the adult mouse brain.

The Na⁺/K⁺/2Cl⁻ cotransporter 1 (NKCC1) is present in the choroidal epithelial cells and has been implicated in CSF production due to its capacity to couple water movement to ion translocation [13, 31, 57, 58]. Previously described in vitro [59], the presence of NKCC1 on endothelial cells has been confirmed in the adult rat brain tissue [60, 61].

Confocal microscopy analysis of histological sections obtained from adult mice brain showed that leptomeningeal blood vessels are immunoreactive for both AQP1 and NKCC1 (Fig. 3a, c–f). Optical sections further revealed that NKCC1 as well as AQP1 are distributed along the smooth muscle cell layer, yet not expressed by endothelial cells of arterioles and/or veins (Fig. 3b). NKCC1 was additionally detected in ependymal cells lining the ventricular walls, and in cerebellar neurons located in the molecular and Purkinje layers (Fig. 3g, h), and as previously described [31, 62], NKCC1 was detected in choroidal epithelial cells, which are also AQP1⁺ (Fig. 3i). Brain sections obtained from NKCC1 KO mice showed no immunoreactivity for NKCC1 (Fig. 3j, k). Blood vessels in the hippocampal fissure are also immunoreactive for AQP1 and NKCC1 (Fig. 3l, m). Leptomeningeal vessels observed in this region presented AQP1 and NKCC1 immunoreactivity along their smooth muscle cell layer (Fig. 3n–r).

An extra-choroidal source of CSF has been postulated to correspond to influx of fluid across the BBB [22]. Specifically, some CSF production has been suggested to occur at the capillary level, mediated by endothelial cells [4, 23]. However, there has been no investigation of this scenario using immunohistochemistry and light microscopy. Therefore, we asked if the endothelia of the leptomeningeal vasculature in the adult rodent CNS express AQP1 and NKCC1. The fine structure of capillaries and venules was accessed in paraffin sections obtained from adult mice (P60; n = 2 animals). We also assessed the distribution of AQP1 and NKCC1 in serial sections obtained from adult mouse brain (Fig. 4a, b). The vascular phenotype was determined in sections stained with haematoxylin (Fig. 4c). Higher magnification micrographs of the SAS, specifically at the cisterna interpeduncularis region, revealed AQP1int (Fig. 4d) and NKCC1 (Fig. 4e) immunoreactive cells in the smooth muscle cell layer of arterioles and in the endothelia of capillaries and venules. Confocal microscopy of brain sections from lectin-perfused mice showed that AQP1⁺ cells are restricted to the smooth muscle cell layer, and are absent in the leptomeningeal arterioles’ endothelia (Fig. 4f–j). In the leptomeningeal vasculature of the spinal cord, both AQP1 (Fig. 5a–c) and NKCC1 (Fig. 5d–f) were detected in the endothelial cells of the leptomeningeal capillaries.