Uncoupling of the Astrocyte Syncytium Differentially Affects AQP4 Isoforms

The water channel protein aquaporin-4 (AQP4) and the gap junction forming proteins connexin-43 (Cx43) and connexin-30 (Cx30) are astrocytic proteins critically involved in brain water and ion homeostasis. While AQP4 is mainly involved in water flux across the astrocytic endfeet membranes, astrocytic gap junctions provide syncytial coupling allowing intercellular exchange of water, ions, and other molecules. We have previously shown that mice with targeted deletion of Aqp4 display enhanced gap junctional coupling between astrocytes. Here, we investigate whether uncoupling of the astrocytic syncytium by deletion of the astrocytic connexins Cx43 and Cx30 affects AQP4 membrane localization and expression. By using quantitative immunogold cytochemistry, we show that deletion of astrocytic connexins leads to a substantial reduction of perivascular AQP4, concomitant with a down-regulation of total AQP4 protein and mRNA. Isoform expression analysis shows that while the level of the predominant AQP4 M23 isoform is reduced in Cx43/Cx30 double deficient hippocampal astrocytes, the levels of M1, and the alternative translation AQP4ex isoform protein levels are increased. These findings reveal a complex interdependence between AQP4 and connexins, which are both significantly involved in homeostatic functions and astrogliopathologies.


Introduction
Astrogliopathology is now emerging as a feature common to several neurological conditions [1,2]. In particular, loss or mislocalization of astrocytic transporter or channel molecules has been associated with Alzheimer's disease, stroke, and various forms of human and experimentally induced epilepsy [3][4][5][6]. The prevailing concept is that astrocytes have important homeostatic functions that critically depend on the specific localization of membrane molecules. Table 1. Antibodies. The following antibodies used in this study.

Immunogold Quantitation and Data Analysis
Quantitative analysis was performed as previously described [36][37][38]. Briefly, to assess the expression of AQP4 and α-syntrophin in the perivascular astrocytic endfeet in parietal cortex and hippocampus, images of 20-30 capillaries were acquired randomly from each section. To check whether the possible changes in the AQP4 perivascular pool is due to mislocalization, around 40 images were Cells 2020, 9, 382 4 of 13 also taken of neuropil without capillaries from each section. All the images were taken at 26500× magnification. The linear densities of gold particles were determined by using analysis software (Soft Imaging Systems (SIS), Münster, Germany). Investigators were blinded regarding animal genotypes during processing and analysis. Linear densities were analyzed semi-automatically and the data was transferred to SPSS Version 22 (SPSS, Chicago, IL, USA) for statistical analysis. Sample groups were statistically compared using ANOVA with Bonferroni's post hoc test. Data are presented as mean SEM. p < 0.05 was considered to be significant.

Preparation of Total Protein Lysates from Brain Regions
Mice were subjected to euthanasia in a CO 2 chamber. Brains were isolated and kept on ice cold petri dishes. Samples were briefly rinsed with PBS. Hippocampi and cortices were dissected quickly. They were snap frozen in liquid nitrogen and stored at −80 • C. Cortex and hippocampus from 8 week old WT (n = 3) and dKO mice (n = 3) were homogenized in RIPA buffer (50 mM Tris-HCl pH 7.4; 150mM NaCl; 5mM EDTA; 1% Triton X-100; 0.5% sodium deoxycholate; 0.1% SDS), with freshly added 1× SigmaFAST protease inhibitor cocktail (Sigma-Aldrich, St. Louis, MO, USA) and 1× PhosSTOP phosphatase inhibitor (Roche Life Science, Basel, Switzerland). Homogenates were prepared by mechanical dissociation using lysing matrix tubes (MP Biomedicals) and incubated on ice for 30 min before centrifugation at 14000 rpm at 4 • C for 15 min. The supernatant was collected as total protein and concentrations were measured using a Pierce™ BCA protein assay kit (Thermo Fisher, Waltham, MA, USA).
Membranes were blocked for 1 h at RT in 5% BSA in 1X Tris-buffered saline (TBS) (BioRad) and washed 3 × 5 min in 1X TBS with 1% Tween-20 (Sigma-Aldrich). Membranes were cut before separate overnight incubation with primary antibodies (Table 1) at 4 • C. Subsequently, membranes were washed in TBS-T before incubation with secondary HRP-conjugated antibodies ( Table 1) for 1 h at RT. Membranes were washed 3 × 10 min in TBS-T before detection of immunoreactive bands by SuperSignal™ West Pico Chemiluminescent Substrate (Thermo Fisher) using a Fujifilm LAS-3000 (Fujifilm, Tokyo, Japan) and ChemiDoc™ Touch (BioRad) imaging system. For these two systems, secondary HRP-conjugated antibody concentrations were diluted 1:5000 and 1:25,000, respectively. The AQP4 M1 isoform antibody was generated by GeneScript towards the predicted antigen MSDRAAARRWGKC within the rat M1-AQP4 specific sequence.
Bands were quantified as arbitrary background-subtracted density units in Image Studio Lite (Ver 5.2, Licor Biosciences, Lincoln, NE, USA). Normalization was performed by dividing intensities of protein bands of interest with the normalizing control α-tubulin band intensity for their respective lane. The obtained values were transferred to SPSS Version 25 (SPSS, Chicago, IL, USA) and compared using the non-parametric Kruskal-Wallis test. Data are presented as percentages of the average control with medians (95% CI). p < 0.05 was considered to be significant.

RNA Isolation and Reverse Transcriptase Quantitative PCR (RT-qPCR)
Total RNA was isolated from cortex and hippocampus using the RNeasy Plus Mini Kit (QIAGEN). The RNA concentration and integrity were determined using a NanoDrop 2000c spectrophotometer (Thermo Scientific) and agarose gel electrophoresis. cDNA was synthesized using 400 ng of RNA from each sample. The reaction was performed by GoScript Reverse Transcription System (Promega) using Oligo (dT) 15 . All the cDNA samples were diluted in 10 mM Tris-HCl (pH 8.0) to 2.5 ng/µL. Copy numbers of Aqp4 were calculated using absolute standards. The qPCR was performed in a total volume of 20 µL, containing the Power SYBR Green PCR Master Mix (Applied Biosystems), forward and reverse primers (10 µM), and 2 µL of the cDNA template. Thermal cycling was performed on the StepOnePlus system (Applied Biosystems) with the following conditions: 95 • C for 10 min, followed by 40 cycles at 95 • C for 15 s and 60 • C for 1 min. Standard curve, No Reverse Transcriptase (NRT) and No Template Control (NTC) were included in the study. Primers used are listed in Table 2. Gapdh, Tbp, Hprt1 and Actb were evaluated for use as internal controls by Norm Finder [39] and a combination of Gapdh and Hprt1 weas selected for double normalization as they showed the best stability value of 0.060. Independent samples t-test was used for data analysis. Data are presented as mean and SEM. p < 0.05 was considered to be significant. Table 2. PCR primers. The following primer pairs used in the study.

Gene Forward Primer Reverse Primer
Aqp4

Deletion of Astroglial Connexins Leads to a Significant Decrease in Perivascular AQP4 and Abolishes the Regional Heterogeneity in AQP4 Distribution
Quantitative high resolution immunogold cytochemistry using an antibody against AQP4 revealed a decrease in the density of perivascular AQP4 immunogold particles in parietal cortex and hippocampus of Cx43/30 dKO mice compared to WT ( Figure 1A-D). In agreement with our previous studies, quantitative analysis of immunogold labeling in WT mice showed about 38% higher AQP4 immunogold linear density in the hippocampus (mean 16.45) compared to parietal cortex (mean 11.93, Figure 1E, p < 0.001). This regional difference was abolished in Cx43/30 dKO mice as the reduction in the perivascular AQP4 immunogold density was more pronounced in hippocampus (34%) than in parietal cortex (20%) ( Figure 1E).
To resolve if the decrease in the perivascular AQP4 immunogold density is due to a redistribution of AQP4, we quantified the AQP4 immunogold labeling in the neuropil outside the perivascular zone. Our analysis showed that deletion of astroglial connexins was associated with a significant decrease in AQP4 immunogold density also in the neuropil, both in hippocampus and parietal cortex ( Figure 1F). (F) Quantitative analysis of AQP4 immunogold labeling in randomly selected micrographs of neuropil outside the perivascular zone in parietal cortex and hippocampus of Cx43/30 dKO and WT mice. AQP4 labeling intensity was significantly lower in both parietal cortex and hippocampus of Cx43/30 dKO compared to WT (p < 0.005). E; endothelial cells, L; vessel lumen. Scale bar: 500 nm. * significant difference according to ANOVA with Bonferroni's post hoc test; n = 5 for each genotype; error bars indicate SEM. p < 0.05.

Reduction in the Perivascular Pool of AQP4 is Independent of α-Syntrophin
To assess whether the reduction observed in the perivascular pool of AQP4 in Cx43/30 dKO mice is due to down-regulation of the AQP4 anchoring molecule, α-syntrophin, we compared the perivascular density of this molecule in both genotypes. Our data did not show any significant differences in the perivascular density of α-syntrophin between the two groups ( Figure 2). (F) Quantitative analysis of AQP4 immunogold labeling in randomly selected micrographs of neuropil outside the perivascular zone in parietal cortex and hippocampus of Cx43/30 dKO and WT mice. AQP4 labeling intensity was significantly lower in both parietal cortex and hippocampus of Cx43/30 dKO compared to WT (p < 0.005). E; endothelial cells, L; vessel lumen. Scale bar: 500 nm. * significant difference according to ANOVA with Bonferroni's post hoc test; n = 5 for each genotype; error bars indicate SEM. p < 0.05.

Reduction in the Perivascular Pool of AQP4 is Independent of α-Syntrophin
To assess whether the reduction observed in the perivascular pool of AQP4 in Cx43/30 dKO mice is due to down-regulation of the AQP4 anchoring molecule, α-syntrophin, we compared the perivascular density of this molecule in both genotypes. Our data did not show any significant differences in the perivascular density of α-syntrophin between the two groups ( Figure 2).

Deletion of Astroglial Connexins Leads to a Decrease in Aqp4 Transcript, but Has a Differential Effect on AQP4 Isoforms at the Protein Level
RT-qPCR analysis demonstrated a significant decrease in Aqp4 transcripts in the parietal cortex and hippocampus of Cx43/Cx30 dKO mice ( Figure 3A). Similarly, Western blots provided evidence for a reduction in the total amounts of total AQP4 protein and the main AQP4 isoform (M23) in the Cx43/Cx30 dKO cortex and hippocampus compared to WT controls ( Figure 3B). These changes were significant for both total AQP4 and AQP4 M23 in both regions ( Figure 3C,D). A band corresponding to the AQP4 M1 isoform was identified ( Figure 3B arrow) but this was not distinct enough to allow for quantitative analysis.

Deletion of Astroglial Connexins Leads to a Decrease in Aqp4 Transcript, but Has a Differential Effect on AQP4 Isoforms at the Protein Level
RT-qPCR analysis demonstrated a significant decrease in Aqp4 transcripts in the parietal cortex and hippocampus of Cx43/Cx30 dKO mice ( Figure 3A). Similarly, Western blots provided evidence for a reduction in the total amounts of total AQP4 protein and the main AQP4 isoform (M23) in the Cx43/Cx30 dKO cortex and hippocampus compared to WT controls ( Figure 3B). These changes were significant for both total AQP4 and AQP4 M23 in both regions ( Figure 3C,D). A band corresponding to the AQP4 M1 isoform was identified ( Figure 3B arrow) but this was not distinct enough to allow for quantitative analysis.
To determine whether the remaining AQP4 isoforms M1 and AQP4-ex were altered in the Cx43/30 dKO animals, we used specific antibodies towards these isoforms. While total AQP4 and the AQP4-M23 isoform showed decreased expression in dKO mice, we observed an up-regulation of the AQP4-M1 and AQP4-ex isoforms in Cx43/30 dKO animals compared to WT ( Figure 4A).
The increase in AQP4-M1 was modest yet significant ( Figure 4B). The AQP4-ex protein level showed a 7-fold increase when comparing the medians of dKO and WT mice ( Figure 4C). and hippocampus of Cx43/Cx30 dKO mice ( Figure 3A). Similarly, Western blots provided evidence for a reduction in the total amounts of total AQP4 protein and the main AQP4 isoform (M23) in the Cx43/Cx30 dKO cortex and hippocampus compared to WT controls ( Figure 3B). These changes were significant for both total AQP4 and AQP4 M23 in both regions ( Figure 3C,D). A band corresponding to the AQP4 M1 isoform was identified ( Figure 3B arrow) but this was not distinct enough to allow for quantitative analysis.  To determine whether the remaining AQP4 isoforms M1 and AQP4-ex were altered in the Cx43/30 dKO animals, we used specific antibodies towards these isoforms. While total AQP4 and the AQP4-M23 isoform showed decreased expression in dKO mice, we observed an up-regulation of the AQP4-M1 and AQP4-ex isoforms in Cx43/30 dKO animals compared to WT ( Figure 4A).
The increase in AQP4-M1 was modest yet significant ( Figure 4B). The AQP4-ex protein level showed a 7-fold increase when comparing the medians of dKO and WT mice ( Figure 4C). (B,C) Graphs illustrate densitometric analysis of AQP4-M1 and AQP4-ex immunoblotts of hippocampus of WT and Cx43/30 dKO mice. A significant increase in AQP4-M1 and AQP4-ex protein isoforms was found in Cx43/30 dKO animals compared to WT. Values are presented as average of the respective wild type regional control *significant difference with non-parametric Kruskal-Wallis test; n = 3 for each genotype; error bars medians with 95% CI, p < 0.05.

Discussion
Astrocytic gap junctions and AQP4 are complementary in terms of their localization and function. While gap junctions are conduits for redistribution of water and other molecules within the astrocytic syncytium, AQP4 serves as an influx and efflux route for water between individual astrocytes and the perivascular space. The question explored in the present paper is whether deletion Graphs illustrate densitometric analysis of AQP4-M1 and AQP4-ex immunoblotts of hippocampus of WT and Cx43/30 dKO mice. A significant increase in AQP4-M1 and AQP4-ex protein isoforms was found in Cx43/30 dKO animals compared to WT. Values are presented as average of the respective wild type regional control *significant difference with non-parametric Kruskal-Wallis test; n = 3 for each genotype; error bars medians with 95% CI, p < 0.05.

Discussion
Astrocytic gap junctions and AQP4 are complementary in terms of their localization and function. While gap junctions are conduits for redistribution of water and other molecules within the astrocytic syncytium, AQP4 serves as an influx and efflux route for water between individual astrocytes and the perivascular space. The question explored in the present paper is whether deletion of connexins-entailing uncoupling of the astrocytic syncytium-affects the expression of specific AQP4 isoforms in specialized membrane domains. Our data provide new insight in how AQP4 expression is regulated and have implications for our understanding of brain pathophysiology as altered expression and function of astroglial connexins is seen in several neurological conditions [40][41][42][43]. The present results indicate that dysregulation of connexins in the context of neurological diseases might be associated with loss or mislocalization of AQP4 water channels.
We have previously shown that there is a regional heterogeneity in the perivascular AQP4 density.Specifically, the perivascular AQP4 density in hippocampus is about 20% higher than the perivascular AQP4 density in parietal cortex [36]. In Cx43/Cx30 dKO animals, the decrease in perivascular AQP4 was more pronounced in hippocampus than in parietal cortex, thus abolishing the regional heterogeneity. Interestingly, previous studies have shown that the size of astrocytic syncytium, assessed by gap junction mediated dye coupling, is larger in hippocampus than in neocortex [44]. Given that clearance of metabolically produced water to the perivascular space is one of the main functions of the perivascular AQP4 channels [23], one can speculate that size of astrocytic syncytium regulates the density of the perivascular AQP4, leading to a higher density of perivascular AQP4 in hippocampus than in neocortex. This might explain why the regional heterogeneity in the perivascular AQP4 density is abolished in dKO mice lacking a functional astrocyte syncytium.
AQP4 M23 is the most predominant AQP4 isoform in brain. This is the isoform that constitutes the orthogonal arrays of proteins (OAPs) that accumulate in astrocytic endfeet around brain microvessels. Here we show that the perivascular expression of this isoform is reduced in Cx43/30 dKO mice, as is the total amount of AQP4 in the brain. The latter finding is consistent with that of Ezan, Andre, Cisternino, Saubamea, Boulay, Doutremer, Thomas, Quenech'du, Giaume and Cohen-Salmon [30]. The reduction of perivascular AQP4 observed in our study, was more pronounced in hippocampus than in parietal cortex. Our data indicate that Cx43/30 dKO mice have a reduced capacity for water transport across astrocytic plasma membranes and specifically between astrocytes and the perivascular space in the hippocampus. In line with this, Lutz, et al. [45] noted swelling of astrocytes in the hippocampus of mice subjected to deletion of Cx43 and Cx30, although in the latter study deletion of astrocytic connexins was under control of the mouse GFAP promoter.
The mechanisms that underlie down-regulation of M23 after Cx43/30 dKO are not known. AQP4 expression is under the influence of several micro-RNAs [46] and further studies are required to determine whether these are involved in the alterations seen in the present study.
As for the astrocytic connexins, the M1 isoform of AQP4 is known to play a role in astrocytic migration. Expression of this isoform reduces the size of OAPs, which predominantly contain the M23 isoform [24]. The up-regulation of M1 after Cx43/30 dKO could therefore be a compensation for the non-channel functions of the connexins, such as migration [24] or process motility [47]. This is particularly relevant for astrogliosis, in which AQP4 and Cx43 might play similar roles [48].
Among the AQP4 isoforms investigated AQP4ex was the one showing the most conspicuous change in dKO mice. The level of this isoform-which constitutes about 10% of the total AQP4 pool-was almost seven-fold higher after connexin deletion. AQP4ex is the longest of the AQP4 isoforms and is formed by translational readthrough. With its long tail, the AQP4ex isoform might play a specific role in anchoring OAPs to the dystrophin complex in astrocytic endfeet [26]. If the up-regulation of AQP4ex translates into a more efficient anchoring of perivascular M23 it would serve to constrain the reduction of the perivascular AQP4 pool secondary to the down-regulation of M23 expression. The long C-terminus of AQP4-ex might also substitute for the C-terminus of Cx43 in interaction with intracellular proteins. This idea will be followed up in future studies.

Conclusions
We previously demonstrated that targeted deletion of Aqp4 leads to an increased number of gap junctions in astrocytic endfoot membrane domains [29] corresponding to an enhanced functional coupling [28]. This increase occurs at the post-translational level as the amounts of Cx43 and Cx30 transcripts and protein were unchanged [29]. Here we show that connexins and AQP4 are mutually interdependent as deletion of Cx43/Cx30 causes a down-regulation of AQP4 M23 and a sizeable up-regulation of the newly discovered AQP4ex. The down-regulation of M23 AQP4 (the predominant AQP4 isoform in brain) is paralleled by a reduced level of Aqp4 mRNA indicating regulation at the transcriptional level. Thus, while astrocytic connexins and AQP4 are mutually interdependent, there is a striking asymmetry when it comes to the mechanisms involved. The present data unravel a complex interaction between Cx43, Cx30, and AQP4 that might be an important feature in astrogliopathology.