A High Throughput Screen Identifies Chemical Modulators of the Laminin-Induced Clustering of Dystroglycan and Aquaporin-4 in Primary Astrocytes

Background Aquaporin-4 (AQP4) constitutes the principal water channel in the brain and is clusteredat the perivascular astrocyte endfeet. This specific distribution of AQP4 plays a major role in maintaining water homeostasis in the brain. A growing body of evidence points to a role ofthe dystroglycan complex and its interaction with perivascular laminin in the clusteringof AQP4 atperivascular astrocyte endfeet. Indeed, mice lacking components of this complex or in which laminin-dystroglycan interaction is disrupted show a delayed onset of brain edema due to a redistribution of AQP4 away from astrocyte endfeet. It is therefore important to identify inhibitory drugs of laminin-dependent AQP4 clustering which may prevent or reduce brain edema. Methodolgy/Principal Findings In the present study we used primary rat astrocyte cultures toscreen a library of >3,500 chemicals and identified 6 drugs that inhibit the laminin-induced clustering of dystroglycan and AQP4. Detailed analysis of the inhibitory drug, chloranil, revealed that its inhibition of the clustering is due to the metalloproteinase-2-mediated ß-dystroglycan shedding and subsequent loss of laminin interaction with dystroglycan. Furthermore, chemical variants of chloranil induced a similar effect on ß-dystroglycan and this was prevented by the antioxidant N-acetylcysteine. Conclusion/Significance These findings reveal the mechanism of action of chloranil in preventing the laminin-induced clustering of dystroglycan and AQP4 and validate the use of high-throughput screening as a tool to identify drugs that modulate AQP4 clustering and that could be tested in models of brain edema.


Introduction
Cerebral edema with excess accumulation of water and cellular swelling is a common consequence of stroke, traumatic brain injury, brain tumor and meningitis. In the normal brain, water is distributed between cerebrospinal fluid, blood, intracellular and interstitial compartments and moves between these compartments in response to osmotic gradients. In pathological conditions, the abnormal accumulation of water in brain parenchyma gives rise to either cytotoxic or vasogenic edema [1]. Cytotoxic edema is seen in early cerebral ischemia and is primarily characterized by an increase in astrocyte volume due to enhanced water flux from the bloodstream into these cells. Vasogenic edema is often seen following brain tumor formation and is characterized by enhanced water entry in the interstitial compartment of the brain due to the disruption of the blood-brain barrier [1].
Aquaporin-4 (AQP4), the principal water channel in the brain, is mainly expressed at the interface between the brain tissue and the blood at the perivascular astrocyte endfeet [2]. Recent studies in the AQP4 knockout or overexpressing mice demonstrated a dual role for AQP4 in the pathophysiology of brain edema [1,3,4,5,6,7]. Indeed, there is increasing evidence that AQP4 deficiency is associated with reduced water entry into the brain and reduced water outflow from the brain parenchyma in edema models that include water intoxication, focal cerebral ischemia, bacterial meningitis, cortical freeze-injury and brain tumor [1,4]. Because AQP4 permits bidirectional water transport it has been shown that it plays a role not only in the early accumulation of water in cytotoxic edema [4] but also in the removal of excess water in vasogenic edema [1,8,9,10]. Therefore, blocking AQP4 or inhibiting its clustering around blood vessels would reduce water entry which may be beneficial in preventing cytotoxic edema at early stages of stroke. On the contrary, activating AQP4 or increasing its clustering around blood vessels would promote the extrusion of the excess water from the brain preventing thereby vasogenic edema.
Several studies have focused their effort in designing drugs that specifically inhibit AQP4 function [11,12,13,14,15,16,17]. To date, only three drugs, tetraethylammonium (TEA), 6-ethoxybenzothiazole-2-sulfamide (EZA) and acetazolamide (AZA) have been identified as potential inhibitors of AQP4 activity. However, the ability of these candidate blockers to inhibit cell swelling upon hypo-osmotic chockhas been disputed [18]. In an attempt to identify other drugs that inhibit the function of AQP4 in water transport, Mola et al [15] developed a functional high-throughput assay based on the measurement of osmotically-induced cell volume changes to screen several libraries of drugs and identified four blockers of AQP4-mediated water transport.
Multiple lines of evidence suggest that the integrity ofthe dystroglycan complex (DGC) is essential for the properlocalization Figure 1. Validation of the Cellomics TM Arrayscan V TI automated fluorescence imaging of the laminin-induced clustering of ß-dystroglycan in primary astrocyte cultures. A. Images of untreated astrocytes and astrocytes treated with 20 nM laminin-1 were acquired on the ArrayScan V TI HCS Reader. The right panels are merged images of Hoechst 33342 (left panels) and ß-DG labelled astrocytes (middle panels). Automatic identification of the cytoplamic maskof individual astrocytes(red) and detected spots (yellow) by the Compartmental Analysis BioApplication. B, C. Dose-response curve of laminin-induced ß-DG clusteringin astrocytes grown in 96 well-plates. Each point represents the mean number of clusters or areas of clusters 6 S.E.M. from three different experiments. Scale bar, 30 mm. doi:10.1371/journal.pone.0017559.g001 and function of AQP4. This multiprotein complex includes numerous intracellular proteins such as syntrophin, dystrobrevin and dystrophin downstream of dystroglycan (DG) [19,20]. DG is post-translationally cleaved into a transmembrane protein, b-DG, and an extracellular protein, a-DG. Alpha-DG binds noncovalently to the short extracellular domain of ß-DG as well as to the extracellular matrix proteins laminin, agrin and perlecan as well as neurexin [21,22,23,24]. Intracellularly, ß-DG binds directly to dystrophin, which in turn binds dystrobrevin and syntrophin. Interestingly, DG, dystrophin, dystrobrevin and syntrophin are co-clustered with AQP4 at perivascular astrocyte endfeet [25,26]. We have shown in astrocyte cultures that this coclusteringis mediated by the interaction of DG with extracellular laminin-1 [27,28]. Likewise, we have demonstrated that a-DG interaction with perivascular laminin is key to the polarized distribution of AQP4 at astrocyte endfeet in brain [25]. Furthermore, the deletion of a-syntrophin or mutations of dystrophinresult in the loss of AQP4 clustering, relocating it away from astrocyte endfeet [29,30,31]. Most importantly, theloss of AQP4 polarized distribution in thesemice resulted in a beneficial effect as it delayed the onset of brain edema [32,33]. Therefore, treatments using drugscapable of disrupting the clustered distribution of the DGC at astrocyte endfeet will ultimately leadto the disruption of AQP4 distributionas well.
In the present study we used a well established assay of laminininduced clustering of the DGC and AQP4 in primary astrocyte cultures to screen a library of .3,500 chemicals and identified 5 enhancers and 6 inhibitorsof this clustering. Here, we focused our analysis on chloranil,one the inhibitory drugs, anddemonstrate that it causes cell surface cleavage of the 43 kDa ß-DGto a 31 kDa form via a metalloproteinase-mediated mechanism. Chemical variants of chloranil also induce ß-DG sheddingleading to the loss of cell surface tethering of a-DG andsubsequent severance of laminin binding,resulting in the inhibition of DG and AQP4 clustering. This study provides evidence that the DGC is a valuable target for the disruption of laminin-induced AQP4 clustering andvalidates the high-throughput screen on primary astrocyte cultures as a powerful tool in the discovery of potential therapeutic drugs for both cytotoxic and vasogenic edema that occur in brain disorders including stroke, tumour, hydrocephalus, infection and traumatic brain injury.

Chemicals
The 3,594 chemicals used in the screen were from the Pretswick, Sigma LOPAC, Microsource Spectrum and Biomol natural products collections, and were provided by the Canadian Chemical Biology Network (www.ccbn-rbc.ca). P-chloranil (referred to as chloranil in the rest of the text), benzoquinone and flunarizine were purchased from Sigma-Aldrich and Hoechst 33342 was from Invitrogen. Prinomastat (AG3340) was purchased from Agouron Pharmaceuticals, the tissue inhibitors of metalloproteinases TIMP1 and TIMP2 were a generous gift from Dr. C.M. Overall (University of British Columbia, Vancouver)and TIMP3was purchased from R&D Systems (Minneapolis, MN, USA). N-acetyl-L-cysteine (NAC) was purchased from Sigma-Aldrich.

Chemical screen for modulators of laminin-induced dystroglycan clustering in astrocytes
Experiments were performed using postnatal day 1 Sprague-Dawley rats (Charles River) in accordance with protocols (#A06-0319)approved by the animal care committee of the University of British Columbia. Primary hippocampal astrocyte cultures were prepared from postnatal day 1Sprague-Dawley rats (Charles River). Hippocampi were dissected, and meninges and choroid plexus were removed. They were then cut into small pieces and incubated for 25 min with trypsin (3.0 mg/ml; Gibco, Burlington, Canada). Dissociated hippocampi were then plated in culture flasks and grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 1% penicillinstreptomycin and 1 mM L-glutamine (Gibco) for 2-3 weeks. To remove microglia and oligodendrocyte progenitors, the flasks were shaken the day following the plating.
The culture medium was changed every 3 days. After trypsinization, the cells were plated in PerkinElmer View 96well plates at 8000 cells per well. One day after plating, the cells were treated for 7 h with 20 nM Engelbreth-Holm-Swarm Sarcoma laminin-1 (Sigma-Aldrich), washed with warn phosphate buffered saline (PBS) and fixed with 4% (w/v) paraformaldehyde in 0.1 M phosphate buffer for 20 min followed by rinsing in PBS, 3615 min. Four hours before fixation, chemicals were transferred from stock plates (5 mM in DMSO) to each well for a final concentration of <15 mM using a Biorobotics Biogrid II robot equipped with a 0.7 mm diameter 96-pin tool that adds,20 nl per well. Fixed cells were incubated for 1 h at room temperature (20-22uC) in a solution containing 2% bovine serum albumin (Sigma) and 0.25% Triton X-100. Immunolabelling was performed by incubating the cells at room temperature for 1 h in the presence of the primary antibody against ß-DG (1/100). Subsequently, cells were rinsed with PBS (3615 min) and incubated with Alexa Fluor 568 goat anti-mouse IgG (1/ 200; Molecular Probes, USA). After several washes with PBS, the cells were incubated with PBS containing 500 ng/ml Hoechst 33342 for 15 min at room temperature. Plates were read in a Cellomics TM Arrayscan V TI automated fluorescence ß-dystroglycan clustering is a suitable assay for detection using the Cellomics TM Arrayscan V TI automated fluorescence imager. ß-DG clustering in laminintreated astrocytes was used as the positive control and the spontaneous clustering occurring in the absence of exogenous laminin was used as the negative control. Z9 = 12((3s max +3s min )/|m max 2m min |), where m max is the mean positive control signal, s max is the standard deviation of the positive control signal, m min is the mean negative control signal, and s min is the standard deviation of the negative control signal. Signal/background (fold increase) = m max /m min . Signal/noise = (m max 2m min )/s min . Signal window = (m max 2m min 23(s max +s min ))/s max [34]. doi:10.1371/journal.pone.0017559.t001 imager. Cells were photographed using a 206 objective in the Hoechst and TRITC(XF-100 filter) channels. The compartment analysis algorithm was used to identify the nuclei, apply a cytoplasmic mask and quantitate TRITC spotsin the TRITC channel fixed at 250 pixel intensity units. Fluorescence intensity in the TRITC channel was gated at 10 average pixel intensity units inside the cytoplasmic mask to select against diffuse ß-DG staining. The total pixel intensity for clustered ß-DG was acquired as 'circ spot total intensity ch2'. Compounds inducing an increase or decrease greater than 1.5 in clustered ß-DG staining were considered as active. Wells showing 3-fold or higher increase in clustered ß-DG staining over control were reexamined to eliminate any false positives resulting from precipitation of fluorescent compounds. Similarly, compounds causing a decrease in cell number wereconsidered as toxic chemicalsand disregarded. The Z-factor of the assay was determined from clustered ß-DG measured in cells treated with DMSO (negative control) or laminin (positive control). Effective drugs identified in the screening assay were further evaluated in a dose-response curve where DMSO concentration was kept at 0.1% in all wells for all experiments.

Cell viability assay
Cells were seeded in 96-well plates at 8000 cells per well and grown for 24 h prior to the treatment with the drugs for 4 h as described above. The drugs and media were then removed and cell viability was measured using the 3-[4,5-dimethylthiazol-2-yl]-2.5-diphenyl-tetrazolium bromide (MTT) assay (Sigma, USA). Cells were incubated for 2 h at 37uC with MTT, then 20% sodium dodecyl sulfate was added and the absorbance at 570 nm was measured after overnight incubation.

Gelatin zymography
Media from treated astrocytes were collected and 10 ml aliquots per well were applied to 10% SDS-PAGE gels co-polymerized with 1% gelatin (Fisher, USA). After electrophoresis under non-reducing conditions, the gels were washed in 2.5% Triton X-100 for 1 h to remove SDS and then incubated for 16 h at 37uC in 20 mM Tris, 150 mM NaCl, 5 mM CaCl 2 and stained with SimplyBlue TM Safestain (Invitrogen, USA). For the standards, we used the prestained protein ladder from Bio-Rad as well as purified gelatinase A (MMP-2) and B (MMP-9). Regions of gelatinolytic activity regions were observed as clear bands against a blue background.

Development of an automated microscopy screenfor chemical modulators of the laminin-induced dystroglycan clustering in primary astrocyte cultures
To screen libraries of compounds for modulators of the laminininduced clustering of ß-DG in primary astrocyte cultures, we used an automated microscopy assay. The high-content screening instrument programmed to detect and quantify fluorescent spots enabled the automatic imaging of ß-DG fluorescent clusters on a Cellomics ArrayScan V TI HCS Reader and analysis by the Compartmental Analysis BioApplication. Representative images obtained with this automated image analysis system are shown in Figure 1A in which each cellis outlined in red and nuclei of each live cell overlaid in blue. Individual clusters associated with each cell are shown in yellow (Fig. 1A). To determine the most suitable conditions for optimum detection of ß-DG clustering in astrocytes cultured in 96 well plates using the Cellomics reader, wetreated these cultures with increasing concentrations of laminin-1for 7 h. In the absence of laminin-1, ß-DG labelling was largely diffuse at the cell surface (Fig. 1A), however in the presence of laminina dose-dependent response both in the number and area of the clusters was observed reaching a plateau at 20 nM of laminin-1as previously reported in Muller glial cells (Fig. 1A-C; 21).
Evaluating assay performance. The values from untreated negative control and treated (20 nM laminin) positive control wells were used to assess the Z9 factor for the number of clusters and the total area covered by the clusters per cell. The values for the number of clusters (0.54) and the total area of clustering per cell (0.62), are greater than 0.5, indicating that thesecriteriaare appropriate for being used in this screening assay ( Table 1). As demonstrated by the automated assay, the addition of laminin-1 for 7 h caused a 9.11 to 18.35-fold increase in the number and area of ß-DG clusters, respectively (Table 1). Other criteria shown in the Table 1 include the assay's signal/noise and signal window, both of which are greater than 2-fold, further demonstrating the robustness of this assay [34].
Screening of compound collections. A collection of 3,584 drugs and pharmaceutically active chemicals was tested at a concentration of 15 mM for the final 4 h of the 7 h laminin incubation. In addition to automatically reporting the number of clusters and their total area, the number of cells (i.e., number of nuclei) and the nuclear area were simultaneously measured. Chemicals causing 80% reduction in cell number were considered cytotoxic and were eliminated from our analysis. Compounds causing over 1.6-fold increase or decrease in the clustering ofß-DG without causing cell death were considered as active (Fig. 2 C and D). Examples of these compounds are illustrated in Figure 2A and Bin which data compiled from three 96-well plates corresponding to approximately 200 compounds are shown. Twelve active chemicals were initially identified as inhibitors of the clustering,but subsequent testing by dose-response proved that only 6 of these were true inhibitors inducing a reduction of ß-DG clustering ranging from 2.2 to 17.1-fold ( Fig. 3 and data not shown). To our knowledge, none of these 6 compounds have been previously reported to reduce laminin-DG interaction or laminin-induced clustering of the DGC. Subsequent analysis has shown that two of the identified inhibitors, namely chloranil and flunarizine, induce a dose-dependent decrease in ß-DG clustering (Fig. 3B). To determine the most efficient concentration of chloranil and flunarizine required to induce a 50% inhibition (EC 50 ) of laminin-induced clustering, astrocytes were incubated with 20 nM laminin-1 and either chloranil of flunarizine at concentrations ranging from 1 to 100 mM (Fig. 3B). Thedata from thisdose-response experimentare represented as curves that were fitted using a sigmoidal function allowing the determination of the EC 50 . Figure 3B shows a dose-dependent inhibition of the laminininduced ß-DG clustering by both drugs, however chloranil presents an EC 50 (,20 mM) lower than that of flunarizine (,30 nM) indicating that chloranil is more potent than flunarizine.

Characterization of the effect of chloranil and flunarizine on ß-dystroglycan and AQP4 clustering
To assess whether AQP4 clustering is also reduced in the presence of chloranil and flunarizine, ß-DG and AQP4 clusters were examined at higher resolution by laser confocal microscopy. The data show that chloranilsignificantly inhibits the laminin-induced AQP4 clustering and as for ß-DG, this inhibition is dose-dependent (Fig. 4). Similar results were found for flunarizine (data not shown).
To ensure that the reduction in ß-DG and AQP4 clustering was not due to either cytotoxicity or reduction in ß-DG or AQP4 expression levels, we assessed cell viability using the MTT assay andß-DG and AQP4levels by immunoblotting. Figure 5 shows that while ß-DGexpression level remains unchanged in the presence of flunarizine that ofAQP4is slightly decreased. Interestingly, chloranil causes the cell surface cleavage of full length 43 kDa ß-DG and consequent formation of the 31 kDa fragment of ß-DG (Fig. 5A). The size of this fragment is consistent with the cleavage of the short extracellular domain of ß-DG that has been reported to be mediated by metalloproteinases. We next investigated the effectof increasing concentrations of chloranil and flunarizine on astrocyte survival. Four hours following the treatment with chloranil and flunarizine, we subjected the astrocytes to the MTT cell viability assay and found that chloranil caused a slight reduction in the total number of actrocytes but only at a very high concentration (100 mM; Fig. 5B). These results demonstrate that the inhibition of ß-DG and AQP4 clustering observed with ,100 mM chloranil is not due to adverse effects on cell survival (Fig. 3 and 4).

Chloranil-induced ß-dystroglycan shedding is mediated by metalloproteinases
Since the formation of the 31 kDa ß-DG fragment in chloraniltreated astrocytes is consistent with a proteolytic cleavage of ß-DG possibly implicating metalloproteinases, we tested whether longer exposure to chloranil results in increased accumulation of the 31 kDaß-DG. The data show a more intense 31 kDa band and weaker 43 kDa band at 7 h compared to 4 h post-treatment (Fig. 6). This time-dependent accumulation of the proteolytic 31 kDaß-DG fragmentisconsistent with proteolytic activity. In- creasing accumulation of the 31 kDa ß-DG fragment was also seen withincreased chloranil concentrationsranging from 15 mM to 50 mM(data not shown).
We next evaluated the role of metalloproteinases in the chloranil-induced ß-DG cleavage using the metalloproteinaseinhibitor, Prinomastat at 15 and 30 mM. Immunoblots of protein extracts of astrocytes co-incubated with 15 mM chloranil and15 or 30 mM prinomastat for 7 h show that prinomastat inhibits the chloranil-induced ß-DG cleavage partially when applied at 15 mM and completely at 30 mM (Fig. 6). These data show conclusively that the chloranil-induced cleavage of ß-DG is mediated by metalloproteinases.
Shedding of ß-DG at the cell surface has been studied in many cell types including tumor cells and glial cells. Several studies have reported that the two metalloproteinases, MMP-2 and MMP-9 areinvolved in this processin brain [35,36]. To determine whether chloranil-induced shedding of ß-DG in astrocytes is mediated by these twometalloproteinases, we investigated their activity via gelatin zymography. Culture media from untreated astrocytes and chloranil-treated astrocytes were collected and analyzed by gelatin zymography. As standards, we used a mix of gelatinase A (MMP-2) and B (MMP-9) from CHO cells or purified gelatinase A. The gelatinolytic regions observed with both untreated and chloraniltreated astrocytes' culture media correspond to MMP-9 and the inactive form of MMP-2, as revealed by the clear bands at ,92 and ,72 kDa, respectively (Fig. 7A). No difference in intensity or activation state of either one of these MMPs was observed between the untreated and chloranil-treated astrocytes (Fig. 7A). Further analysis was conducted using the tissue inhibitors of metalloproteinases, TIMP1 (110 nM), and TIMP2(100 mM) that inhibit selectively MMP-2 and MMP-9 but not the ADAMs (a disintegrin and a metalloproteinase) as well as TIMP3 (100 mM) that has a broader inhibition profile which includes MMPs, the tumor necrosis factor a-converting enzyme (TACE or ADAM17) and the related enzyme ADAM10 [37]. Astrocytes treated for 4 h with chloranil alone, chloranil with prinomastat or chloranil with TIMP1, TIMP2 or TIMP3 were analyzed by immunoblotting. Unlike prinomastat, TIMP1 did not inhibit the chloranil-induced shedding of ß-DG (Fig. 7B). However, both TIMP2 and TIMP3 were able to inhibit this effect (Fig. 7B).

ß-dystroglycan sheddingby chloranil and its chemical variants is mediated by reactive oxygen species
Chloranil is a general electron acceptor that has a role in radical ion formation and has been shown to be most effective at producingreactive oxygen species (ROS) by accepting electrons from oxygen [38]. Based on this property of chloranil, its reported inability to interact or regulate collagenase activity [39] and the fact that metalloproteinases can be activated by ROS, we hypothesized that choranil inducesß-DG shedding by producing ROS that in turn activate specific metalloproteinases. To test this hypothesis,astrocytes were incubated either with chloranil alone or chloranil plus the ROS scavenger, N-acetyl-cysteine (NAC). Interestingly, NAC completely inhibited the proteolytic cleavage of ß-DG by chloranil (Fig. 8A).
Similar to chloranil,chemical variants of chloranil such as 2,3dichloro-5,6-dimethyl (1,4) benzoquinoneand benzoquinone, induced ß-DG cell surface cleavage and a dose-dependent decrease both in the number and area of ß-DG clusters as assessedby immunoblotting (Fig. 8A) and immunofluorescence analyses (Fig. 8B). The effect of chloranil and its variants is most likely mediated by the benzoquinone which is common to these three aromatic compounds. The observation that these variants induce a similar effect to chloranil and that chloranil acts via the production of ROS suggest strongly that chloranil variants increase ROS production as well leading to the activation of metalloproteinases.

Discussion
Several studies have supported the role of the DGC and its interaction with extracellular laminin in the polarized distribution of AQP4 at astrocyte endfeet abutting blood vessels. Indeed, the severance of the DG interaction with laminin in the DGglycosylation deficient Large myd mouse, results in the loss of AQP4 localization at astrocyte endfeet without affecting its expressionlevel [25]. We have also shown that the laminininduced clustering of AQP4 in astrocyte cultures requires DG [27,28]. In addition, the deletion of a-syntrophin or mutations in the dystrophin gene result in themislocalization of AQP4, distributing it away from astrocyte endfeet [29,30,31]. It is noteworthy that AQP4 mislocalization in these mice results in a better outcome as it delays the onset of brain edema [32,40]. A study in the AQP4-null mouse reported brain swelling in models of vasogenic edema where excess fluid accumulates in the extracellular space, due to impairment of the AQP4-dependent water clearance [41]. On the contrary, another study in the AQP4 null mouse reported reduced brain swelling and improved neurological outcome following water intoxication and focal cerebral ischemia [4]. Thisdefines a role for AQP4 in the development of cytotoxic (cellular) cerebral edema where excess fluid accumulates mainly in the astrocyte foot processes [4]. Modulators of AQP4 function, expression or perivascularclusteringmay therefore have a therapeutic application to reduce cytotoxic edema by blocking water influx in astrocytes at early stages of stroke. Indeed, recent studies have identified potential inhibitors of AQP4-mediated watertran- sport in vitro [12,14,15], however the data by Huber et al [12,14] have been disputed [18].
In the present study we implemented a high-throughput screen in primary astrocyte cultures toidentifycompounds that modulate the laminin-induced clustering of ß-DG and AQP4. First, we showed that the CellomicsTMArrayscan VTI automated fluorescence imaging is suitable for efficient detection of the laminin-induced clustering of ß-DG (Fig. 1). Indeed, the Z9 factor (0.54),signal/background,signal/ noise and signal window valuesdemonstrate that the number and area of the clusters are suitable parameters for the Cellomics-based screening ( Table 1). Screening of the Pretswick, Sigma LOPAC, Microsource Spectrum and Biomol natural products collections identified a number of chemically active molecules, 6 of which inhibited ß-DG clustering. Here, we focused our study onflunarizine  and chloranil, two inhibitors of the clustering, and show based on the EC 50 that chloranil (EC 50 ,20 mM) is more potent than flunarizine (EC 50 ,30 mM). The effect of these drugs on ß-DG clustering was reversible (data not shown) and neither of them compromised astrocyte viability at the concentration used throughout the experiments. These are generally considered favourable properties for drugs with a therapeutic potential.
However, studies have reported that chloranil is able to generate DNA single-strand brakes in V79 chinese Hamster cells [42] and various levels of toxicity depending on the cell type [38,43]. Indeed, amongst several quinones tested, chloranil was the only one to induce a weak toxicity in platelets [43] and was less toxic in PC12 cells than in primary hepatocytes [38]. We noted minimal toxicity on primary astrocytes at doses that induced a significant inhibition on ß-DG and AQP4 clustering, suggesting that the degree of chloranil toxicity is cell type-dependent.
Flunarizine plays the role of a calcium channel blocker in astrocytes [44,45] and protects rats from severe ischemic cell death [46]. Its role on brain edema has been investigated and despite a remarkable recovery of the electroencephalographic activity in a rat model of cerebral ischemia [47] and an increased recovery rate of the electrical potential in a cat model of spinal contusion injury [48], flunarizine did not induce a significant inhibition of brain swelling. Furthermore, flunarizine has proven to be effective in preventing ischemia in subarachnoid hemmorrhage, although its effect on cerebral infarction has not been proven yet [49]. Further in vivo experiments aiming at investigating the role of flunarizine in models of cytotoxic brain edema addressing particularly its effect on DG and AQP4 perivascular distribution are warranted.
Chloranil is commonly used as an oxidizing agent in organic syntheses. Here we show that it induces the shedding ofß-DG ectodomain resulting in the formation of the 31 kDa ß-DG fragment and a similar effect isseenwith other oxidizing agents such as 2,3-dichloro-5,6-dimethyl-(1,4)benzoquinoneand benzoquinone, suggesting that the effect of chloranil and its variants on ß-DG is due toROS production. This is further substantiated by the observation that chloranil-induced ß-DG shedding is prevented by the anti-oxidant, NAC (Fig. 8). ß-DG shedding is typical of a metalloproteinase-mediated cell surface cleavage of ß-DG as previously reported [36,50,51]. Indeed, our data show that prinomastat, an inhibitor with selectivity for MMP-2, 9, 13, and 14,completely blocks the chloranil-induced cell surface shedding of ß-DG (Fig. 7B). These findings together with the established role of ROS in MMP activation [52,53,54] indicate that chloranil induces the production of ROS that in turn activates metalloproteinases leading to the cleavage of ß-DG ectodomain. This is in accordance with the report that cerebral ischemia seen after oxygen-glucose deprivation leads to a metalloproteolytic cleavage of ß-DG in astrocytes [55] and that prinomastat protects against ischemic brain injury as shown in hippocampal organotypic slices [56].
A number of studies have shown that ß-DG is a substrate for MMP-2 and 9 but not MMP-3, 8 or 13 [36,50,57,58,59]. Our data showing a lack of inhibition of the chloranil effect by TIMP1 argue against an MMP-9-mediated ß-DG shedding (Fig. 7B). However, the fact that both TIMP2 and 3 inhibit the chloranilinduced ß-DG shedding indicate that MMP-2 is one of the metalloproteinases targeted by chloranil. Although TIMP3 has the ability to inhibit not only gelatinases including MMP-2 but also the tumor necrosis factor-a (TNF-a)-containing enzyme (TACE or ADAM17) and the related enzyme ADAM10 [60], ADAM17 and 10 are unlikely targets of chloranil since prinomastat, which is a potent inhibitor of MMP-2 [61] but not ADAMs can completely inhibit the chloranil-induced ß-DG shedding. This, together with the TIMP2 inhibition, supports further the implication of MMP-2 in the chloranil-induced effect. It is noteworthy that the pro-MMP-2 (inactive MMP2; Fig. 7A) but not the active MMP-2 is detected in the culture medium harvested from astrocytes treated with chloranil, as reflected by the gelatine zymography assay. It is therefore possible that the activation of MMP-2 by chloranil occurs at or near the cell membrane where membrane-type MMPs (MT-MMPs) play a role in tethering MMP activity [62], which would explain the absence of active MMP2 in the culture medium. This suggests that TIMP2 and TIMP3 inhibition of MMP-2 activation by chloranil occurs at the cell surface. Indeed,membrane-bound pro-MMP2 processing is inhibited in the presence of high concentrations of TIMP2 [63] and TIMP3 [64].
In light of the fact that ROS-induced activation of MMPs is involved in blood brain barrier (BBB) disruption leading to vasogenic edema after stroke [65], the use of chloranil would be deleterious as it would result in the breakdown of the BBB. While the detailed analysis of chloranil allows its elimination as a potential drug that could be used in models of brain edema, several other drugs that inhibit both ß-DG and AQP4 clustering in a metalloproteinase-independent manner have been identified in the present study using the high throughput screen on primary astrocyte cultures. Further investigations will determine the effectiveness of these drugs in reducing the clustering of ß-DG and AQP4 at perivascular astrocyte endfeet in vivo and in preventing cytotoxic brain edema.