Suppression of SNARE‐dependent exocytosis in retinal glial cells and its effect on ischemia‐induced neurodegeneration

Abstract Nervous tissue is characterized by a tight structural association between glial cells and neurons. It is well known that glial cells support neuronal functions, but their role under pathologic conditions is less well understood. Here, we addressed this question in vivo using an experimental model of retinal ischemia and transgenic mice for glia‐specific inhibition of soluble N‐ethylmaleimide‐sensitive factor attachment protein receptor (SNARE)‐dependent exocytosis. Transgene expression reduced glutamate, but not ATP release from single Müller cells, impaired glial volume regulation under normal conditions and reduced neuronal dysfunction and death in the inner retina during the early stages of ischemia. Our study reveals that the SNARE‐dependent exocytosis in glial cells contributes to neurotoxicity during ischemia in vivo and suggests glial exocytosis as a target for therapeutic approaches.

Studies within the last decade have shown that glial cells can release transmitters (which are called gliotransmitters) in a calcium-and soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE)-dependent manner (Marchaland et al., 2008;Martineau, 2013;Montana, Malarkey, Verderio, Matteoli, & Parpura, 2006;Slezak et al., 2012). The physiologic relevance of this process is debated: it may contribute to glial volume regulation  and modulate neuronal activity Chen et al., 2012;Perez-Alvarez, Navarrete, Covelo, Martin, & Araque, 2014;Takata et al., 2011). Here, we studied the contribution of glial SNARE-dependent exocytosis to neurodegeneration in the retina. To this end, we used an established model of transient ischemia that triggers robust neuronal dysfunction and degeneration (Osborne et al., 2004;Pannicke et al., 2014). To interfere with glial exocytosis, we used a mouse line enabling inducible expression of a dominant-negative domain of synaptobrevin (dnSNARE mice; Pascual et al., 2005). Our results reveal that gliaspecific inhibition of SNARE-dependent signaling protects retinal neurons from dysfunction and degeneration.

| Animals
All experiments were performed in accordance with the European Community Council Directive 2010/63/EU after approval by the local authorities (TVV 54/12 approved by the state directorate Leipzig). Animals were maintained with free access to water and food in an airconditioned room on a 12-hr light-dark cycle. Transgenic dnSNARE mice were bred as described earlier (Pascual et al., 2005). These mice use the Tet-Off system to express a dominant negative subdomain of synaptobrevin (VAMP2) and enhanced green fluorescent protein (EGFP) under the control of a human GFAP promoter fragment. To prevent transgene expression during development, the animals received doxycycline until weaning (25 lg/ml in drinking water).

| Measurement of transmitter release in single M€ uller cells
Transmitter release from acutely isolated M€ uller cells was measured using a fluorimetric enzyme assay as described .
Briefly, retinae were treated with papain (0.2 mg/ml; Roche Molecular Biochemicals) for 30 min at 378C in the dark in Ca 21 -and Mg 21 -free extracellular solution (140 mM NaCl, 3 mM KCl, 10 mM HEPES, 11 mM glucose, pH 7.4). Before detecting glutamate release, glutamine (0.25 mM) and glutamate (0.5 mM) were supplemented to maintain high intracellular glutamate levels. After several washes, retinae were triturated in extracellular solution (with 1 mM MgCl 2 and 2 mM CaCl 2 ) containing all components of the Amplex® Red Glutamic Acid kit, D,Lthreo-beta-benzyloxyaspartate (TBOA, 200 mM) to block glial glutamate uptake and NP-EGTA (10 mM) or NPE-ATP at varying concentrations. Subsequently, the cell suspension was mixed with 1.5% agarose and incubated for 30 min in the recording chamber at 378C. At each recording session, we measured cells from one wild type mouse and one dnSNARE mouse using the same assay solution freshly prepared for the actual experiment. This rules out confounding effects of differences in the assay composition. If blockers were tested, they were coapplied with the gel-enzyme-mixture and also preincubated for 30 min. were taken from the endfeet of M€ uller cells, which were identified based on their unique morphology. To detect ATP release, glutamatespecific enzymes (L-glutamate oxidase, L-glutamate-pyruvate transaminase), TBOA, and alanine were replaced by glycerol kinase (10 U/ml), glycerol-3-phosphate oxidase (10 U/ml) and glycerol (500 mM) from the Amplex® Red assay. The combination of enzymes and substrates generates H 2 O 2 (and finally resorufin) only in the presence of ATP (Dale & Frenguelli, 2012).
To trigger calcium-dependent release, intracellular calcium rises were induced by four UV-pulses (351 nm/364 nm Enterprise UV Laser, 500 ms at maximal intensity) to uncage calcium from NP-EGTA or to release ATP from NPE-ATP. Peak amplitudes were calculated as difference between mean fluorescence intensity across four time points acquired before and after the UV pulses. To validate concentration-dependent detection of glutamate or ATP, we performed the enzymatic assay under cell-free conditions. To detect glutamate, extracellular solution with all enzymatic components of the Amplex® Red Glutamic Acid kit, TBOA (200 mM), and 1.5% agarose but without cells was filled into the recording chamber. Extracellular solution containing defined concentrations of glutamate was added to the recording chamber. Similarly, we validated the ATP detection assay using a similar procedure, but implementing glycerol kinase, glycerol-3phosphate oxidase and glycerol. We recorded resorufin fluorescence under cell-free conditions on a microplate reader (HTS7000 Bio Assay Reader Perkin-Elmer, Rodgau J€ ugesheim, Germany).

| Calcium measurements on retinal whole-mounts
Retinae were prepared from male mice (P23-P90, C57BL/6) and transferred onto a net-insert in a microdish filled with 1 ml artificial cerebrospinal fluid (ACSF) containing in mM: 125 NaCl, 26 NaHCO 3 , 1.25 NaH 2 PO 4 , 20 glucose, 2.5 KCl, 1 MgCl 2 , and 2 CaCl 2 . The solution was aerated with carbogen (95%O 2 , 5% CO 2 ). Whole-mounts were incubated in the dark for 1 hr at room temperature (228C) with the fluorescent calcium indicator Fluo-4 AM (15 mM, Invitrogen, Carlsbad, CA) added to the ACSF to selectively load M€ uller cells. VEGF (30 ng/ml) was bath-applied in a closed perfusion circuit with a total volume of 14 ml. For calcium imaging the whole-mount was transferred to a submerged chamber perfused with ACSF and held in place by a nylon grid with the ganglion cell layer (GCL) facing toward the water immersion objective (NA 1.0; Nikon Instruments, Tokyo, Japan). Retinal cells were visualized by gradient contrast using a Femto-2D-uncage microscope (Femtonics, Budapest, Hungary) controlled by MES v4.5.613 software (Femtonics). A tunable, Verdi-pumped Ti:Sa laser (Chameleon Ultra, Coherent, Santa Clara, CA) was used for excitation of Fluo-4 AM at 840 nm. Green fluorescence images were collected both in the epiand transfluorescence mode. M€ uller cell endfeet (MCE) were imaged in frame scan mode for 2 min (rate: 20 Hz; size: 40 3 83 mm corresponding to 9 endfeet; image resolution: 1.6 mm per pixel. For the quantitative analysis, endfeet of M€ uller cells were marked as region of interest (ROI). Average or integral of pixels in these ROIs were averaged and changes in calcium were measured as DF/F and compared between different conditions. Control experiments included consecutive frame scans without drug application. Offline data analysis was performed using custom macros written in IGOR Pro (Wavemetrics, Lake Oswego, OR) and MES (Femtonics) and custom MATLAB (MathWorks, Ismaning, Germany) scripts.

| Retinal ischemia
The high intraocular pressure (HIOP) method was used to induce transient ischemia in the retina as reported (Pannicke et al., 2014). Briefly, animals were anesthetized by intraperitoneal (ip) injections of ketamine (100 mg/kg body weight; Ratiopharm, Ulm, Germany), xylazine (5 mg/ kg; Bayer Vital, Leverkusen, Germany), and atropine sulfate (100 mg/ kg; Braun, Melsungen; Germany). The anterior chamber of one eye was cannulated from the pars plana by a 30-gauge infusion needle connected to Deltajonin® (Deltaselect, Dreieich, Germany) bottle. The contralateral eye remained untreated and served as control. To interrupt retinal blood supply, the intraocular pressure was raised to 160 mm Hg for 90 min by elevating the bottle. Animals were sacrificed 14 hr, 1 or 7 days postoperation with carbon dioxide.

| Volume regulation
Volume changes in retinal M€ uller cells were measured as described . Briefly, retinal slices were loaded with the vital dye Mitotracker Orange (10 mM, excitation: 543 nm, emission: 560 nm long-pass filter; Life Technologies), which is preferentially taken up by M€ uller cells (Uckermann et al., 2004). Slices were exposed to hypotonic solution (60% of control osmolarity using distilled water) for 4 min with or without test substances. Somata of labelled M€ uller cells were imaged using confocal microscopy (LSM 510 Meta, Zeiss, Oberkochen, Germany) and their cross-sectional areas were measured (Zeiss LSM Image Examiner Version 3.2.0.70). In dnSNARE mice, EGFP1 and EGFP-cells were imaged and analyzed separately (excitation: 488 nm; emission: 505 nm long-pass filter).

| Histological and immunohistochemical staining
Retinae were immersion-fixed (4% paraformaldehyde for 2 hr), washed with phosphate-buffered saline (PBS), embedded in PBS containing 3% agarose (w/v) and cut in 70 mm thick sections using a vibratome. Retinal sections were permeabilized (0.3% Triton X-100 plus 1.0% DMSO in PBS) and blocked (5% normal goat serum with 0.3% Triton X-100 and 1.0% DMSO in PBS) for 2 hr at room temperature. Primary antibodies were incubated overnight at 48C. Sections were washed (1% bovine serum albumin in PBS) and incubated with secondary antibodies (2 hr at room temperature). Cell nuclei were labeled with TO-PRO-3 (1:1000; Life Technologies). Retinal whole-mounts were labeled using a similar protocol, except that tissue was permeabilized by higher concentrations of Triton X-100 and DMSO (0.3% Triton X-100 plus 1.0% DMSO in PBS) and secondary antibodies were also incubated at 48C overnight. Control experiments without primary antibodies showed no unspecific labeling except for the goat-anti-mouse secondary antibody which labeled blood vessels (not shown). Images were acquired using confocal microscopy (LSM 510 Meta, Zeiss). Cell nuclei were counted in three retinal layers in 100 mm-wide areas of the central retina close to the optic nerve head (optical slice thickness, 1.5 mm). The TUNEL assay was performed according to the manufacturer's protocol (Roche Molecular Biochemicals). Free-floating retinal sections were subjected to a brief microwave treatment in citrate buffer (pH 6.0, 0.1 M) to enhance tissue penetration. After several washing steps, sections were permeabilized (10 min in 4% Triton X-100), incubated in the labeling solution (90 min), counterstained with TO-PRO-3 and mounted on glass slides.

| qRT-PCR
Total RNA was isolated from whole retinae and from enriched cell populations using the RNeasy Micro Kit (Qiagen, Hilden, Germany). A DNase digestion step was included to remove genomic DNA (Roche).
First-strand cDNAs from 50 ng of total RNA were synthesized using the RevertAid H Minus First-Strand cDNA Synthesis Kit (Fermentas by Thermo Fisher Scientific, Schwerte, Germany). Primers were desgined using the Universal ProbeLibrary Assay Design Center (Roche). The primers to detect dnSNARE expression covered partially the sequence of the dnSNARE domain and the SV40 polyA tail of the expression construct. Transcript levels of candidate genes were measured by qRT-PCR using cDNA with the TaqMan hPSC Scorecard TM Panel (384 well, ViiA7, Life Technologies, Darmstadt, Germany) according to the company's guidelines and its cloud-based online analysis software.

| Full-field electroretinography
Mice were dark adapted for at least 12 hr before recordings and anesthetized by subcutaneous injection of ketamine (65 mg/kg) and xylazine (13 mg/kg). Pupils were dilated with tropicamide eyedrops (Mydriaticum Stulln; Pharma Stulln). Silver needle electrodes served as reference (forehead) and ground (tail) and gold wire ring electrodes as active electrodes. Corneregel (Bausch & Lomb, Berlin, Germany) was applied to keep the eye hydrated and to maintain good electrical contact. ERGs were recorded using a Ganzfeld bowl (Ganzfeld QC450 SCX, Roland Consult, Brandenburg, Germany) and an amplifier with a recording unit (RETI-Port, Roland Consult). ERGs were recorded from both eyes simultaneously, band-pass filtered (1-300 Hz) and averaged.
Single flash scotopic (dark adapted) responses to a series of ten LEDflash intensities ranging from 23.5 to 1.0 log cd.s/m 2 with an interstimulus interval of 2 up to 20 s for the highest intensity were recorded.
After 10 min of adaptation to a white background illumination (20 cd/ m 2 ) single flash photopic (light adapted) responses to three Xenon-flash intensities (1, 2, and 3 log cd.s/m 2 ) were recorded. Responses were quantified based on mean waveform peak amplitude and implicit time.
All analysis and plotting was carried out with R 3.2.1 (The R Foundation for Statistical Computing) and ggplot2 2.1.0.

| Statistics
All data are expressed as mean 6 standard error (SEM) unless stated otherwise. Statistical analyses were performed using Prism (Graphpad Software, San Diego, CA). Unless stated otherwise the significance was determined by the nonparametric Mann-Whitney U test.

| Validation of glia-specific dnSNARE expression and its impact on transmitter release and volume regulation in M€ uller cells
We first examined the expression of the dnSNARE transgene and its impact on M€ uller cells using different approaches. Immunohistochemical staining of retinal sections and whole-mounts from transgenic mice revealed expression of EGFP in M€ uller cells and astrocytes, but not in neurons or microglial cells (Figure 1a,b). We investigated whether dnSNARE expression impacted the calcium-induced release of the gliotransmitters glutamate and ATP in single acutely isolated M€ uller cells using enzymatic assays . UV photolysis-induced calcium transients evoked the release of glutamate from M€ uller cells isolated from wildtype mice after loading with the photolabile chelator NP-EGTA. In M€ uller cells from transgenic mice, the release was significantly reduced (Figure 2a Since photolysis of NP-EGTA may induce nonphysiological levels of intracellular calcium, we tested next, whether extracellular ATP triggers purinergic receptor-dependent glutamate release from M€ uller glia as observed previously in astrocytes (Domercq et al., 2006;Jourdain et al., 2007). Photolysis of caged ATP caused a P2Y 1 receptordependent release of glutamate from M€ uller cells: it was attenuated by the P2Y 1 antagonist MRS2179 (30 mM) and it was reduced in M€ uller cells isolated from P2Y 1 KO mice (Figure 2e,f). M€ uller cells from dnSNARE mice showed reduced P2Y-dependent glutamate release but calcium transients similar to those from wildtype controls (Figures 2g and 6).
Previous findings showing SNARE-dependent ATP release from astrocytes prompted us to establish a modified enzymatic assay to detect ATP release from M€ uller cells (Figure 3a). UV light induced equal ATP signals in cells from wildtype and dnSNARE mice whereas removal of the respective enzymes prevented the ATP signals (Figure 3b,c).
Addition of the connexin hemichannel blockers, 18-a-glycyrrhetinic acid (50 mM) and carbenoxolone (CBX) (200 mM) significantly reduced the ATP signals. The data suggest that ATP is released from M€ uller cells in a SNARE-independent manner, probably via hemichannels.
Next, we tested how the dnSNARE transgene affected volume regulation in M€ uller cells. Previous studies revealed that this process relies on a VEGF-triggered signaling pathway comprising calciumdependent glutamate release, metabotropic glutamate receptors and the nonvesicular release of ATP and adenosine (Br€ uckner et al., 2012;Slezak et al., 2012;Wurm, Pannicke, Wiedemann, Reichenbach, & Bringmann, 2008). As shown in Figure 4a Taken together, these results validated that dnSNARE expression in the retina was specific to M€ uller cells, that it inhibited a fraction of their calcium-dependent glutamate release but not of their calciumdependent ATP release, and that it impaired their volume regulation.

| Ischemia-induced changes in retinal M€ uller cells
To determine whether and how calcium-dependent release from M€ uller cells contributes to pathologic changes ensuing transient Next, we analyzed whether and how ischemia affected the ATPinduced glutamate release from M€ uller cells. In wildtype animals, the glial glutamate release was only transiently reduced one day after the insult and returned to a normal level at 7 d ( Figure 6). In mice bearing the dnSNARE transgene, glial ATP-induced glutamate release was constantly lower compared with wildtype mice regardless of ischemia (Figure 6) except at 1 d after ischemia. The ischemic insult did not affect the ATP-induced calcium response in M€ uller cells from wildtype or transgenic mice (Figure 6).

| Inhibition of glial SNARE-dependent release reduces postischemic loss of neurons
Next, we studied how dnSNARE expression influenced pathologic changes in postischemic retinae. A cardinal feature of the ischemia/ reperfusion model is the robust loss of neurons as shown by the reduction of rhodopsin (Figure 5b) and the thinning of retinal layers (Figure 7a,   b). While all morphological parameters (e.g., cell counts and IPL thickness) investigated were not significantly different in untreated retinae of wild type and dnSNARE mice (data not shown), our quantitative analysis revealed that dnSNARE expression reduced the ischemia-induced loss of cells in the nuclear layers and diminished the thinning of the inner plexiform layer (IPL) at different time points after surgery compared with wildtype animals (Figure 7a,b). These findings were confirmed by TUNEL staining of apoptotic cells. One day after surgery, when the neuronal cell loss peaks (Kuroiwa et al., 1998)

| DISCUSSION
In this study, we used a transgenic mouse model to investigate the impact of glial SNARE-dependent exocytosis on postischemic neurodegeneration in the retina in vivo. We show that glia-specific inhibition of SNARE-dependent signaling by dnSNARE expression reduces exocytotic release of glutamate from M€ uller cells, impairs glial volume regulation and protects retinal neurons from postischemic dysfunction and degeneration, whereas ATP release by M€ uller cells remained unaltered.

| Glia-specific dnSNARE expression inhibits exocytotic transmitter release and impairs volume regulation
Our findings validate dnSNARE mice as a tool to interfere with SNARE-dependent processes in retinal glial cells. Immunohistochemical staining of retinal sections ( Figure 1) and transcript profiling revealed that the transgene is exclusively present in glial cells (Figure 5b). The human GFAP promoter fragment driving transgene expression faithfully recapitulated the activity of the endogenous GFAP promoter.
Notably, both promoters were activated by ischemia and caused a parallel increase of transcripts encoding GFAP and dnSNARE/EGFP (Figure 5b). Neuronal transgene expression previously reported in cortex and hippocampus of dnSNARE mice (Fujita et al., 2014) was not detectable in the retina indicating regional differences in the targeting efficacy of transgenic lines .
Expression of the dnSNARE transgene reduced the amount of glutamate release from M€ uller cells regardless of its induction by calcium (36% less release) or by ATP (48% reduction) (Figure 2a,f). These results are similar to what was observed previously after glial expression of Botulinum neurotoxin serotype B (38% less) or pharmacological blockade of vesicle loading with glutamate by application of bafilomycin A1 (reduced by 31%; Slezak et al., 2012). In the end, a very consistent reduction of glutamate release was measured in the three independent experimental settings. This suggests that all approaches (a) were similarly effective (quite likely close to 100% given the high consistency of the results), (b) could hence serve as valuable tools to analyze the impact of an absent glial exocytotic release. Moreover, they corroborate the presence of glial glutamate release that is SNARE-independent and mediated by alternative pathways (Hamilton & Attwell, 2010;Montero & Orellana, 2015;Woo et al., 2012).
In concordance with Slezak et al. (2012), we also found that interference with SNARE-dependent signaling impaired the ability of M€ uller cells to maintain their volume when challenged by hypotonic solution ( Figure   4). Glutamate has been demonstrated to evoke release of ATP from M€ uller cells within the volume regulatory signaling cascade (Wurm et al., 2008). Therefore, the block of swelling of EGFP-positive cells by glutamate ( Figure 4a) suggests that the release of ATP is not affected in dnSNARE mice. This is in agreement with our direct measurements of ATP release by single M€ uller cells (Figure 3). The defect in volume regulation was absent in M€ uller cells that did not express the dnSNARE and the EGFP transgenes. This particular pool of cells may prevent major morphological or functional alterations in retinae from transgenic mice. In particular, their presence may explain why light-evoked responses in ERGs appear fairly normal in dnSNARE mice as well as in iBot mice (Slezak et al., 2012) compared with wildtype animals. Taken together, two independent genetic approaches to inhibit SNARE-dependent processes in glial cells provide converging results, a partial inhibition of calcium-dependent glutamate release from individual cells and impaired volume regulation.  (1) from mice with indicated genotypes (n 5 5-12 cells from 2 mice per group). **p < .01; *p < .05 and inner retinal cell types (b-wave) following ischemia. Obviously, the functional integrity of retinal cells is highly sensitive to ischemic damage and impaired long before cell loss. Our observation that transgene expression protected neurons in the inner retina but not photoreceptors underline previous observations that among the retinal neurons, ganglion and amacrine cells are most susceptible to glutamate-induced hyperexcitation in the adult stage (Adachi et al., 1998;Lam, Abler, & Tso, 1999). Under normal conditions, M€ uller cells take up extracellular glutamate by dedicated transporters such as the glutamate-aspartate transporter (Bringmann et al., 2013;Rauen, 2000), which depend on a negative membrane potential of M€ uller glia (Brew & Attwell, 1987;Bringmann et al., 2013;Sarantis & Attwell, 1990). Depolarization of M€ uller cell membranes by high extracellular potassium concentrations in ischemic tissue impairs glutamate uptake (Billups & Attwell, 1996;Maguire et al., 1998;Szatkowski, Barbour, & Attwell, 1990) ultimately increasing extracellular glutamate levels. Cells highly susceptible to glutamate-induced neurotoxicity die and thereby release large amounts of ATP, which in turn aggravates the disastrous rise of extracellular glutamate by triggering its release from M€ uller glia. This process is probably interrupted by reduced glutamate release from M€ uller cells expressing the dnSNARE transgene. It has to be considered that release of glutamate from glial cells is small compared with release from presynaptic terminals. This is probably one of the reasons why blockade of SNARE-dependent processes in glial cells cannot prevent ultimate neuronal death in our ischemic model. Nevertheless, we observe a transient amelioration in neuronal function and a delayed loss of specific cells in dnSNARE mice revealing a contribution of SNARE-dependent processes in glial cells to postischemic neuronal dysfunction and degeneration. At present, we cannot exclude that SNARE-dependent processes in glial cells other than glutamate release have contributed to reduced neurodegeneration. We have demonstrated that ATP release by M€ uller cells is unaltered by dnSNARE expression, however, a potential release of other gliotransmitters, for example, D-serine, remains to be investigated.